diff --git "a/checkpoint-63330/trainer_state.json" "b/checkpoint-63330/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-63330/trainer_state.json"
@@ -0,0 +1,380086 @@
+{
+  "best_metric": 1.7745146751403809,
+  "best_model_checkpoint": "bird_sounds_classification_3/checkpoint-63330",
+  "epoch": 10.0,
+  "global_step": 63330,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-08,
+      "loss": 4.1956,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2000000000000002e-07,
+      "loss": 4.2052,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-07,
+      "loss": 4.1917,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4000000000000003e-07,
+      "loss": 4.2166,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 4.2089,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6e-07,
+      "loss": 4.2651,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2000000000000006e-07,
+      "loss": 4.2202,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.800000000000001e-07,
+      "loss": 4.1985,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.4e-07,
+      "loss": 4.1721,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 4.2011,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.6e-07,
+      "loss": 4.2079,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2e-07,
+      "loss": 4.1908,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.8e-07,
+      "loss": 4.2088,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 4.2098,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-07,
+      "loss": 4.213,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.600000000000001e-07,
+      "loss": 4.1878,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0200000000000002e-06,
+      "loss": 4.2019,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.08e-06,
+      "loss": 4.1992,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.14e-06,
+      "loss": 4.2035,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 4.1393,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.26e-06,
+      "loss": 4.227,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.32e-06,
+      "loss": 4.1888,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3800000000000001e-06,
+      "loss": 4.2055,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.44e-06,
+      "loss": 4.2111,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-06,
+      "loss": 4.2014,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.56e-06,
+      "loss": 4.2023,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6200000000000002e-06,
+      "loss": 4.1937,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6800000000000002e-06,
+      "loss": 4.2173,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7399999999999999e-06,
+      "loss": 4.2096,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8e-06,
+      "loss": 4.1926,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.86e-06,
+      "loss": 4.1967,
+      "step": 31
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9200000000000003e-06,
+      "loss": 4.1975,
+      "step": 32
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.98e-06,
+      "loss": 4.2035,
+      "step": 33
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0400000000000004e-06,
+      "loss": 4.2068,
+      "step": 34
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1e-06,
+      "loss": 4.1873,
+      "step": 35
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.16e-06,
+      "loss": 4.2326,
+      "step": 36
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.22e-06,
+      "loss": 4.1942,
+      "step": 37
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.28e-06,
+      "loss": 4.2021,
+      "step": 38
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.34e-06,
+      "loss": 4.1926,
+      "step": 39
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 4.2094,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4599999999999997e-06,
+      "loss": 4.1761,
+      "step": 41
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.52e-06,
+      "loss": 4.2041,
+      "step": 42
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.58e-06,
+      "loss": 4.2202,
+      "step": 43
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.64e-06,
+      "loss": 4.2019,
+      "step": 44
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7e-06,
+      "loss": 4.1926,
+      "step": 45
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7600000000000003e-06,
+      "loss": 4.1422,
+      "step": 46
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.82e-06,
+      "loss": 4.2075,
+      "step": 47
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.88e-06,
+      "loss": 4.212,
+      "step": 48
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9400000000000002e-06,
+      "loss": 4.1846,
+      "step": 49
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-06,
+      "loss": 4.1998,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9999525916561314e-06,
+      "loss": 4.2065,
+      "step": 51
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999905183312263e-06,
+      "loss": 4.2078,
+      "step": 52
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9998577749683943e-06,
+      "loss": 4.1951,
+      "step": 53
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999810366624526e-06,
+      "loss": 4.1921,
+      "step": 54
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9997629582806577e-06,
+      "loss": 4.1951,
+      "step": 55
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999715549936789e-06,
+      "loss": 4.2243,
+      "step": 56
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9996681415929203e-06,
+      "loss": 4.1948,
+      "step": 57
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999620733249052e-06,
+      "loss": 4.2348,
+      "step": 58
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9995733249051833e-06,
+      "loss": 4.2033,
+      "step": 59
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999525916561315e-06,
+      "loss": 4.1976,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9994785082174463e-06,
+      "loss": 4.192,
+      "step": 61
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999431099873578e-06,
+      "loss": 4.2169,
+      "step": 62
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9993836915297093e-06,
+      "loss": 4.2185,
+      "step": 63
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9993362831858405e-06,
+      "loss": 4.1971,
+      "step": 64
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9992888748419722e-06,
+      "loss": 4.1976,
+      "step": 65
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9992414664981035e-06,
+      "loss": 4.1919,
+      "step": 66
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9991940581542352e-06,
+      "loss": 4.1972,
+      "step": 67
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999146649810367e-06,
+      "loss": 4.3965,
+      "step": 68
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9990992414664982e-06,
+      "loss": 4.2029,
+      "step": 69
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9990518331226295e-06,
+      "loss": 4.1892,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.999004424778761e-06,
+      "loss": 4.1912,
+      "step": 71
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9989570164348925e-06,
+      "loss": 4.1901,
+      "step": 72
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.998909608091024e-06,
+      "loss": 4.2003,
+      "step": 73
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.998862199747156e-06,
+      "loss": 4.1859,
+      "step": 74
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.998814791403287e-06,
+      "loss": 4.226,
+      "step": 75
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9987673830594184e-06,
+      "loss": 4.2104,
+      "step": 76
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.99871997471555e-06,
+      "loss": 4.1974,
+      "step": 77
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9986725663716814e-06,
+      "loss": 4.1959,
+      "step": 78
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9986251580278127e-06,
+      "loss": 4.1953,
+      "step": 79
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9985777496839444e-06,
+      "loss": 4.223,
+      "step": 80
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.998530341340076e-06,
+      "loss": 4.2155,
+      "step": 81
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9984829329962074e-06,
+      "loss": 4.2109,
+      "step": 82
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9984355246523387e-06,
+      "loss": 4.1754,
+      "step": 83
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9983881163084704e-06,
+      "loss": 4.216,
+      "step": 84
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9983407079646017e-06,
+      "loss": 4.2024,
+      "step": 85
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9982932996207334e-06,
+      "loss": 4.2069,
+      "step": 86
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.998245891276865e-06,
+      "loss": 4.1907,
+      "step": 87
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9981984829329964e-06,
+      "loss": 4.2044,
+      "step": 88
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9981510745891276e-06,
+      "loss": 4.2152,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9981036662452593e-06,
+      "loss": 4.2351,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9980562579013906e-06,
+      "loss": 4.2041,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.998008849557522e-06,
+      "loss": 4.2445,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9979614412136536e-06,
+      "loss": 4.2041,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9979140328697853e-06,
+      "loss": 4.2063,
+      "step": 94
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9978666245259166e-06,
+      "loss": 4.1789,
+      "step": 95
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9978192161820483e-06,
+      "loss": 4.1999,
+      "step": 96
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9977718078381796e-06,
+      "loss": 4.1982,
+      "step": 97
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.997724399494311e-06,
+      "loss": 4.1986,
+      "step": 98
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9976769911504426e-06,
+      "loss": 4.2001,
+      "step": 99
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9976295828065743e-06,
+      "loss": 4.1914,
+      "step": 100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9975821744627055e-06,
+      "loss": 4.1988,
+      "step": 101
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9975347661188372e-06,
+      "loss": 4.1896,
+      "step": 102
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9974873577749685e-06,
+      "loss": 4.2081,
+      "step": 103
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9974399494311e-06,
+      "loss": 4.2013,
+      "step": 104
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.997392541087231e-06,
+      "loss": 4.1814,
+      "step": 105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.997345132743363e-06,
+      "loss": 4.2121,
+      "step": 106
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9972977243994945e-06,
+      "loss": 4.2085,
+      "step": 107
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9972503160556258e-06,
+      "loss": 4.1984,
+      "step": 108
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9972029077117575e-06,
+      "loss": 4.1793,
+      "step": 109
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9971554993678888e-06,
+      "loss": 4.2268,
+      "step": 110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.99710809102402e-06,
+      "loss": 4.2066,
+      "step": 111
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9970606826801517e-06,
+      "loss": 4.2033,
+      "step": 112
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9970132743362834e-06,
+      "loss": 4.17,
+      "step": 113
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9969658659924147e-06,
+      "loss": 4.1828,
+      "step": 114
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9969184576485464e-06,
+      "loss": 4.1948,
+      "step": 115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9968710493046777e-06,
+      "loss": 4.2058,
+      "step": 116
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996823640960809e-06,
+      "loss": 4.2127,
+      "step": 117
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9967762326169407e-06,
+      "loss": 4.1971,
+      "step": 118
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996728824273072e-06,
+      "loss": 4.2137,
+      "step": 119
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9966814159292037e-06,
+      "loss": 4.2061,
+      "step": 120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9966340075853354e-06,
+      "loss": 4.2148,
+      "step": 121
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9965865992414667e-06,
+      "loss": 4.1837,
+      "step": 122
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996539190897598e-06,
+      "loss": 4.2178,
+      "step": 123
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9964917825537297e-06,
+      "loss": 4.1697,
+      "step": 124
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996444374209861e-06,
+      "loss": 4.1885,
+      "step": 125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9963969658659926e-06,
+      "loss": 4.1965,
+      "step": 126
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996349557522124e-06,
+      "loss": 4.2134,
+      "step": 127
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9963021491782556e-06,
+      "loss": 4.196,
+      "step": 128
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996254740834387e-06,
+      "loss": 4.1947,
+      "step": 129
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996207332490518e-06,
+      "loss": 4.2122,
+      "step": 130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.99615992414665e-06,
+      "loss": 4.2133,
+      "step": 131
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996112515802781e-06,
+      "loss": 4.1947,
+      "step": 132
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.996065107458913e-06,
+      "loss": 4.2096,
+      "step": 133
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9960176991150446e-06,
+      "loss": 4.2174,
+      "step": 134
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995970290771176e-06,
+      "loss": 4.2041,
+      "step": 135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995922882427307e-06,
+      "loss": 4.2183,
+      "step": 136
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995875474083439e-06,
+      "loss": 4.1756,
+      "step": 137
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.99582806573957e-06,
+      "loss": 4.1864,
+      "step": 138
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995780657395702e-06,
+      "loss": 4.199,
+      "step": 139
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9957332490518335e-06,
+      "loss": 4.1906,
+      "step": 140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995685840707965e-06,
+      "loss": 4.2104,
+      "step": 141
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995638432364096e-06,
+      "loss": 4.1944,
+      "step": 142
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995591024020228e-06,
+      "loss": 4.1964,
+      "step": 143
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995543615676359e-06,
+      "loss": 4.1884,
+      "step": 144
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9954962073324904e-06,
+      "loss": 4.2127,
+      "step": 145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995448798988622e-06,
+      "loss": 4.193,
+      "step": 146
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9954013906447538e-06,
+      "loss": 4.193,
+      "step": 147
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995353982300885e-06,
+      "loss": 4.2151,
+      "step": 148
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9953065739570163e-06,
+      "loss": 4.1885,
+      "step": 149
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995259165613148e-06,
+      "loss": 4.1988,
+      "step": 150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9952117572692793e-06,
+      "loss": 4.2213,
+      "step": 151
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995164348925411e-06,
+      "loss": 4.1902,
+      "step": 152
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9951169405815427e-06,
+      "loss": 4.1889,
+      "step": 153
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.995069532237674e-06,
+      "loss": 4.1929,
+      "step": 154
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9950221238938053e-06,
+      "loss": 4.2093,
+      "step": 155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.994974715549937e-06,
+      "loss": 4.1841,
+      "step": 156
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9949273072060683e-06,
+      "loss": 4.1834,
+      "step": 157
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9948798988621995e-06,
+      "loss": 4.1983,
+      "step": 158
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9948324905183312e-06,
+      "loss": 4.1784,
+      "step": 159
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994785082174463e-06,
+      "loss": 4.2071,
+      "step": 160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9947376738305942e-06,
+      "loss": 4.1913,
+      "step": 161
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994690265486726e-06,
+      "loss": 4.2005,
+      "step": 162
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994642857142857e-06,
+      "loss": 4.1986,
+      "step": 163
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9945954487989885e-06,
+      "loss": 4.193,
+      "step": 164
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.99454804045512e-06,
+      "loss": 4.215,
+      "step": 165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994500632111252e-06,
+      "loss": 4.1994,
+      "step": 166
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994453223767383e-06,
+      "loss": 4.216,
+      "step": 167
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994405815423515e-06,
+      "loss": 4.1681,
+      "step": 168
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994358407079646e-06,
+      "loss": 4.1881,
+      "step": 169
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9943109987357774e-06,
+      "loss": 4.1896,
+      "step": 170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9942635903919087e-06,
+      "loss": 4.2001,
+      "step": 171
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9942161820480404e-06,
+      "loss": 4.2226,
+      "step": 172
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994168773704172e-06,
+      "loss": 4.1824,
+      "step": 173
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9941213653603034e-06,
+      "loss": 4.1919,
+      "step": 174
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.994073957016435e-06,
+      "loss": 4.2016,
+      "step": 175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9940265486725664e-06,
+      "loss": 4.1914,
+      "step": 176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9939791403286977e-06,
+      "loss": 4.1932,
+      "step": 177
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9939317319848294e-06,
+      "loss": 4.2206,
+      "step": 178
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.993884323640961e-06,
+      "loss": 4.2118,
+      "step": 179
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9938369152970924e-06,
+      "loss": 4.2216,
+      "step": 180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.993789506953224e-06,
+      "loss": 4.2359,
+      "step": 181
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9937420986093553e-06,
+      "loss": 4.1974,
+      "step": 182
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9936946902654866e-06,
+      "loss": 4.1904,
+      "step": 183
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9936472819216183e-06,
+      "loss": 4.205,
+      "step": 184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9935998735777496e-06,
+      "loss": 4.1923,
+      "step": 185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9935524652338813e-06,
+      "loss": 4.2102,
+      "step": 186
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.993505056890013e-06,
+      "loss": 4.2005,
+      "step": 187
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9934576485461443e-06,
+      "loss": 4.1921,
+      "step": 188
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9934102402022756e-06,
+      "loss": 4.1811,
+      "step": 189
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9933628318584073e-06,
+      "loss": 4.1897,
+      "step": 190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9933154235145386e-06,
+      "loss": 4.1988,
+      "step": 191
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.99326801517067e-06,
+      "loss": 4.1704,
+      "step": 192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9932206068268016e-06,
+      "loss": 4.1741,
+      "step": 193
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9931731984829333e-06,
+      "loss": 4.1866,
+      "step": 194
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9931257901390645e-06,
+      "loss": 4.1691,
+      "step": 195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.993078381795196e-06,
+      "loss": 4.1643,
+      "step": 196
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9930309734513275e-06,
+      "loss": 4.2071,
+      "step": 197
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.992983565107459e-06,
+      "loss": 4.2009,
+      "step": 198
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9929361567635905e-06,
+      "loss": 4.1723,
+      "step": 199
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.992888748419722e-06,
+      "loss": 4.1701,
+      "step": 200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9928413400758535e-06,
+      "loss": 4.2053,
+      "step": 201
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9927939317319848e-06,
+      "loss": 4.1898,
+      "step": 202
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9927465233881165e-06,
+      "loss": 4.1792,
+      "step": 203
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9926991150442478e-06,
+      "loss": 4.192,
+      "step": 204
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.992651706700379e-06,
+      "loss": 4.1689,
+      "step": 205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9926042983565107e-06,
+      "loss": 4.1995,
+      "step": 206
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9925568900126424e-06,
+      "loss": 4.2076,
+      "step": 207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9925094816687737e-06,
+      "loss": 4.187,
+      "step": 208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9924620733249054e-06,
+      "loss": 4.1744,
+      "step": 209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9924146649810367e-06,
+      "loss": 4.2036,
+      "step": 210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.992367256637168e-06,
+      "loss": 4.2096,
+      "step": 211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9923198482932997e-06,
+      "loss": 4.1944,
+      "step": 212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9922724399494314e-06,
+      "loss": 4.1699,
+      "step": 213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9922250316055627e-06,
+      "loss": 4.2432,
+      "step": 214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.992177623261694e-06,
+      "loss": 4.1894,
+      "step": 215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9921302149178257e-06,
+      "loss": 4.2353,
+      "step": 216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.992082806573957e-06,
+      "loss": 4.1814,
+      "step": 217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9920353982300882e-06,
+      "loss": 4.2149,
+      "step": 218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.99198798988622e-06,
+      "loss": 4.1788,
+      "step": 219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9919405815423516e-06,
+      "loss": 4.1961,
+      "step": 220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.991893173198483e-06,
+      "loss": 4.1788,
+      "step": 221
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9918457648546146e-06,
+      "loss": 4.1955,
+      "step": 222
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991798356510746e-06,
+      "loss": 4.2003,
+      "step": 223
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991750948166877e-06,
+      "loss": 4.1795,
+      "step": 224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991703539823009e-06,
+      "loss": 4.1952,
+      "step": 225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9916561314791406e-06,
+      "loss": 4.1907,
+      "step": 226
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991608723135272e-06,
+      "loss": 4.2143,
+      "step": 227
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9915613147914036e-06,
+      "loss": 4.1828,
+      "step": 228
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991513906447535e-06,
+      "loss": 4.209,
+      "step": 229
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991466498103666e-06,
+      "loss": 4.1842,
+      "step": 230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991419089759798e-06,
+      "loss": 4.1936,
+      "step": 231
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991371681415929e-06,
+      "loss": 4.1824,
+      "step": 232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991324273072061e-06,
+      "loss": 4.2207,
+      "step": 233
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9912768647281925e-06,
+      "loss": 4.2407,
+      "step": 234
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991229456384324e-06,
+      "loss": 4.2061,
+      "step": 235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991182048040455e-06,
+      "loss": 4.2007,
+      "step": 236
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9911346396965864e-06,
+      "loss": 4.1823,
+      "step": 237
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.991087231352718e-06,
+      "loss": 4.2213,
+      "step": 238
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9910398230088498e-06,
+      "loss": 4.1843,
+      "step": 239
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990992414664981e-06,
+      "loss": 4.1464,
+      "step": 240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9909450063211128e-06,
+      "loss": 4.1905,
+      "step": 241
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990897597977244e-06,
+      "loss": 4.2116,
+      "step": 242
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9908501896333753e-06,
+      "loss": 4.1944,
+      "step": 243
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990802781289507e-06,
+      "loss": 4.1792,
+      "step": 244
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9907553729456383e-06,
+      "loss": 4.1989,
+      "step": 245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.99070796460177e-06,
+      "loss": 4.2105,
+      "step": 246
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9906605562579017e-06,
+      "loss": 4.2325,
+      "step": 247
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990613147914033e-06,
+      "loss": 4.2052,
+      "step": 248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9905657395701643e-06,
+      "loss": 4.233,
+      "step": 249
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990518331226296e-06,
+      "loss": 4.2287,
+      "step": 250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9904709228824273e-06,
+      "loss": 4.1732,
+      "step": 251
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990423514538559e-06,
+      "loss": 4.1692,
+      "step": 252
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9903761061946907e-06,
+      "loss": 4.1981,
+      "step": 253
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990328697850822e-06,
+      "loss": 4.2139,
+      "step": 254
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9902812895069532e-06,
+      "loss": 4.1715,
+      "step": 255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990233881163085e-06,
+      "loss": 4.177,
+      "step": 256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990186472819216e-06,
+      "loss": 4.1998,
+      "step": 257
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9901390644753475e-06,
+      "loss": 4.192,
+      "step": 258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990091656131479e-06,
+      "loss": 4.203,
+      "step": 259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.990044247787611e-06,
+      "loss": 4.1845,
+      "step": 260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989996839443742e-06,
+      "loss": 4.2116,
+      "step": 261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9899494310998735e-06,
+      "loss": 4.1601,
+      "step": 262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989902022756005e-06,
+      "loss": 4.1729,
+      "step": 263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9898546144121364e-06,
+      "loss": 4.2026,
+      "step": 264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989807206068268e-06,
+      "loss": 4.1637,
+      "step": 265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9897597977244e-06,
+      "loss": 4.1733,
+      "step": 266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989712389380531e-06,
+      "loss": 4.2217,
+      "step": 267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9896649810366624e-06,
+      "loss": 4.1805,
+      "step": 268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989617572692794e-06,
+      "loss": 4.1438,
+      "step": 269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9895701643489254e-06,
+      "loss": 4.1658,
+      "step": 270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9895227560050567e-06,
+      "loss": 4.1682,
+      "step": 271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9894753476611884e-06,
+      "loss": 4.1908,
+      "step": 272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.98942793931732e-06,
+      "loss": 4.1972,
+      "step": 273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9893805309734514e-06,
+      "loss": 4.1857,
+      "step": 274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989333122629583e-06,
+      "loss": 4.1521,
+      "step": 275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9892857142857143e-06,
+      "loss": 4.1951,
+      "step": 276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9892383059418456e-06,
+      "loss": 4.215,
+      "step": 277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9891908975979773e-06,
+      "loss": 4.1643,
+      "step": 278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.989143489254109e-06,
+      "loss": 4.1856,
+      "step": 279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9890960809102403e-06,
+      "loss": 4.1886,
+      "step": 280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9890486725663716e-06,
+      "loss": 4.1864,
+      "step": 281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9890012642225033e-06,
+      "loss": 4.1593,
+      "step": 282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.9889538558786346e-06,
+      "loss": 4.1735,
+      "step": 283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 2.988906447534766e-06,
+      "loss": 4.19,
+      "step": 284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9888590391908976e-06,
+      "loss": 4.1966,
+      "step": 285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9888116308470293e-06,
+      "loss": 4.1824,
+      "step": 286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9887642225031605e-06,
+      "loss": 4.2032,
+      "step": 287
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9887168141592922e-06,
+      "loss": 4.1837,
+      "step": 288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9886694058154235e-06,
+      "loss": 4.1669,
+      "step": 289
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.988621997471555e-06,
+      "loss": 4.1583,
+      "step": 290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9885745891276865e-06,
+      "loss": 4.2308,
+      "step": 291
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9885271807838182e-06,
+      "loss": 4.1336,
+      "step": 292
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9884797724399495e-06,
+      "loss": 4.1981,
+      "step": 293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.988432364096081e-06,
+      "loss": 4.1878,
+      "step": 294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9883849557522125e-06,
+      "loss": 4.2046,
+      "step": 295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9883375474083438e-06,
+      "loss": 4.1691,
+      "step": 296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9882901390644755e-06,
+      "loss": 4.1787,
+      "step": 297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9882427307206067e-06,
+      "loss": 4.1625,
+      "step": 298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9881953223767385e-06,
+      "loss": 4.2156,
+      "step": 299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.98814791403287e-06,
+      "loss": 4.1917,
+      "step": 300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9881005056890014e-06,
+      "loss": 4.2394,
+      "step": 301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9880530973451327e-06,
+      "loss": 4.2321,
+      "step": 302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.988005689001264e-06,
+      "loss": 4.2101,
+      "step": 303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9879582806573957e-06,
+      "loss": 4.2049,
+      "step": 304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9879108723135274e-06,
+      "loss": 4.1727,
+      "step": 305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9878634639696587e-06,
+      "loss": 4.1654,
+      "step": 306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9878160556257904e-06,
+      "loss": 4.1771,
+      "step": 307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9877686472819217e-06,
+      "loss": 4.2284,
+      "step": 308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987721238938053e-06,
+      "loss": 4.1484,
+      "step": 309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9876738305941847e-06,
+      "loss": 4.2101,
+      "step": 310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987626422250316e-06,
+      "loss": 4.1584,
+      "step": 311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9875790139064476e-06,
+      "loss": 4.1484,
+      "step": 312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9875316055625793e-06,
+      "loss": 4.1931,
+      "step": 313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9874841972187106e-06,
+      "loss": 4.2211,
+      "step": 314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987436788874842e-06,
+      "loss": 4.1575,
+      "step": 315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9873893805309736e-06,
+      "loss": 4.1833,
+      "step": 316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987341972187105e-06,
+      "loss": 4.1987,
+      "step": 317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9872945638432366e-06,
+      "loss": 4.1774,
+      "step": 318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9872471554993683e-06,
+      "loss": 4.2081,
+      "step": 319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9871997471554996e-06,
+      "loss": 4.1451,
+      "step": 320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987152338811631e-06,
+      "loss": 4.2088,
+      "step": 321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9871049304677626e-06,
+      "loss": 4.1785,
+      "step": 322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987057522123894e-06,
+      "loss": 4.1186,
+      "step": 323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.987010113780025e-06,
+      "loss": 4.241,
+      "step": 324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986962705436157e-06,
+      "loss": 4.1845,
+      "step": 325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9869152970922885e-06,
+      "loss": 4.2066,
+      "step": 326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.98686788874842e-06,
+      "loss": 4.1865,
+      "step": 327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986820480404551e-06,
+      "loss": 4.1913,
+      "step": 328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986773072060683e-06,
+      "loss": 4.249,
+      "step": 329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986725663716814e-06,
+      "loss": 4.1827,
+      "step": 330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9866782553729458e-06,
+      "loss": 4.1724,
+      "step": 331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9866308470290775e-06,
+      "loss": 4.2123,
+      "step": 332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9865834386852088e-06,
+      "loss": 4.1907,
+      "step": 333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.98653603034134e-06,
+      "loss": 4.1948,
+      "step": 334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9864886219974717e-06,
+      "loss": 4.1861,
+      "step": 335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986441213653603e-06,
+      "loss": 4.1964,
+      "step": 336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9863938053097343e-06,
+      "loss": 4.1051,
+      "step": 337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986346396965866e-06,
+      "loss": 4.1684,
+      "step": 338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9862989886219977e-06,
+      "loss": 4.1611,
+      "step": 339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986251580278129e-06,
+      "loss": 4.1911,
+      "step": 340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9862041719342607e-06,
+      "loss": 4.2024,
+      "step": 341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986156763590392e-06,
+      "loss": 4.1333,
+      "step": 342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9861093552465233e-06,
+      "loss": 4.2238,
+      "step": 343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.986061946902655e-06,
+      "loss": 4.1433,
+      "step": 344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9860145385587867e-06,
+      "loss": 4.1869,
+      "step": 345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.985967130214918e-06,
+      "loss": 4.1883,
+      "step": 346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.9859197218710492e-06,
+      "loss": 4.1659,
+      "step": 347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.985872313527181e-06,
+      "loss": 4.1543,
+      "step": 348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9858249051833122e-06,
+      "loss": 4.1436,
+      "step": 349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9857774968394435e-06,
+      "loss": 4.2044,
+      "step": 350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985730088495575e-06,
+      "loss": 4.1946,
+      "step": 351
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985682680151707e-06,
+      "loss": 4.1829,
+      "step": 352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985635271807838e-06,
+      "loss": 4.1399,
+      "step": 353
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.98558786346397e-06,
+      "loss": 4.1995,
+      "step": 354
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985540455120101e-06,
+      "loss": 4.1826,
+      "step": 355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9854930467762324e-06,
+      "loss": 4.1748,
+      "step": 356
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985445638432364e-06,
+      "loss": 4.1939,
+      "step": 357
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985398230088496e-06,
+      "loss": 4.1678,
+      "step": 358
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985350821744627e-06,
+      "loss": 4.2145,
+      "step": 359
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985303413400759e-06,
+      "loss": 4.1437,
+      "step": 360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.98525600505689e-06,
+      "loss": 4.1837,
+      "step": 361
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9852085967130214e-06,
+      "loss": 4.131,
+      "step": 362
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985161188369153e-06,
+      "loss": 4.1501,
+      "step": 363
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9851137800252844e-06,
+      "loss": 4.1858,
+      "step": 364
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985066371681416e-06,
+      "loss": 4.1486,
+      "step": 365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.985018963337548e-06,
+      "loss": 4.2098,
+      "step": 366
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.984971554993679e-06,
+      "loss": 4.1835,
+      "step": 367
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9849241466498104e-06,
+      "loss": 4.1776,
+      "step": 368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9848767383059416e-06,
+      "loss": 4.1808,
+      "step": 369
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9848293299620733e-06,
+      "loss": 4.189,
+      "step": 370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.984781921618205e-06,
+      "loss": 4.1568,
+      "step": 371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9847345132743363e-06,
+      "loss": 4.2184,
+      "step": 372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.984687104930468e-06,
+      "loss": 4.1957,
+      "step": 373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9846396965865993e-06,
+      "loss": 4.1833,
+      "step": 374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9845922882427306e-06,
+      "loss": 4.2394,
+      "step": 375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9845448798988623e-06,
+      "loss": 4.1476,
+      "step": 376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9844974715549936e-06,
+      "loss": 4.23,
+      "step": 377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9844500632111253e-06,
+      "loss": 4.1253,
+      "step": 378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.984402654867257e-06,
+      "loss": 4.1187,
+      "step": 379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9843552465233883e-06,
+      "loss": 4.2079,
+      "step": 380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9843078381795195e-06,
+      "loss": 4.1525,
+      "step": 381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9842604298356512e-06,
+      "loss": 4.1898,
+      "step": 382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9842130214917825e-06,
+      "loss": 4.153,
+      "step": 383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9841656131479142e-06,
+      "loss": 4.2029,
+      "step": 384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.984118204804046e-06,
+      "loss": 4.1064,
+      "step": 385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9840707964601772e-06,
+      "loss": 4.1858,
+      "step": 386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9840233881163085e-06,
+      "loss": 4.1835,
+      "step": 387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.98397597977244e-06,
+      "loss": 4.2309,
+      "step": 388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9839285714285715e-06,
+      "loss": 4.0723,
+      "step": 389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9838811630847028e-06,
+      "loss": 4.1744,
+      "step": 390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9838337547408345e-06,
+      "loss": 4.1309,
+      "step": 391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.983786346396966e-06,
+      "loss": 4.1688,
+      "step": 392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9837389380530974e-06,
+      "loss": 4.0949,
+      "step": 393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9836915297092287e-06,
+      "loss": 4.1344,
+      "step": 394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9836441213653604e-06,
+      "loss": 4.1705,
+      "step": 395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9835967130214917e-06,
+      "loss": 4.1876,
+      "step": 396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9835493046776234e-06,
+      "loss": 4.1477,
+      "step": 397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.983501896333755e-06,
+      "loss": 4.1769,
+      "step": 398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9834544879898864e-06,
+      "loss": 4.1496,
+      "step": 399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9834070796460177e-06,
+      "loss": 4.229,
+      "step": 400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9833596713021494e-06,
+      "loss": 4.2021,
+      "step": 401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9833122629582807e-06,
+      "loss": 4.112,
+      "step": 402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.983264854614412e-06,
+      "loss": 4.2059,
+      "step": 403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9832174462705436e-06,
+      "loss": 4.1492,
+      "step": 404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9831700379266754e-06,
+      "loss": 4.0994,
+      "step": 405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9831226295828066e-06,
+      "loss": 4.2018,
+      "step": 406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9830752212389383e-06,
+      "loss": 4.1943,
+      "step": 407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9830278128950696e-06,
+      "loss": 4.165,
+      "step": 408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.982980404551201e-06,
+      "loss": 4.2055,
+      "step": 409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9829329962073326e-06,
+      "loss": 4.221,
+      "step": 410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.9828855878634643e-06,
+      "loss": 4.1694,
+      "step": 411
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9828381795195956e-06,
+      "loss": 4.1873,
+      "step": 412
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.982790771175727e-06,
+      "loss": 4.1905,
+      "step": 413
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9827433628318586e-06,
+      "loss": 4.1786,
+      "step": 414
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98269595448799e-06,
+      "loss": 4.1978,
+      "step": 415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.982648546144121e-06,
+      "loss": 4.0658,
+      "step": 416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.982601137800253e-06,
+      "loss": 4.1992,
+      "step": 417
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9825537294563845e-06,
+      "loss": 4.1437,
+      "step": 418
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.982506321112516e-06,
+      "loss": 4.112,
+      "step": 419
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9824589127686475e-06,
+      "loss": 4.1243,
+      "step": 420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.982411504424779e-06,
+      "loss": 4.1791,
+      "step": 421
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98236409608091e-06,
+      "loss": 4.1413,
+      "step": 422
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9823166877370418e-06,
+      "loss": 4.2345,
+      "step": 423
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9822692793931735e-06,
+      "loss": 4.2283,
+      "step": 424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9822218710493048e-06,
+      "loss": 4.1339,
+      "step": 425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9821744627054365e-06,
+      "loss": 4.0822,
+      "step": 426
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9821270543615678e-06,
+      "loss": 4.2026,
+      "step": 427
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.982079646017699e-06,
+      "loss": 4.2454,
+      "step": 428
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9820322376738307e-06,
+      "loss": 4.1764,
+      "step": 429
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981984829329962e-06,
+      "loss": 4.1277,
+      "step": 430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9819374209860937e-06,
+      "loss": 4.1528,
+      "step": 431
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9818900126422254e-06,
+      "loss": 4.1906,
+      "step": 432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9818426042983567e-06,
+      "loss": 4.127,
+      "step": 433
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981795195954488e-06,
+      "loss": 4.1953,
+      "step": 434
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9817477876106193e-06,
+      "loss": 4.2825,
+      "step": 435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981700379266751e-06,
+      "loss": 4.1005,
+      "step": 436
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9816529709228827e-06,
+      "loss": 4.2528,
+      "step": 437
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981605562579014e-06,
+      "loss": 4.1785,
+      "step": 438
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9815581542351457e-06,
+      "loss": 4.2438,
+      "step": 439
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981510745891277e-06,
+      "loss": 4.1771,
+      "step": 440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9814633375474082e-06,
+      "loss": 4.1536,
+      "step": 441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98141592920354e-06,
+      "loss": 4.1662,
+      "step": 442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981368520859671e-06,
+      "loss": 4.2246,
+      "step": 443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981321112515803e-06,
+      "loss": 4.1985,
+      "step": 444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9812737041719346e-06,
+      "loss": 4.1433,
+      "step": 445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981226295828066e-06,
+      "loss": 4.1829,
+      "step": 446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981178887484197e-06,
+      "loss": 4.1606,
+      "step": 447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981131479140329e-06,
+      "loss": 4.1632,
+      "step": 448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.98108407079646e-06,
+      "loss": 4.2034,
+      "step": 449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.981036662452592e-06,
+      "loss": 4.1424,
+      "step": 450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9809892541087236e-06,
+      "loss": 4.1367,
+      "step": 451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980941845764855e-06,
+      "loss": 4.1897,
+      "step": 452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980894437420986e-06,
+      "loss": 4.162,
+      "step": 453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980847029077118e-06,
+      "loss": 4.1701,
+      "step": 454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980799620733249e-06,
+      "loss": 4.1648,
+      "step": 455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9807522123893804e-06,
+      "loss": 4.1821,
+      "step": 456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980704804045512e-06,
+      "loss": 4.1339,
+      "step": 457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980657395701644e-06,
+      "loss": 4.2018,
+      "step": 458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980609987357775e-06,
+      "loss": 4.1644,
+      "step": 459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9805625790139064e-06,
+      "loss": 4.2438,
+      "step": 460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980515170670038e-06,
+      "loss": 4.1402,
+      "step": 461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9804677623261693e-06,
+      "loss": 4.185,
+      "step": 462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9804203539823006e-06,
+      "loss": 4.2554,
+      "step": 463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9803729456384328e-06,
+      "loss": 4.1323,
+      "step": 464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980325537294564e-06,
+      "loss": 4.1919,
+      "step": 465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9802781289506953e-06,
+      "loss": 4.1151,
+      "step": 466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980230720606827e-06,
+      "loss": 4.187,
+      "step": 467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9801833122629583e-06,
+      "loss": 4.1775,
+      "step": 468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9801359039190896e-06,
+      "loss": 4.2097,
+      "step": 469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9800884955752213e-06,
+      "loss": 4.0254,
+      "step": 470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.980041087231353e-06,
+      "loss": 4.1456,
+      "step": 471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9799936788874843e-06,
+      "loss": 4.1497,
+      "step": 472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.979946270543616e-06,
+      "loss": 4.2362,
+      "step": 473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.9798988621997473e-06,
+      "loss": 4.1997,
+      "step": 474
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9798514538558785e-06,
+      "loss": 4.1184,
+      "step": 475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9798040455120102e-06,
+      "loss": 4.1765,
+      "step": 476
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9797566371681415e-06,
+      "loss": 4.1548,
+      "step": 477
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9797092288242732e-06,
+      "loss": 4.1395,
+      "step": 478
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9796618204804045e-06,
+      "loss": 4.1455,
+      "step": 479
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979614412136536e-06,
+      "loss": 4.1635,
+      "step": 480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9795670037926675e-06,
+      "loss": 4.1955,
+      "step": 481
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9795195954487988e-06,
+      "loss": 4.2075,
+      "step": 482
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9794721871049305e-06,
+      "loss": 4.0665,
+      "step": 483
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979424778761062e-06,
+      "loss": 4.0864,
+      "step": 484
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9793773704171935e-06,
+      "loss": 4.1236,
+      "step": 485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979329962073325e-06,
+      "loss": 4.1752,
+      "step": 486
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9792825537294564e-06,
+      "loss": 4.0905,
+      "step": 487
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9792351453855877e-06,
+      "loss": 4.1091,
+      "step": 488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9791877370417194e-06,
+      "loss": 4.1951,
+      "step": 489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9791403286978507e-06,
+      "loss": 4.0991,
+      "step": 490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9790929203539824e-06,
+      "loss": 4.1985,
+      "step": 491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.979045512010114e-06,
+      "loss": 4.1065,
+      "step": 492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9789981036662454e-06,
+      "loss": 4.12,
+      "step": 493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9789506953223767e-06,
+      "loss": 4.1988,
+      "step": 494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9789032869785084e-06,
+      "loss": 4.1021,
+      "step": 495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9788558786346397e-06,
+      "loss": 4.0778,
+      "step": 496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9788084702907714e-06,
+      "loss": 4.1669,
+      "step": 497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9787610619469026e-06,
+      "loss": 4.2307,
+      "step": 498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9787136536030343e-06,
+      "loss": 4.1688,
+      "step": 499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9786662452591656e-06,
+      "loss": 4.1921,
+      "step": 500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978618836915297e-06,
+      "loss": 4.1574,
+      "step": 501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9785714285714286e-06,
+      "loss": 4.1527,
+      "step": 502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.97852402022756e-06,
+      "loss": 4.1478,
+      "step": 503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9784766118836916e-06,
+      "loss": 4.1857,
+      "step": 504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9784292035398233e-06,
+      "loss": 4.0791,
+      "step": 505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9783817951959546e-06,
+      "loss": 4.1316,
+      "step": 506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978334386852086e-06,
+      "loss": 4.1896,
+      "step": 507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9782869785082176e-06,
+      "loss": 4.1143,
+      "step": 508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978239570164349e-06,
+      "loss": 4.1305,
+      "step": 509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9781921618204805e-06,
+      "loss": 4.19,
+      "step": 510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9781447534766123e-06,
+      "loss": 4.1257,
+      "step": 511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9780973451327435e-06,
+      "loss": 4.2065,
+      "step": 512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.978049936788875e-06,
+      "loss": 4.0345,
+      "step": 513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9780025284450065e-06,
+      "loss": 4.1113,
+      "step": 514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977955120101138e-06,
+      "loss": 4.2307,
+      "step": 515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977907711757269e-06,
+      "loss": 4.167,
+      "step": 516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9778603034134008e-06,
+      "loss": 4.1037,
+      "step": 517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9778128950695325e-06,
+      "loss": 4.2088,
+      "step": 518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9777654867256638e-06,
+      "loss": 4.1002,
+      "step": 519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977718078381795e-06,
+      "loss": 4.1416,
+      "step": 520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9776706700379268e-06,
+      "loss": 4.159,
+      "step": 521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977623261694058e-06,
+      "loss": 4.1206,
+      "step": 522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9775758533501897e-06,
+      "loss": 4.1357,
+      "step": 523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9775284450063214e-06,
+      "loss": 4.1307,
+      "step": 524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9774810366624527e-06,
+      "loss": 4.1035,
+      "step": 525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977433628318584e-06,
+      "loss": 4.1199,
+      "step": 526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9773862199747157e-06,
+      "loss": 4.1872,
+      "step": 527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977338811630847e-06,
+      "loss": 4.1663,
+      "step": 528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9772914032869783e-06,
+      "loss": 4.1622,
+      "step": 529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.97724399494311e-06,
+      "loss": 4.2222,
+      "step": 530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9771965865992417e-06,
+      "loss": 4.1846,
+      "step": 531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977149178255373e-06,
+      "loss": 4.1762,
+      "step": 532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9771017699115047e-06,
+      "loss": 4.1398,
+      "step": 533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.977054361567636e-06,
+      "loss": 4.0521,
+      "step": 534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9770069532237672e-06,
+      "loss": 4.1781,
+      "step": 535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.976959544879899e-06,
+      "loss": 4.1979,
+      "step": 536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.9769121365360306e-06,
+      "loss": 4.1084,
+      "step": 537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.976864728192162e-06,
+      "loss": 4.184,
+      "step": 538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9768173198482936e-06,
+      "loss": 4.2424,
+      "step": 539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976769911504425e-06,
+      "loss": 4.1224,
+      "step": 540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976722503160556e-06,
+      "loss": 4.2378,
+      "step": 541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9766750948166875e-06,
+      "loss": 4.1441,
+      "step": 542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976627686472819e-06,
+      "loss": 4.1852,
+      "step": 543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976580278128951e-06,
+      "loss": 4.1822,
+      "step": 544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976532869785082e-06,
+      "loss": 4.1541,
+      "step": 545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976485461441214e-06,
+      "loss": 4.1459,
+      "step": 546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976438053097345e-06,
+      "loss": 4.154,
+      "step": 547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9763906447534764e-06,
+      "loss": 4.0462,
+      "step": 548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976343236409608e-06,
+      "loss": 4.1827,
+      "step": 549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97629582806574e-06,
+      "loss": 4.1342,
+      "step": 550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976248419721871e-06,
+      "loss": 4.2242,
+      "step": 551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976201011378003e-06,
+      "loss": 4.1036,
+      "step": 552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976153603034134e-06,
+      "loss": 4.2647,
+      "step": 553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9761061946902654e-06,
+      "loss": 4.0999,
+      "step": 554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.976058786346397e-06,
+      "loss": 4.1245,
+      "step": 555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9760113780025283e-06,
+      "loss": 4.1253,
+      "step": 556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97596396965866e-06,
+      "loss": 4.0531,
+      "step": 557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9759165613147917e-06,
+      "loss": 4.272,
+      "step": 558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975869152970923e-06,
+      "loss": 4.1792,
+      "step": 559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9758217446270543e-06,
+      "loss": 4.2295,
+      "step": 560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975774336283186e-06,
+      "loss": 4.0807,
+      "step": 561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9757269279393173e-06,
+      "loss": 4.1305,
+      "step": 562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975679519595449e-06,
+      "loss": 4.0471,
+      "step": 563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9756321112515803e-06,
+      "loss": 4.1514,
+      "step": 564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975584702907712e-06,
+      "loss": 4.1866,
+      "step": 565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9755372945638433e-06,
+      "loss": 4.1372,
+      "step": 566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9754898862199745e-06,
+      "loss": 4.1069,
+      "step": 567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9754424778761062e-06,
+      "loss": 4.1024,
+      "step": 568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9753950695322375e-06,
+      "loss": 4.1244,
+      "step": 569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9753476611883692e-06,
+      "loss": 4.1067,
+      "step": 570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975300252844501e-06,
+      "loss": 4.0103,
+      "step": 571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9752528445006322e-06,
+      "loss": 4.1241,
+      "step": 572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9752054361567635e-06,
+      "loss": 4.2342,
+      "step": 573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975158027812895e-06,
+      "loss": 4.1251,
+      "step": 574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9751106194690265e-06,
+      "loss": 4.1698,
+      "step": 575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.975063211125158e-06,
+      "loss": 4.262,
+      "step": 576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97501580278129e-06,
+      "loss": 4.1456,
+      "step": 577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974968394437421e-06,
+      "loss": 4.1331,
+      "step": 578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9749209860935524e-06,
+      "loss": 4.2021,
+      "step": 579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974873577749684e-06,
+      "loss": 4.1347,
+      "step": 580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9748261694058154e-06,
+      "loss": 4.0337,
+      "step": 581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9747787610619467e-06,
+      "loss": 4.1773,
+      "step": 582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9747313527180784e-06,
+      "loss": 4.1499,
+      "step": 583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.97468394437421e-06,
+      "loss": 4.1303,
+      "step": 584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9746365360303414e-06,
+      "loss": 4.1882,
+      "step": 585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9745891276864727e-06,
+      "loss": 4.1297,
+      "step": 586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9745417193426044e-06,
+      "loss": 4.111,
+      "step": 587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9744943109987357e-06,
+      "loss": 4.0826,
+      "step": 588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9744469026548674e-06,
+      "loss": 4.1174,
+      "step": 589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974399494310999e-06,
+      "loss": 4.215,
+      "step": 590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9743520859671304e-06,
+      "loss": 4.1832,
+      "step": 591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9743046776232616e-06,
+      "loss": 4.1849,
+      "step": 592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9742572692793933e-06,
+      "loss": 4.1015,
+      "step": 593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9742098609355246e-06,
+      "loss": 4.1236,
+      "step": 594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.974162452591656e-06,
+      "loss": 4.1111,
+      "step": 595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9741150442477876e-06,
+      "loss": 4.0755,
+      "step": 596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9740676359039193e-06,
+      "loss": 4.0908,
+      "step": 597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9740202275600506e-06,
+      "loss": 4.2475,
+      "step": 598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9739728192161823e-06,
+      "loss": 4.1811,
+      "step": 599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9739254108723136e-06,
+      "loss": 4.2103,
+      "step": 600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.973878002528445e-06,
+      "loss": 4.1204,
+      "step": 601
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9738305941845766e-06,
+      "loss": 4.0583,
+      "step": 602
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9737831858407083e-06,
+      "loss": 4.2141,
+      "step": 603
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9737357774968395e-06,
+      "loss": 4.2282,
+      "step": 604
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9736883691529712e-06,
+      "loss": 4.0535,
+      "step": 605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9736409608091025e-06,
+      "loss": 4.1605,
+      "step": 606
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.973593552465234e-06,
+      "loss": 4.0589,
+      "step": 607
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.973546144121365e-06,
+      "loss": 4.1712,
+      "step": 608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.973498735777497e-06,
+      "loss": 4.1629,
+      "step": 609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9734513274336285e-06,
+      "loss": 4.1055,
+      "step": 610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9734039190897598e-06,
+      "loss": 4.1377,
+      "step": 611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9733565107458915e-06,
+      "loss": 4.1331,
+      "step": 612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9733091024020228e-06,
+      "loss": 4.2389,
+      "step": 613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.973261694058154e-06,
+      "loss": 4.2113,
+      "step": 614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9732142857142857e-06,
+      "loss": 4.1799,
+      "step": 615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9731668773704174e-06,
+      "loss": 4.2091,
+      "step": 616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9731194690265487e-06,
+      "loss": 4.1363,
+      "step": 617
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9730720606826804e-06,
+      "loss": 4.1357,
+      "step": 618
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9730246523388117e-06,
+      "loss": 4.0005,
+      "step": 619
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972977243994943e-06,
+      "loss": 4.1449,
+      "step": 620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9729298356510747e-06,
+      "loss": 4.1861,
+      "step": 621
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972882427307206e-06,
+      "loss": 4.2079,
+      "step": 622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9728350189633377e-06,
+      "loss": 4.1206,
+      "step": 623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9727876106194694e-06,
+      "loss": 4.1256,
+      "step": 624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9727402022756007e-06,
+      "loss": 4.1149,
+      "step": 625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972692793931732e-06,
+      "loss": 4.1962,
+      "step": 626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9726453855878637e-06,
+      "loss": 4.1302,
+      "step": 627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972597977243995e-06,
+      "loss": 4.2274,
+      "step": 628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9725505689001266e-06,
+      "loss": 4.1616,
+      "step": 629
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972503160556258e-06,
+      "loss": 4.2021,
+      "step": 630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9724557522123896e-06,
+      "loss": 4.0752,
+      "step": 631
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972408343868521e-06,
+      "loss": 4.2029,
+      "step": 632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972360935524652e-06,
+      "loss": 4.2589,
+      "step": 633
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972313527180784e-06,
+      "loss": 4.0996,
+      "step": 634
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972266118836915e-06,
+      "loss": 4.1781,
+      "step": 635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972218710493047e-06,
+      "loss": 4.1071,
+      "step": 636
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9721713021491786e-06,
+      "loss": 4.1633,
+      "step": 637
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.97212389380531e-06,
+      "loss": 4.145,
+      "step": 638
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972076485461441e-06,
+      "loss": 4.032,
+      "step": 639
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.972029077117573e-06,
+      "loss": 4.1258,
+      "step": 640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971981668773704e-06,
+      "loss": 4.1183,
+      "step": 641
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971934260429836e-06,
+      "loss": 4.1146,
+      "step": 642
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9718868520859675e-06,
+      "loss": 4.1751,
+      "step": 643
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971839443742099e-06,
+      "loss": 4.0869,
+      "step": 644
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.97179203539823e-06,
+      "loss": 4.182,
+      "step": 645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971744627054362e-06,
+      "loss": 4.1702,
+      "step": 646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971697218710493e-06,
+      "loss": 4.2076,
+      "step": 647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9716498103666244e-06,
+      "loss": 4.0573,
+      "step": 648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971602402022756e-06,
+      "loss": 4.1628,
+      "step": 649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9715549936788878e-06,
+      "loss": 4.1462,
+      "step": 650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971507585335019e-06,
+      "loss": 4.143,
+      "step": 651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9714601769911503e-06,
+      "loss": 4.1473,
+      "step": 652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971412768647282e-06,
+      "loss": 4.2109,
+      "step": 653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9713653603034133e-06,
+      "loss": 4.2001,
+      "step": 654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971317951959545e-06,
+      "loss": 4.2108,
+      "step": 655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9712705436156767e-06,
+      "loss": 4.1024,
+      "step": 656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971223135271808e-06,
+      "loss": 4.1677,
+      "step": 657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9711757269279393e-06,
+      "loss": 4.0904,
+      "step": 658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.971128318584071e-06,
+      "loss": 4.1826,
+      "step": 659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9710809102402023e-06,
+      "loss": 4.1509,
+      "step": 660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9710335018963335e-06,
+      "loss": 4.1917,
+      "step": 661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9709860935524652e-06,
+      "loss": 4.1926,
+      "step": 662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.970938685208597e-06,
+      "loss": 4.147,
+      "step": 663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.9708912768647282e-06,
+      "loss": 4.1568,
+      "step": 664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.97084386852086e-06,
+      "loss": 4.1172,
+      "step": 665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970796460176991e-06,
+      "loss": 4.2197,
+      "step": 666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9707490518331225e-06,
+      "loss": 4.0884,
+      "step": 667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970701643489254e-06,
+      "loss": 4.0898,
+      "step": 668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970654235145386e-06,
+      "loss": 4.0898,
+      "step": 669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970606826801517e-06,
+      "loss": 4.2013,
+      "step": 670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970559418457649e-06,
+      "loss": 4.1254,
+      "step": 671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.97051201011378e-06,
+      "loss": 4.1736,
+      "step": 672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9704646017699114e-06,
+      "loss": 4.1389,
+      "step": 673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9704171934260427e-06,
+      "loss": 4.1409,
+      "step": 674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9703697850821744e-06,
+      "loss": 4.1563,
+      "step": 675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970322376738306e-06,
+      "loss": 4.1817,
+      "step": 676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9702749683944374e-06,
+      "loss": 4.2206,
+      "step": 677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970227560050569e-06,
+      "loss": 4.1112,
+      "step": 678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9701801517067004e-06,
+      "loss": 4.2082,
+      "step": 679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9701327433628317e-06,
+      "loss": 4.2322,
+      "step": 680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9700853350189634e-06,
+      "loss": 4.1921,
+      "step": 681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.970037926675095e-06,
+      "loss": 4.184,
+      "step": 682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9699905183312264e-06,
+      "loss": 4.1486,
+      "step": 683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.969943109987358e-06,
+      "loss": 4.0907,
+      "step": 684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9698957016434893e-06,
+      "loss": 4.1581,
+      "step": 685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9698482932996206e-06,
+      "loss": 3.9372,
+      "step": 686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9698008849557523e-06,
+      "loss": 4.2184,
+      "step": 687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9697534766118836e-06,
+      "loss": 4.1643,
+      "step": 688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9697060682680153e-06,
+      "loss": 4.1272,
+      "step": 689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.969658659924147e-06,
+      "loss": 4.0522,
+      "step": 690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9696112515802783e-06,
+      "loss": 4.0472,
+      "step": 691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9695638432364096e-06,
+      "loss": 4.165,
+      "step": 692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9695164348925413e-06,
+      "loss": 4.1364,
+      "step": 693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9694690265486726e-06,
+      "loss": 4.2168,
+      "step": 694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9694216182048043e-06,
+      "loss": 4.014,
+      "step": 695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9693742098609356e-06,
+      "loss": 4.2092,
+      "step": 696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9693268015170673e-06,
+      "loss": 4.0986,
+      "step": 697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9692793931731985e-06,
+      "loss": 4.1862,
+      "step": 698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.96923198482933e-06,
+      "loss": 4.1703,
+      "step": 699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9691845764854615e-06,
+      "loss": 4.2156,
+      "step": 700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.969137168141593e-06,
+      "loss": 4.1807,
+      "step": 701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9690897597977245e-06,
+      "loss": 4.2119,
+      "step": 702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.969042351453856e-06,
+      "loss": 4.1728,
+      "step": 703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9689949431099875e-06,
+      "loss": 4.1444,
+      "step": 704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9689475347661188e-06,
+      "loss": 4.1581,
+      "step": 705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9689001264222505e-06,
+      "loss": 4.0822,
+      "step": 706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9688527180783818e-06,
+      "loss": 4.0567,
+      "step": 707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9688053097345135e-06,
+      "loss": 3.9664,
+      "step": 708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.968757901390645e-06,
+      "loss": 4.1769,
+      "step": 709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9687104930467764e-06,
+      "loss": 4.0455,
+      "step": 710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9686630847029077e-06,
+      "loss": 4.2078,
+      "step": 711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9686156763590394e-06,
+      "loss": 4.215,
+      "step": 712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9685682680151707e-06,
+      "loss": 4.0586,
+      "step": 713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.968520859671302e-06,
+      "loss": 3.975,
+      "step": 714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9684734513274337e-06,
+      "loss": 4.2667,
+      "step": 715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9684260429835654e-06,
+      "loss": 4.1406,
+      "step": 716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9683786346396967e-06,
+      "loss": 4.1913,
+      "step": 717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.968331226295828e-06,
+      "loss": 4.0731,
+      "step": 718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9682838179519597e-06,
+      "loss": 4.0921,
+      "step": 719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.968236409608091e-06,
+      "loss": 4.045,
+      "step": 720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9681890012642226e-06,
+      "loss": 4.13,
+      "step": 721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9681415929203543e-06,
+      "loss": 4.1561,
+      "step": 722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9680941845764856e-06,
+      "loss": 4.151,
+      "step": 723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.968046776232617e-06,
+      "loss": 4.1824,
+      "step": 724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9679993678887486e-06,
+      "loss": 4.1732,
+      "step": 725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.96795195954488e-06,
+      "loss": 4.1088,
+      "step": 726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.967904551201011e-06,
+      "loss": 4.2285,
+      "step": 727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.967857142857143e-06,
+      "loss": 4.0745,
+      "step": 728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9678097345132746e-06,
+      "loss": 4.0689,
+      "step": 729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967762326169406e-06,
+      "loss": 4.2749,
+      "step": 730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9677149178255376e-06,
+      "loss": 4.058,
+      "step": 731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967667509481669e-06,
+      "loss": 4.1153,
+      "step": 732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9676201011378e-06,
+      "loss": 4.0998,
+      "step": 733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967572692793932e-06,
+      "loss": 4.1133,
+      "step": 734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9675252844500635e-06,
+      "loss": 4.2184,
+      "step": 735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967477876106195e-06,
+      "loss": 4.0189,
+      "step": 736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9674304677623265e-06,
+      "loss": 4.1375,
+      "step": 737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967383059418458e-06,
+      "loss": 4.2268,
+      "step": 738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967335651074589e-06,
+      "loss": 4.0172,
+      "step": 739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9672882427307204e-06,
+      "loss": 4.0761,
+      "step": 740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967240834386852e-06,
+      "loss": 4.1239,
+      "step": 741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9671934260429838e-06,
+      "loss": 4.1286,
+      "step": 742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967146017699115e-06,
+      "loss": 4.1576,
+      "step": 743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9670986093552468e-06,
+      "loss": 4.0342,
+      "step": 744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.967051201011378e-06,
+      "loss": 4.0401,
+      "step": 745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9670037926675093e-06,
+      "loss": 4.1138,
+      "step": 746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966956384323641e-06,
+      "loss": 4.1435,
+      "step": 747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9669089759797723e-06,
+      "loss": 4.2148,
+      "step": 748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966861567635904e-06,
+      "loss": 4.1192,
+      "step": 749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9668141592920357e-06,
+      "loss": 4.2015,
+      "step": 750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966766750948167e-06,
+      "loss": 3.9613,
+      "step": 751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9667193426042983e-06,
+      "loss": 3.9748,
+      "step": 752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.96667193426043e-06,
+      "loss": 4.2649,
+      "step": 753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9666245259165613e-06,
+      "loss": 3.9855,
+      "step": 754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966577117572693e-06,
+      "loss": 4.2138,
+      "step": 755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9665297092288247e-06,
+      "loss": 4.2156,
+      "step": 756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966482300884956e-06,
+      "loss": 4.0944,
+      "step": 757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9664348925410872e-06,
+      "loss": 4.1732,
+      "step": 758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966387484197219e-06,
+      "loss": 4.0846,
+      "step": 759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.96634007585335e-06,
+      "loss": 4.0968,
+      "step": 760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9662926675094815e-06,
+      "loss": 4.0724,
+      "step": 761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966245259165613e-06,
+      "loss": 4.0561,
+      "step": 762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966197850821745e-06,
+      "loss": 4.0344,
+      "step": 763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966150442477876e-06,
+      "loss": 4.1537,
+      "step": 764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9661030341340075e-06,
+      "loss": 4.1782,
+      "step": 765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.966055625790139e-06,
+      "loss": 4.0934,
+      "step": 766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9660082174462704e-06,
+      "loss": 4.173,
+      "step": 767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965960809102402e-06,
+      "loss": 4.1499,
+      "step": 768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965913400758534e-06,
+      "loss": 4.1307,
+      "step": 769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965865992414665e-06,
+      "loss": 4.1337,
+      "step": 770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9658185840707964e-06,
+      "loss": 4.224,
+      "step": 771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965771175726928e-06,
+      "loss": 4.213,
+      "step": 772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9657237673830594e-06,
+      "loss": 4.0218,
+      "step": 773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9656763590391907e-06,
+      "loss": 4.1703,
+      "step": 774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9656289506953224e-06,
+      "loss": 4.1566,
+      "step": 775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965581542351454e-06,
+      "loss": 4.1867,
+      "step": 776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9655341340075854e-06,
+      "loss": 4.1494,
+      "step": 777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965486725663717e-06,
+      "loss": 4.1502,
+      "step": 778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9654393173198483e-06,
+      "loss": 4.205,
+      "step": 779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9653919089759796e-06,
+      "loss": 4.1627,
+      "step": 780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9653445006321113e-06,
+      "loss": 4.1027,
+      "step": 781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.965297092288243e-06,
+      "loss": 4.1603,
+      "step": 782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9652496839443743e-06,
+      "loss": 4.0646,
+      "step": 783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9652022756005056e-06,
+      "loss": 4.226,
+      "step": 784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9651548672566373e-06,
+      "loss": 4.0879,
+      "step": 785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9651074589127686e-06,
+      "loss": 4.1961,
+      "step": 786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9650600505689e-06,
+      "loss": 4.1201,
+      "step": 787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9650126422250316e-06,
+      "loss": 4.1498,
+      "step": 788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9649652338811633e-06,
+      "loss": 4.0993,
+      "step": 789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9649178255372945e-06,
+      "loss": 4.1056,
+      "step": 790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9648704171934262e-06,
+      "loss": 4.1578,
+      "step": 791
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9648230088495575e-06,
+      "loss": 4.1918,
+      "step": 792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.964775600505689e-06,
+      "loss": 4.2491,
+      "step": 793
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9647281921618205e-06,
+      "loss": 4.0093,
+      "step": 794
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9646807838179522e-06,
+      "loss": 4.1308,
+      "step": 795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9646333754740835e-06,
+      "loss": 4.2455,
+      "step": 796
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.964585967130215e-06,
+      "loss": 4.1396,
+      "step": 797
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9645385587863465e-06,
+      "loss": 4.1379,
+      "step": 798
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9644911504424778e-06,
+      "loss": 4.2133,
+      "step": 799
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9644437420986095e-06,
+      "loss": 4.0292,
+      "step": 800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9643963337547407e-06,
+      "loss": 4.1358,
+      "step": 801
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9643489254108725e-06,
+      "loss": 4.2542,
+      "step": 802
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.964301517067004e-06,
+      "loss": 4.2154,
+      "step": 803
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9642541087231354e-06,
+      "loss": 4.1536,
+      "step": 804
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9642067003792667e-06,
+      "loss": 4.1568,
+      "step": 805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.964159292035398e-06,
+      "loss": 4.1496,
+      "step": 806
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9641118836915297e-06,
+      "loss": 4.1589,
+      "step": 807
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9640644753476614e-06,
+      "loss": 4.192,
+      "step": 808
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9640170670037927e-06,
+      "loss": 4.0671,
+      "step": 809
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9639696586599244e-06,
+      "loss": 4.1171,
+      "step": 810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9639222503160557e-06,
+      "loss": 4.2151,
+      "step": 811
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963874841972187e-06,
+      "loss": 4.1831,
+      "step": 812
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9638274336283187e-06,
+      "loss": 4.0896,
+      "step": 813
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.96378002528445e-06,
+      "loss": 4.2439,
+      "step": 814
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9637326169405816e-06,
+      "loss": 3.9894,
+      "step": 815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9636852085967133e-06,
+      "loss": 4.2469,
+      "step": 816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9636378002528446e-06,
+      "loss": 4.1891,
+      "step": 817
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963590391908976e-06,
+      "loss": 4.035,
+      "step": 818
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9635429835651076e-06,
+      "loss": 4.1405,
+      "step": 819
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963495575221239e-06,
+      "loss": 4.1034,
+      "step": 820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9634481668773706e-06,
+      "loss": 4.2191,
+      "step": 821
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9634007585335023e-06,
+      "loss": 4.1699,
+      "step": 822
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9633533501896336e-06,
+      "loss": 4.1181,
+      "step": 823
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963305941845765e-06,
+      "loss": 4.1375,
+      "step": 824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9632585335018966e-06,
+      "loss": 4.0706,
+      "step": 825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963211125158028e-06,
+      "loss": 4.1834,
+      "step": 826
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963163716814159e-06,
+      "loss": 4.1463,
+      "step": 827
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963116308470291e-06,
+      "loss": 4.1897,
+      "step": 828
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9630689001264225e-06,
+      "loss": 4.0136,
+      "step": 829
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.963021491782554e-06,
+      "loss": 4.1469,
+      "step": 830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962974083438685e-06,
+      "loss": 4.1078,
+      "step": 831
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962926675094817e-06,
+      "loss": 3.9845,
+      "step": 832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962879266750948e-06,
+      "loss": 4.1165,
+      "step": 833
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9628318584070798e-06,
+      "loss": 4.1084,
+      "step": 834
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9627844500632115e-06,
+      "loss": 4.0721,
+      "step": 835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9627370417193428e-06,
+      "loss": 4.1025,
+      "step": 836
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962689633375474e-06,
+      "loss": 4.1232,
+      "step": 837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9626422250316057e-06,
+      "loss": 4.1264,
+      "step": 838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962594816687737e-06,
+      "loss": 4.0966,
+      "step": 839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9625474083438683e-06,
+      "loss": 4.146,
+      "step": 840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9625e-06,
+      "loss": 4.1733,
+      "step": 841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9624525916561317e-06,
+      "loss": 4.0885,
+      "step": 842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962405183312263e-06,
+      "loss": 4.0573,
+      "step": 843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9623577749683947e-06,
+      "loss": 4.2391,
+      "step": 844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962310366624526e-06,
+      "loss": 4.244,
+      "step": 845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9622629582806573e-06,
+      "loss": 4.1009,
+      "step": 846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962215549936789e-06,
+      "loss": 4.3184,
+      "step": 847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9621681415929207e-06,
+      "loss": 4.1935,
+      "step": 848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962120733249052e-06,
+      "loss": 4.1209,
+      "step": 849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9620733249051832e-06,
+      "loss": 4.1975,
+      "step": 850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.962025916561315e-06,
+      "loss": 4.174,
+      "step": 851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9619785082174462e-06,
+      "loss": 4.1066,
+      "step": 852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.9619310998735775e-06,
+      "loss": 3.947,
+      "step": 853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.961883691529709e-06,
+      "loss": 4.2256,
+      "step": 854
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961836283185841e-06,
+      "loss": 4.2316,
+      "step": 855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961788874841972e-06,
+      "loss": 4.1844,
+      "step": 856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961741466498104e-06,
+      "loss": 4.238,
+      "step": 857
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961694058154235e-06,
+      "loss": 4.055,
+      "step": 858
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9616466498103664e-06,
+      "loss": 4.0379,
+      "step": 859
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961599241466498e-06,
+      "loss": 4.0116,
+      "step": 860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.96155183312263e-06,
+      "loss": 4.1706,
+      "step": 861
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961504424778761e-06,
+      "loss": 4.0714,
+      "step": 862
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961457016434893e-06,
+      "loss": 4.085,
+      "step": 863
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961409608091024e-06,
+      "loss": 4.0785,
+      "step": 864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9613621997471554e-06,
+      "loss": 4.1329,
+      "step": 865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961314791403287e-06,
+      "loss": 4.0246,
+      "step": 866
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9612673830594184e-06,
+      "loss": 4.269,
+      "step": 867
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.96121997471555e-06,
+      "loss": 4.0927,
+      "step": 868
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961172566371682e-06,
+      "loss": 4.0739,
+      "step": 869
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.961125158027813e-06,
+      "loss": 4.1761,
+      "step": 870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9610777496839444e-06,
+      "loss": 4.0962,
+      "step": 871
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9610303413400756e-06,
+      "loss": 4.1076,
+      "step": 872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9609829329962073e-06,
+      "loss": 4.1361,
+      "step": 873
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.960935524652339e-06,
+      "loss": 4.1799,
+      "step": 874
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9608881163084703e-06,
+      "loss": 4.1136,
+      "step": 875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.960840707964602e-06,
+      "loss": 4.2225,
+      "step": 876
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9607932996207333e-06,
+      "loss": 4.2053,
+      "step": 877
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9607458912768646e-06,
+      "loss": 4.034,
+      "step": 878
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9606984829329963e-06,
+      "loss": 4.0098,
+      "step": 879
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9606510745891276e-06,
+      "loss": 4.1265,
+      "step": 880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9606036662452593e-06,
+      "loss": 4.2038,
+      "step": 881
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.960556257901391e-06,
+      "loss": 3.9704,
+      "step": 882
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9605088495575223e-06,
+      "loss": 4.0547,
+      "step": 883
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9604614412136535e-06,
+      "loss": 4.2139,
+      "step": 884
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9604140328697852e-06,
+      "loss": 4.111,
+      "step": 885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9603666245259165e-06,
+      "loss": 4.2521,
+      "step": 886
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9603192161820482e-06,
+      "loss": 4.2332,
+      "step": 887
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.96027180783818e-06,
+      "loss": 3.969,
+      "step": 888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9602243994943112e-06,
+      "loss": 4.0653,
+      "step": 889
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9601769911504425e-06,
+      "loss": 3.9923,
+      "step": 890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.960129582806574e-06,
+      "loss": 4.2419,
+      "step": 891
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9600821744627055e-06,
+      "loss": 4.1889,
+      "step": 892
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9600347661188368e-06,
+      "loss": 4.2489,
+      "step": 893
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9599873577749685e-06,
+      "loss": 3.9089,
+      "step": 894
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9599399494311e-06,
+      "loss": 4.2477,
+      "step": 895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9598925410872314e-06,
+      "loss": 4.2279,
+      "step": 896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9598451327433627e-06,
+      "loss": 4.1546,
+      "step": 897
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9597977243994944e-06,
+      "loss": 4.1733,
+      "step": 898
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9597503160556257e-06,
+      "loss": 4.1714,
+      "step": 899
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9597029077117574e-06,
+      "loss": 4.1353,
+      "step": 900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.959655499367889e-06,
+      "loss": 4.2044,
+      "step": 901
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9596080910240204e-06,
+      "loss": 4.1463,
+      "step": 902
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9595606826801517e-06,
+      "loss": 4.1423,
+      "step": 903
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9595132743362834e-06,
+      "loss": 4.1637,
+      "step": 904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9594658659924147e-06,
+      "loss": 4.1564,
+      "step": 905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.959418457648546e-06,
+      "loss": 4.1354,
+      "step": 906
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9593710493046776e-06,
+      "loss": 3.9916,
+      "step": 907
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9593236409608094e-06,
+      "loss": 4.099,
+      "step": 908
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9592762326169406e-06,
+      "loss": 3.9682,
+      "step": 909
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9592288242730723e-06,
+      "loss": 4.161,
+      "step": 910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9591814159292036e-06,
+      "loss": 4.3077,
+      "step": 911
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.959134007585335e-06,
+      "loss": 4.1247,
+      "step": 912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9590865992414666e-06,
+      "loss": 4.0638,
+      "step": 913
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9590391908975983e-06,
+      "loss": 4.1002,
+      "step": 914
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9589917825537296e-06,
+      "loss": 3.9512,
+      "step": 915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.958944374209861e-06,
+      "loss": 4.0646,
+      "step": 916
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.9588969658659926e-06,
+      "loss": 4.1784,
+      "step": 917
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2.958849557522124e-06,
+      "loss": 4.1795,
+      "step": 918
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958802149178255e-06,
+      "loss": 4.1051,
+      "step": 919
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958754740834387e-06,
+      "loss": 4.1154,
+      "step": 920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9587073324905185e-06,
+      "loss": 4.0997,
+      "step": 921
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.95865992414665e-06,
+      "loss": 4.1309,
+      "step": 922
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9586125158027815e-06,
+      "loss": 4.0507,
+      "step": 923
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958565107458913e-06,
+      "loss": 3.9621,
+      "step": 924
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958517699115044e-06,
+      "loss": 4.2448,
+      "step": 925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9584702907711758e-06,
+      "loss": 4.1252,
+      "step": 926
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9584228824273075e-06,
+      "loss": 4.0603,
+      "step": 927
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9583754740834388e-06,
+      "loss": 3.9533,
+      "step": 928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9583280657395705e-06,
+      "loss": 4.1729,
+      "step": 929
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9582806573957018e-06,
+      "loss": 4.0378,
+      "step": 930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958233249051833e-06,
+      "loss": 4.2578,
+      "step": 931
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9581858407079647e-06,
+      "loss": 4.153,
+      "step": 932
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958138432364096e-06,
+      "loss": 4.2523,
+      "step": 933
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9580910240202277e-06,
+      "loss": 4.1556,
+      "step": 934
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.958043615676359e-06,
+      "loss": 4.1434,
+      "step": 935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9579962073324907e-06,
+      "loss": 4.2064,
+      "step": 936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957948798988622e-06,
+      "loss": 4.155,
+      "step": 937
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9579013906447533e-06,
+      "loss": 4.1558,
+      "step": 938
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957853982300885e-06,
+      "loss": 4.2462,
+      "step": 939
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9578065739570167e-06,
+      "loss": 4.1921,
+      "step": 940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957759165613148e-06,
+      "loss": 4.2363,
+      "step": 941
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9577117572692797e-06,
+      "loss": 4.0859,
+      "step": 942
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957664348925411e-06,
+      "loss": 4.1331,
+      "step": 943
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9576169405815422e-06,
+      "loss": 4.1972,
+      "step": 944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957569532237674e-06,
+      "loss": 4.0868,
+      "step": 945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957522123893805e-06,
+      "loss": 4.1375,
+      "step": 946
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957474715549937e-06,
+      "loss": 4.101,
+      "step": 947
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9574273072060686e-06,
+      "loss": 4.2038,
+      "step": 948
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9573798988622e-06,
+      "loss": 4.1246,
+      "step": 949
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957332490518331e-06,
+      "loss": 4.0924,
+      "step": 950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957285082174463e-06,
+      "loss": 4.0906,
+      "step": 951
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957237673830594e-06,
+      "loss": 4.2355,
+      "step": 952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957190265486726e-06,
+      "loss": 4.2059,
+      "step": 953
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9571428571428576e-06,
+      "loss": 4.0624,
+      "step": 954
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.957095448798989e-06,
+      "loss": 4.1029,
+      "step": 955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.95704804045512e-06,
+      "loss": 4.0863,
+      "step": 956
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9570006321112514e-06,
+      "loss": 3.9507,
+      "step": 957
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956953223767383e-06,
+      "loss": 4.017,
+      "step": 958
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9569058154235144e-06,
+      "loss": 4.1312,
+      "step": 959
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956858407079646e-06,
+      "loss": 4.2201,
+      "step": 960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956810998735778e-06,
+      "loss": 4.2141,
+      "step": 961
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956763590391909e-06,
+      "loss": 4.1666,
+      "step": 962
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9567161820480404e-06,
+      "loss": 4.2411,
+      "step": 963
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956668773704172e-06,
+      "loss": 4.0101,
+      "step": 964
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9566213653603033e-06,
+      "loss": 4.3808,
+      "step": 965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956573957016435e-06,
+      "loss": 4.0583,
+      "step": 966
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9565265486725668e-06,
+      "loss": 4.0658,
+      "step": 967
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956479140328698e-06,
+      "loss": 4.2621,
+      "step": 968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9564317319848293e-06,
+      "loss": 4.1883,
+      "step": 969
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956384323640961e-06,
+      "loss": 4.0354,
+      "step": 970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9563369152970923e-06,
+      "loss": 3.9603,
+      "step": 971
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9562895069532236e-06,
+      "loss": 4.2701,
+      "step": 972
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9562420986093553e-06,
+      "loss": 4.0983,
+      "step": 973
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.956194690265487e-06,
+      "loss": 4.1347,
+      "step": 974
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9561472819216183e-06,
+      "loss": 4.2272,
+      "step": 975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.95609987357775e-06,
+      "loss": 4.0638,
+      "step": 976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9560524652338813e-06,
+      "loss": 4.0949,
+      "step": 977
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9560050568900125e-06,
+      "loss": 4.2283,
+      "step": 978
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9559576485461442e-06,
+      "loss": 4.1615,
+      "step": 979
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.955910240202276e-06,
+      "loss": 4.0219,
+      "step": 980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.9558628318584072e-06,
+      "loss": 4.1478,
+      "step": 981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9558154235145385e-06,
+      "loss": 4.1703,
+      "step": 982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.95576801517067e-06,
+      "loss": 4.0745,
+      "step": 983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9557206068268015e-06,
+      "loss": 4.2458,
+      "step": 984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9556731984829328e-06,
+      "loss": 4.0679,
+      "step": 985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9556257901390645e-06,
+      "loss": 4.0221,
+      "step": 986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.955578381795196e-06,
+      "loss": 4.0602,
+      "step": 987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9555309734513275e-06,
+      "loss": 4.1082,
+      "step": 988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.955483565107459e-06,
+      "loss": 4.1165,
+      "step": 989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9554361567635904e-06,
+      "loss": 4.143,
+      "step": 990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9553887484197217e-06,
+      "loss": 4.0439,
+      "step": 991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9553413400758534e-06,
+      "loss": 4.0821,
+      "step": 992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.955293931731985e-06,
+      "loss": 4.1558,
+      "step": 993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9552465233881164e-06,
+      "loss": 4.0492,
+      "step": 994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.955199115044248e-06,
+      "loss": 4.0878,
+      "step": 995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9551517067003794e-06,
+      "loss": 4.2323,
+      "step": 996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9551042983565107e-06,
+      "loss": 4.148,
+      "step": 997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9550568900126424e-06,
+      "loss": 4.2224,
+      "step": 998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9550094816687737e-06,
+      "loss": 4.1953,
+      "step": 999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9549620733249054e-06,
+      "loss": 4.2408,
+      "step": 1000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9549146649810366e-06,
+      "loss": 4.1788,
+      "step": 1001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9548672566371683e-06,
+      "loss": 4.0469,
+      "step": 1002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9548198482932996e-06,
+      "loss": 4.1802,
+      "step": 1003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.954772439949431e-06,
+      "loss": 4.1799,
+      "step": 1004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9547250316055626e-06,
+      "loss": 4.1339,
+      "step": 1005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9546776232616943e-06,
+      "loss": 4.0346,
+      "step": 1006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9546302149178256e-06,
+      "loss": 4.0743,
+      "step": 1007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9545828065739573e-06,
+      "loss": 4.1106,
+      "step": 1008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9545353982300886e-06,
+      "loss": 4.1059,
+      "step": 1009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.95448798988622e-06,
+      "loss": 4.1761,
+      "step": 1010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9544405815423516e-06,
+      "loss": 4.2026,
+      "step": 1011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.954393173198483e-06,
+      "loss": 4.0751,
+      "step": 1012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9543457648546145e-06,
+      "loss": 4.2,
+      "step": 1013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9542983565107463e-06,
+      "loss": 4.1497,
+      "step": 1014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9542509481668775e-06,
+      "loss": 4.1164,
+      "step": 1015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.954203539823009e-06,
+      "loss": 4.1105,
+      "step": 1016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9541561314791405e-06,
+      "loss": 4.0535,
+      "step": 1017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.954108723135272e-06,
+      "loss": 4.187,
+      "step": 1018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.954061314791403e-06,
+      "loss": 4.1433,
+      "step": 1019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.954013906447535e-06,
+      "loss": 4.0144,
+      "step": 1020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9539664981036665e-06,
+      "loss": 4.1664,
+      "step": 1021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9539190897597978e-06,
+      "loss": 4.1696,
+      "step": 1022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953871681415929e-06,
+      "loss": 4.2269,
+      "step": 1023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9538242730720608e-06,
+      "loss": 4.3179,
+      "step": 1024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953776864728192e-06,
+      "loss": 4.1556,
+      "step": 1025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9537294563843237e-06,
+      "loss": 4.1037,
+      "step": 1026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9536820480404554e-06,
+      "loss": 4.244,
+      "step": 1027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9536346396965867e-06,
+      "loss": 4.1022,
+      "step": 1028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953587231352718e-06,
+      "loss": 4.3252,
+      "step": 1029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9535398230088497e-06,
+      "loss": 4.1875,
+      "step": 1030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953492414664981e-06,
+      "loss": 4.0544,
+      "step": 1031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9534450063211123e-06,
+      "loss": 4.1132,
+      "step": 1032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953397597977244e-06,
+      "loss": 4.1353,
+      "step": 1033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9533501896333757e-06,
+      "loss": 4.2515,
+      "step": 1034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953302781289507e-06,
+      "loss": 4.0042,
+      "step": 1035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9532553729456387e-06,
+      "loss": 4.064,
+      "step": 1036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.95320796460177e-06,
+      "loss": 4.1846,
+      "step": 1037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9531605562579012e-06,
+      "loss": 4.1187,
+      "step": 1038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953113147914033e-06,
+      "loss": 4.1799,
+      "step": 1039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9530657395701646e-06,
+      "loss": 4.1401,
+      "step": 1040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.953018331226296e-06,
+      "loss": 4.2199,
+      "step": 1041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.9529709228824276e-06,
+      "loss": 4.145,
+      "step": 1042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.952923514538559e-06,
+      "loss": 4.109,
+      "step": 1043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.95287610619469e-06,
+      "loss": 4.0945,
+      "step": 1044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9528286978508215e-06,
+      "loss": 4.0816,
+      "step": 1045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952781289506953e-06,
+      "loss": 4.128,
+      "step": 1046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952733881163085e-06,
+      "loss": 4.1514,
+      "step": 1047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952686472819216e-06,
+      "loss": 4.0429,
+      "step": 1048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952639064475348e-06,
+      "loss": 4.0581,
+      "step": 1049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952591656131479e-06,
+      "loss": 4.2348,
+      "step": 1050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9525442477876104e-06,
+      "loss": 4.1063,
+      "step": 1051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952496839443742e-06,
+      "loss": 4.2269,
+      "step": 1052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952449431099874e-06,
+      "loss": 4.1183,
+      "step": 1053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952402022756005e-06,
+      "loss": 4.2416,
+      "step": 1054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952354614412137e-06,
+      "loss": 4.1494,
+      "step": 1055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952307206068268e-06,
+      "loss": 4.0932,
+      "step": 1056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9522597977243994e-06,
+      "loss": 3.9054,
+      "step": 1057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952212389380531e-06,
+      "loss": 4.0775,
+      "step": 1058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9521649810366623e-06,
+      "loss": 4.1595,
+      "step": 1059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952117572692794e-06,
+      "loss": 4.0093,
+      "step": 1060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9520701643489257e-06,
+      "loss": 4.0916,
+      "step": 1061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.952022756005057e-06,
+      "loss": 4.1245,
+      "step": 1062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9519753476611883e-06,
+      "loss": 4.2535,
+      "step": 1063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.95192793931732e-06,
+      "loss": 4.128,
+      "step": 1064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9518805309734513e-06,
+      "loss": 4.1508,
+      "step": 1065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951833122629583e-06,
+      "loss": 4.0608,
+      "step": 1066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9517857142857143e-06,
+      "loss": 4.1242,
+      "step": 1067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951738305941846e-06,
+      "loss": 4.1472,
+      "step": 1068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9516908975979773e-06,
+      "loss": 4.0789,
+      "step": 1069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9516434892541085e-06,
+      "loss": 4.1408,
+      "step": 1070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9515960809102402e-06,
+      "loss": 4.0846,
+      "step": 1071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9515486725663715e-06,
+      "loss": 4.1987,
+      "step": 1072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9515012642225032e-06,
+      "loss": 4.0286,
+      "step": 1073
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951453855878635e-06,
+      "loss": 4.2638,
+      "step": 1074
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9514064475347662e-06,
+      "loss": 3.8771,
+      "step": 1075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9513590391908975e-06,
+      "loss": 4.1209,
+      "step": 1076
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951311630847029e-06,
+      "loss": 4.0531,
+      "step": 1077
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9512642225031605e-06,
+      "loss": 4.0611,
+      "step": 1078
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951216814159292e-06,
+      "loss": 4.2038,
+      "step": 1079
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951169405815424e-06,
+      "loss": 4.2119,
+      "step": 1080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951121997471555e-06,
+      "loss": 4.179,
+      "step": 1081
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9510745891276864e-06,
+      "loss": 4.2031,
+      "step": 1082
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.951027180783818e-06,
+      "loss": 4.1386,
+      "step": 1083
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9509797724399494e-06,
+      "loss": 4.1032,
+      "step": 1084
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9509323640960807e-06,
+      "loss": 4.1057,
+      "step": 1085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9508849557522124e-06,
+      "loss": 4.0879,
+      "step": 1086
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.950837547408344e-06,
+      "loss": 4.1402,
+      "step": 1087
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9507901390644754e-06,
+      "loss": 4.13,
+      "step": 1088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9507427307206067e-06,
+      "loss": 4.191,
+      "step": 1089
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9506953223767384e-06,
+      "loss": 4.1434,
+      "step": 1090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9506479140328697e-06,
+      "loss": 4.1433,
+      "step": 1091
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9506005056890014e-06,
+      "loss": 4.1641,
+      "step": 1092
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.950553097345133e-06,
+      "loss": 4.1756,
+      "step": 1093
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9505056890012644e-06,
+      "loss": 4.2193,
+      "step": 1094
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9504582806573956e-06,
+      "loss": 4.2065,
+      "step": 1095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9504108723135273e-06,
+      "loss": 4.2104,
+      "step": 1096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9503634639696586e-06,
+      "loss": 4.088,
+      "step": 1097
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.95031605562579e-06,
+      "loss": 4.096,
+      "step": 1098
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9502686472819216e-06,
+      "loss": 3.9897,
+      "step": 1099
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9502212389380533e-06,
+      "loss": 4.1465,
+      "step": 1100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9501738305941846e-06,
+      "loss": 4.0578,
+      "step": 1101
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9501264222503163e-06,
+      "loss": 4.2139,
+      "step": 1102
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9500790139064476e-06,
+      "loss": 4.1746,
+      "step": 1103
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.950031605562579e-06,
+      "loss": 4.0181,
+      "step": 1104
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9499841972187106e-06,
+      "loss": 4.2555,
+      "step": 1105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9499367888748423e-06,
+      "loss": 4.1208,
+      "step": 1106
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9498893805309735e-06,
+      "loss": 4.0778,
+      "step": 1107
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.9498419721871052e-06,
+      "loss": 4.1079,
+      "step": 1108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9497945638432365e-06,
+      "loss": 4.2077,
+      "step": 1109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.949747155499368e-06,
+      "loss": 4.0324,
+      "step": 1110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.949699747155499e-06,
+      "loss": 4.0099,
+      "step": 1111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.949652338811631e-06,
+      "loss": 4.0942,
+      "step": 1112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9496049304677625e-06,
+      "loss": 4.107,
+      "step": 1113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9495575221238938e-06,
+      "loss": 4.1787,
+      "step": 1114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9495101137800255e-06,
+      "loss": 3.9308,
+      "step": 1115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9494627054361568e-06,
+      "loss": 4.1087,
+      "step": 1116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.949415297092288e-06,
+      "loss": 4.2667,
+      "step": 1117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9493678887484197e-06,
+      "loss": 4.1759,
+      "step": 1118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9493204804045514e-06,
+      "loss": 4.1332,
+      "step": 1119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9492730720606827e-06,
+      "loss": 4.043,
+      "step": 1120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9492256637168144e-06,
+      "loss": 4.0883,
+      "step": 1121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9491782553729457e-06,
+      "loss": 4.0924,
+      "step": 1122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.949130847029077e-06,
+      "loss": 4.0908,
+      "step": 1123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9490834386852087e-06,
+      "loss": 4.0893,
+      "step": 1124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.94903603034134e-06,
+      "loss": 4.1866,
+      "step": 1125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9489886219974717e-06,
+      "loss": 4.2972,
+      "step": 1126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9489412136536034e-06,
+      "loss": 4.2909,
+      "step": 1127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9488938053097347e-06,
+      "loss": 4.0896,
+      "step": 1128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948846396965866e-06,
+      "loss": 4.2206,
+      "step": 1129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9487989886219977e-06,
+      "loss": 4.2769,
+      "step": 1130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948751580278129e-06,
+      "loss": 3.886,
+      "step": 1131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9487041719342606e-06,
+      "loss": 4.09,
+      "step": 1132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948656763590392e-06,
+      "loss": 4.1531,
+      "step": 1133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9486093552465236e-06,
+      "loss": 4.227,
+      "step": 1134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948561946902655e-06,
+      "loss": 4.1162,
+      "step": 1135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948514538558786e-06,
+      "loss": 4.1445,
+      "step": 1136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948467130214918e-06,
+      "loss": 4.2361,
+      "step": 1137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948419721871049e-06,
+      "loss": 4.2103,
+      "step": 1138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948372313527181e-06,
+      "loss": 4.1027,
+      "step": 1139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9483249051833126e-06,
+      "loss": 4.2554,
+      "step": 1140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948277496839444e-06,
+      "loss": 3.9034,
+      "step": 1141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948230088495575e-06,
+      "loss": 4.088,
+      "step": 1142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948182680151707e-06,
+      "loss": 4.1717,
+      "step": 1143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.948135271807838e-06,
+      "loss": 4.0625,
+      "step": 1144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.94808786346397e-06,
+      "loss": 3.9939,
+      "step": 1145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9480404551201015e-06,
+      "loss": 3.9764,
+      "step": 1146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947993046776233e-06,
+      "loss": 4.1496,
+      "step": 1147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947945638432364e-06,
+      "loss": 4.1398,
+      "step": 1148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947898230088496e-06,
+      "loss": 4.041,
+      "step": 1149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947850821744627e-06,
+      "loss": 4.1842,
+      "step": 1150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9478034134007584e-06,
+      "loss": 4.2345,
+      "step": 1151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.94775600505689e-06,
+      "loss": 4.0556,
+      "step": 1152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9477085967130218e-06,
+      "loss": 4.2192,
+      "step": 1153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947661188369153e-06,
+      "loss": 4.0027,
+      "step": 1154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9476137800252843e-06,
+      "loss": 4.1098,
+      "step": 1155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947566371681416e-06,
+      "loss": 4.1981,
+      "step": 1156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9475189633375473e-06,
+      "loss": 4.2517,
+      "step": 1157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947471554993679e-06,
+      "loss": 4.1072,
+      "step": 1158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9474241466498107e-06,
+      "loss": 4.2569,
+      "step": 1159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947376738305942e-06,
+      "loss": 4.1466,
+      "step": 1160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9473293299620733e-06,
+      "loss": 3.936,
+      "step": 1161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947281921618205e-06,
+      "loss": 4.1987,
+      "step": 1162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9472345132743363e-06,
+      "loss": 4.2727,
+      "step": 1163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9471871049304675e-06,
+      "loss": 4.1549,
+      "step": 1164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9471396965865992e-06,
+      "loss": 4.112,
+      "step": 1165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.947092288242731e-06,
+      "loss": 3.8876,
+      "step": 1166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9470448798988622e-06,
+      "loss": 4.105,
+      "step": 1167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.946997471554994e-06,
+      "loss": 4.078,
+      "step": 1168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.946950063211125e-06,
+      "loss": 3.9327,
+      "step": 1169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.9469026548672565e-06,
+      "loss": 4.0708,
+      "step": 1170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.946855246523388e-06,
+      "loss": 4.117,
+      "step": 1171
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.94680783817952e-06,
+      "loss": 4.1916,
+      "step": 1172
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.946760429835651e-06,
+      "loss": 4.2102,
+      "step": 1173
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.946713021491783e-06,
+      "loss": 4.1461,
+      "step": 1174
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.946665613147914e-06,
+      "loss": 4.241,
+      "step": 1175
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9466182048040454e-06,
+      "loss": 3.9119,
+      "step": 1176
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9465707964601767e-06,
+      "loss": 4.2108,
+      "step": 1177
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9465233881163084e-06,
+      "loss": 4.317,
+      "step": 1178
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.94647597977244e-06,
+      "loss": 4.0473,
+      "step": 1179
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9464285714285714e-06,
+      "loss": 4.1737,
+      "step": 1180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.946381163084703e-06,
+      "loss": 4.1221,
+      "step": 1181
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9463337547408344e-06,
+      "loss": 4.2551,
+      "step": 1182
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9462863463969657e-06,
+      "loss": 4.1291,
+      "step": 1183
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9462389380530974e-06,
+      "loss": 4.1602,
+      "step": 1184
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.946191529709229e-06,
+      "loss": 4.2326,
+      "step": 1185
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9461441213653604e-06,
+      "loss": 4.1916,
+      "step": 1186
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.946096713021492e-06,
+      "loss": 4.2524,
+      "step": 1187
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9460493046776233e-06,
+      "loss": 4.2248,
+      "step": 1188
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9460018963337546e-06,
+      "loss": 4.1642,
+      "step": 1189
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9459544879898863e-06,
+      "loss": 3.9662,
+      "step": 1190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9459070796460176e-06,
+      "loss": 4.0883,
+      "step": 1191
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9458596713021493e-06,
+      "loss": 4.2126,
+      "step": 1192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.945812262958281e-06,
+      "loss": 4.01,
+      "step": 1193
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9457648546144123e-06,
+      "loss": 4.0343,
+      "step": 1194
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9457174462705436e-06,
+      "loss": 3.9068,
+      "step": 1195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9456700379266753e-06,
+      "loss": 4.1922,
+      "step": 1196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9456226295828066e-06,
+      "loss": 4.1842,
+      "step": 1197
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9455752212389383e-06,
+      "loss": 4.0741,
+      "step": 1198
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9455278128950696e-06,
+      "loss": 4.0221,
+      "step": 1199
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9454804045512013e-06,
+      "loss": 4.2238,
+      "step": 1200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9454329962073325e-06,
+      "loss": 4.1008,
+      "step": 1201
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.945385587863464e-06,
+      "loss": 4.0316,
+      "step": 1202
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9453381795195955e-06,
+      "loss": 4.0547,
+      "step": 1203
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.945290771175727e-06,
+      "loss": 4.113,
+      "step": 1204
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9452433628318585e-06,
+      "loss": 4.1698,
+      "step": 1205
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.94519595448799e-06,
+      "loss": 4.1726,
+      "step": 1206
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9451485461441215e-06,
+      "loss": 3.8852,
+      "step": 1207
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9451011378002528e-06,
+      "loss": 4.0406,
+      "step": 1208
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9450537294563845e-06,
+      "loss": 4.2049,
+      "step": 1209
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9450063211125158e-06,
+      "loss": 4.2912,
+      "step": 1210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9449589127686475e-06,
+      "loss": 4.3508,
+      "step": 1211
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944911504424779e-06,
+      "loss": 4.1948,
+      "step": 1212
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9448640960809104e-06,
+      "loss": 4.1045,
+      "step": 1213
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9448166877370417e-06,
+      "loss": 4.1026,
+      "step": 1214
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9447692793931734e-06,
+      "loss": 4.1628,
+      "step": 1215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9447218710493047e-06,
+      "loss": 4.2152,
+      "step": 1216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944674462705436e-06,
+      "loss": 3.9058,
+      "step": 1217
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9446270543615677e-06,
+      "loss": 3.9499,
+      "step": 1218
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9445796460176994e-06,
+      "loss": 4.1891,
+      "step": 1219
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9445322376738307e-06,
+      "loss": 4.0802,
+      "step": 1220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944484829329962e-06,
+      "loss": 4.2184,
+      "step": 1221
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9444374209860937e-06,
+      "loss": 4.2363,
+      "step": 1222
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944390012642225e-06,
+      "loss": 4.063,
+      "step": 1223
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9443426042983566e-06,
+      "loss": 4.1479,
+      "step": 1224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9442951959544883e-06,
+      "loss": 3.996,
+      "step": 1225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9442477876106196e-06,
+      "loss": 4.2449,
+      "step": 1226
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944200379266751e-06,
+      "loss": 4.0945,
+      "step": 1227
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9441529709228826e-06,
+      "loss": 4.2015,
+      "step": 1228
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944105562579014e-06,
+      "loss": 3.9093,
+      "step": 1229
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944058154235145e-06,
+      "loss": 4.142,
+      "step": 1230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.944010745891277e-06,
+      "loss": 4.2857,
+      "step": 1231
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9439633375474086e-06,
+      "loss": 3.9994,
+      "step": 1232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.94391592920354e-06,
+      "loss": 4.1323,
+      "step": 1233
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.9438685208596716e-06,
+      "loss": 4.0642,
+      "step": 1234
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943821112515803e-06,
+      "loss": 4.2018,
+      "step": 1235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943773704171934e-06,
+      "loss": 4.1037,
+      "step": 1236
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943726295828066e-06,
+      "loss": 4.1207,
+      "step": 1237
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9436788874841975e-06,
+      "loss": 3.9775,
+      "step": 1238
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943631479140329e-06,
+      "loss": 4.2825,
+      "step": 1239
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9435840707964605e-06,
+      "loss": 4.1183,
+      "step": 1240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943536662452592e-06,
+      "loss": 4.1432,
+      "step": 1241
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943489254108723e-06,
+      "loss": 4.3211,
+      "step": 1242
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9434418457648544e-06,
+      "loss": 4.1636,
+      "step": 1243
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943394437420986e-06,
+      "loss": 4.2305,
+      "step": 1244
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9433470290771178e-06,
+      "loss": 4.1547,
+      "step": 1245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943299620733249e-06,
+      "loss": 4.0787,
+      "step": 1246
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9432522123893808e-06,
+      "loss": 4.1464,
+      "step": 1247
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943204804045512e-06,
+      "loss": 4.1046,
+      "step": 1248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9431573957016433e-06,
+      "loss": 4.1577,
+      "step": 1249
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943109987357775e-06,
+      "loss": 4.0263,
+      "step": 1250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9430625790139067e-06,
+      "loss": 4.1121,
+      "step": 1251
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.943015170670038e-06,
+      "loss": 4.0204,
+      "step": 1252
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9429677623261697e-06,
+      "loss": 4.1455,
+      "step": 1253
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942920353982301e-06,
+      "loss": 4.0989,
+      "step": 1254
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9428729456384323e-06,
+      "loss": 4.1321,
+      "step": 1255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942825537294564e-06,
+      "loss": 4.1946,
+      "step": 1256
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9427781289506953e-06,
+      "loss": 4.0078,
+      "step": 1257
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942730720606827e-06,
+      "loss": 3.9287,
+      "step": 1258
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9426833122629587e-06,
+      "loss": 4.0251,
+      "step": 1259
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.94263590391909e-06,
+      "loss": 4.1703,
+      "step": 1260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9425884955752212e-06,
+      "loss": 4.2937,
+      "step": 1261
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942541087231353e-06,
+      "loss": 4.0869,
+      "step": 1262
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942493678887484e-06,
+      "loss": 4.1192,
+      "step": 1263
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942446270543616e-06,
+      "loss": 4.1989,
+      "step": 1264
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942398862199747e-06,
+      "loss": 4.0255,
+      "step": 1265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942351453855879e-06,
+      "loss": 4.2018,
+      "step": 1266
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.94230404551201e-06,
+      "loss": 4.2714,
+      "step": 1267
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9422566371681415e-06,
+      "loss": 3.947,
+      "step": 1268
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942209228824273e-06,
+      "loss": 4.1971,
+      "step": 1269
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9421618204804044e-06,
+      "loss": 4.2023,
+      "step": 1270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942114412136536e-06,
+      "loss": 4.0372,
+      "step": 1271
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942067003792668e-06,
+      "loss": 4.139,
+      "step": 1272
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.942019595448799e-06,
+      "loss": 4.1854,
+      "step": 1273
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9419721871049304e-06,
+      "loss": 4.1531,
+      "step": 1274
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941924778761062e-06,
+      "loss": 4.1938,
+      "step": 1275
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9418773704171934e-06,
+      "loss": 4.271,
+      "step": 1276
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941829962073325e-06,
+      "loss": 4.0484,
+      "step": 1277
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941782553729457e-06,
+      "loss": 4.1825,
+      "step": 1278
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941735145385588e-06,
+      "loss": 4.1326,
+      "step": 1279
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9416877370417194e-06,
+      "loss": 4.0847,
+      "step": 1280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941640328697851e-06,
+      "loss": 4.0629,
+      "step": 1281
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9415929203539823e-06,
+      "loss": 3.9992,
+      "step": 1282
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9415455120101136e-06,
+      "loss": 3.9901,
+      "step": 1283
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9414981036662453e-06,
+      "loss": 4.1442,
+      "step": 1284
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941450695322377e-06,
+      "loss": 4.2112,
+      "step": 1285
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9414032869785083e-06,
+      "loss": 4.0489,
+      "step": 1286
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9413558786346396e-06,
+      "loss": 4.0181,
+      "step": 1287
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9413084702907713e-06,
+      "loss": 4.0134,
+      "step": 1288
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9412610619469026e-06,
+      "loss": 4.2009,
+      "step": 1289
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941213653603034e-06,
+      "loss": 3.9739,
+      "step": 1290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.941166245259166e-06,
+      "loss": 4.059,
+      "step": 1291
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9411188369152973e-06,
+      "loss": 4.0344,
+      "step": 1292
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9410714285714285e-06,
+      "loss": 3.9755,
+      "step": 1293
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9410240202275603e-06,
+      "loss": 4.1,
+      "step": 1294
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9409766118836915e-06,
+      "loss": 4.1722,
+      "step": 1295
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.940929203539823e-06,
+      "loss": 4.0879,
+      "step": 1296
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9408817951959545e-06,
+      "loss": 4.226,
+      "step": 1297
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.9408343868520862e-06,
+      "loss": 4.0585,
+      "step": 1298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9407869785082175e-06,
+      "loss": 4.132,
+      "step": 1299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.940739570164349e-06,
+      "loss": 4.0705,
+      "step": 1300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9406921618204805e-06,
+      "loss": 4.0956,
+      "step": 1301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9406447534766118e-06,
+      "loss": 4.1851,
+      "step": 1302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9405973451327435e-06,
+      "loss": 3.9349,
+      "step": 1303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9405499367888747e-06,
+      "loss": 4.0462,
+      "step": 1304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9405025284450065e-06,
+      "loss": 4.1056,
+      "step": 1305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.940455120101138e-06,
+      "loss": 4.0661,
+      "step": 1306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9404077117572694e-06,
+      "loss": 4.2112,
+      "step": 1307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9403603034134007e-06,
+      "loss": 4.2179,
+      "step": 1308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.940312895069532e-06,
+      "loss": 4.0738,
+      "step": 1309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9402654867256637e-06,
+      "loss": 3.977,
+      "step": 1310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9402180783817954e-06,
+      "loss": 4.0249,
+      "step": 1311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9401706700379267e-06,
+      "loss": 4.2187,
+      "step": 1312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9401232616940584e-06,
+      "loss": 3.9658,
+      "step": 1313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9400758533501897e-06,
+      "loss": 4.0353,
+      "step": 1314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.940028445006321e-06,
+      "loss": 4.0957,
+      "step": 1315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9399810366624527e-06,
+      "loss": 4.1013,
+      "step": 1316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939933628318584e-06,
+      "loss": 4.1049,
+      "step": 1317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9398862199747156e-06,
+      "loss": 3.9827,
+      "step": 1318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9398388116308473e-06,
+      "loss": 4.2653,
+      "step": 1319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9397914032869786e-06,
+      "loss": 4.0123,
+      "step": 1320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.93974399494311e-06,
+      "loss": 4.1813,
+      "step": 1321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9396965865992416e-06,
+      "loss": 4.1906,
+      "step": 1322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939649178255373e-06,
+      "loss": 4.2771,
+      "step": 1323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9396017699115046e-06,
+      "loss": 4.1083,
+      "step": 1324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9395543615676363e-06,
+      "loss": 4.115,
+      "step": 1325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9395069532237676e-06,
+      "loss": 4.1048,
+      "step": 1326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939459544879899e-06,
+      "loss": 4.1408,
+      "step": 1327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9394121365360306e-06,
+      "loss": 4.0939,
+      "step": 1328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939364728192162e-06,
+      "loss": 4.1559,
+      "step": 1329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939317319848293e-06,
+      "loss": 4.0831,
+      "step": 1330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939269911504425e-06,
+      "loss": 4.1674,
+      "step": 1331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9392225031605565e-06,
+      "loss": 4.2985,
+      "step": 1332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939175094816688e-06,
+      "loss": 3.8415,
+      "step": 1333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939127686472819e-06,
+      "loss": 4.1935,
+      "step": 1334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939080278128951e-06,
+      "loss": 4.0226,
+      "step": 1335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.939032869785082e-06,
+      "loss": 4.0388,
+      "step": 1336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9389854614412138e-06,
+      "loss": 3.9691,
+      "step": 1337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9389380530973455e-06,
+      "loss": 4.0574,
+      "step": 1338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9388906447534768e-06,
+      "loss": 4.1859,
+      "step": 1339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938843236409608e-06,
+      "loss": 3.9354,
+      "step": 1340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9387958280657397e-06,
+      "loss": 4.1216,
+      "step": 1341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938748419721871e-06,
+      "loss": 4.1584,
+      "step": 1342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9387010113780023e-06,
+      "loss": 4.0555,
+      "step": 1343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938653603034134e-06,
+      "loss": 4.2101,
+      "step": 1344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9386061946902657e-06,
+      "loss": 4.2356,
+      "step": 1345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938558786346397e-06,
+      "loss": 4.2553,
+      "step": 1346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9385113780025287e-06,
+      "loss": 4.0717,
+      "step": 1347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.93846396965866e-06,
+      "loss": 3.882,
+      "step": 1348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9384165613147913e-06,
+      "loss": 4.2853,
+      "step": 1349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938369152970923e-06,
+      "loss": 4.0393,
+      "step": 1350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9383217446270547e-06,
+      "loss": 4.0863,
+      "step": 1351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938274336283186e-06,
+      "loss": 4.1697,
+      "step": 1352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9382269279393172e-06,
+      "loss": 4.2013,
+      "step": 1353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938179519595449e-06,
+      "loss": 4.2924,
+      "step": 1354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9381321112515802e-06,
+      "loss": 4.0417,
+      "step": 1355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.9380847029077115e-06,
+      "loss": 4.1356,
+      "step": 1356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.938037294563843e-06,
+      "loss": 4.1386,
+      "step": 1357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.937989886219975e-06,
+      "loss": 4.0238,
+      "step": 1358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.937942477876106e-06,
+      "loss": 4.0606,
+      "step": 1359
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.937895069532238e-06,
+      "loss": 4.0031,
+      "step": 1360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.937847661188369e-06,
+      "loss": 4.0225,
+      "step": 1361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9378002528445004e-06,
+      "loss": 4.1537,
+      "step": 1362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937752844500632e-06,
+      "loss": 4.1393,
+      "step": 1363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937705436156764e-06,
+      "loss": 4.0448,
+      "step": 1364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937658027812895e-06,
+      "loss": 4.1717,
+      "step": 1365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937610619469027e-06,
+      "loss": 4.0951,
+      "step": 1366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937563211125158e-06,
+      "loss": 4.0386,
+      "step": 1367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9375158027812894e-06,
+      "loss": 4.1369,
+      "step": 1368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937468394437421e-06,
+      "loss": 4.1824,
+      "step": 1369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9374209860935524e-06,
+      "loss": 3.9406,
+      "step": 1370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937373577749684e-06,
+      "loss": 4.1684,
+      "step": 1371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9373261694058154e-06,
+      "loss": 4.0726,
+      "step": 1372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937278761061947e-06,
+      "loss": 4.0164,
+      "step": 1373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9372313527180784e-06,
+      "loss": 4.1031,
+      "step": 1374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9371839443742096e-06,
+      "loss": 4.2397,
+      "step": 1375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9371365360303413e-06,
+      "loss": 3.9912,
+      "step": 1376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.937089127686473e-06,
+      "loss": 4.248,
+      "step": 1377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9370417193426043e-06,
+      "loss": 3.7427,
+      "step": 1378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.936994310998736e-06,
+      "loss": 4.0758,
+      "step": 1379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9369469026548673e-06,
+      "loss": 3.9651,
+      "step": 1380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9368994943109986e-06,
+      "loss": 4.1592,
+      "step": 1381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9368520859671303e-06,
+      "loss": 4.1963,
+      "step": 1382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9368046776232616e-06,
+      "loss": 4.0043,
+      "step": 1383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9367572692793933e-06,
+      "loss": 3.8387,
+      "step": 1384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.936709860935525e-06,
+      "loss": 4.0669,
+      "step": 1385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9366624525916563e-06,
+      "loss": 3.9493,
+      "step": 1386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9366150442477875e-06,
+      "loss": 4.2635,
+      "step": 1387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9365676359039192e-06,
+      "loss": 4.0407,
+      "step": 1388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9365202275600505e-06,
+      "loss": 4.2039,
+      "step": 1389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9364728192161822e-06,
+      "loss": 4.074,
+      "step": 1390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.936425410872314e-06,
+      "loss": 3.885,
+      "step": 1391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9363780025284452e-06,
+      "loss": 3.9729,
+      "step": 1392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9363305941845765e-06,
+      "loss": 4.0099,
+      "step": 1393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9362831858407078e-06,
+      "loss": 4.0802,
+      "step": 1394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9362357774968395e-06,
+      "loss": 4.3388,
+      "step": 1395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9361883691529708e-06,
+      "loss": 3.8594,
+      "step": 1396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9361409608091025e-06,
+      "loss": 3.907,
+      "step": 1397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.936093552465234e-06,
+      "loss": 3.9233,
+      "step": 1398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9360461441213654e-06,
+      "loss": 3.9407,
+      "step": 1399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9359987357774967e-06,
+      "loss": 4.1745,
+      "step": 1400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9359513274336284e-06,
+      "loss": 4.0953,
+      "step": 1401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9359039190897597e-06,
+      "loss": 4.1624,
+      "step": 1402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9358565107458914e-06,
+      "loss": 4.2931,
+      "step": 1403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.935809102402023e-06,
+      "loss": 4.0015,
+      "step": 1404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9357616940581544e-06,
+      "loss": 4.0018,
+      "step": 1405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9357142857142857e-06,
+      "loss": 4.0176,
+      "step": 1406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9356668773704174e-06,
+      "loss": 4.0143,
+      "step": 1407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9356194690265487e-06,
+      "loss": 3.9281,
+      "step": 1408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.93557206068268e-06,
+      "loss": 4.1367,
+      "step": 1409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9355246523388116e-06,
+      "loss": 4.0331,
+      "step": 1410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9354772439949434e-06,
+      "loss": 4.1817,
+      "step": 1411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9354298356510746e-06,
+      "loss": 4.1349,
+      "step": 1412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9353824273072063e-06,
+      "loss": 4.0046,
+      "step": 1413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9353350189633376e-06,
+      "loss": 4.2036,
+      "step": 1414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.935287610619469e-06,
+      "loss": 3.9457,
+      "step": 1415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9352402022756006e-06,
+      "loss": 4.1458,
+      "step": 1416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9351927939317323e-06,
+      "loss": 4.114,
+      "step": 1417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9351453855878636e-06,
+      "loss": 4.1212,
+      "step": 1418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.935097977243995e-06,
+      "loss": 4.067,
+      "step": 1419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9350505689001266e-06,
+      "loss": 4.1998,
+      "step": 1420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.935003160556258e-06,
+      "loss": 3.9938,
+      "step": 1421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.934955752212389e-06,
+      "loss": 3.9821,
+      "step": 1422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.934908343868521e-06,
+      "loss": 4.196,
+      "step": 1423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.9348609355246525e-06,
+      "loss": 4.2141,
+      "step": 1424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934813527180784e-06,
+      "loss": 4.2844,
+      "step": 1425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9347661188369155e-06,
+      "loss": 4.2387,
+      "step": 1426
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934718710493047e-06,
+      "loss": 4.2739,
+      "step": 1427
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934671302149178e-06,
+      "loss": 4.1828,
+      "step": 1428
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9346238938053098e-06,
+      "loss": 3.9212,
+      "step": 1429
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9345764854614415e-06,
+      "loss": 4.1459,
+      "step": 1430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9345290771175728e-06,
+      "loss": 4.101,
+      "step": 1431
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9344816687737045e-06,
+      "loss": 4.1561,
+      "step": 1432
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9344342604298358e-06,
+      "loss": 4.2515,
+      "step": 1433
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934386852085967e-06,
+      "loss": 4.0615,
+      "step": 1434
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9343394437420987e-06,
+      "loss": 4.0907,
+      "step": 1435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.93429203539823e-06,
+      "loss": 4.1146,
+      "step": 1436
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9342446270543617e-06,
+      "loss": 4.072,
+      "step": 1437
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934197218710493e-06,
+      "loss": 4.0467,
+      "step": 1438
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9341498103666247e-06,
+      "loss": 4.077,
+      "step": 1439
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934102402022756e-06,
+      "loss": 4.0709,
+      "step": 1440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9340549936788873e-06,
+      "loss": 3.9266,
+      "step": 1441
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.934007585335019e-06,
+      "loss": 4.1902,
+      "step": 1442
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9339601769911507e-06,
+      "loss": 4.0083,
+      "step": 1443
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933912768647282e-06,
+      "loss": 4.1918,
+      "step": 1444
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9338653603034137e-06,
+      "loss": 4.0686,
+      "step": 1445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933817951959545e-06,
+      "loss": 4.0703,
+      "step": 1446
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9337705436156762e-06,
+      "loss": 4.1165,
+      "step": 1447
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933723135271808e-06,
+      "loss": 4.1277,
+      "step": 1448
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933675726927939e-06,
+      "loss": 4.1341,
+      "step": 1449
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933628318584071e-06,
+      "loss": 4.0061,
+      "step": 1450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9335809102402026e-06,
+      "loss": 4.07,
+      "step": 1451
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933533501896334e-06,
+      "loss": 4.1117,
+      "step": 1452
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933486093552465e-06,
+      "loss": 4.0045,
+      "step": 1453
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933438685208597e-06,
+      "loss": 3.8781,
+      "step": 1454
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933391276864728e-06,
+      "loss": 4.126,
+      "step": 1455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.93334386852086e-06,
+      "loss": 4.0711,
+      "step": 1456
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9332964601769916e-06,
+      "loss": 4.0591,
+      "step": 1457
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933249051833123e-06,
+      "loss": 4.1818,
+      "step": 1458
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933201643489254e-06,
+      "loss": 4.0331,
+      "step": 1459
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9331542351453854e-06,
+      "loss": 3.9132,
+      "step": 1460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.933106826801517e-06,
+      "loss": 4.1462,
+      "step": 1461
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9330594184576484e-06,
+      "loss": 3.8732,
+      "step": 1462
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.93301201011378e-06,
+      "loss": 4.199,
+      "step": 1463
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932964601769912e-06,
+      "loss": 4.1738,
+      "step": 1464
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932917193426043e-06,
+      "loss": 4.1244,
+      "step": 1465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9328697850821744e-06,
+      "loss": 4.2393,
+      "step": 1466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932822376738306e-06,
+      "loss": 3.9654,
+      "step": 1467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9327749683944373e-06,
+      "loss": 4.0486,
+      "step": 1468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932727560050569e-06,
+      "loss": 4.0552,
+      "step": 1469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9326801517067008e-06,
+      "loss": 4.0365,
+      "step": 1470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932632743362832e-06,
+      "loss": 4.1778,
+      "step": 1471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9325853350189633e-06,
+      "loss": 3.9708,
+      "step": 1472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932537926675095e-06,
+      "loss": 4.1387,
+      "step": 1473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9324905183312263e-06,
+      "loss": 4.1178,
+      "step": 1474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9324431099873576e-06,
+      "loss": 4.1992,
+      "step": 1475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9323957016434893e-06,
+      "loss": 4.0579,
+      "step": 1476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932348293299621e-06,
+      "loss": 4.0772,
+      "step": 1477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9323008849557523e-06,
+      "loss": 3.9589,
+      "step": 1478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.932253476611884e-06,
+      "loss": 4.1615,
+      "step": 1479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9322060682680153e-06,
+      "loss": 4.0483,
+      "step": 1480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9321586599241465e-06,
+      "loss": 4.0367,
+      "step": 1481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9321112515802782e-06,
+      "loss": 3.9103,
+      "step": 1482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.93206384323641e-06,
+      "loss": 3.8405,
+      "step": 1483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9320164348925412e-06,
+      "loss": 4.2079,
+      "step": 1484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9319690265486725e-06,
+      "loss": 3.9313,
+      "step": 1485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.931921618204804e-06,
+      "loss": 3.9915,
+      "step": 1486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9318742098609355e-06,
+      "loss": 3.9369,
+      "step": 1487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.9318268015170668e-06,
+      "loss": 3.9615,
+      "step": 1488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9317793931731985e-06,
+      "loss": 4.0487,
+      "step": 1489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.93173198482933e-06,
+      "loss": 4.1135,
+      "step": 1490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9316845764854615e-06,
+      "loss": 4.0601,
+      "step": 1491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.931637168141593e-06,
+      "loss": 4.0889,
+      "step": 1492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9315897597977244e-06,
+      "loss": 4.2796,
+      "step": 1493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9315423514538557e-06,
+      "loss": 4.062,
+      "step": 1494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9314949431099874e-06,
+      "loss": 3.9481,
+      "step": 1495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.931447534766119e-06,
+      "loss": 3.9877,
+      "step": 1496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9314001264222504e-06,
+      "loss": 4.1689,
+      "step": 1497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.931352718078382e-06,
+      "loss": 4.0932,
+      "step": 1498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9313053097345134e-06,
+      "loss": 4.1131,
+      "step": 1499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9312579013906447e-06,
+      "loss": 3.8751,
+      "step": 1500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9312104930467764e-06,
+      "loss": 3.8196,
+      "step": 1501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9311630847029077e-06,
+      "loss": 4.0222,
+      "step": 1502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9311156763590394e-06,
+      "loss": 4.0615,
+      "step": 1503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9310682680151706e-06,
+      "loss": 4.2003,
+      "step": 1504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9310208596713023e-06,
+      "loss": 3.8816,
+      "step": 1505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9309734513274336e-06,
+      "loss": 4.1748,
+      "step": 1506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930926042983565e-06,
+      "loss": 4.1637,
+      "step": 1507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9308786346396966e-06,
+      "loss": 4.1162,
+      "step": 1508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9308312262958283e-06,
+      "loss": 4.1454,
+      "step": 1509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9307838179519596e-06,
+      "loss": 3.9986,
+      "step": 1510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9307364096080913e-06,
+      "loss": 4.2082,
+      "step": 1511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9306890012642226e-06,
+      "loss": 4.1051,
+      "step": 1512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930641592920354e-06,
+      "loss": 4.0977,
+      "step": 1513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9305941845764856e-06,
+      "loss": 3.8796,
+      "step": 1514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930546776232617e-06,
+      "loss": 4.0348,
+      "step": 1515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9304993678887485e-06,
+      "loss": 4.156,
+      "step": 1516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9304519595448803e-06,
+      "loss": 4.1649,
+      "step": 1517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9304045512010115e-06,
+      "loss": 4.0769,
+      "step": 1518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930357142857143e-06,
+      "loss": 4.1721,
+      "step": 1519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9303097345132745e-06,
+      "loss": 4.1971,
+      "step": 1520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930262326169406e-06,
+      "loss": 3.9533,
+      "step": 1521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9302149178255375e-06,
+      "loss": 3.9649,
+      "step": 1522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930167509481669e-06,
+      "loss": 4.0872,
+      "step": 1523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9301201011378005e-06,
+      "loss": 4.0484,
+      "step": 1524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9300726927939318e-06,
+      "loss": 4.0938,
+      "step": 1525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.930025284450063e-06,
+      "loss": 4.1229,
+      "step": 1526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9299778761061948e-06,
+      "loss": 4.2167,
+      "step": 1527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929930467762326e-06,
+      "loss": 4.2054,
+      "step": 1528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9298830594184577e-06,
+      "loss": 3.8781,
+      "step": 1529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9298356510745894e-06,
+      "loss": 3.8269,
+      "step": 1530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9297882427307207e-06,
+      "loss": 4.0778,
+      "step": 1531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929740834386852e-06,
+      "loss": 4.0622,
+      "step": 1532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9296934260429837e-06,
+      "loss": 4.0121,
+      "step": 1533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929646017699115e-06,
+      "loss": 4.1127,
+      "step": 1534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9295986093552467e-06,
+      "loss": 4.1216,
+      "step": 1535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9295512010113784e-06,
+      "loss": 4.1256,
+      "step": 1536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9295037926675097e-06,
+      "loss": 3.8752,
+      "step": 1537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929456384323641e-06,
+      "loss": 4.0122,
+      "step": 1538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9294089759797727e-06,
+      "loss": 4.1905,
+      "step": 1539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929361567635904e-06,
+      "loss": 4.0628,
+      "step": 1540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9293141592920352e-06,
+      "loss": 4.001,
+      "step": 1541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929266750948167e-06,
+      "loss": 4.1659,
+      "step": 1542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9292193426042986e-06,
+      "loss": 4.0825,
+      "step": 1543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.92917193426043e-06,
+      "loss": 3.9164,
+      "step": 1544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9291245259165616e-06,
+      "loss": 3.9541,
+      "step": 1545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929077117572693e-06,
+      "loss": 3.9482,
+      "step": 1546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.929029709228824e-06,
+      "loss": 3.9846,
+      "step": 1547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.928982300884956e-06,
+      "loss": 4.0892,
+      "step": 1548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.9289348925410876e-06,
+      "loss": 3.925,
+      "step": 1549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.928887484197219e-06,
+      "loss": 4.0835,
+      "step": 1550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.92884007585335e-06,
+      "loss": 3.9395,
+      "step": 1551
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928792667509482e-06,
+      "loss": 4.1436,
+      "step": 1552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928745259165613e-06,
+      "loss": 4.1154,
+      "step": 1553
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9286978508217444e-06,
+      "loss": 3.8451,
+      "step": 1554
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928650442477876e-06,
+      "loss": 4.0612,
+      "step": 1555
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928603034134008e-06,
+      "loss": 4.0575,
+      "step": 1556
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928555625790139e-06,
+      "loss": 4.1557,
+      "step": 1557
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928508217446271e-06,
+      "loss": 4.2769,
+      "step": 1558
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928460809102402e-06,
+      "loss": 4.0099,
+      "step": 1559
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9284134007585334e-06,
+      "loss": 4.279,
+      "step": 1560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928365992414665e-06,
+      "loss": 4.2163,
+      "step": 1561
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9283185840707968e-06,
+      "loss": 4.2183,
+      "step": 1562
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928271175726928e-06,
+      "loss": 4.0534,
+      "step": 1563
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9282237673830597e-06,
+      "loss": 4.1781,
+      "step": 1564
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928176359039191e-06,
+      "loss": 4.1045,
+      "step": 1565
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9281289506953223e-06,
+      "loss": 4.0635,
+      "step": 1566
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.928081542351454e-06,
+      "loss": 4.0528,
+      "step": 1567
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9280341340075853e-06,
+      "loss": 4.0234,
+      "step": 1568
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927986725663717e-06,
+      "loss": 3.9064,
+      "step": 1569
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9279393173198483e-06,
+      "loss": 4.1737,
+      "step": 1570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.92789190897598e-06,
+      "loss": 4.0933,
+      "step": 1571
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9278445006321113e-06,
+      "loss": 4.0081,
+      "step": 1572
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9277970922882425e-06,
+      "loss": 4.0991,
+      "step": 1573
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9277496839443742e-06,
+      "loss": 4.1019,
+      "step": 1574
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9277022756005055e-06,
+      "loss": 3.9113,
+      "step": 1575
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9276548672566372e-06,
+      "loss": 3.9294,
+      "step": 1576
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927607458912769e-06,
+      "loss": 4.1675,
+      "step": 1577
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9275600505689002e-06,
+      "loss": 4.2106,
+      "step": 1578
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9275126422250315e-06,
+      "loss": 4.1287,
+      "step": 1579
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927465233881163e-06,
+      "loss": 3.9913,
+      "step": 1580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9274178255372945e-06,
+      "loss": 3.9136,
+      "step": 1581
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927370417193426e-06,
+      "loss": 4.0814,
+      "step": 1582
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927323008849558e-06,
+      "loss": 4.1763,
+      "step": 1583
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927275600505689e-06,
+      "loss": 4.2487,
+      "step": 1584
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9272281921618205e-06,
+      "loss": 3.8891,
+      "step": 1585
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.927180783817952e-06,
+      "loss": 4.0152,
+      "step": 1586
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9271333754740834e-06,
+      "loss": 4.2051,
+      "step": 1587
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9270859671302147e-06,
+      "loss": 3.8414,
+      "step": 1588
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9270385587863464e-06,
+      "loss": 4.2541,
+      "step": 1589
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.926991150442478e-06,
+      "loss": 4.2101,
+      "step": 1590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9269437420986094e-06,
+      "loss": 4.1633,
+      "step": 1591
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9268963337547407e-06,
+      "loss": 4.2458,
+      "step": 1592
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9268489254108724e-06,
+      "loss": 4.2085,
+      "step": 1593
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9268015170670037e-06,
+      "loss": 3.9243,
+      "step": 1594
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9267541087231354e-06,
+      "loss": 4.198,
+      "step": 1595
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.926706700379267e-06,
+      "loss": 4.1727,
+      "step": 1596
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9266592920353984e-06,
+      "loss": 4.0814,
+      "step": 1597
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9266118836915296e-06,
+      "loss": 4.1798,
+      "step": 1598
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9265644753476613e-06,
+      "loss": 4.0625,
+      "step": 1599
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9265170670037926e-06,
+      "loss": 4.0124,
+      "step": 1600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.926469658659924e-06,
+      "loss": 4.1087,
+      "step": 1601
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9264222503160556e-06,
+      "loss": 4.0395,
+      "step": 1602
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9263748419721873e-06,
+      "loss": 4.288,
+      "step": 1603
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9263274336283186e-06,
+      "loss": 4.1966,
+      "step": 1604
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9262800252844503e-06,
+      "loss": 3.7007,
+      "step": 1605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9262326169405816e-06,
+      "loss": 4.0949,
+      "step": 1606
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.926185208596713e-06,
+      "loss": 4.1268,
+      "step": 1607
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9261378002528446e-06,
+      "loss": 4.2539,
+      "step": 1608
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9260903919089763e-06,
+      "loss": 4.1059,
+      "step": 1609
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9260429835651075e-06,
+      "loss": 4.0585,
+      "step": 1610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9259955752212392e-06,
+      "loss": 4.0508,
+      "step": 1611
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.9259481668773705e-06,
+      "loss": 4.02,
+      "step": 1612
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.925900758533502e-06,
+      "loss": 3.9654,
+      "step": 1613
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2.925853350189633e-06,
+      "loss": 4.0129,
+      "step": 1614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.925805941845765e-06,
+      "loss": 3.9963,
+      "step": 1615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9257585335018965e-06,
+      "loss": 4.1445,
+      "step": 1616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9257111251580278e-06,
+      "loss": 3.903,
+      "step": 1617
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9256637168141595e-06,
+      "loss": 4.1731,
+      "step": 1618
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9256163084702908e-06,
+      "loss": 4.03,
+      "step": 1619
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.925568900126422e-06,
+      "loss": 4.0474,
+      "step": 1620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9255214917825537e-06,
+      "loss": 3.9673,
+      "step": 1621
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9254740834386854e-06,
+      "loss": 4.0096,
+      "step": 1622
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9254266750948167e-06,
+      "loss": 4.0611,
+      "step": 1623
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9253792667509484e-06,
+      "loss": 4.0079,
+      "step": 1624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9253318584070797e-06,
+      "loss": 3.9632,
+      "step": 1625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.925284450063211e-06,
+      "loss": 4.0998,
+      "step": 1626
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9252370417193427e-06,
+      "loss": 3.9883,
+      "step": 1627
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.925189633375474e-06,
+      "loss": 3.9441,
+      "step": 1628
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9251422250316057e-06,
+      "loss": 3.9813,
+      "step": 1629
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9250948166877374e-06,
+      "loss": 3.8351,
+      "step": 1630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9250474083438687e-06,
+      "loss": 3.7786,
+      "step": 1631
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.925e-06,
+      "loss": 3.9306,
+      "step": 1632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9249525916561317e-06,
+      "loss": 4.1217,
+      "step": 1633
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924905183312263e-06,
+      "loss": 4.2989,
+      "step": 1634
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9248577749683946e-06,
+      "loss": 3.8814,
+      "step": 1635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924810366624526e-06,
+      "loss": 4.1631,
+      "step": 1636
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9247629582806576e-06,
+      "loss": 4.0859,
+      "step": 1637
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924715549936789e-06,
+      "loss": 4.2024,
+      "step": 1638
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.92466814159292e-06,
+      "loss": 4.0534,
+      "step": 1639
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924620733249052e-06,
+      "loss": 3.7832,
+      "step": 1640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924573324905183e-06,
+      "loss": 4.1624,
+      "step": 1641
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924525916561315e-06,
+      "loss": 4.0472,
+      "step": 1642
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9244785082174466e-06,
+      "loss": 4.1964,
+      "step": 1643
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924431099873578e-06,
+      "loss": 4.0373,
+      "step": 1644
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924383691529709e-06,
+      "loss": 3.9554,
+      "step": 1645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924336283185841e-06,
+      "loss": 3.9946,
+      "step": 1646
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924288874841972e-06,
+      "loss": 3.944,
+      "step": 1647
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924241466498104e-06,
+      "loss": 4.1761,
+      "step": 1648
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9241940581542355e-06,
+      "loss": 3.962,
+      "step": 1649
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924146649810367e-06,
+      "loss": 4.068,
+      "step": 1650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924099241466498e-06,
+      "loss": 4.2026,
+      "step": 1651
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.92405183312263e-06,
+      "loss": 4.0179,
+      "step": 1652
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.924004424778761e-06,
+      "loss": 4.1236,
+      "step": 1653
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9239570164348924e-06,
+      "loss": 3.9591,
+      "step": 1654
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923909608091024e-06,
+      "loss": 3.9981,
+      "step": 1655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9238621997471558e-06,
+      "loss": 4.1085,
+      "step": 1656
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923814791403287e-06,
+      "loss": 3.9886,
+      "step": 1657
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9237673830594183e-06,
+      "loss": 4.2233,
+      "step": 1658
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.92371997471555e-06,
+      "loss": 4.2918,
+      "step": 1659
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9236725663716813e-06,
+      "loss": 3.9949,
+      "step": 1660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923625158027813e-06,
+      "loss": 4.0637,
+      "step": 1661
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9235777496839447e-06,
+      "loss": 4.0471,
+      "step": 1662
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923530341340076e-06,
+      "loss": 4.1445,
+      "step": 1663
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9234829329962073e-06,
+      "loss": 3.8474,
+      "step": 1664
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923435524652339e-06,
+      "loss": 3.9584,
+      "step": 1665
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9233881163084703e-06,
+      "loss": 4.1891,
+      "step": 1666
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9233407079646015e-06,
+      "loss": 4.1979,
+      "step": 1667
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9232932996207332e-06,
+      "loss": 4.2564,
+      "step": 1668
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923245891276865e-06,
+      "loss": 4.1048,
+      "step": 1669
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9231984829329962e-06,
+      "loss": 4.1457,
+      "step": 1670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923151074589128e-06,
+      "loss": 4.2543,
+      "step": 1671
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923103666245259e-06,
+      "loss": 4.0447,
+      "step": 1672
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.9230562579013905e-06,
+      "loss": 4.2817,
+      "step": 1673
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.923008849557522e-06,
+      "loss": 4.1649,
+      "step": 1674
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.922961441213654e-06,
+      "loss": 4.1694,
+      "step": 1675
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.922914032869785e-06,
+      "loss": 4.0807,
+      "step": 1676
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.922866624525917e-06,
+      "loss": 3.8256,
+      "step": 1677
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2.922819216182048e-06,
+      "loss": 4.0296,
+      "step": 1678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9227718078381794e-06,
+      "loss": 4.0915,
+      "step": 1679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9227243994943107e-06,
+      "loss": 4.0726,
+      "step": 1680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9226769911504424e-06,
+      "loss": 4.178,
+      "step": 1681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.922629582806574e-06,
+      "loss": 3.9535,
+      "step": 1682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9225821744627054e-06,
+      "loss": 3.8731,
+      "step": 1683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.922534766118837e-06,
+      "loss": 4.191,
+      "step": 1684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9224873577749684e-06,
+      "loss": 3.8027,
+      "step": 1685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9224399494310997e-06,
+      "loss": 4.0907,
+      "step": 1686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9223925410872314e-06,
+      "loss": 3.7998,
+      "step": 1687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.922345132743363e-06,
+      "loss": 4.2242,
+      "step": 1688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9222977243994944e-06,
+      "loss": 4.2225,
+      "step": 1689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.922250316055626e-06,
+      "loss": 3.8911,
+      "step": 1690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9222029077117574e-06,
+      "loss": 4.2283,
+      "step": 1691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9221554993678886e-06,
+      "loss": 4.1881,
+      "step": 1692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9221080910240203e-06,
+      "loss": 4.2122,
+      "step": 1693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9220606826801516e-06,
+      "loss": 4.1137,
+      "step": 1694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9220132743362833e-06,
+      "loss": 4.1499,
+      "step": 1695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.921965865992415e-06,
+      "loss": 4.1639,
+      "step": 1696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9219184576485463e-06,
+      "loss": 4.1184,
+      "step": 1697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9218710493046776e-06,
+      "loss": 4.2769,
+      "step": 1698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9218236409608093e-06,
+      "loss": 3.7502,
+      "step": 1699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9217762326169406e-06,
+      "loss": 4.0214,
+      "step": 1700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9217288242730723e-06,
+      "loss": 4.261,
+      "step": 1701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9216814159292036e-06,
+      "loss": 4.0651,
+      "step": 1702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9216340075853353e-06,
+      "loss": 4.1646,
+      "step": 1703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9215865992414665e-06,
+      "loss": 4.0247,
+      "step": 1704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.921539190897598e-06,
+      "loss": 4.0533,
+      "step": 1705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9214917825537295e-06,
+      "loss": 4.1344,
+      "step": 1706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.921444374209861e-06,
+      "loss": 3.9021,
+      "step": 1707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9213969658659925e-06,
+      "loss": 4.0814,
+      "step": 1708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.921349557522124e-06,
+      "loss": 3.9887,
+      "step": 1709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9213021491782555e-06,
+      "loss": 4.0173,
+      "step": 1710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9212547408343868e-06,
+      "loss": 4.1727,
+      "step": 1711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9212073324905185e-06,
+      "loss": 4.1394,
+      "step": 1712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9211599241466498e-06,
+      "loss": 4.2392,
+      "step": 1713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9211125158027815e-06,
+      "loss": 4.0751,
+      "step": 1714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.921065107458913e-06,
+      "loss": 3.9955,
+      "step": 1715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9210176991150444e-06,
+      "loss": 4.108,
+      "step": 1716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9209702907711757e-06,
+      "loss": 4.0014,
+      "step": 1717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9209228824273074e-06,
+      "loss": 4.1344,
+      "step": 1718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9208754740834387e-06,
+      "loss": 3.9675,
+      "step": 1719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.92082806573957e-06,
+      "loss": 4.1697,
+      "step": 1720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9207806573957017e-06,
+      "loss": 4.2507,
+      "step": 1721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9207332490518334e-06,
+      "loss": 4.1583,
+      "step": 1722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9206858407079647e-06,
+      "loss": 4.1029,
+      "step": 1723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920638432364096e-06,
+      "loss": 4.0037,
+      "step": 1724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9205910240202277e-06,
+      "loss": 3.917,
+      "step": 1725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920543615676359e-06,
+      "loss": 4.0017,
+      "step": 1726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9204962073324906e-06,
+      "loss": 3.926,
+      "step": 1727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9204487989886223e-06,
+      "loss": 4.0029,
+      "step": 1728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9204013906447536e-06,
+      "loss": 3.7955,
+      "step": 1729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920353982300885e-06,
+      "loss": 4.2811,
+      "step": 1730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9203065739570166e-06,
+      "loss": 4.2806,
+      "step": 1731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920259165613148e-06,
+      "loss": 3.9586,
+      "step": 1732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920211757269279e-06,
+      "loss": 4.137,
+      "step": 1733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920164348925411e-06,
+      "loss": 3.9104,
+      "step": 1734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9201169405815426e-06,
+      "loss": 3.9685,
+      "step": 1735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.920069532237674e-06,
+      "loss": 4.2656,
+      "step": 1736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9200221238938056e-06,
+      "loss": 3.867,
+      "step": 1737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.919974715549937e-06,
+      "loss": 4.295,
+      "step": 1738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.919927307206068e-06,
+      "loss": 4.1714,
+      "step": 1739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9198798988622e-06,
+      "loss": 4.2114,
+      "step": 1740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.9198324905183315e-06,
+      "loss": 4.2338,
+      "step": 1741
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919785082174463e-06,
+      "loss": 3.9567,
+      "step": 1742
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9197376738305945e-06,
+      "loss": 3.9378,
+      "step": 1743
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919690265486726e-06,
+      "loss": 3.8931,
+      "step": 1744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919642857142857e-06,
+      "loss": 4.2045,
+      "step": 1745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9195954487989884e-06,
+      "loss": 4.059,
+      "step": 1746
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.91954804045512e-06,
+      "loss": 4.0523,
+      "step": 1747
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9195006321112518e-06,
+      "loss": 3.9225,
+      "step": 1748
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919453223767383e-06,
+      "loss": 4.2917,
+      "step": 1749
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9194058154235148e-06,
+      "loss": 4.0667,
+      "step": 1750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919358407079646e-06,
+      "loss": 3.9272,
+      "step": 1751
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9193109987357773e-06,
+      "loss": 3.8886,
+      "step": 1752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919263590391909e-06,
+      "loss": 3.9439,
+      "step": 1753
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9192161820480407e-06,
+      "loss": 4.3145,
+      "step": 1754
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919168773704172e-06,
+      "loss": 3.9663,
+      "step": 1755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9191213653603037e-06,
+      "loss": 3.9966,
+      "step": 1756
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.919073957016435e-06,
+      "loss": 3.773,
+      "step": 1757
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9190265486725663e-06,
+      "loss": 3.7502,
+      "step": 1758
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918979140328698e-06,
+      "loss": 4.3497,
+      "step": 1759
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9189317319848293e-06,
+      "loss": 4.1144,
+      "step": 1760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918884323640961e-06,
+      "loss": 4.0619,
+      "step": 1761
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9188369152970927e-06,
+      "loss": 4.2065,
+      "step": 1762
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918789506953224e-06,
+      "loss": 3.9566,
+      "step": 1763
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9187420986093552e-06,
+      "loss": 4.0169,
+      "step": 1764
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918694690265487e-06,
+      "loss": 4.277,
+      "step": 1765
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918647281921618e-06,
+      "loss": 3.9873,
+      "step": 1766
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.91859987357775e-06,
+      "loss": 4.2733,
+      "step": 1767
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918552465233881e-06,
+      "loss": 3.9936,
+      "step": 1768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918505056890013e-06,
+      "loss": 4.1087,
+      "step": 1769
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918457648546144e-06,
+      "loss": 3.9423,
+      "step": 1770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9184102402022755e-06,
+      "loss": 4.276,
+      "step": 1771
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918362831858407e-06,
+      "loss": 4.0809,
+      "step": 1772
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9183154235145384e-06,
+      "loss": 3.958,
+      "step": 1773
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.91826801517067e-06,
+      "loss": 3.9047,
+      "step": 1774
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918220606826802e-06,
+      "loss": 3.9816,
+      "step": 1775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918173198482933e-06,
+      "loss": 3.8424,
+      "step": 1776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9181257901390644e-06,
+      "loss": 4.1668,
+      "step": 1777
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.918078381795196e-06,
+      "loss": 4.3516,
+      "step": 1778
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9180309734513274e-06,
+      "loss": 4.1376,
+      "step": 1779
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.917983565107459e-06,
+      "loss": 4.1781,
+      "step": 1780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.917936156763591e-06,
+      "loss": 3.9821,
+      "step": 1781
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.917888748419722e-06,
+      "loss": 3.9551,
+      "step": 1782
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9178413400758534e-06,
+      "loss": 4.2266,
+      "step": 1783
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.917793931731985e-06,
+      "loss": 3.951,
+      "step": 1784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9177465233881163e-06,
+      "loss": 4.094,
+      "step": 1785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9176991150442476e-06,
+      "loss": 4.17,
+      "step": 1786
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9176517067003793e-06,
+      "loss": 4.1923,
+      "step": 1787
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.917604298356511e-06,
+      "loss": 4.0661,
+      "step": 1788
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9175568900126423e-06,
+      "loss": 3.9207,
+      "step": 1789
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9175094816687736e-06,
+      "loss": 4.0876,
+      "step": 1790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9174620733249053e-06,
+      "loss": 3.9656,
+      "step": 1791
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9174146649810366e-06,
+      "loss": 4.0986,
+      "step": 1792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9173672566371683e-06,
+      "loss": 3.9058,
+      "step": 1793
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9173198482933e-06,
+      "loss": 3.9717,
+      "step": 1794
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9172724399494313e-06,
+      "loss": 3.9711,
+      "step": 1795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9172250316055625e-06,
+      "loss": 4.1895,
+      "step": 1796
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9171776232616943e-06,
+      "loss": 4.1856,
+      "step": 1797
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9171302149178255e-06,
+      "loss": 4.0791,
+      "step": 1798
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.917082806573957e-06,
+      "loss": 4.1102,
+      "step": 1799
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9170353982300885e-06,
+      "loss": 3.9317,
+      "step": 1800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9169879898862202e-06,
+      "loss": 4.1138,
+      "step": 1801
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9169405815423515e-06,
+      "loss": 4.0539,
+      "step": 1802
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.916893173198483e-06,
+      "loss": 3.8762,
+      "step": 1803
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2.9168457648546145e-06,
+      "loss": 4.216,
+      "step": 1804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9167983565107458e-06,
+      "loss": 3.9284,
+      "step": 1805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9167509481668775e-06,
+      "loss": 4.0984,
+      "step": 1806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.916703539823009e-06,
+      "loss": 4.161,
+      "step": 1807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9166561314791405e-06,
+      "loss": 4.181,
+      "step": 1808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9166087231352717e-06,
+      "loss": 4.055,
+      "step": 1809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9165613147914034e-06,
+      "loss": 4.0941,
+      "step": 1810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9165139064475347e-06,
+      "loss": 4.3141,
+      "step": 1811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.916466498103666e-06,
+      "loss": 4.0376,
+      "step": 1812
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9164190897597977e-06,
+      "loss": 4.0642,
+      "step": 1813
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9163716814159294e-06,
+      "loss": 3.8347,
+      "step": 1814
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9163242730720607e-06,
+      "loss": 4.1238,
+      "step": 1815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9162768647281924e-06,
+      "loss": 4.3768,
+      "step": 1816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9162294563843237e-06,
+      "loss": 3.9648,
+      "step": 1817
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.916182048040455e-06,
+      "loss": 4.1498,
+      "step": 1818
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9161346396965867e-06,
+      "loss": 4.1869,
+      "step": 1819
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9160872313527184e-06,
+      "loss": 3.8973,
+      "step": 1820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9160398230088496e-06,
+      "loss": 4.053,
+      "step": 1821
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9159924146649813e-06,
+      "loss": 4.3146,
+      "step": 1822
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9159450063211126e-06,
+      "loss": 4.2696,
+      "step": 1823
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915897597977244e-06,
+      "loss": 4.0841,
+      "step": 1824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9158501896333756e-06,
+      "loss": 4.2665,
+      "step": 1825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915802781289507e-06,
+      "loss": 4.0001,
+      "step": 1826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9157553729456386e-06,
+      "loss": 4.1462,
+      "step": 1827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9157079646017703e-06,
+      "loss": 4.1675,
+      "step": 1828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9156605562579016e-06,
+      "loss": 3.9871,
+      "step": 1829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915613147914033e-06,
+      "loss": 4.0959,
+      "step": 1830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915565739570164e-06,
+      "loss": 4.1106,
+      "step": 1831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915518331226296e-06,
+      "loss": 3.9428,
+      "step": 1832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9154709228824275e-06,
+      "loss": 3.9713,
+      "step": 1833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915423514538559e-06,
+      "loss": 4.3012,
+      "step": 1834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9153761061946905e-06,
+      "loss": 4.2066,
+      "step": 1835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915328697850822e-06,
+      "loss": 4.1886,
+      "step": 1836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915281289506953e-06,
+      "loss": 4.0395,
+      "step": 1837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915233881163085e-06,
+      "loss": 4.2044,
+      "step": 1838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.915186472819216e-06,
+      "loss": 4.099,
+      "step": 1839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9151390644753478e-06,
+      "loss": 4.0413,
+      "step": 1840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9150916561314795e-06,
+      "loss": 3.9761,
+      "step": 1841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9150442477876108e-06,
+      "loss": 4.0831,
+      "step": 1842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914996839443742e-06,
+      "loss": 4.0665,
+      "step": 1843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9149494310998737e-06,
+      "loss": 3.9004,
+      "step": 1844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914902022756005e-06,
+      "loss": 4.1127,
+      "step": 1845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9148546144121367e-06,
+      "loss": 4.1846,
+      "step": 1846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9148072060682684e-06,
+      "loss": 4.1462,
+      "step": 1847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9147597977243997e-06,
+      "loss": 3.9103,
+      "step": 1848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914712389380531e-06,
+      "loss": 3.891,
+      "step": 1849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9146649810366627e-06,
+      "loss": 3.815,
+      "step": 1850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914617572692794e-06,
+      "loss": 4.01,
+      "step": 1851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9145701643489253e-06,
+      "loss": 3.9164,
+      "step": 1852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914522756005057e-06,
+      "loss": 3.8432,
+      "step": 1853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9144753476611887e-06,
+      "loss": 3.9945,
+      "step": 1854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.91442793931732e-06,
+      "loss": 4.1616,
+      "step": 1855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9143805309734512e-06,
+      "loss": 3.9861,
+      "step": 1856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914333122629583e-06,
+      "loss": 3.911,
+      "step": 1857
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9142857142857142e-06,
+      "loss": 4.072,
+      "step": 1858
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9142383059418455e-06,
+      "loss": 3.948,
+      "step": 1859
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9141908975979776e-06,
+      "loss": 3.9158,
+      "step": 1860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914143489254109e-06,
+      "loss": 3.7377,
+      "step": 1861
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.91409608091024e-06,
+      "loss": 3.9698,
+      "step": 1862
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914048672566372e-06,
+      "loss": 3.8836,
+      "step": 1863
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.914001264222503e-06,
+      "loss": 4.0802,
+      "step": 1864
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.9139538558786344e-06,
+      "loss": 4.1221,
+      "step": 1865
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.913906447534766e-06,
+      "loss": 4.1804,
+      "step": 1866
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.913859039190898e-06,
+      "loss": 4.1697,
+      "step": 1867
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2.913811630847029e-06,
+      "loss": 4.2467,
+      "step": 1868
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.913764222503161e-06,
+      "loss": 3.9838,
+      "step": 1869
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.913716814159292e-06,
+      "loss": 3.9081,
+      "step": 1870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9136694058154234e-06,
+      "loss": 4.1755,
+      "step": 1871
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.913621997471555e-06,
+      "loss": 4.1833,
+      "step": 1872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9135745891276864e-06,
+      "loss": 3.8736,
+      "step": 1873
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.913527180783818e-06,
+      "loss": 4.0826,
+      "step": 1874
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9134797724399494e-06,
+      "loss": 4.2258,
+      "step": 1875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.913432364096081e-06,
+      "loss": 3.9607,
+      "step": 1876
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9133849557522124e-06,
+      "loss": 4.0759,
+      "step": 1877
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9133375474083436e-06,
+      "loss": 4.2524,
+      "step": 1878
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9132901390644753e-06,
+      "loss": 4.0123,
+      "step": 1879
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.913242730720607e-06,
+      "loss": 3.8727,
+      "step": 1880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9131953223767383e-06,
+      "loss": 4.0066,
+      "step": 1881
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.91314791403287e-06,
+      "loss": 4.1067,
+      "step": 1882
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9131005056890013e-06,
+      "loss": 4.3717,
+      "step": 1883
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9130530973451326e-06,
+      "loss": 4.1703,
+      "step": 1884
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9130056890012643e-06,
+      "loss": 3.9574,
+      "step": 1885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9129582806573956e-06,
+      "loss": 4.1786,
+      "step": 1886
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9129108723135273e-06,
+      "loss": 3.9422,
+      "step": 1887
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.912863463969659e-06,
+      "loss": 4.1412,
+      "step": 1888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9128160556257903e-06,
+      "loss": 4.2543,
+      "step": 1889
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9127686472819215e-06,
+      "loss": 3.8727,
+      "step": 1890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9127212389380532e-06,
+      "loss": 3.8226,
+      "step": 1891
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9126738305941845e-06,
+      "loss": 4.1116,
+      "step": 1892
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9126264222503162e-06,
+      "loss": 4.1174,
+      "step": 1893
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.912579013906448e-06,
+      "loss": 3.9632,
+      "step": 1894
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9125316055625792e-06,
+      "loss": 4.1695,
+      "step": 1895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9124841972187105e-06,
+      "loss": 4.2596,
+      "step": 1896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9124367888748418e-06,
+      "loss": 4.2269,
+      "step": 1897
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9123893805309735e-06,
+      "loss": 3.9789,
+      "step": 1898
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9123419721871048e-06,
+      "loss": 3.9258,
+      "step": 1899
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9122945638432365e-06,
+      "loss": 4.1272,
+      "step": 1900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.912247155499368e-06,
+      "loss": 4.3055,
+      "step": 1901
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9121997471554994e-06,
+      "loss": 4.154,
+      "step": 1902
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9121523388116307e-06,
+      "loss": 4.2024,
+      "step": 1903
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9121049304677624e-06,
+      "loss": 4.2733,
+      "step": 1904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9120575221238937e-06,
+      "loss": 4.2499,
+      "step": 1905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9120101137800254e-06,
+      "loss": 4.1629,
+      "step": 1906
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911962705436157e-06,
+      "loss": 3.9012,
+      "step": 1907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9119152970922884e-06,
+      "loss": 4.264,
+      "step": 1908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9118678887484197e-06,
+      "loss": 4.169,
+      "step": 1909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9118204804045514e-06,
+      "loss": 4.2046,
+      "step": 1910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9117730720606827e-06,
+      "loss": 4.0875,
+      "step": 1911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911725663716814e-06,
+      "loss": 4.2242,
+      "step": 1912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9116782553729456e-06,
+      "loss": 4.1748,
+      "step": 1913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9116308470290774e-06,
+      "loss": 4.2049,
+      "step": 1914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9115834386852086e-06,
+      "loss": 4.1209,
+      "step": 1915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9115360303413403e-06,
+      "loss": 3.8595,
+      "step": 1916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9114886219974716e-06,
+      "loss": 4.0149,
+      "step": 1917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911441213653603e-06,
+      "loss": 4.3518,
+      "step": 1918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9113938053097346e-06,
+      "loss": 4.082,
+      "step": 1919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9113463969658663e-06,
+      "loss": 3.9593,
+      "step": 1920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9112989886219976e-06,
+      "loss": 4.202,
+      "step": 1921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911251580278129e-06,
+      "loss": 4.0244,
+      "step": 1922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9112041719342606e-06,
+      "loss": 4.0008,
+      "step": 1923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911156763590392e-06,
+      "loss": 4.1127,
+      "step": 1924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911109355246523e-06,
+      "loss": 4.0874,
+      "step": 1925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.911061946902655e-06,
+      "loss": 4.2621,
+      "step": 1926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9110145385587865e-06,
+      "loss": 4.1557,
+      "step": 1927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.910967130214918e-06,
+      "loss": 4.0916,
+      "step": 1928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.9109197218710495e-06,
+      "loss": 4.1389,
+      "step": 1929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.910872313527181e-06,
+      "loss": 3.9051,
+      "step": 1930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2.910824905183312e-06,
+      "loss": 4.0391,
+      "step": 1931
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.910777496839444e-06,
+      "loss": 3.8418,
+      "step": 1932
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9107300884955755e-06,
+      "loss": 3.7633,
+      "step": 1933
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9106826801517068e-06,
+      "loss": 3.9421,
+      "step": 1934
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9106352718078385e-06,
+      "loss": 3.9776,
+      "step": 1935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9105878634639698e-06,
+      "loss": 4.1757,
+      "step": 1936
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.910540455120101e-06,
+      "loss": 4.2244,
+      "step": 1937
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9104930467762327e-06,
+      "loss": 3.8967,
+      "step": 1938
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.910445638432364e-06,
+      "loss": 4.1876,
+      "step": 1939
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9103982300884957e-06,
+      "loss": 4.2962,
+      "step": 1940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.910350821744627e-06,
+      "loss": 4.0257,
+      "step": 1941
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9103034134007587e-06,
+      "loss": 4.011,
+      "step": 1942
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.91025600505689e-06,
+      "loss": 4.2698,
+      "step": 1943
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9102085967130213e-06,
+      "loss": 4.028,
+      "step": 1944
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.910161188369153e-06,
+      "loss": 3.9248,
+      "step": 1945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9101137800252847e-06,
+      "loss": 4.0784,
+      "step": 1946
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.910066371681416e-06,
+      "loss": 3.9997,
+      "step": 1947
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9100189633375477e-06,
+      "loss": 3.9998,
+      "step": 1948
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909971554993679e-06,
+      "loss": 4.01,
+      "step": 1949
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9099241466498102e-06,
+      "loss": 3.8583,
+      "step": 1950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909876738305942e-06,
+      "loss": 4.0258,
+      "step": 1951
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909829329962073e-06,
+      "loss": 4.1177,
+      "step": 1952
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909781921618205e-06,
+      "loss": 3.9245,
+      "step": 1953
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9097345132743366e-06,
+      "loss": 4.1463,
+      "step": 1954
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909687104930468e-06,
+      "loss": 3.9665,
+      "step": 1955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909639696586599e-06,
+      "loss": 4.0715,
+      "step": 1956
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909592288242731e-06,
+      "loss": 4.1583,
+      "step": 1957
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909544879898862e-06,
+      "loss": 4.0502,
+      "step": 1958
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909497471554994e-06,
+      "loss": 3.6647,
+      "step": 1959
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9094500632111256e-06,
+      "loss": 4.1325,
+      "step": 1960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909402654867257e-06,
+      "loss": 4.0531,
+      "step": 1961
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909355246523388e-06,
+      "loss": 4.0425,
+      "step": 1962
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9093078381795194e-06,
+      "loss": 3.9976,
+      "step": 1963
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909260429835651e-06,
+      "loss": 4.0052,
+      "step": 1964
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9092130214917824e-06,
+      "loss": 4.038,
+      "step": 1965
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909165613147914e-06,
+      "loss": 4.1353,
+      "step": 1966
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909118204804046e-06,
+      "loss": 4.2497,
+      "step": 1967
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.909070796460177e-06,
+      "loss": 4.1822,
+      "step": 1968
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9090233881163084e-06,
+      "loss": 4.0952,
+      "step": 1969
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.90897597977244e-06,
+      "loss": 4.1827,
+      "step": 1970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9089285714285713e-06,
+      "loss": 4.0179,
+      "step": 1971
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908881163084703e-06,
+      "loss": 3.9156,
+      "step": 1972
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9088337547408348e-06,
+      "loss": 3.9998,
+      "step": 1973
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908786346396966e-06,
+      "loss": 4.1397,
+      "step": 1974
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9087389380530973e-06,
+      "loss": 4.1802,
+      "step": 1975
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908691529709229e-06,
+      "loss": 3.8552,
+      "step": 1976
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9086441213653603e-06,
+      "loss": 3.8022,
+      "step": 1977
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9085967130214916e-06,
+      "loss": 3.9179,
+      "step": 1978
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9085493046776233e-06,
+      "loss": 3.942,
+      "step": 1979
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908501896333755e-06,
+      "loss": 3.9299,
+      "step": 1980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9084544879898863e-06,
+      "loss": 4.0496,
+      "step": 1981
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908407079646018e-06,
+      "loss": 3.9668,
+      "step": 1982
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9083596713021493e-06,
+      "loss": 4.1018,
+      "step": 1983
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9083122629582805e-06,
+      "loss": 4.2607,
+      "step": 1984
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9082648546144122e-06,
+      "loss": 3.9916,
+      "step": 1985
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908217446270544e-06,
+      "loss": 4.1369,
+      "step": 1986
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9081700379266752e-06,
+      "loss": 3.8464,
+      "step": 1987
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9081226295828065e-06,
+      "loss": 4.3273,
+      "step": 1988
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.908075221238938e-06,
+      "loss": 4.174,
+      "step": 1989
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9080278128950695e-06,
+      "loss": 3.7604,
+      "step": 1990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9079804045512008e-06,
+      "loss": 4.1052,
+      "step": 1991
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9079329962073325e-06,
+      "loss": 3.9121,
+      "step": 1992
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.907885587863464e-06,
+      "loss": 4.2155,
+      "step": 1993
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2.9078381795195955e-06,
+      "loss": 4.0077,
+      "step": 1994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.907790771175727e-06,
+      "loss": 4.0123,
+      "step": 1995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9077433628318584e-06,
+      "loss": 3.9191,
+      "step": 1996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9076959544879897e-06,
+      "loss": 3.8777,
+      "step": 1997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9076485461441214e-06,
+      "loss": 4.1471,
+      "step": 1998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.907601137800253e-06,
+      "loss": 3.7764,
+      "step": 1999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9075537294563844e-06,
+      "loss": 3.9045,
+      "step": 2000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.907506321112516e-06,
+      "loss": 3.9859,
+      "step": 2001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9074589127686474e-06,
+      "loss": 4.0244,
+      "step": 2002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9074115044247787e-06,
+      "loss": 4.2619,
+      "step": 2003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9073640960809104e-06,
+      "loss": 4.0791,
+      "step": 2004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9073166877370417e-06,
+      "loss": 3.9237,
+      "step": 2005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9072692793931734e-06,
+      "loss": 3.6987,
+      "step": 2006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9072218710493046e-06,
+      "loss": 3.9855,
+      "step": 2007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9071744627054363e-06,
+      "loss": 3.8714,
+      "step": 2008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9071270543615676e-06,
+      "loss": 3.9522,
+      "step": 2009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.907079646017699e-06,
+      "loss": 4.3028,
+      "step": 2010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9070322376738306e-06,
+      "loss": 4.1197,
+      "step": 2011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9069848293299623e-06,
+      "loss": 4.018,
+      "step": 2012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9069374209860936e-06,
+      "loss": 3.9237,
+      "step": 2013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9068900126422253e-06,
+      "loss": 4.0458,
+      "step": 2014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9068426042983566e-06,
+      "loss": 3.9747,
+      "step": 2015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.906795195954488e-06,
+      "loss": 4.1047,
+      "step": 2016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9067477876106196e-06,
+      "loss": 3.9154,
+      "step": 2017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.906700379266751e-06,
+      "loss": 3.8834,
+      "step": 2018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9066529709228825e-06,
+      "loss": 4.1313,
+      "step": 2019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9066055625790143e-06,
+      "loss": 3.9005,
+      "step": 2020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9065581542351455e-06,
+      "loss": 4.2069,
+      "step": 2021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.906510745891277e-06,
+      "loss": 4.1474,
+      "step": 2022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9064633375474085e-06,
+      "loss": 4.107,
+      "step": 2023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.90641592920354e-06,
+      "loss": 3.9577,
+      "step": 2024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9063685208596715e-06,
+      "loss": 3.9959,
+      "step": 2025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.906321112515803e-06,
+      "loss": 4.195,
+      "step": 2026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9062737041719345e-06,
+      "loss": 3.8412,
+      "step": 2027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9062262958280658e-06,
+      "loss": 4.0561,
+      "step": 2028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.906178887484197e-06,
+      "loss": 3.8688,
+      "step": 2029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9061314791403288e-06,
+      "loss": 4.0781,
+      "step": 2030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.90608407079646e-06,
+      "loss": 4.0639,
+      "step": 2031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9060366624525917e-06,
+      "loss": 3.8669,
+      "step": 2032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9059892541087234e-06,
+      "loss": 3.993,
+      "step": 2033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9059418457648547e-06,
+      "loss": 4.2369,
+      "step": 2034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905894437420986e-06,
+      "loss": 4.1355,
+      "step": 2035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9058470290771177e-06,
+      "loss": 3.9703,
+      "step": 2036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905799620733249e-06,
+      "loss": 4.0791,
+      "step": 2037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9057522123893807e-06,
+      "loss": 4.1057,
+      "step": 2038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9057048040455124e-06,
+      "loss": 4.0018,
+      "step": 2039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9056573957016437e-06,
+      "loss": 3.7935,
+      "step": 2040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905609987357775e-06,
+      "loss": 4.244,
+      "step": 2041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9055625790139067e-06,
+      "loss": 4.2741,
+      "step": 2042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905515170670038e-06,
+      "loss": 4.2726,
+      "step": 2043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9054677623261692e-06,
+      "loss": 3.9207,
+      "step": 2044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905420353982301e-06,
+      "loss": 3.9185,
+      "step": 2045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9053729456384326e-06,
+      "loss": 4.3352,
+      "step": 2046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905325537294564e-06,
+      "loss": 4.0792,
+      "step": 2047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9052781289506956e-06,
+      "loss": 4.2594,
+      "step": 2048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905230720606827e-06,
+      "loss": 4.0015,
+      "step": 2049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905183312262958e-06,
+      "loss": 4.0252,
+      "step": 2050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.90513590391909e-06,
+      "loss": 3.9085,
+      "step": 2051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9050884955752216e-06,
+      "loss": 4.1477,
+      "step": 2052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.905041087231353e-06,
+      "loss": 4.1462,
+      "step": 2053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.904993678887484e-06,
+      "loss": 3.7387,
+      "step": 2054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.904946270543616e-06,
+      "loss": 4.0356,
+      "step": 2055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.904898862199747e-06,
+      "loss": 4.2833,
+      "step": 2056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.9048514538558784e-06,
+      "loss": 3.8355,
+      "step": 2057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2.90480404551201e-06,
+      "loss": 4.0428,
+      "step": 2058
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904756637168142e-06,
+      "loss": 4.192,
+      "step": 2059
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904709228824273e-06,
+      "loss": 4.2278,
+      "step": 2060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904661820480405e-06,
+      "loss": 4.0437,
+      "step": 2061
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904614412136536e-06,
+      "loss": 4.0119,
+      "step": 2062
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9045670037926674e-06,
+      "loss": 3.8696,
+      "step": 2063
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904519595448799e-06,
+      "loss": 4.1344,
+      "step": 2064
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9044721871049308e-06,
+      "loss": 4.0725,
+      "step": 2065
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904424778761062e-06,
+      "loss": 4.1872,
+      "step": 2066
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9043773704171937e-06,
+      "loss": 4.1943,
+      "step": 2067
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904329962073325e-06,
+      "loss": 3.9842,
+      "step": 2068
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9042825537294563e-06,
+      "loss": 4.1841,
+      "step": 2069
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904235145385588e-06,
+      "loss": 4.2432,
+      "step": 2070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9041877370417193e-06,
+      "loss": 4.1942,
+      "step": 2071
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904140328697851e-06,
+      "loss": 4.3065,
+      "step": 2072
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9040929203539823e-06,
+      "loss": 3.989,
+      "step": 2073
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.904045512010114e-06,
+      "loss": 4.0459,
+      "step": 2074
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9039981036662453e-06,
+      "loss": 3.8859,
+      "step": 2075
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9039506953223765e-06,
+      "loss": 4.0858,
+      "step": 2076
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9039032869785082e-06,
+      "loss": 4.0196,
+      "step": 2077
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.90385587863464e-06,
+      "loss": 3.9882,
+      "step": 2078
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9038084702907712e-06,
+      "loss": 4.1663,
+      "step": 2079
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903761061946903e-06,
+      "loss": 3.9965,
+      "step": 2080
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9037136536030342e-06,
+      "loss": 4.1452,
+      "step": 2081
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9036662452591655e-06,
+      "loss": 4.1147,
+      "step": 2082
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903618836915297e-06,
+      "loss": 4.0536,
+      "step": 2083
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9035714285714285e-06,
+      "loss": 4.0,
+      "step": 2084
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.90352402022756e-06,
+      "loss": 4.1656,
+      "step": 2085
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903476611883692e-06,
+      "loss": 4.037,
+      "step": 2086
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903429203539823e-06,
+      "loss": 3.9359,
+      "step": 2087
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9033817951959545e-06,
+      "loss": 3.7621,
+      "step": 2088
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903334386852086e-06,
+      "loss": 3.793,
+      "step": 2089
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9032869785082174e-06,
+      "loss": 4.2858,
+      "step": 2090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903239570164349e-06,
+      "loss": 4.1445,
+      "step": 2091
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903192161820481e-06,
+      "loss": 3.7971,
+      "step": 2092
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.903144753476612e-06,
+      "loss": 3.9123,
+      "step": 2093
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9030973451327434e-06,
+      "loss": 3.8996,
+      "step": 2094
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9030499367888747e-06,
+      "loss": 4.2218,
+      "step": 2095
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9030025284450064e-06,
+      "loss": 4.1857,
+      "step": 2096
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9029551201011377e-06,
+      "loss": 4.1014,
+      "step": 2097
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9029077117572694e-06,
+      "loss": 4.1559,
+      "step": 2098
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.902860303413401e-06,
+      "loss": 3.8675,
+      "step": 2099
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9028128950695324e-06,
+      "loss": 4.1796,
+      "step": 2100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9027654867256636e-06,
+      "loss": 4.0399,
+      "step": 2101
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9027180783817953e-06,
+      "loss": 4.0493,
+      "step": 2102
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9026706700379266e-06,
+      "loss": 4.0785,
+      "step": 2103
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9026232616940583e-06,
+      "loss": 3.9788,
+      "step": 2104
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.90257585335019e-06,
+      "loss": 4.0221,
+      "step": 2105
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9025284450063213e-06,
+      "loss": 4.1626,
+      "step": 2106
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9024810366624526e-06,
+      "loss": 3.9351,
+      "step": 2107
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9024336283185843e-06,
+      "loss": 4.0067,
+      "step": 2108
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9023862199747156e-06,
+      "loss": 4.0925,
+      "step": 2109
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.902338811630847e-06,
+      "loss": 4.1152,
+      "step": 2110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9022914032869786e-06,
+      "loss": 4.0412,
+      "step": 2111
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9022439949431103e-06,
+      "loss": 3.951,
+      "step": 2112
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9021965865992415e-06,
+      "loss": 4.2001,
+      "step": 2113
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9021491782553732e-06,
+      "loss": 3.8443,
+      "step": 2114
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9021017699115045e-06,
+      "loss": 4.0933,
+      "step": 2115
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.902054361567636e-06,
+      "loss": 4.0343,
+      "step": 2116
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9020069532237675e-06,
+      "loss": 4.1151,
+      "step": 2117
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9019595448798992e-06,
+      "loss": 4.1089,
+      "step": 2118
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9019121365360305e-06,
+      "loss": 3.9146,
+      "step": 2119
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9018647281921618e-06,
+      "loss": 4.0204,
+      "step": 2120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2.9018173198482935e-06,
+      "loss": 4.0631,
+      "step": 2121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9017699115044248e-06,
+      "loss": 4.0522,
+      "step": 2122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.901722503160556e-06,
+      "loss": 4.2901,
+      "step": 2123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9016750948166877e-06,
+      "loss": 4.0547,
+      "step": 2124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9016276864728194e-06,
+      "loss": 3.8402,
+      "step": 2125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9015802781289507e-06,
+      "loss": 4.3234,
+      "step": 2126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9015328697850824e-06,
+      "loss": 4.1122,
+      "step": 2127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9014854614412137e-06,
+      "loss": 4.1861,
+      "step": 2128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.901438053097345e-06,
+      "loss": 4.0762,
+      "step": 2129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9013906447534767e-06,
+      "loss": 4.0847,
+      "step": 2130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9013432364096084e-06,
+      "loss": 3.9976,
+      "step": 2131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9012958280657397e-06,
+      "loss": 3.9308,
+      "step": 2132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9012484197218714e-06,
+      "loss": 4.0948,
+      "step": 2133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9012010113780027e-06,
+      "loss": 3.9635,
+      "step": 2134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.901153603034134e-06,
+      "loss": 4.0452,
+      "step": 2135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9011061946902657e-06,
+      "loss": 4.0404,
+      "step": 2136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.901058786346397e-06,
+      "loss": 4.2698,
+      "step": 2137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9010113780025286e-06,
+      "loss": 4.1407,
+      "step": 2138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.90096396965866e-06,
+      "loss": 4.293,
+      "step": 2139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9009165613147916e-06,
+      "loss": 4.1029,
+      "step": 2140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900869152970923e-06,
+      "loss": 4.1476,
+      "step": 2141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900821744627054e-06,
+      "loss": 3.9025,
+      "step": 2142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900774336283186e-06,
+      "loss": 4.1905,
+      "step": 2143
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900726927939317e-06,
+      "loss": 4.0563,
+      "step": 2144
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900679519595449e-06,
+      "loss": 3.8321,
+      "step": 2145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9006321112515806e-06,
+      "loss": 4.145,
+      "step": 2146
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900584702907712e-06,
+      "loss": 4.1956,
+      "step": 2147
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900537294563843e-06,
+      "loss": 4.2698,
+      "step": 2148
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900489886219975e-06,
+      "loss": 4.1192,
+      "step": 2149
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900442477876106e-06,
+      "loss": 4.1467,
+      "step": 2150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900395069532238e-06,
+      "loss": 4.0326,
+      "step": 2151
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9003476611883695e-06,
+      "loss": 3.8905,
+      "step": 2152
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900300252844501e-06,
+      "loss": 4.1361,
+      "step": 2153
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900252844500632e-06,
+      "loss": 3.9002,
+      "step": 2154
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900205436156764e-06,
+      "loss": 3.9593,
+      "step": 2155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900158027812895e-06,
+      "loss": 4.0343,
+      "step": 2156
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9001106194690264e-06,
+      "loss": 3.9203,
+      "step": 2157
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.900063211125158e-06,
+      "loss": 4.2866,
+      "step": 2158
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.9000158027812898e-06,
+      "loss": 3.8636,
+      "step": 2159
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899968394437421e-06,
+      "loss": 4.0054,
+      "step": 2160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8999209860935523e-06,
+      "loss": 3.8791,
+      "step": 2161
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899873577749684e-06,
+      "loss": 4.0784,
+      "step": 2162
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8998261694058153e-06,
+      "loss": 4.1499,
+      "step": 2163
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899778761061947e-06,
+      "loss": 3.9307,
+      "step": 2164
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8997313527180787e-06,
+      "loss": 4.1471,
+      "step": 2165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.89968394437421e-06,
+      "loss": 3.9397,
+      "step": 2166
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8996365360303413e-06,
+      "loss": 4.0821,
+      "step": 2167
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899589127686473e-06,
+      "loss": 3.981,
+      "step": 2168
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8995417193426043e-06,
+      "loss": 4.1392,
+      "step": 2169
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8994943109987355e-06,
+      "loss": 3.9366,
+      "step": 2170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8994469026548672e-06,
+      "loss": 4.2043,
+      "step": 2171
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899399494310999e-06,
+      "loss": 4.2848,
+      "step": 2172
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8993520859671302e-06,
+      "loss": 4.0854,
+      "step": 2173
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899304677623262e-06,
+      "loss": 4.2183,
+      "step": 2174
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899257269279393e-06,
+      "loss": 3.798,
+      "step": 2175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8992098609355245e-06,
+      "loss": 4.1404,
+      "step": 2176
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899162452591656e-06,
+      "loss": 4.103,
+      "step": 2177
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899115044247788e-06,
+      "loss": 4.114,
+      "step": 2178
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899067635903919e-06,
+      "loss": 4.087,
+      "step": 2179
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.899020227560051e-06,
+      "loss": 4.0357,
+      "step": 2180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.898972819216182e-06,
+      "loss": 4.1484,
+      "step": 2181
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8989254108723134e-06,
+      "loss": 4.1328,
+      "step": 2182
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8988780025284447e-06,
+      "loss": 4.1523,
+      "step": 2183
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2.8988305941845764e-06,
+      "loss": 4.2836,
+      "step": 2184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.898783185840708e-06,
+      "loss": 3.9078,
+      "step": 2185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8987357774968394e-06,
+      "loss": 4.1002,
+      "step": 2186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.898688369152971e-06,
+      "loss": 3.9532,
+      "step": 2187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8986409608091024e-06,
+      "loss": 3.8322,
+      "step": 2188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8985935524652337e-06,
+      "loss": 4.1768,
+      "step": 2189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8985461441213654e-06,
+      "loss": 4.2153,
+      "step": 2190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.898498735777497e-06,
+      "loss": 4.1793,
+      "step": 2191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8984513274336284e-06,
+      "loss": 4.0911,
+      "step": 2192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.89840391908976e-06,
+      "loss": 4.0023,
+      "step": 2193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8983565107458914e-06,
+      "loss": 4.0098,
+      "step": 2194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8983091024020226e-06,
+      "loss": 3.7876,
+      "step": 2195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8982616940581543e-06,
+      "loss": 4.1798,
+      "step": 2196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8982142857142856e-06,
+      "loss": 4.0478,
+      "step": 2197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8981668773704173e-06,
+      "loss": 3.9325,
+      "step": 2198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.898119469026549e-06,
+      "loss": 4.2048,
+      "step": 2199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8980720606826803e-06,
+      "loss": 3.8238,
+      "step": 2200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8980246523388116e-06,
+      "loss": 4.0252,
+      "step": 2201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8979772439949433e-06,
+      "loss": 3.7557,
+      "step": 2202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8979298356510746e-06,
+      "loss": 4.2536,
+      "step": 2203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8978824273072063e-06,
+      "loss": 4.1024,
+      "step": 2204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8978350189633376e-06,
+      "loss": 4.2503,
+      "step": 2205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8977876106194693e-06,
+      "loss": 4.0,
+      "step": 2206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8977402022756005e-06,
+      "loss": 4.2759,
+      "step": 2207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.897692793931732e-06,
+      "loss": 4.1265,
+      "step": 2208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8976453855878635e-06,
+      "loss": 4.0882,
+      "step": 2209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.897597977243995e-06,
+      "loss": 4.1756,
+      "step": 2210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8975505689001265e-06,
+      "loss": 3.8909,
+      "step": 2211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.897503160556258e-06,
+      "loss": 3.7633,
+      "step": 2212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8974557522123895e-06,
+      "loss": 4.2184,
+      "step": 2213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8974083438685208e-06,
+      "loss": 3.8547,
+      "step": 2214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8973609355246525e-06,
+      "loss": 3.9545,
+      "step": 2215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8973135271807838e-06,
+      "loss": 4.0892,
+      "step": 2216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8972661188369155e-06,
+      "loss": 3.6606,
+      "step": 2217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.897218710493047e-06,
+      "loss": 4.1727,
+      "step": 2218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8971713021491784e-06,
+      "loss": 3.9108,
+      "step": 2219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8971238938053097e-06,
+      "loss": 4.1438,
+      "step": 2220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8970764854614414e-06,
+      "loss": 3.8148,
+      "step": 2221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8970290771175727e-06,
+      "loss": 3.8882,
+      "step": 2222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896981668773704e-06,
+      "loss": 3.9914,
+      "step": 2223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8969342604298357e-06,
+      "loss": 3.8472,
+      "step": 2224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8968868520859674e-06,
+      "loss": 3.9821,
+      "step": 2225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8968394437420987e-06,
+      "loss": 4.239,
+      "step": 2226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.89679203539823e-06,
+      "loss": 4.2966,
+      "step": 2227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8967446270543617e-06,
+      "loss": 4.1565,
+      "step": 2228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896697218710493e-06,
+      "loss": 4.0114,
+      "step": 2229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8966498103666246e-06,
+      "loss": 4.1087,
+      "step": 2230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8966024020227563e-06,
+      "loss": 3.8434,
+      "step": 2231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8965549936788876e-06,
+      "loss": 3.8075,
+      "step": 2232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896507585335019e-06,
+      "loss": 3.9516,
+      "step": 2233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8964601769911506e-06,
+      "loss": 4.013,
+      "step": 2234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896412768647282e-06,
+      "loss": 4.1738,
+      "step": 2235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896365360303413e-06,
+      "loss": 4.1595,
+      "step": 2236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896317951959545e-06,
+      "loss": 3.9955,
+      "step": 2237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8962705436156766e-06,
+      "loss": 3.9164,
+      "step": 2238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896223135271808e-06,
+      "loss": 4.1762,
+      "step": 2239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8961757269279396e-06,
+      "loss": 3.9894,
+      "step": 2240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896128318584071e-06,
+      "loss": 4.2504,
+      "step": 2241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896080910240202e-06,
+      "loss": 3.8835,
+      "step": 2242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.896033501896334e-06,
+      "loss": 3.5579,
+      "step": 2243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8959860935524655e-06,
+      "loss": 4.0993,
+      "step": 2244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.895938685208597e-06,
+      "loss": 4.0182,
+      "step": 2245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.8958912768647285e-06,
+      "loss": 4.0718,
+      "step": 2246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.89584386852086e-06,
+      "loss": 3.9996,
+      "step": 2247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2.895796460176991e-06,
+      "loss": 4.0732,
+      "step": 2248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8957490518331224e-06,
+      "loss": 4.1424,
+      "step": 2249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895701643489254e-06,
+      "loss": 3.9757,
+      "step": 2250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8956542351453858e-06,
+      "loss": 3.8543,
+      "step": 2251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895606826801517e-06,
+      "loss": 4.0718,
+      "step": 2252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8955594184576488e-06,
+      "loss": 4.1007,
+      "step": 2253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.89551201011378e-06,
+      "loss": 3.7422,
+      "step": 2254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8954646017699113e-06,
+      "loss": 3.697,
+      "step": 2255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895417193426043e-06,
+      "loss": 4.0417,
+      "step": 2256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8953697850821747e-06,
+      "loss": 4.1272,
+      "step": 2257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895322376738306e-06,
+      "loss": 3.7722,
+      "step": 2258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8952749683944377e-06,
+      "loss": 3.6354,
+      "step": 2259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895227560050569e-06,
+      "loss": 3.7935,
+      "step": 2260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8951801517067003e-06,
+      "loss": 3.9307,
+      "step": 2261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895132743362832e-06,
+      "loss": 4.2356,
+      "step": 2262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8950853350189633e-06,
+      "loss": 4.097,
+      "step": 2263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.895037926675095e-06,
+      "loss": 3.912,
+      "step": 2264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8949905183312267e-06,
+      "loss": 3.8264,
+      "step": 2265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894943109987358e-06,
+      "loss": 3.8432,
+      "step": 2266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8948957016434892e-06,
+      "loss": 4.1542,
+      "step": 2267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8948482932996205e-06,
+      "loss": 3.8682,
+      "step": 2268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894800884955752e-06,
+      "loss": 4.1298,
+      "step": 2269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894753476611884e-06,
+      "loss": 4.2873,
+      "step": 2270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894706068268015e-06,
+      "loss": 3.6553,
+      "step": 2271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894658659924147e-06,
+      "loss": 4.0625,
+      "step": 2272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894611251580278e-06,
+      "loss": 4.154,
+      "step": 2273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8945638432364095e-06,
+      "loss": 4.0359,
+      "step": 2274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894516434892541e-06,
+      "loss": 3.9959,
+      "step": 2275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8944690265486724e-06,
+      "loss": 4.0941,
+      "step": 2276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894421618204804e-06,
+      "loss": 3.8886,
+      "step": 2277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894374209860936e-06,
+      "loss": 3.9207,
+      "step": 2278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894326801517067e-06,
+      "loss": 3.6948,
+      "step": 2279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8942793931731984e-06,
+      "loss": 4.2524,
+      "step": 2280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.89423198482933e-06,
+      "loss": 4.1445,
+      "step": 2281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8941845764854614e-06,
+      "loss": 4.1086,
+      "step": 2282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894137168141593e-06,
+      "loss": 3.9021,
+      "step": 2283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894089759797725e-06,
+      "loss": 3.9282,
+      "step": 2284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.894042351453856e-06,
+      "loss": 4.17,
+      "step": 2285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8939949431099874e-06,
+      "loss": 4.3137,
+      "step": 2286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.893947534766119e-06,
+      "loss": 4.1802,
+      "step": 2287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8939001264222503e-06,
+      "loss": 4.074,
+      "step": 2288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8938527180783816e-06,
+      "loss": 4.1897,
+      "step": 2289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8938053097345133e-06,
+      "loss": 4.1125,
+      "step": 2290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.893757901390645e-06,
+      "loss": 4.0788,
+      "step": 2291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8937104930467763e-06,
+      "loss": 3.9459,
+      "step": 2292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8936630847029076e-06,
+      "loss": 3.9665,
+      "step": 2293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8936156763590393e-06,
+      "loss": 4.0169,
+      "step": 2294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8935682680151706e-06,
+      "loss": 4.0822,
+      "step": 2295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8935208596713023e-06,
+      "loss": 3.9613,
+      "step": 2296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.893473451327434e-06,
+      "loss": 3.8807,
+      "step": 2297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8934260429835653e-06,
+      "loss": 4.0797,
+      "step": 2298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8933786346396965e-06,
+      "loss": 4.0056,
+      "step": 2299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8933312262958283e-06,
+      "loss": 4.0134,
+      "step": 2300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8932838179519595e-06,
+      "loss": 4.2875,
+      "step": 2301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.893236409608091e-06,
+      "loss": 3.897,
+      "step": 2302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8931890012642225e-06,
+      "loss": 4.1101,
+      "step": 2303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8931415929203542e-06,
+      "loss": 3.726,
+      "step": 2304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8930941845764855e-06,
+      "loss": 4.1309,
+      "step": 2305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.893046776232617e-06,
+      "loss": 4.0169,
+      "step": 2306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8929993678887485e-06,
+      "loss": 4.1154,
+      "step": 2307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8929519595448798e-06,
+      "loss": 4.0987,
+      "step": 2308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8929045512010115e-06,
+      "loss": 3.7835,
+      "step": 2309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.892857142857143e-06,
+      "loss": 3.9969,
+      "step": 2310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2.8928097345132745e-06,
+      "loss": 4.1145,
+      "step": 2311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8927623261694057e-06,
+      "loss": 4.1356,
+      "step": 2312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8927149178255374e-06,
+      "loss": 4.0857,
+      "step": 2313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8926675094816687e-06,
+      "loss": 3.8853,
+      "step": 2314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8926201011378e-06,
+      "loss": 4.0675,
+      "step": 2315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8925726927939317e-06,
+      "loss": 4.1189,
+      "step": 2316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8925252844500634e-06,
+      "loss": 4.2164,
+      "step": 2317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8924778761061947e-06,
+      "loss": 4.0976,
+      "step": 2318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8924304677623264e-06,
+      "loss": 4.2145,
+      "step": 2319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8923830594184577e-06,
+      "loss": 4.0661,
+      "step": 2320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.892335651074589e-06,
+      "loss": 4.0886,
+      "step": 2321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8922882427307207e-06,
+      "loss": 3.9906,
+      "step": 2322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8922408343868524e-06,
+      "loss": 3.9166,
+      "step": 2323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8921934260429836e-06,
+      "loss": 4.2683,
+      "step": 2324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8921460176991153e-06,
+      "loss": 3.9794,
+      "step": 2325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8920986093552466e-06,
+      "loss": 4.1462,
+      "step": 2326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.892051201011378e-06,
+      "loss": 3.7436,
+      "step": 2327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8920037926675096e-06,
+      "loss": 4.2805,
+      "step": 2328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891956384323641e-06,
+      "loss": 4.1112,
+      "step": 2329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8919089759797726e-06,
+      "loss": 4.1423,
+      "step": 2330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8918615676359043e-06,
+      "loss": 4.2397,
+      "step": 2331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8918141592920356e-06,
+      "loss": 4.0923,
+      "step": 2332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891766750948167e-06,
+      "loss": 4.3527,
+      "step": 2333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891719342604298e-06,
+      "loss": 3.849,
+      "step": 2334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.89167193426043e-06,
+      "loss": 4.1575,
+      "step": 2335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8916245259165615e-06,
+      "loss": 4.1082,
+      "step": 2336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891577117572693e-06,
+      "loss": 3.9649,
+      "step": 2337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8915297092288245e-06,
+      "loss": 4.0744,
+      "step": 2338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891482300884956e-06,
+      "loss": 4.0818,
+      "step": 2339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891434892541087e-06,
+      "loss": 3.8913,
+      "step": 2340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891387484197219e-06,
+      "loss": 3.7929,
+      "step": 2341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.89134007585335e-06,
+      "loss": 4.1014,
+      "step": 2342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8912926675094818e-06,
+      "loss": 3.8656,
+      "step": 2343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8912452591656135e-06,
+      "loss": 4.2405,
+      "step": 2344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8911978508217448e-06,
+      "loss": 4.1706,
+      "step": 2345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891150442477876e-06,
+      "loss": 4.1602,
+      "step": 2346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8911030341340077e-06,
+      "loss": 4.0462,
+      "step": 2347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.891055625790139e-06,
+      "loss": 4.122,
+      "step": 2348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8910082174462707e-06,
+      "loss": 3.8925,
+      "step": 2349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8909608091024024e-06,
+      "loss": 3.9621,
+      "step": 2350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8909134007585337e-06,
+      "loss": 4.0738,
+      "step": 2351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890865992414665e-06,
+      "loss": 4.0401,
+      "step": 2352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8908185840707967e-06,
+      "loss": 3.883,
+      "step": 2353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890771175726928e-06,
+      "loss": 4.1239,
+      "step": 2354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8907237673830593e-06,
+      "loss": 3.9737,
+      "step": 2355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890676359039191e-06,
+      "loss": 4.1714,
+      "step": 2356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8906289506953227e-06,
+      "loss": 3.8501,
+      "step": 2357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890581542351454e-06,
+      "loss": 4.1664,
+      "step": 2358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8905341340075852e-06,
+      "loss": 3.9299,
+      "step": 2359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890486725663717e-06,
+      "loss": 3.8727,
+      "step": 2360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8904393173198482e-06,
+      "loss": 3.9543,
+      "step": 2361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.89039190897598e-06,
+      "loss": 4.0292,
+      "step": 2362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8903445006321116e-06,
+      "loss": 4.4052,
+      "step": 2363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890297092288243e-06,
+      "loss": 4.0475,
+      "step": 2364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890249683944374e-06,
+      "loss": 3.9781,
+      "step": 2365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890202275600506e-06,
+      "loss": 4.0351,
+      "step": 2366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890154867256637e-06,
+      "loss": 4.3102,
+      "step": 2367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8901074589127684e-06,
+      "loss": 4.3195,
+      "step": 2368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8900600505689e-06,
+      "loss": 4.1284,
+      "step": 2369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.890012642225032e-06,
+      "loss": 3.9527,
+      "step": 2370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.889965233881163e-06,
+      "loss": 4.1825,
+      "step": 2371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.889917825537295e-06,
+      "loss": 4.1825,
+      "step": 2372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.889870417193426e-06,
+      "loss": 3.7871,
+      "step": 2373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2.8898230088495574e-06,
+      "loss": 4.0725,
+      "step": 2374
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889775600505689e-06,
+      "loss": 4.1485,
+      "step": 2375
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889728192161821e-06,
+      "loss": 4.0158,
+      "step": 2376
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889680783817952e-06,
+      "loss": 4.0776,
+      "step": 2377
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8896333754740834e-06,
+      "loss": 4.026,
+      "step": 2378
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889585967130215e-06,
+      "loss": 4.109,
+      "step": 2379
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8895385587863464e-06,
+      "loss": 3.9545,
+      "step": 2380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8894911504424776e-06,
+      "loss": 4.0413,
+      "step": 2381
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8894437420986093e-06,
+      "loss": 4.1733,
+      "step": 2382
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889396333754741e-06,
+      "loss": 4.0775,
+      "step": 2383
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8893489254108723e-06,
+      "loss": 4.207,
+      "step": 2384
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889301517067004e-06,
+      "loss": 3.9956,
+      "step": 2385
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8892541087231353e-06,
+      "loss": 4.1322,
+      "step": 2386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8892067003792666e-06,
+      "loss": 4.1558,
+      "step": 2387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8891592920353983e-06,
+      "loss": 3.9123,
+      "step": 2388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.88911188369153e-06,
+      "loss": 4.3154,
+      "step": 2389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8890644753476613e-06,
+      "loss": 3.8516,
+      "step": 2390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.889017067003793e-06,
+      "loss": 4.1713,
+      "step": 2391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8889696586599243e-06,
+      "loss": 4.057,
+      "step": 2392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8889222503160555e-06,
+      "loss": 4.2118,
+      "step": 2393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8888748419721872e-06,
+      "loss": 4.3065,
+      "step": 2394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8888274336283185e-06,
+      "loss": 4.3289,
+      "step": 2395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8887800252844502e-06,
+      "loss": 3.8366,
+      "step": 2396
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.888732616940582e-06,
+      "loss": 4.2093,
+      "step": 2397
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8886852085967132e-06,
+      "loss": 4.1808,
+      "step": 2398
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8886378002528445e-06,
+      "loss": 4.0497,
+      "step": 2399
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8885903919089758e-06,
+      "loss": 3.9412,
+      "step": 2400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8885429835651075e-06,
+      "loss": 4.1484,
+      "step": 2401
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.888495575221239e-06,
+      "loss": 4.0823,
+      "step": 2402
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8884481668773705e-06,
+      "loss": 3.8867,
+      "step": 2403
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.888400758533502e-06,
+      "loss": 4.0015,
+      "step": 2404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8883533501896334e-06,
+      "loss": 3.9956,
+      "step": 2405
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8883059418457647e-06,
+      "loss": 3.9923,
+      "step": 2406
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8882585335018964e-06,
+      "loss": 4.1126,
+      "step": 2407
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8882111251580277e-06,
+      "loss": 4.0851,
+      "step": 2408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8881637168141594e-06,
+      "loss": 4.0028,
+      "step": 2409
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.888116308470291e-06,
+      "loss": 3.9805,
+      "step": 2410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8880689001264224e-06,
+      "loss": 4.0902,
+      "step": 2411
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8880214917825537e-06,
+      "loss": 4.1863,
+      "step": 2412
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8879740834386854e-06,
+      "loss": 4.1335,
+      "step": 2413
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8879266750948167e-06,
+      "loss": 4.3263,
+      "step": 2414
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887879266750948e-06,
+      "loss": 4.2563,
+      "step": 2415
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.88783185840708e-06,
+      "loss": 4.1468,
+      "step": 2416
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8877844500632114e-06,
+      "loss": 3.9677,
+      "step": 2417
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8877370417193426e-06,
+      "loss": 3.9931,
+      "step": 2418
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8876896333754743e-06,
+      "loss": 4.252,
+      "step": 2419
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8876422250316056e-06,
+      "loss": 4.0711,
+      "step": 2420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887594816687737e-06,
+      "loss": 4.0534,
+      "step": 2421
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8875474083438686e-06,
+      "loss": 4.0001,
+      "step": 2422
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8875000000000003e-06,
+      "loss": 4.1607,
+      "step": 2423
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8874525916561316e-06,
+      "loss": 4.1407,
+      "step": 2424
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887405183312263e-06,
+      "loss": 3.8783,
+      "step": 2425
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8873577749683946e-06,
+      "loss": 3.8794,
+      "step": 2426
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887310366624526e-06,
+      "loss": 4.2578,
+      "step": 2427
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887262958280657e-06,
+      "loss": 4.1843,
+      "step": 2428
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887215549936789e-06,
+      "loss": 3.9453,
+      "step": 2429
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8871681415929205e-06,
+      "loss": 4.0113,
+      "step": 2430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887120733249052e-06,
+      "loss": 4.2229,
+      "step": 2431
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8870733249051835e-06,
+      "loss": 4.1088,
+      "step": 2432
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.887025916561315e-06,
+      "loss": 4.1178,
+      "step": 2433
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.886978508217446e-06,
+      "loss": 3.94,
+      "step": 2434
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.886931099873578e-06,
+      "loss": 4.0821,
+      "step": 2435
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8868836915297095e-06,
+      "loss": 4.1694,
+      "step": 2436
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8868362831858408e-06,
+      "loss": 3.7217,
+      "step": 2437
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2.8867888748419725e-06,
+      "loss": 3.9554,
+      "step": 2438
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8867414664981038e-06,
+      "loss": 3.9406,
+      "step": 2439
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886694058154235e-06,
+      "loss": 3.9669,
+      "step": 2440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8866466498103667e-06,
+      "loss": 4.3465,
+      "step": 2441
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886599241466498e-06,
+      "loss": 4.1504,
+      "step": 2442
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8865518331226297e-06,
+      "loss": 4.2264,
+      "step": 2443
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886504424778761e-06,
+      "loss": 3.9814,
+      "step": 2444
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8864570164348927e-06,
+      "loss": 3.988,
+      "step": 2445
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886409608091024e-06,
+      "loss": 4.0062,
+      "step": 2446
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8863621997471553e-06,
+      "loss": 4.0953,
+      "step": 2447
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886314791403287e-06,
+      "loss": 4.2357,
+      "step": 2448
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8862673830594187e-06,
+      "loss": 4.1541,
+      "step": 2449
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.88621997471555e-06,
+      "loss": 3.9079,
+      "step": 2450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8861725663716817e-06,
+      "loss": 3.7685,
+      "step": 2451
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886125158027813e-06,
+      "loss": 4.1031,
+      "step": 2452
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8860777496839442e-06,
+      "loss": 4.3565,
+      "step": 2453
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.886030341340076e-06,
+      "loss": 4.3121,
+      "step": 2454
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885982932996207e-06,
+      "loss": 3.9615,
+      "step": 2455
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885935524652339e-06,
+      "loss": 3.8851,
+      "step": 2456
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8858881163084706e-06,
+      "loss": 4.1916,
+      "step": 2457
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885840707964602e-06,
+      "loss": 4.1412,
+      "step": 2458
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885793299620733e-06,
+      "loss": 4.0686,
+      "step": 2459
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885745891276865e-06,
+      "loss": 4.2003,
+      "step": 2460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885698482932996e-06,
+      "loss": 3.9649,
+      "step": 2461
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885651074589128e-06,
+      "loss": 4.009,
+      "step": 2462
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8856036662452596e-06,
+      "loss": 4.1957,
+      "step": 2463
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885556257901391e-06,
+      "loss": 4.0946,
+      "step": 2464
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885508849557522e-06,
+      "loss": 4.0137,
+      "step": 2465
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8854614412136534e-06,
+      "loss": 3.956,
+      "step": 2466
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885414032869785e-06,
+      "loss": 4.241,
+      "step": 2467
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8853666245259164e-06,
+      "loss": 4.1492,
+      "step": 2468
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885319216182048e-06,
+      "loss": 4.1339,
+      "step": 2469
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.88527180783818e-06,
+      "loss": 4.0526,
+      "step": 2470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885224399494311e-06,
+      "loss": 4.1252,
+      "step": 2471
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8851769911504424e-06,
+      "loss": 3.9505,
+      "step": 2472
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885129582806574e-06,
+      "loss": 3.948,
+      "step": 2473
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8850821744627053e-06,
+      "loss": 3.9102,
+      "step": 2474
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.885034766118837e-06,
+      "loss": 4.2833,
+      "step": 2475
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8849873577749688e-06,
+      "loss": 3.8592,
+      "step": 2476
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8849399494311e-06,
+      "loss": 3.5743,
+      "step": 2477
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8848925410872313e-06,
+      "loss": 4.173,
+      "step": 2478
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.884845132743363e-06,
+      "loss": 3.97,
+      "step": 2479
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8847977243994943e-06,
+      "loss": 4.1673,
+      "step": 2480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8847503160556256e-06,
+      "loss": 4.1279,
+      "step": 2481
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8847029077117573e-06,
+      "loss": 4.0873,
+      "step": 2482
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.884655499367889e-06,
+      "loss": 4.2463,
+      "step": 2483
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8846080910240203e-06,
+      "loss": 4.1134,
+      "step": 2484
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.884560682680152e-06,
+      "loss": 4.204,
+      "step": 2485
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8845132743362833e-06,
+      "loss": 4.0169,
+      "step": 2486
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8844658659924145e-06,
+      "loss": 3.6356,
+      "step": 2487
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8844184576485462e-06,
+      "loss": 3.9496,
+      "step": 2488
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.884371049304678e-06,
+      "loss": 4.0607,
+      "step": 2489
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8843236409608092e-06,
+      "loss": 4.0945,
+      "step": 2490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8842762326169405e-06,
+      "loss": 4.1586,
+      "step": 2491
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.884228824273072e-06,
+      "loss": 4.0026,
+      "step": 2492
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8841814159292035e-06,
+      "loss": 3.9844,
+      "step": 2493
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8841340075853348e-06,
+      "loss": 4.1055,
+      "step": 2494
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8840865992414665e-06,
+      "loss": 3.8794,
+      "step": 2495
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.884039190897598e-06,
+      "loss": 3.9564,
+      "step": 2496
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8839917825537295e-06,
+      "loss": 4.0921,
+      "step": 2497
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.883944374209861e-06,
+      "loss": 3.8704,
+      "step": 2498
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8838969658659924e-06,
+      "loss": 3.8848,
+      "step": 2499
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8838495575221237e-06,
+      "loss": 4.0322,
+      "step": 2500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2.8838021491782554e-06,
+      "loss": 3.8558,
+      "step": 2501
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.883754740834387e-06,
+      "loss": 3.8895,
+      "step": 2502
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8837073324905184e-06,
+      "loss": 3.7696,
+      "step": 2503
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.88365992414665e-06,
+      "loss": 3.8325,
+      "step": 2504
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8836125158027814e-06,
+      "loss": 3.977,
+      "step": 2505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8835651074589127e-06,
+      "loss": 4.1885,
+      "step": 2506
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8835176991150444e-06,
+      "loss": 4.0998,
+      "step": 2507
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8834702907711757e-06,
+      "loss": 3.809,
+      "step": 2508
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8834228824273074e-06,
+      "loss": 4.1374,
+      "step": 2509
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8833754740834386e-06,
+      "loss": 4.0061,
+      "step": 2510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8833280657395703e-06,
+      "loss": 4.3296,
+      "step": 2511
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8832806573957016e-06,
+      "loss": 4.115,
+      "step": 2512
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.883233249051833e-06,
+      "loss": 4.0952,
+      "step": 2513
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8831858407079646e-06,
+      "loss": 3.877,
+      "step": 2514
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8831384323640963e-06,
+      "loss": 4.2751,
+      "step": 2515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8830910240202276e-06,
+      "loss": 4.1701,
+      "step": 2516
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8830436156763593e-06,
+      "loss": 4.2447,
+      "step": 2517
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8829962073324906e-06,
+      "loss": 4.017,
+      "step": 2518
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882948798988622e-06,
+      "loss": 4.2687,
+      "step": 2519
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8829013906447536e-06,
+      "loss": 4.2606,
+      "step": 2520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882853982300885e-06,
+      "loss": 3.9079,
+      "step": 2521
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8828065739570165e-06,
+      "loss": 4.0584,
+      "step": 2522
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8827591656131483e-06,
+      "loss": 4.033,
+      "step": 2523
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8827117572692795e-06,
+      "loss": 4.2489,
+      "step": 2524
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882664348925411e-06,
+      "loss": 3.9729,
+      "step": 2525
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8826169405815425e-06,
+      "loss": 4.0545,
+      "step": 2526
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882569532237674e-06,
+      "loss": 4.1876,
+      "step": 2527
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8825221238938055e-06,
+      "loss": 3.9033,
+      "step": 2528
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882474715549937e-06,
+      "loss": 4.0766,
+      "step": 2529
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8824273072060685e-06,
+      "loss": 4.1852,
+      "step": 2530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8823798988621998e-06,
+      "loss": 3.8556,
+      "step": 2531
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882332490518331e-06,
+      "loss": 4.0218,
+      "step": 2532
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8822850821744628e-06,
+      "loss": 3.9709,
+      "step": 2533
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.882237673830594e-06,
+      "loss": 3.5756,
+      "step": 2534
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8821902654867257e-06,
+      "loss": 3.9951,
+      "step": 2535
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8821428571428574e-06,
+      "loss": 3.9001,
+      "step": 2536
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8820954487989887e-06,
+      "loss": 4.0107,
+      "step": 2537
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.88204804045512e-06,
+      "loss": 4.2111,
+      "step": 2538
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8820006321112517e-06,
+      "loss": 4.1224,
+      "step": 2539
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881953223767383e-06,
+      "loss": 4.0632,
+      "step": 2540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8819058154235147e-06,
+      "loss": 4.0742,
+      "step": 2541
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8818584070796464e-06,
+      "loss": 4.077,
+      "step": 2542
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8818109987357777e-06,
+      "loss": 4.399,
+      "step": 2543
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881763590391909e-06,
+      "loss": 3.9268,
+      "step": 2544
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8817161820480407e-06,
+      "loss": 3.8889,
+      "step": 2545
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881668773704172e-06,
+      "loss": 4.1228,
+      "step": 2546
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8816213653603032e-06,
+      "loss": 4.2384,
+      "step": 2547
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881573957016435e-06,
+      "loss": 4.3173,
+      "step": 2548
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8815265486725666e-06,
+      "loss": 4.0536,
+      "step": 2549
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881479140328698e-06,
+      "loss": 4.0622,
+      "step": 2550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8814317319848296e-06,
+      "loss": 3.8703,
+      "step": 2551
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881384323640961e-06,
+      "loss": 4.3271,
+      "step": 2552
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881336915297092e-06,
+      "loss": 4.1691,
+      "step": 2553
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881289506953224e-06,
+      "loss": 3.9505,
+      "step": 2554
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8812420986093556e-06,
+      "loss": 3.8419,
+      "step": 2555
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881194690265487e-06,
+      "loss": 4.0361,
+      "step": 2556
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881147281921618e-06,
+      "loss": 4.0096,
+      "step": 2557
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.88109987357775e-06,
+      "loss": 3.7856,
+      "step": 2558
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.881052465233881e-06,
+      "loss": 4.0165,
+      "step": 2559
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.8810050568900124e-06,
+      "loss": 4.1843,
+      "step": 2560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.880957648546144e-06,
+      "loss": 3.5593,
+      "step": 2561
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.880910240202276e-06,
+      "loss": 3.7422,
+      "step": 2562
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.880862831858407e-06,
+      "loss": 4.1928,
+      "step": 2563
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2.880815423514539e-06,
+      "loss": 3.8456,
+      "step": 2564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.88076801517067e-06,
+      "loss": 4.1253,
+      "step": 2565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8807206068268014e-06,
+      "loss": 4.0213,
+      "step": 2566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880673198482933e-06,
+      "loss": 4.2287,
+      "step": 2567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8806257901390648e-06,
+      "loss": 3.971,
+      "step": 2568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880578381795196e-06,
+      "loss": 4.0769,
+      "step": 2569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8805309734513278e-06,
+      "loss": 4.0978,
+      "step": 2570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880483565107459e-06,
+      "loss": 4.1364,
+      "step": 2571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8804361567635903e-06,
+      "loss": 3.8699,
+      "step": 2572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880388748419722e-06,
+      "loss": 3.8547,
+      "step": 2573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8803413400758533e-06,
+      "loss": 4.2414,
+      "step": 2574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880293931731985e-06,
+      "loss": 4.1296,
+      "step": 2575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8802465233881163e-06,
+      "loss": 3.7392,
+      "step": 2576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880199115044248e-06,
+      "loss": 4.0006,
+      "step": 2577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8801517067003793e-06,
+      "loss": 4.0259,
+      "step": 2578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8801042983565105e-06,
+      "loss": 3.9096,
+      "step": 2579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8800568900126422e-06,
+      "loss": 3.8988,
+      "step": 2580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.880009481668774e-06,
+      "loss": 3.8833,
+      "step": 2581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8799620733249052e-06,
+      "loss": 4.1224,
+      "step": 2582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879914664981037e-06,
+      "loss": 3.8841,
+      "step": 2583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8798672566371682e-06,
+      "loss": 3.9215,
+      "step": 2584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8798198482932995e-06,
+      "loss": 3.9937,
+      "step": 2585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879772439949431e-06,
+      "loss": 4.2793,
+      "step": 2586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8797250316055625e-06,
+      "loss": 3.7739,
+      "step": 2587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879677623261694e-06,
+      "loss": 4.0124,
+      "step": 2588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879630214917826e-06,
+      "loss": 3.776,
+      "step": 2589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879582806573957e-06,
+      "loss": 3.8276,
+      "step": 2590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8795353982300885e-06,
+      "loss": 4.0398,
+      "step": 2591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.87948798988622e-06,
+      "loss": 3.718,
+      "step": 2592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8794405815423514e-06,
+      "loss": 4.1301,
+      "step": 2593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879393173198483e-06,
+      "loss": 4.0828,
+      "step": 2594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879345764854615e-06,
+      "loss": 3.975,
+      "step": 2595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879298356510746e-06,
+      "loss": 4.1444,
+      "step": 2596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8792509481668774e-06,
+      "loss": 4.0797,
+      "step": 2597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8792035398230087e-06,
+      "loss": 4.0227,
+      "step": 2598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8791561314791404e-06,
+      "loss": 4.2076,
+      "step": 2599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8791087231352717e-06,
+      "loss": 4.1666,
+      "step": 2600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8790613147914034e-06,
+      "loss": 4.2333,
+      "step": 2601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.879013906447535e-06,
+      "loss": 3.759,
+      "step": 2602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8789664981036664e-06,
+      "loss": 4.2372,
+      "step": 2603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8789190897597976e-06,
+      "loss": 4.1181,
+      "step": 2604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8788716814159293e-06,
+      "loss": 4.1894,
+      "step": 2605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8788242730720606e-06,
+      "loss": 4.0605,
+      "step": 2606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8787768647281923e-06,
+      "loss": 3.8754,
+      "step": 2607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.878729456384324e-06,
+      "loss": 4.1781,
+      "step": 2608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8786820480404553e-06,
+      "loss": 4.1117,
+      "step": 2609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8786346396965866e-06,
+      "loss": 3.9631,
+      "step": 2610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8785872313527183e-06,
+      "loss": 4.0623,
+      "step": 2611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8785398230088496e-06,
+      "loss": 3.7141,
+      "step": 2612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.878492414664981e-06,
+      "loss": 4.1941,
+      "step": 2613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8784450063211126e-06,
+      "loss": 4.0841,
+      "step": 2614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8783975979772443e-06,
+      "loss": 4.0229,
+      "step": 2615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8783501896333755e-06,
+      "loss": 4.1273,
+      "step": 2616
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8783027812895072e-06,
+      "loss": 4.0567,
+      "step": 2617
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8782553729456385e-06,
+      "loss": 3.9051,
+      "step": 2618
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.87820796460177e-06,
+      "loss": 4.1159,
+      "step": 2619
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8781605562579015e-06,
+      "loss": 3.7803,
+      "step": 2620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8781131479140332e-06,
+      "loss": 4.1967,
+      "step": 2621
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8780657395701645e-06,
+      "loss": 3.9526,
+      "step": 2622
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8780183312262958e-06,
+      "loss": 4.2365,
+      "step": 2623
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8779709228824275e-06,
+      "loss": 4.1013,
+      "step": 2624
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8779235145385588e-06,
+      "loss": 4.1829,
+      "step": 2625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.87787610619469e-06,
+      "loss": 3.8676,
+      "step": 2626
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8778286978508217e-06,
+      "loss": 3.8778,
+      "step": 2627
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2.8777812895069534e-06,
+      "loss": 4.1891,
+      "step": 2628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8777338811630847e-06,
+      "loss": 3.9717,
+      "step": 2629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8776864728192164e-06,
+      "loss": 4.1466,
+      "step": 2630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8776390644753477e-06,
+      "loss": 3.7351,
+      "step": 2631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.877591656131479e-06,
+      "loss": 4.2665,
+      "step": 2632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8775442477876107e-06,
+      "loss": 3.995,
+      "step": 2633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8774968394437424e-06,
+      "loss": 4.1519,
+      "step": 2634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8774494310998737e-06,
+      "loss": 4.0471,
+      "step": 2635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8774020227560054e-06,
+      "loss": 4.2774,
+      "step": 2636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8773546144121367e-06,
+      "loss": 4.1997,
+      "step": 2637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.877307206068268e-06,
+      "loss": 4.1475,
+      "step": 2638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8772597977243997e-06,
+      "loss": 4.2615,
+      "step": 2639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.877212389380531e-06,
+      "loss": 4.2581,
+      "step": 2640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8771649810366626e-06,
+      "loss": 4.0188,
+      "step": 2641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.877117572692794e-06,
+      "loss": 4.1677,
+      "step": 2642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8770701643489256e-06,
+      "loss": 3.9388,
+      "step": 2643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.877022756005057e-06,
+      "loss": 4.232,
+      "step": 2644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876975347661188e-06,
+      "loss": 4.0191,
+      "step": 2645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.87692793931732e-06,
+      "loss": 4.3288,
+      "step": 2646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8768805309734516e-06,
+      "loss": 4.2719,
+      "step": 2647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876833122629583e-06,
+      "loss": 4.2193,
+      "step": 2648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8767857142857146e-06,
+      "loss": 3.6876,
+      "step": 2649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876738305941846e-06,
+      "loss": 3.6803,
+      "step": 2650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876690897597977e-06,
+      "loss": 4.1481,
+      "step": 2651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876643489254109e-06,
+      "loss": 3.6261,
+      "step": 2652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.87659608091024e-06,
+      "loss": 4.0777,
+      "step": 2653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876548672566372e-06,
+      "loss": 4.1248,
+      "step": 2654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8765012642225035e-06,
+      "loss": 4.1198,
+      "step": 2655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876453855878635e-06,
+      "loss": 4.1308,
+      "step": 2656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876406447534766e-06,
+      "loss": 3.8298,
+      "step": 2657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876359039190898e-06,
+      "loss": 4.0386,
+      "step": 2658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876311630847029e-06,
+      "loss": 3.9759,
+      "step": 2659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8762642225031608e-06,
+      "loss": 3.8941,
+      "step": 2660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8762168141592925e-06,
+      "loss": 4.075,
+      "step": 2661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8761694058154238e-06,
+      "loss": 4.1793,
+      "step": 2662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876121997471555e-06,
+      "loss": 4.0346,
+      "step": 2663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8760745891276863e-06,
+      "loss": 4.0775,
+      "step": 2664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.876027180783818e-06,
+      "loss": 3.9872,
+      "step": 2665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8759797724399493e-06,
+      "loss": 3.8426,
+      "step": 2666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875932364096081e-06,
+      "loss": 3.8717,
+      "step": 2667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8758849557522127e-06,
+      "loss": 3.7822,
+      "step": 2668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875837547408344e-06,
+      "loss": 3.8376,
+      "step": 2669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8757901390644753e-06,
+      "loss": 4.1997,
+      "step": 2670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875742730720607e-06,
+      "loss": 3.718,
+      "step": 2671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8756953223767383e-06,
+      "loss": 4.2149,
+      "step": 2672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.87564791403287e-06,
+      "loss": 4.0196,
+      "step": 2673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8756005056890017e-06,
+      "loss": 3.9681,
+      "step": 2674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875553097345133e-06,
+      "loss": 4.0013,
+      "step": 2675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8755056890012642e-06,
+      "loss": 3.7598,
+      "step": 2676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875458280657396e-06,
+      "loss": 3.9554,
+      "step": 2677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875410872313527e-06,
+      "loss": 4.0826,
+      "step": 2678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8753634639696585e-06,
+      "loss": 3.9837,
+      "step": 2679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.87531605562579e-06,
+      "loss": 3.9287,
+      "step": 2680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875268647281922e-06,
+      "loss": 3.6888,
+      "step": 2681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875221238938053e-06,
+      "loss": 4.2467,
+      "step": 2682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875173830594185e-06,
+      "loss": 4.0091,
+      "step": 2683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.875126422250316e-06,
+      "loss": 4.1127,
+      "step": 2684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8750790139064474e-06,
+      "loss": 3.9785,
+      "step": 2685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8750316055625787e-06,
+      "loss": 3.9826,
+      "step": 2686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.874984197218711e-06,
+      "loss": 4.0919,
+      "step": 2687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.874936788874842e-06,
+      "loss": 3.8326,
+      "step": 2688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8748893805309734e-06,
+      "loss": 4.2514,
+      "step": 2689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.874841972187105e-06,
+      "loss": 4.1953,
+      "step": 2690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2.8747945638432364e-06,
+      "loss": 3.912,
+      "step": 2691
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8747471554993677e-06,
+      "loss": 4.1638,
+      "step": 2692
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8746997471554994e-06,
+      "loss": 4.0475,
+      "step": 2693
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.874652338811631e-06,
+      "loss": 4.0866,
+      "step": 2694
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8746049304677624e-06,
+      "loss": 3.8221,
+      "step": 2695
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.874557522123894e-06,
+      "loss": 4.0046,
+      "step": 2696
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8745101137800254e-06,
+      "loss": 4.1705,
+      "step": 2697
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8744627054361566e-06,
+      "loss": 4.1367,
+      "step": 2698
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8744152970922883e-06,
+      "loss": 3.8189,
+      "step": 2699
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8743678887484196e-06,
+      "loss": 3.8912,
+      "step": 2700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8743204804045513e-06,
+      "loss": 4.0133,
+      "step": 2701
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.874273072060683e-06,
+      "loss": 4.2732,
+      "step": 2702
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8742256637168143e-06,
+      "loss": 3.8176,
+      "step": 2703
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8741782553729456e-06,
+      "loss": 4.2613,
+      "step": 2704
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.874130847029077e-06,
+      "loss": 3.9209,
+      "step": 2705
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8740834386852086e-06,
+      "loss": 4.0712,
+      "step": 2706
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8740360303413403e-06,
+      "loss": 4.0072,
+      "step": 2707
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8739886219974716e-06,
+      "loss": 4.084,
+      "step": 2708
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8739412136536033e-06,
+      "loss": 4.1993,
+      "step": 2709
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8738938053097345e-06,
+      "loss": 4.2726,
+      "step": 2710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.873846396965866e-06,
+      "loss": 4.1763,
+      "step": 2711
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8737989886219975e-06,
+      "loss": 4.1713,
+      "step": 2712
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.873751580278129e-06,
+      "loss": 3.9403,
+      "step": 2713
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8737041719342605e-06,
+      "loss": 4.0949,
+      "step": 2714
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.873656763590392e-06,
+      "loss": 4.1612,
+      "step": 2715
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8736093552465235e-06,
+      "loss": 3.9412,
+      "step": 2716
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8735619469026548e-06,
+      "loss": 3.7004,
+      "step": 2717
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8735145385587865e-06,
+      "loss": 3.7643,
+      "step": 2718
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8734671302149178e-06,
+      "loss": 4.1237,
+      "step": 2719
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8734197218710495e-06,
+      "loss": 4.1256,
+      "step": 2720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.873372313527181e-06,
+      "loss": 4.1724,
+      "step": 2721
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8733249051833124e-06,
+      "loss": 4.0736,
+      "step": 2722
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8732774968394437e-06,
+      "loss": 4.1571,
+      "step": 2723
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8732300884955754e-06,
+      "loss": 3.9325,
+      "step": 2724
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8731826801517067e-06,
+      "loss": 4.1252,
+      "step": 2725
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.873135271807838e-06,
+      "loss": 3.9917,
+      "step": 2726
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8730878634639697e-06,
+      "loss": 4.2411,
+      "step": 2727
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8730404551201014e-06,
+      "loss": 3.5981,
+      "step": 2728
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8729930467762327e-06,
+      "loss": 3.9447,
+      "step": 2729
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872945638432364e-06,
+      "loss": 4.0087,
+      "step": 2730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8728982300884957e-06,
+      "loss": 3.8531,
+      "step": 2731
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872850821744627e-06,
+      "loss": 3.9159,
+      "step": 2732
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8728034134007586e-06,
+      "loss": 4.1836,
+      "step": 2733
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8727560050568903e-06,
+      "loss": 4.0334,
+      "step": 2734
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8727085967130216e-06,
+      "loss": 4.1536,
+      "step": 2735
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872661188369153e-06,
+      "loss": 4.3673,
+      "step": 2736
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8726137800252846e-06,
+      "loss": 4.1623,
+      "step": 2737
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872566371681416e-06,
+      "loss": 4.179,
+      "step": 2738
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872518963337547e-06,
+      "loss": 3.8717,
+      "step": 2739
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872471554993679e-06,
+      "loss": 4.2264,
+      "step": 2740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8724241466498106e-06,
+      "loss": 3.8525,
+      "step": 2741
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872376738305942e-06,
+      "loss": 4.0537,
+      "step": 2742
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8723293299620736e-06,
+      "loss": 4.0703,
+      "step": 2743
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872281921618205e-06,
+      "loss": 4.2095,
+      "step": 2744
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872234513274336e-06,
+      "loss": 4.1026,
+      "step": 2745
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872187104930468e-06,
+      "loss": 4.2406,
+      "step": 2746
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8721396965865995e-06,
+      "loss": 3.9472,
+      "step": 2747
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872092288242731e-06,
+      "loss": 4.2144,
+      "step": 2748
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.872044879898862e-06,
+      "loss": 4.1378,
+      "step": 2749
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.871997471554994e-06,
+      "loss": 3.7495,
+      "step": 2750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.871950063211125e-06,
+      "loss": 3.879,
+      "step": 2751
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8719026548672564e-06,
+      "loss": 4.1275,
+      "step": 2752
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.871855246523388e-06,
+      "loss": 4.1285,
+      "step": 2753
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2.8718078381795198e-06,
+      "loss": 3.9765,
+      "step": 2754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871760429835651e-06,
+      "loss": 3.9442,
+      "step": 2755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8717130214917828e-06,
+      "loss": 4.0035,
+      "step": 2756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871665613147914e-06,
+      "loss": 3.9645,
+      "step": 2757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8716182048040453e-06,
+      "loss": 3.8745,
+      "step": 2758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871570796460177e-06,
+      "loss": 3.9975,
+      "step": 2759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8715233881163087e-06,
+      "loss": 3.8439,
+      "step": 2760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.87147597977244e-06,
+      "loss": 3.8748,
+      "step": 2761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8714285714285717e-06,
+      "loss": 3.9578,
+      "step": 2762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871381163084703e-06,
+      "loss": 3.6414,
+      "step": 2763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8713337547408343e-06,
+      "loss": 4.0313,
+      "step": 2764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871286346396966e-06,
+      "loss": 3.96,
+      "step": 2765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8712389380530973e-06,
+      "loss": 4.1399,
+      "step": 2766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871191529709229e-06,
+      "loss": 4.2899,
+      "step": 2767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8711441213653607e-06,
+      "loss": 3.5843,
+      "step": 2768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.871096713021492e-06,
+      "loss": 3.979,
+      "step": 2769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8710493046776232e-06,
+      "loss": 4.0456,
+      "step": 2770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8710018963337545e-06,
+      "loss": 3.9452,
+      "step": 2771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870954487989886e-06,
+      "loss": 3.8862,
+      "step": 2772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870907079646018e-06,
+      "loss": 3.9398,
+      "step": 2773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870859671302149e-06,
+      "loss": 4.0502,
+      "step": 2774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870812262958281e-06,
+      "loss": 4.1283,
+      "step": 2775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870764854614412e-06,
+      "loss": 3.9782,
+      "step": 2776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8707174462705435e-06,
+      "loss": 3.8203,
+      "step": 2777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870670037926675e-06,
+      "loss": 3.6543,
+      "step": 2778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8706226295828064e-06,
+      "loss": 4.1478,
+      "step": 2779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870575221238938e-06,
+      "loss": 4.1479,
+      "step": 2780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.87052781289507e-06,
+      "loss": 3.9965,
+      "step": 2781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870480404551201e-06,
+      "loss": 4.1084,
+      "step": 2782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8704329962073324e-06,
+      "loss": 4.208,
+      "step": 2783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870385587863464e-06,
+      "loss": 3.9104,
+      "step": 2784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8703381795195954e-06,
+      "loss": 4.1907,
+      "step": 2785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870290771175727e-06,
+      "loss": 4.3345,
+      "step": 2786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870243362831859e-06,
+      "loss": 4.0411,
+      "step": 2787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.87019595448799e-06,
+      "loss": 3.9335,
+      "step": 2788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8701485461441214e-06,
+      "loss": 3.9241,
+      "step": 2789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.870101137800253e-06,
+      "loss": 3.8485,
+      "step": 2790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8700537294563843e-06,
+      "loss": 4.0812,
+      "step": 2791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8700063211125156e-06,
+      "loss": 4.0411,
+      "step": 2792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8699589127686473e-06,
+      "loss": 4.0389,
+      "step": 2793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.869911504424779e-06,
+      "loss": 3.9953,
+      "step": 2794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8698640960809103e-06,
+      "loss": 4.1337,
+      "step": 2795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8698166877370416e-06,
+      "loss": 4.294,
+      "step": 2796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8697692793931733e-06,
+      "loss": 4.2699,
+      "step": 2797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8697218710493046e-06,
+      "loss": 3.6423,
+      "step": 2798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8696744627054363e-06,
+      "loss": 3.8142,
+      "step": 2799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.869627054361568e-06,
+      "loss": 4.0778,
+      "step": 2800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8695796460176993e-06,
+      "loss": 3.9104,
+      "step": 2801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8695322376738305e-06,
+      "loss": 4.0351,
+      "step": 2802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8694848293299623e-06,
+      "loss": 4.05,
+      "step": 2803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8694374209860935e-06,
+      "loss": 3.9479,
+      "step": 2804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.869390012642225e-06,
+      "loss": 3.9772,
+      "step": 2805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8693426042983565e-06,
+      "loss": 4.0531,
+      "step": 2806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8692951959544882e-06,
+      "loss": 3.8307,
+      "step": 2807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8692477876106195e-06,
+      "loss": 4.1337,
+      "step": 2808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.869200379266751e-06,
+      "loss": 4.0522,
+      "step": 2809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8691529709228825e-06,
+      "loss": 4.2824,
+      "step": 2810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8691055625790138e-06,
+      "loss": 4.0918,
+      "step": 2811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8690581542351455e-06,
+      "loss": 4.1431,
+      "step": 2812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.869010745891277e-06,
+      "loss": 4.0373,
+      "step": 2813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8689633375474085e-06,
+      "loss": 3.968,
+      "step": 2814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8689159292035397e-06,
+      "loss": 3.9396,
+      "step": 2815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8688685208596714e-06,
+      "loss": 4.03,
+      "step": 2816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.8688211125158027e-06,
+      "loss": 4.0008,
+      "step": 2817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2.868773704171934e-06,
+      "loss": 3.8314,
+      "step": 2818
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8687262958280657e-06,
+      "loss": 4.1984,
+      "step": 2819
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8686788874841974e-06,
+      "loss": 4.1988,
+      "step": 2820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8686314791403287e-06,
+      "loss": 3.7979,
+      "step": 2821
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8685840707964604e-06,
+      "loss": 4.1634,
+      "step": 2822
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8685366624525917e-06,
+      "loss": 4.2366,
+      "step": 2823
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.868489254108723e-06,
+      "loss": 4.1383,
+      "step": 2824
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8684418457648547e-06,
+      "loss": 3.9734,
+      "step": 2825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8683944374209864e-06,
+      "loss": 3.8541,
+      "step": 2826
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8683470290771176e-06,
+      "loss": 3.7856,
+      "step": 2827
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8682996207332493e-06,
+      "loss": 4.1093,
+      "step": 2828
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8682522123893806e-06,
+      "loss": 4.2023,
+      "step": 2829
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.868204804045512e-06,
+      "loss": 4.0762,
+      "step": 2830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8681573957016436e-06,
+      "loss": 3.702,
+      "step": 2831
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.868109987357775e-06,
+      "loss": 4.2091,
+      "step": 2832
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8680625790139066e-06,
+      "loss": 4.2524,
+      "step": 2833
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8680151706700383e-06,
+      "loss": 3.9137,
+      "step": 2834
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8679677623261696e-06,
+      "loss": 4.1753,
+      "step": 2835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867920353982301e-06,
+      "loss": 3.9273,
+      "step": 2836
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867872945638432e-06,
+      "loss": 4.132,
+      "step": 2837
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867825537294564e-06,
+      "loss": 3.9654,
+      "step": 2838
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8677781289506955e-06,
+      "loss": 4.1221,
+      "step": 2839
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867730720606827e-06,
+      "loss": 3.8352,
+      "step": 2840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8676833122629585e-06,
+      "loss": 4.1644,
+      "step": 2841
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.86763590391909e-06,
+      "loss": 3.5838,
+      "step": 2842
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867588495575221e-06,
+      "loss": 4.0571,
+      "step": 2843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867541087231353e-06,
+      "loss": 4.0277,
+      "step": 2844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867493678887484e-06,
+      "loss": 3.9813,
+      "step": 2845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8674462705436158e-06,
+      "loss": 4.0885,
+      "step": 2846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8673988621997475e-06,
+      "loss": 4.1709,
+      "step": 2847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8673514538558788e-06,
+      "loss": 4.1903,
+      "step": 2848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.86730404551201e-06,
+      "loss": 4.0817,
+      "step": 2849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8672566371681417e-06,
+      "loss": 3.8272,
+      "step": 2850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867209228824273e-06,
+      "loss": 4.0944,
+      "step": 2851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8671618204804047e-06,
+      "loss": 4.2108,
+      "step": 2852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8671144121365364e-06,
+      "loss": 4.0574,
+      "step": 2853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8670670037926677e-06,
+      "loss": 4.1022,
+      "step": 2854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.867019595448799e-06,
+      "loss": 3.9467,
+      "step": 2855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8669721871049307e-06,
+      "loss": 3.9765,
+      "step": 2856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866924778761062e-06,
+      "loss": 4.0209,
+      "step": 2857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8668773704171933e-06,
+      "loss": 3.9664,
+      "step": 2858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866829962073325e-06,
+      "loss": 3.9147,
+      "step": 2859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8667825537294567e-06,
+      "loss": 3.8519,
+      "step": 2860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866735145385588e-06,
+      "loss": 3.9958,
+      "step": 2861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8666877370417192e-06,
+      "loss": 4.0127,
+      "step": 2862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866640328697851e-06,
+      "loss": 4.2081,
+      "step": 2863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8665929203539822e-06,
+      "loss": 3.9335,
+      "step": 2864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866545512010114e-06,
+      "loss": 3.8566,
+      "step": 2865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8664981036662456e-06,
+      "loss": 4.324,
+      "step": 2866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866450695322377e-06,
+      "loss": 3.829,
+      "step": 2867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866403286978508e-06,
+      "loss": 4.1201,
+      "step": 2868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.86635587863464e-06,
+      "loss": 4.0746,
+      "step": 2869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866308470290771e-06,
+      "loss": 3.9239,
+      "step": 2870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8662610619469024e-06,
+      "loss": 3.9273,
+      "step": 2871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866213653603034e-06,
+      "loss": 3.9572,
+      "step": 2872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866166245259166e-06,
+      "loss": 3.7474,
+      "step": 2873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866118836915297e-06,
+      "loss": 3.6255,
+      "step": 2874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.866071428571429e-06,
+      "loss": 3.8012,
+      "step": 2875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.86602402022756e-06,
+      "loss": 3.9699,
+      "step": 2876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8659766118836914e-06,
+      "loss": 4.0077,
+      "step": 2877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.865929203539823e-06,
+      "loss": 4.0736,
+      "step": 2878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.865881795195955e-06,
+      "loss": 3.9258,
+      "step": 2879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.865834386852086e-06,
+      "loss": 3.7916,
+      "step": 2880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.8657869785082174e-06,
+      "loss": 3.8368,
+      "step": 2881
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.865739570164349e-06,
+      "loss": 4.1611,
+      "step": 2882
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8656921618204804e-06,
+      "loss": 4.1241,
+      "step": 2883
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8656447534766116e-06,
+      "loss": 4.2154,
+      "step": 2884
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8655973451327433e-06,
+      "loss": 4.1331,
+      "step": 2885
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.865549936788875e-06,
+      "loss": 3.8311,
+      "step": 2886
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8655025284450063e-06,
+      "loss": 4.0021,
+      "step": 2887
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.865455120101138e-06,
+      "loss": 3.5823,
+      "step": 2888
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8654077117572693e-06,
+      "loss": 3.9938,
+      "step": 2889
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8653603034134006e-06,
+      "loss": 4.0622,
+      "step": 2890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8653128950695323e-06,
+      "loss": 4.1655,
+      "step": 2891
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.865265486725664e-06,
+      "loss": 3.8297,
+      "step": 2892
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8652180783817953e-06,
+      "loss": 3.8076,
+      "step": 2893
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.865170670037927e-06,
+      "loss": 4.1968,
+      "step": 2894
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8651232616940583e-06,
+      "loss": 4.0085,
+      "step": 2895
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8650758533501895e-06,
+      "loss": 4.0774,
+      "step": 2896
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8650284450063212e-06,
+      "loss": 4.1234,
+      "step": 2897
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8649810366624525e-06,
+      "loss": 3.8696,
+      "step": 2898
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8649336283185842e-06,
+      "loss": 4.1614,
+      "step": 2899
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.864886219974716e-06,
+      "loss": 4.2252,
+      "step": 2900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8648388116308472e-06,
+      "loss": 3.901,
+      "step": 2901
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8647914032869785e-06,
+      "loss": 4.1618,
+      "step": 2902
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8647439949431098e-06,
+      "loss": 4.252,
+      "step": 2903
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8646965865992415e-06,
+      "loss": 4.4322,
+      "step": 2904
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.864649178255373e-06,
+      "loss": 4.2257,
+      "step": 2905
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8646017699115045e-06,
+      "loss": 3.6499,
+      "step": 2906
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.864554361567636e-06,
+      "loss": 4.1462,
+      "step": 2907
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8645069532237674e-06,
+      "loss": 4.2224,
+      "step": 2908
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8644595448798987e-06,
+      "loss": 4.0206,
+      "step": 2909
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8644121365360304e-06,
+      "loss": 3.9184,
+      "step": 2910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8643647281921617e-06,
+      "loss": 3.8181,
+      "step": 2911
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8643173198482934e-06,
+      "loss": 3.9421,
+      "step": 2912
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.864269911504425e-06,
+      "loss": 3.98,
+      "step": 2913
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8642225031605564e-06,
+      "loss": 4.0604,
+      "step": 2914
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8641750948166877e-06,
+      "loss": 4.0514,
+      "step": 2915
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8641276864728194e-06,
+      "loss": 4.0775,
+      "step": 2916
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8640802781289507e-06,
+      "loss": 4.2616,
+      "step": 2917
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8640328697850824e-06,
+      "loss": 4.2171,
+      "step": 2918
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.863985461441214e-06,
+      "loss": 4.2606,
+      "step": 2919
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8639380530973454e-06,
+      "loss": 3.8691,
+      "step": 2920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8638906447534766e-06,
+      "loss": 4.0278,
+      "step": 2921
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8638432364096083e-06,
+      "loss": 4.1898,
+      "step": 2922
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8637958280657396e-06,
+      "loss": 4.0986,
+      "step": 2923
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.863748419721871e-06,
+      "loss": 4.2802,
+      "step": 2924
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8637010113780026e-06,
+      "loss": 4.1972,
+      "step": 2925
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8636536030341343e-06,
+      "loss": 3.8693,
+      "step": 2926
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8636061946902656e-06,
+      "loss": 3.9671,
+      "step": 2927
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.863558786346397e-06,
+      "loss": 4.0034,
+      "step": 2928
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8635113780025286e-06,
+      "loss": 4.2621,
+      "step": 2929
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.86346396965866e-06,
+      "loss": 3.8516,
+      "step": 2930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8634165613147916e-06,
+      "loss": 3.717,
+      "step": 2931
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8633691529709233e-06,
+      "loss": 4.0035,
+      "step": 2932
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8633217446270545e-06,
+      "loss": 4.1942,
+      "step": 2933
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.863274336283186e-06,
+      "loss": 3.9547,
+      "step": 2934
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8632269279393175e-06,
+      "loss": 3.8902,
+      "step": 2935
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.863179519595449e-06,
+      "loss": 3.7345,
+      "step": 2936
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.86313211125158e-06,
+      "loss": 3.8105,
+      "step": 2937
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.863084702907712e-06,
+      "loss": 3.915,
+      "step": 2938
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8630372945638435e-06,
+      "loss": 3.9199,
+      "step": 2939
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8629898862199748e-06,
+      "loss": 4.1235,
+      "step": 2940
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8629424778761065e-06,
+      "loss": 4.0972,
+      "step": 2941
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8628950695322378e-06,
+      "loss": 3.9405,
+      "step": 2942
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.862847661188369e-06,
+      "loss": 4.2348,
+      "step": 2943
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.8628002528445007e-06,
+      "loss": 4.1885,
+      "step": 2944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8627528445006324e-06,
+      "loss": 3.5938,
+      "step": 2945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8627054361567637e-06,
+      "loss": 4.1114,
+      "step": 2946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.862658027812895e-06,
+      "loss": 4.0451,
+      "step": 2947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8626106194690267e-06,
+      "loss": 4.1618,
+      "step": 2948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.862563211125158e-06,
+      "loss": 3.8939,
+      "step": 2949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8625158027812893e-06,
+      "loss": 4.1848,
+      "step": 2950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.862468394437421e-06,
+      "loss": 3.6705,
+      "step": 2951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8624209860935527e-06,
+      "loss": 4.185,
+      "step": 2952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.862373577749684e-06,
+      "loss": 4.2313,
+      "step": 2953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8623261694058157e-06,
+      "loss": 3.9601,
+      "step": 2954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.862278761061947e-06,
+      "loss": 4.0951,
+      "step": 2955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8622313527180782e-06,
+      "loss": 3.9952,
+      "step": 2956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.86218394437421e-06,
+      "loss": 3.972,
+      "step": 2957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8621365360303416e-06,
+      "loss": 4.0443,
+      "step": 2958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.862089127686473e-06,
+      "loss": 3.6817,
+      "step": 2959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8620417193426046e-06,
+      "loss": 4.1783,
+      "step": 2960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861994310998736e-06,
+      "loss": 3.9899,
+      "step": 2961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861946902654867e-06,
+      "loss": 4.0906,
+      "step": 2962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861899494310999e-06,
+      "loss": 3.944,
+      "step": 2963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.86185208596713e-06,
+      "loss": 4.1551,
+      "step": 2964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861804677623262e-06,
+      "loss": 4.1265,
+      "step": 2965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8617572692793936e-06,
+      "loss": 3.8227,
+      "step": 2966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861709860935525e-06,
+      "loss": 3.9182,
+      "step": 2967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861662452591656e-06,
+      "loss": 3.9301,
+      "step": 2968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8616150442477874e-06,
+      "loss": 3.9523,
+      "step": 2969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861567635903919e-06,
+      "loss": 4.3004,
+      "step": 2970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8615202275600504e-06,
+      "loss": 4.0801,
+      "step": 2971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861472819216182e-06,
+      "loss": 3.9781,
+      "step": 2972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861425410872314e-06,
+      "loss": 4.1238,
+      "step": 2973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861378002528445e-06,
+      "loss": 4.0485,
+      "step": 2974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8613305941845764e-06,
+      "loss": 3.8719,
+      "step": 2975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861283185840708e-06,
+      "loss": 3.7585,
+      "step": 2976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8612357774968393e-06,
+      "loss": 3.9796,
+      "step": 2977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861188369152971e-06,
+      "loss": 3.9795,
+      "step": 2978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8611409608091028e-06,
+      "loss": 3.8527,
+      "step": 2979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.861093552465234e-06,
+      "loss": 3.6809,
+      "step": 2980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8610461441213653e-06,
+      "loss": 4.0457,
+      "step": 2981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860998735777497e-06,
+      "loss": 3.6343,
+      "step": 2982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8609513274336283e-06,
+      "loss": 4.1268,
+      "step": 2983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8609039190897596e-06,
+      "loss": 4.1734,
+      "step": 2984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8608565107458913e-06,
+      "loss": 3.9306,
+      "step": 2985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860809102402023e-06,
+      "loss": 4.0558,
+      "step": 2986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8607616940581543e-06,
+      "loss": 3.801,
+      "step": 2987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860714285714286e-06,
+      "loss": 3.9165,
+      "step": 2988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8606668773704173e-06,
+      "loss": 4.0585,
+      "step": 2989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8606194690265485e-06,
+      "loss": 3.6645,
+      "step": 2990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8605720606826802e-06,
+      "loss": 4.0061,
+      "step": 2991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860524652338812e-06,
+      "loss": 4.2403,
+      "step": 2992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8604772439949432e-06,
+      "loss": 3.9895,
+      "step": 2993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8604298356510745e-06,
+      "loss": 3.9402,
+      "step": 2994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860382427307206e-06,
+      "loss": 4.0973,
+      "step": 2995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8603350189633375e-06,
+      "loss": 4.1102,
+      "step": 2996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8602876106194688e-06,
+      "loss": 3.6444,
+      "step": 2997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8602402022756005e-06,
+      "loss": 4.0173,
+      "step": 2998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860192793931732e-06,
+      "loss": 4.2167,
+      "step": 2999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8601453855878635e-06,
+      "loss": 4.0719,
+      "step": 3000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.860097977243995e-06,
+      "loss": 4.0277,
+      "step": 3001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8600505689001264e-06,
+      "loss": 3.9863,
+      "step": 3002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8600031605562577e-06,
+      "loss": 3.7255,
+      "step": 3003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8599557522123894e-06,
+      "loss": 4.095,
+      "step": 3004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.859908343868521e-06,
+      "loss": 3.9823,
+      "step": 3005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8598609355246524e-06,
+      "loss": 4.2585,
+      "step": 3006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.859813527180784e-06,
+      "loss": 3.9437,
+      "step": 3007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8597661188369154e-06,
+      "loss": 3.9051,
+      "step": 3008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8597187104930467e-06,
+      "loss": 4.0757,
+      "step": 3009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8596713021491784e-06,
+      "loss": 4.2758,
+      "step": 3010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8596238938053097e-06,
+      "loss": 3.9383,
+      "step": 3011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8595764854614414e-06,
+      "loss": 4.1384,
+      "step": 3012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8595290771175726e-06,
+      "loss": 3.7946,
+      "step": 3013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8594816687737043e-06,
+      "loss": 4.0279,
+      "step": 3014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8594342604298356e-06,
+      "loss": 4.2601,
+      "step": 3015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.859386852085967e-06,
+      "loss": 4.0006,
+      "step": 3016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8593394437420986e-06,
+      "loss": 4.2661,
+      "step": 3017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8592920353982303e-06,
+      "loss": 4.2182,
+      "step": 3018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8592446270543616e-06,
+      "loss": 3.7959,
+      "step": 3019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8591972187104933e-06,
+      "loss": 3.7054,
+      "step": 3020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8591498103666246e-06,
+      "loss": 3.9194,
+      "step": 3021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.859102402022756e-06,
+      "loss": 4.1976,
+      "step": 3022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8590549936788876e-06,
+      "loss": 3.9461,
+      "step": 3023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.859007585335019e-06,
+      "loss": 3.749,
+      "step": 3024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8589601769911505e-06,
+      "loss": 3.9743,
+      "step": 3025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8589127686472823e-06,
+      "loss": 3.8838,
+      "step": 3026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8588653603034135e-06,
+      "loss": 4.1294,
+      "step": 3027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858817951959545e-06,
+      "loss": 4.3134,
+      "step": 3028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8587705436156765e-06,
+      "loss": 3.7914,
+      "step": 3029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858723135271808e-06,
+      "loss": 4.2488,
+      "step": 3030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8586757269279395e-06,
+      "loss": 4.095,
+      "step": 3031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858628318584071e-06,
+      "loss": 4.0482,
+      "step": 3032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8585809102402025e-06,
+      "loss": 3.8829,
+      "step": 3033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8585335018963338e-06,
+      "loss": 4.0657,
+      "step": 3034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858486093552465e-06,
+      "loss": 4.0343,
+      "step": 3035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8584386852085968e-06,
+      "loss": 3.947,
+      "step": 3036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858391276864728e-06,
+      "loss": 3.6687,
+      "step": 3037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8583438685208597e-06,
+      "loss": 4.2309,
+      "step": 3038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8582964601769914e-06,
+      "loss": 4.0759,
+      "step": 3039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8582490518331227e-06,
+      "loss": 3.7359,
+      "step": 3040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858201643489254e-06,
+      "loss": 4.06,
+      "step": 3041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8581542351453857e-06,
+      "loss": 3.9922,
+      "step": 3042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.858106826801517e-06,
+      "loss": 3.9322,
+      "step": 3043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8580594184576487e-06,
+      "loss": 3.8433,
+      "step": 3044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8580120101137804e-06,
+      "loss": 4.0887,
+      "step": 3045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8579646017699117e-06,
+      "loss": 4.1592,
+      "step": 3046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857917193426043e-06,
+      "loss": 4.1929,
+      "step": 3047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8578697850821747e-06,
+      "loss": 4.1814,
+      "step": 3048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857822376738306e-06,
+      "loss": 3.9201,
+      "step": 3049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8577749683944372e-06,
+      "loss": 4.1019,
+      "step": 3050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857727560050569e-06,
+      "loss": 4.0542,
+      "step": 3051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8576801517067006e-06,
+      "loss": 3.9758,
+      "step": 3052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857632743362832e-06,
+      "loss": 4.3264,
+      "step": 3053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8575853350189636e-06,
+      "loss": 4.119,
+      "step": 3054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857537926675095e-06,
+      "loss": 4.0828,
+      "step": 3055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857490518331226e-06,
+      "loss": 4.1067,
+      "step": 3056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857443109987358e-06,
+      "loss": 4.0645,
+      "step": 3057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8573957016434896e-06,
+      "loss": 4.0713,
+      "step": 3058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857348293299621e-06,
+      "loss": 4.2577,
+      "step": 3059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857300884955752e-06,
+      "loss": 4.1799,
+      "step": 3060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857253476611884e-06,
+      "loss": 4.1709,
+      "step": 3061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857206068268015e-06,
+      "loss": 4.3132,
+      "step": 3062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8571586599241464e-06,
+      "loss": 4.072,
+      "step": 3063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857111251580278e-06,
+      "loss": 4.0637,
+      "step": 3064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.85706384323641e-06,
+      "loss": 4.2223,
+      "step": 3065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.857016434892541e-06,
+      "loss": 4.1759,
+      "step": 3066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.856969026548673e-06,
+      "loss": 4.0027,
+      "step": 3067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.856921618204804e-06,
+      "loss": 3.8894,
+      "step": 3068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8568742098609354e-06,
+      "loss": 3.7294,
+      "step": 3069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.856826801517067e-06,
+      "loss": 4.2707,
+      "step": 3070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8567793931731988e-06,
+      "loss": 4.0247,
+      "step": 3071
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.85673198482933e-06,
+      "loss": 3.8098,
+      "step": 3072
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8566845764854618e-06,
+      "loss": 4.1001,
+      "step": 3073
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.856637168141593e-06,
+      "loss": 4.096,
+      "step": 3074
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8565897597977243e-06,
+      "loss": 4.1113,
+      "step": 3075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.856542351453856e-06,
+      "loss": 3.9639,
+      "step": 3076
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8564949431099873e-06,
+      "loss": 4.229,
+      "step": 3077
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.856447534766119e-06,
+      "loss": 4.0942,
+      "step": 3078
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8564001264222503e-06,
+      "loss": 4.2815,
+      "step": 3079
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.856352718078382e-06,
+      "loss": 3.8282,
+      "step": 3080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8563053097345133e-06,
+      "loss": 4.1139,
+      "step": 3081
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8562579013906445e-06,
+      "loss": 3.7086,
+      "step": 3082
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8562104930467762e-06,
+      "loss": 4.2995,
+      "step": 3083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.856163084702908e-06,
+      "loss": 3.6128,
+      "step": 3084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8561156763590392e-06,
+      "loss": 3.9386,
+      "step": 3085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.856068268015171e-06,
+      "loss": 4.1189,
+      "step": 3086
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8560208596713022e-06,
+      "loss": 3.8327,
+      "step": 3087
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8559734513274335e-06,
+      "loss": 3.7788,
+      "step": 3088
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855926042983565e-06,
+      "loss": 4.1204,
+      "step": 3089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8558786346396965e-06,
+      "loss": 3.7971,
+      "step": 3090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855831226295828e-06,
+      "loss": 3.8288,
+      "step": 3091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.85578381795196e-06,
+      "loss": 3.7077,
+      "step": 3092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855736409608091e-06,
+      "loss": 3.6075,
+      "step": 3093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8556890012642225e-06,
+      "loss": 4.0919,
+      "step": 3094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855641592920354e-06,
+      "loss": 3.7305,
+      "step": 3095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8555941845764854e-06,
+      "loss": 4.2314,
+      "step": 3096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855546776232617e-06,
+      "loss": 4.0003,
+      "step": 3097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855499367888749e-06,
+      "loss": 4.2135,
+      "step": 3098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.85545195954488e-06,
+      "loss": 3.9371,
+      "step": 3099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8554045512010114e-06,
+      "loss": 4.0959,
+      "step": 3100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8553571428571427e-06,
+      "loss": 4.0632,
+      "step": 3101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8553097345132744e-06,
+      "loss": 4.2459,
+      "step": 3102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8552623261694057e-06,
+      "loss": 3.8769,
+      "step": 3103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8552149178255374e-06,
+      "loss": 4.0718,
+      "step": 3104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.855167509481669e-06,
+      "loss": 3.9232,
+      "step": 3105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8551201011378004e-06,
+      "loss": 4.0864,
+      "step": 3106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8550726927939316e-06,
+      "loss": 4.2805,
+      "step": 3107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8550252844500633e-06,
+      "loss": 4.1721,
+      "step": 3108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8549778761061946e-06,
+      "loss": 3.8893,
+      "step": 3109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8549304677623263e-06,
+      "loss": 3.7938,
+      "step": 3110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.854883059418458e-06,
+      "loss": 4.219,
+      "step": 3111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8548356510745893e-06,
+      "loss": 4.1575,
+      "step": 3112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8547882427307206e-06,
+      "loss": 4.0526,
+      "step": 3113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8547408343868523e-06,
+      "loss": 4.2396,
+      "step": 3114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8546934260429836e-06,
+      "loss": 3.6221,
+      "step": 3115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.854646017699115e-06,
+      "loss": 3.8976,
+      "step": 3116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8545986093552466e-06,
+      "loss": 4.193,
+      "step": 3117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8545512010113783e-06,
+      "loss": 3.8973,
+      "step": 3118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8545037926675095e-06,
+      "loss": 3.8354,
+      "step": 3119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8544563843236412e-06,
+      "loss": 4.2533,
+      "step": 3120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8544089759797725e-06,
+      "loss": 4.0926,
+      "step": 3121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.854361567635904e-06,
+      "loss": 3.8706,
+      "step": 3122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8543141592920355e-06,
+      "loss": 4.2646,
+      "step": 3123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8542667509481672e-06,
+      "loss": 4.2086,
+      "step": 3124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8542193426042985e-06,
+      "loss": 4.0307,
+      "step": 3125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8541719342604298e-06,
+      "loss": 4.0951,
+      "step": 3126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8541245259165615e-06,
+      "loss": 3.6341,
+      "step": 3127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8540771175726928e-06,
+      "loss": 4.1283,
+      "step": 3128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.854029709228824e-06,
+      "loss": 3.7967,
+      "step": 3129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8539823008849557e-06,
+      "loss": 3.6758,
+      "step": 3130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8539348925410874e-06,
+      "loss": 3.9357,
+      "step": 3131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8538874841972187e-06,
+      "loss": 3.4633,
+      "step": 3132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8538400758533504e-06,
+      "loss": 4.2125,
+      "step": 3133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.8537926675094817e-06,
+      "loss": 4.2199,
+      "step": 3134
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853745259165613e-06,
+      "loss": 3.8968,
+      "step": 3135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8536978508217447e-06,
+      "loss": 3.9216,
+      "step": 3136
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8536504424778764e-06,
+      "loss": 4.0311,
+      "step": 3137
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8536030341340077e-06,
+      "loss": 4.3033,
+      "step": 3138
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8535556257901394e-06,
+      "loss": 4.2187,
+      "step": 3139
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8535082174462707e-06,
+      "loss": 3.8161,
+      "step": 3140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853460809102402e-06,
+      "loss": 4.1661,
+      "step": 3141
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8534134007585332e-06,
+      "loss": 3.9505,
+      "step": 3142
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853365992414665e-06,
+      "loss": 3.961,
+      "step": 3143
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8533185840707966e-06,
+      "loss": 4.0626,
+      "step": 3144
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853271175726928e-06,
+      "loss": 3.8862,
+      "step": 3145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8532237673830596e-06,
+      "loss": 3.889,
+      "step": 3146
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853176359039191e-06,
+      "loss": 4.2257,
+      "step": 3147
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853128950695322e-06,
+      "loss": 3.9495,
+      "step": 3148
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.853081542351454e-06,
+      "loss": 3.9226,
+      "step": 3149
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8530341340075856e-06,
+      "loss": 3.9828,
+      "step": 3150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852986725663717e-06,
+      "loss": 3.5929,
+      "step": 3151
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8529393173198486e-06,
+      "loss": 3.631,
+      "step": 3152
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.85289190897598e-06,
+      "loss": 4.0281,
+      "step": 3153
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852844500632111e-06,
+      "loss": 4.1517,
+      "step": 3154
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852797092288243e-06,
+      "loss": 4.0848,
+      "step": 3155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852749683944374e-06,
+      "loss": 4.3812,
+      "step": 3156
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852702275600506e-06,
+      "loss": 3.8708,
+      "step": 3157
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8526548672566375e-06,
+      "loss": 3.8151,
+      "step": 3158
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852607458912769e-06,
+      "loss": 3.7619,
+      "step": 3159
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8525600505689e-06,
+      "loss": 3.9403,
+      "step": 3160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852512642225032e-06,
+      "loss": 3.8403,
+      "step": 3161
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852465233881163e-06,
+      "loss": 3.7695,
+      "step": 3162
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8524178255372948e-06,
+      "loss": 3.9751,
+      "step": 3163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852370417193426e-06,
+      "loss": 4.0364,
+      "step": 3164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8523230088495578e-06,
+      "loss": 3.7275,
+      "step": 3165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852275600505689e-06,
+      "loss": 3.8097,
+      "step": 3166
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8522281921618203e-06,
+      "loss": 4.1026,
+      "step": 3167
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852180783817952e-06,
+      "loss": 4.2685,
+      "step": 3168
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8521333754740833e-06,
+      "loss": 3.8939,
+      "step": 3169
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.852085967130215e-06,
+      "loss": 3.9162,
+      "step": 3170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8520385587863467e-06,
+      "loss": 4.0161,
+      "step": 3171
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851991150442478e-06,
+      "loss": 3.8835,
+      "step": 3172
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8519437420986093e-06,
+      "loss": 3.9733,
+      "step": 3173
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851896333754741e-06,
+      "loss": 4.2583,
+      "step": 3174
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8518489254108723e-06,
+      "loss": 4.1458,
+      "step": 3175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851801517067004e-06,
+      "loss": 3.9872,
+      "step": 3176
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8517541087231357e-06,
+      "loss": 4.054,
+      "step": 3177
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851706700379267e-06,
+      "loss": 4.007,
+      "step": 3178
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8516592920353982e-06,
+      "loss": 3.5869,
+      "step": 3179
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.85161188369153e-06,
+      "loss": 4.1818,
+      "step": 3180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851564475347661e-06,
+      "loss": 4.0634,
+      "step": 3181
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8515170670037925e-06,
+      "loss": 4.0291,
+      "step": 3182
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851469658659924e-06,
+      "loss": 4.1569,
+      "step": 3183
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851422250316056e-06,
+      "loss": 4.0713,
+      "step": 3184
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851374841972187e-06,
+      "loss": 3.9249,
+      "step": 3185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8513274336283185e-06,
+      "loss": 4.2745,
+      "step": 3186
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.85128002528445e-06,
+      "loss": 3.8626,
+      "step": 3187
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8512326169405814e-06,
+      "loss": 4.1425,
+      "step": 3188
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851185208596713e-06,
+      "loss": 4.0933,
+      "step": 3189
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851137800252845e-06,
+      "loss": 4.0085,
+      "step": 3190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.851090391908976e-06,
+      "loss": 3.962,
+      "step": 3191
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8510429835651074e-06,
+      "loss": 3.8084,
+      "step": 3192
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.850995575221239e-06,
+      "loss": 4.1058,
+      "step": 3193
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8509481668773704e-06,
+      "loss": 4.0925,
+      "step": 3194
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8509007585335017e-06,
+      "loss": 3.9475,
+      "step": 3195
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8508533501896334e-06,
+      "loss": 3.8658,
+      "step": 3196
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.850805941845765e-06,
+      "loss": 3.8319,
+      "step": 3197
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.8507585335018964e-06,
+      "loss": 3.8538,
+      "step": 3198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.850711125158028e-06,
+      "loss": 3.8406,
+      "step": 3199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8506637168141594e-06,
+      "loss": 4.3007,
+      "step": 3200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8506163084702906e-06,
+      "loss": 4.0736,
+      "step": 3201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8505689001264223e-06,
+      "loss": 3.9136,
+      "step": 3202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.850521491782554e-06,
+      "loss": 4.0481,
+      "step": 3203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8504740834386853e-06,
+      "loss": 4.1305,
+      "step": 3204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.850426675094817e-06,
+      "loss": 3.7877,
+      "step": 3205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8503792667509483e-06,
+      "loss": 3.9601,
+      "step": 3206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8503318584070796e-06,
+      "loss": 3.7432,
+      "step": 3207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.850284450063211e-06,
+      "loss": 3.7546,
+      "step": 3208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8502370417193426e-06,
+      "loss": 4.0511,
+      "step": 3209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8501896333754743e-06,
+      "loss": 4.1773,
+      "step": 3210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8501422250316056e-06,
+      "loss": 3.7765,
+      "step": 3211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8500948166877373e-06,
+      "loss": 3.6061,
+      "step": 3212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8500474083438685e-06,
+      "loss": 4.175,
+      "step": 3213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.85e-06,
+      "loss": 4.0584,
+      "step": 3214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8499525916561315e-06,
+      "loss": 3.7986,
+      "step": 3215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8499051833122632e-06,
+      "loss": 3.7029,
+      "step": 3216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8498577749683945e-06,
+      "loss": 4.0174,
+      "step": 3217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.849810366624526e-06,
+      "loss": 3.7426,
+      "step": 3218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8497629582806575e-06,
+      "loss": 3.9288,
+      "step": 3219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8497155499367888e-06,
+      "loss": 3.7305,
+      "step": 3220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8496681415929205e-06,
+      "loss": 4.1365,
+      "step": 3221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8496207332490518e-06,
+      "loss": 4.1998,
+      "step": 3222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8495733249051835e-06,
+      "loss": 3.5829,
+      "step": 3223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.849525916561315e-06,
+      "loss": 3.9924,
+      "step": 3224
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8494785082174464e-06,
+      "loss": 4.0471,
+      "step": 3225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8494310998735777e-06,
+      "loss": 3.7225,
+      "step": 3226
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8493836915297094e-06,
+      "loss": 3.9505,
+      "step": 3227
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8493362831858407e-06,
+      "loss": 4.1424,
+      "step": 3228
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8492888748419724e-06,
+      "loss": 3.8492,
+      "step": 3229
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8492414664981037e-06,
+      "loss": 4.0047,
+      "step": 3230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8491940581542354e-06,
+      "loss": 4.2507,
+      "step": 3231
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8491466498103667e-06,
+      "loss": 4.0507,
+      "step": 3232
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.849099241466498e-06,
+      "loss": 3.8502,
+      "step": 3233
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8490518331226297e-06,
+      "loss": 3.7363,
+      "step": 3234
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.849004424778761e-06,
+      "loss": 3.6699,
+      "step": 3235
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8489570164348926e-06,
+      "loss": 3.6437,
+      "step": 3236
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8489096080910243e-06,
+      "loss": 3.7652,
+      "step": 3237
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8488621997471556e-06,
+      "loss": 3.9255,
+      "step": 3238
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848814791403287e-06,
+      "loss": 4.0045,
+      "step": 3239
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8487673830594186e-06,
+      "loss": 4.0913,
+      "step": 3240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.84871997471555e-06,
+      "loss": 3.7035,
+      "step": 3241
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848672566371681e-06,
+      "loss": 4.0873,
+      "step": 3242
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8486251580278133e-06,
+      "loss": 4.1216,
+      "step": 3243
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8485777496839446e-06,
+      "loss": 4.1489,
+      "step": 3244
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848530341340076e-06,
+      "loss": 4.3612,
+      "step": 3245
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8484829329962076e-06,
+      "loss": 3.9168,
+      "step": 3246
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848435524652339e-06,
+      "loss": 3.9658,
+      "step": 3247
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.84838811630847e-06,
+      "loss": 4.1823,
+      "step": 3248
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848340707964602e-06,
+      "loss": 4.0717,
+      "step": 3249
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8482932996207335e-06,
+      "loss": 4.2257,
+      "step": 3250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848245891276865e-06,
+      "loss": 4.2034,
+      "step": 3251
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848198482932996e-06,
+      "loss": 3.7795,
+      "step": 3252
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848151074589128e-06,
+      "loss": 3.5487,
+      "step": 3253
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848103666245259e-06,
+      "loss": 4.0096,
+      "step": 3254
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8480562579013904e-06,
+      "loss": 4.0075,
+      "step": 3255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.848008849557522e-06,
+      "loss": 3.7353,
+      "step": 3256
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8479614412136538e-06,
+      "loss": 4.0849,
+      "step": 3257
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.847914032869785e-06,
+      "loss": 4.3473,
+      "step": 3258
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8478666245259168e-06,
+      "loss": 4.1106,
+      "step": 3259
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.847819216182048e-06,
+      "loss": 3.9335,
+      "step": 3260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.8477718078381793e-06,
+      "loss": 3.8164,
+      "step": 3261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847724399494311e-06,
+      "loss": 4.2724,
+      "step": 3262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8476769911504427e-06,
+      "loss": 4.019,
+      "step": 3263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847629582806574e-06,
+      "loss": 4.062,
+      "step": 3264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8475821744627057e-06,
+      "loss": 4.0173,
+      "step": 3265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847534766118837e-06,
+      "loss": 4.2325,
+      "step": 3266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8474873577749683e-06,
+      "loss": 4.2004,
+      "step": 3267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8474399494311e-06,
+      "loss": 4.3,
+      "step": 3268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8473925410872313e-06,
+      "loss": 4.0985,
+      "step": 3269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847345132743363e-06,
+      "loss": 3.8515,
+      "step": 3270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8472977243994947e-06,
+      "loss": 3.951,
+      "step": 3271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847250316055626e-06,
+      "loss": 4.1328,
+      "step": 3272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8472029077117572e-06,
+      "loss": 3.4791,
+      "step": 3273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8471554993678885e-06,
+      "loss": 4.0012,
+      "step": 3274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.84710809102402e-06,
+      "loss": 4.065,
+      "step": 3275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847060682680152e-06,
+      "loss": 4.4587,
+      "step": 3276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.847013274336283e-06,
+      "loss": 3.9945,
+      "step": 3277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846965865992415e-06,
+      "loss": 3.9592,
+      "step": 3278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846918457648546e-06,
+      "loss": 3.895,
+      "step": 3279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8468710493046775e-06,
+      "loss": 3.8862,
+      "step": 3280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846823640960809e-06,
+      "loss": 4.0777,
+      "step": 3281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8467762326169404e-06,
+      "loss": 4.1041,
+      "step": 3282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846728824273072e-06,
+      "loss": 3.988,
+      "step": 3283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846681415929204e-06,
+      "loss": 4.1973,
+      "step": 3284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846634007585335e-06,
+      "loss": 4.0899,
+      "step": 3285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8465865992414664e-06,
+      "loss": 4.0553,
+      "step": 3286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846539190897598e-06,
+      "loss": 4.2318,
+      "step": 3287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8464917825537294e-06,
+      "loss": 3.9194,
+      "step": 3288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846444374209861e-06,
+      "loss": 4.2093,
+      "step": 3289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846396965865993e-06,
+      "loss": 4.0903,
+      "step": 3290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846349557522124e-06,
+      "loss": 4.1863,
+      "step": 3291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8463021491782554e-06,
+      "loss": 3.9888,
+      "step": 3292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846254740834387e-06,
+      "loss": 4.1797,
+      "step": 3293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8462073324905183e-06,
+      "loss": 4.1728,
+      "step": 3294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8461599241466496e-06,
+      "loss": 4.0784,
+      "step": 3295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8461125158027813e-06,
+      "loss": 3.9285,
+      "step": 3296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.846065107458913e-06,
+      "loss": 4.0238,
+      "step": 3297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8460176991150443e-06,
+      "loss": 4.0313,
+      "step": 3298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8459702907711756e-06,
+      "loss": 3.6896,
+      "step": 3299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8459228824273073e-06,
+      "loss": 4.1374,
+      "step": 3300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8458754740834386e-06,
+      "loss": 3.61,
+      "step": 3301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8458280657395703e-06,
+      "loss": 4.2554,
+      "step": 3302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.845780657395702e-06,
+      "loss": 4.1162,
+      "step": 3303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8457332490518333e-06,
+      "loss": 4.0419,
+      "step": 3304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8456858407079645e-06,
+      "loss": 4.2755,
+      "step": 3305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8456384323640963e-06,
+      "loss": 4.1305,
+      "step": 3306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8455910240202275e-06,
+      "loss": 4.0028,
+      "step": 3307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.845543615676359e-06,
+      "loss": 4.1019,
+      "step": 3308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8454962073324905e-06,
+      "loss": 3.8756,
+      "step": 3309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8454487989886222e-06,
+      "loss": 4.1182,
+      "step": 3310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8454013906447535e-06,
+      "loss": 3.9198,
+      "step": 3311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.845353982300885e-06,
+      "loss": 4.0989,
+      "step": 3312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8453065739570165e-06,
+      "loss": 3.9643,
+      "step": 3313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8452591656131478e-06,
+      "loss": 4.1609,
+      "step": 3314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8452117572692795e-06,
+      "loss": 4.0594,
+      "step": 3315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.845164348925411e-06,
+      "loss": 3.9928,
+      "step": 3316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8451169405815425e-06,
+      "loss": 3.9806,
+      "step": 3317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8450695322376737e-06,
+      "loss": 4.2611,
+      "step": 3318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8450221238938054e-06,
+      "loss": 3.9456,
+      "step": 3319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8449747155499367e-06,
+      "loss": 3.8596,
+      "step": 3320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.844927307206068e-06,
+      "loss": 4.0101,
+      "step": 3321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8448798988621997e-06,
+      "loss": 3.8591,
+      "step": 3322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8448324905183314e-06,
+      "loss": 4.0952,
+      "step": 3323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.8447850821744627e-06,
+      "loss": 4.1039,
+      "step": 3324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8447376738305944e-06,
+      "loss": 3.8993,
+      "step": 3325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8446902654867257e-06,
+      "loss": 4.2299,
+      "step": 3326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.844642857142857e-06,
+      "loss": 4.2097,
+      "step": 3327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8445954487989887e-06,
+      "loss": 3.9081,
+      "step": 3328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8445480404551204e-06,
+      "loss": 3.8714,
+      "step": 3329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8445006321112516e-06,
+      "loss": 4.1885,
+      "step": 3330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8444532237673833e-06,
+      "loss": 4.0365,
+      "step": 3331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8444058154235146e-06,
+      "loss": 4.2606,
+      "step": 3332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.844358407079646e-06,
+      "loss": 4.0517,
+      "step": 3333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8443109987357776e-06,
+      "loss": 3.9856,
+      "step": 3334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.844263590391909e-06,
+      "loss": 4.1609,
+      "step": 3335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8442161820480406e-06,
+      "loss": 4.1131,
+      "step": 3336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8441687737041723e-06,
+      "loss": 4.0208,
+      "step": 3337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8441213653603036e-06,
+      "loss": 4.0128,
+      "step": 3338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.844073957016435e-06,
+      "loss": 3.9152,
+      "step": 3339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.844026548672566e-06,
+      "loss": 3.6427,
+      "step": 3340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843979140328698e-06,
+      "loss": 4.0652,
+      "step": 3341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8439317319848295e-06,
+      "loss": 4.2037,
+      "step": 3342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843884323640961e-06,
+      "loss": 3.9756,
+      "step": 3343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8438369152970925e-06,
+      "loss": 3.9854,
+      "step": 3344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843789506953224e-06,
+      "loss": 4.2244,
+      "step": 3345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843742098609355e-06,
+      "loss": 3.9394,
+      "step": 3346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843694690265487e-06,
+      "loss": 3.8377,
+      "step": 3347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843647281921618e-06,
+      "loss": 3.9436,
+      "step": 3348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8435998735777498e-06,
+      "loss": 4.1421,
+      "step": 3349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8435524652338815e-06,
+      "loss": 4.3264,
+      "step": 3350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8435050568900128e-06,
+      "loss": 3.4291,
+      "step": 3351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843457648546144e-06,
+      "loss": 3.9527,
+      "step": 3352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8434102402022757e-06,
+      "loss": 4.1288,
+      "step": 3353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843362831858407e-06,
+      "loss": 4.1094,
+      "step": 3354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8433154235145387e-06,
+      "loss": 4.0545,
+      "step": 3355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8432680151706704e-06,
+      "loss": 4.2958,
+      "step": 3356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8432206068268017e-06,
+      "loss": 4.1574,
+      "step": 3357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843173198482933e-06,
+      "loss": 3.7805,
+      "step": 3358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8431257901390647e-06,
+      "loss": 3.6336,
+      "step": 3359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.843078381795196e-06,
+      "loss": 3.8951,
+      "step": 3360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8430309734513273e-06,
+      "loss": 4.1496,
+      "step": 3361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842983565107459e-06,
+      "loss": 3.7484,
+      "step": 3362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8429361567635907e-06,
+      "loss": 4.0261,
+      "step": 3363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842888748419722e-06,
+      "loss": 3.9991,
+      "step": 3364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8428413400758532e-06,
+      "loss": 3.8121,
+      "step": 3365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842793931731985e-06,
+      "loss": 4.0848,
+      "step": 3366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8427465233881162e-06,
+      "loss": 4.1614,
+      "step": 3367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842699115044248e-06,
+      "loss": 3.9061,
+      "step": 3368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8426517067003796e-06,
+      "loss": 3.8432,
+      "step": 3369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842604298356511e-06,
+      "loss": 4.1099,
+      "step": 3370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842556890012642e-06,
+      "loss": 3.9017,
+      "step": 3371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842509481668774e-06,
+      "loss": 3.7342,
+      "step": 3372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842462073324905e-06,
+      "loss": 3.8959,
+      "step": 3373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8424146649810364e-06,
+      "loss": 4.107,
+      "step": 3374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842367256637168e-06,
+      "loss": 4.2061,
+      "step": 3375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8423198482933e-06,
+      "loss": 3.8755,
+      "step": 3376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842272439949431e-06,
+      "loss": 4.1551,
+      "step": 3377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842225031605563e-06,
+      "loss": 3.979,
+      "step": 3378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842177623261694e-06,
+      "loss": 4.0231,
+      "step": 3379
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8421302149178254e-06,
+      "loss": 4.2985,
+      "step": 3380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842082806573957e-06,
+      "loss": 3.9869,
+      "step": 3381
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.842035398230089e-06,
+      "loss": 4.1726,
+      "step": 3382
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.84198798988622e-06,
+      "loss": 4.1094,
+      "step": 3383
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8419405815423514e-06,
+      "loss": 3.7078,
+      "step": 3384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.841893173198483e-06,
+      "loss": 3.8654,
+      "step": 3385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8418457648546144e-06,
+      "loss": 3.9801,
+      "step": 3386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8417983565107456e-06,
+      "loss": 4.0401,
+      "step": 3387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.8417509481668773e-06,
+      "loss": 4.0572,
+      "step": 3388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.841703539823009e-06,
+      "loss": 4.13,
+      "step": 3389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8416561314791403e-06,
+      "loss": 3.9613,
+      "step": 3390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.841608723135272e-06,
+      "loss": 4.1832,
+      "step": 3391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8415613147914033e-06,
+      "loss": 4.0637,
+      "step": 3392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8415139064475346e-06,
+      "loss": 3.8359,
+      "step": 3393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8414664981036663e-06,
+      "loss": 4.127,
+      "step": 3394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.841419089759798e-06,
+      "loss": 3.7345,
+      "step": 3395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8413716814159293e-06,
+      "loss": 3.8717,
+      "step": 3396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.841324273072061e-06,
+      "loss": 3.6571,
+      "step": 3397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8412768647281923e-06,
+      "loss": 4.0781,
+      "step": 3398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8412294563843235e-06,
+      "loss": 4.0092,
+      "step": 3399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8411820480404552e-06,
+      "loss": 4.0409,
+      "step": 3400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8411346396965865e-06,
+      "loss": 4.0681,
+      "step": 3401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8410872313527182e-06,
+      "loss": 4.1567,
+      "step": 3402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.84103982300885e-06,
+      "loss": 4.0077,
+      "step": 3403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8409924146649812e-06,
+      "loss": 3.7169,
+      "step": 3404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8409450063211125e-06,
+      "loss": 3.8264,
+      "step": 3405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8408975979772438e-06,
+      "loss": 3.8046,
+      "step": 3406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8408501896333755e-06,
+      "loss": 4.278,
+      "step": 3407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.840802781289507e-06,
+      "loss": 4.2323,
+      "step": 3408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8407553729456385e-06,
+      "loss": 4.1547,
+      "step": 3409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.84070796460177e-06,
+      "loss": 4.0135,
+      "step": 3410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8406605562579014e-06,
+      "loss": 3.7407,
+      "step": 3411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8406131479140327e-06,
+      "loss": 4.1592,
+      "step": 3412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8405657395701644e-06,
+      "loss": 3.9146,
+      "step": 3413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8405183312262957e-06,
+      "loss": 3.9191,
+      "step": 3414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8404709228824274e-06,
+      "loss": 4.0978,
+      "step": 3415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.840423514538559e-06,
+      "loss": 3.9251,
+      "step": 3416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8403761061946904e-06,
+      "loss": 3.8471,
+      "step": 3417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8403286978508217e-06,
+      "loss": 3.8769,
+      "step": 3418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8402812895069534e-06,
+      "loss": 3.9544,
+      "step": 3419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8402338811630847e-06,
+      "loss": 4.0952,
+      "step": 3420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8401864728192164e-06,
+      "loss": 3.9771,
+      "step": 3421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.840139064475348e-06,
+      "loss": 4.0728,
+      "step": 3422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8400916561314794e-06,
+      "loss": 3.5582,
+      "step": 3423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8400442477876106e-06,
+      "loss": 3.7196,
+      "step": 3424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8399968394437423e-06,
+      "loss": 4.0098,
+      "step": 3425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8399494310998736e-06,
+      "loss": 3.8148,
+      "step": 3426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839902022756005e-06,
+      "loss": 3.8621,
+      "step": 3427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8398546144121366e-06,
+      "loss": 4.2179,
+      "step": 3428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8398072060682683e-06,
+      "loss": 4.286,
+      "step": 3429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8397597977243996e-06,
+      "loss": 4.1319,
+      "step": 3430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839712389380531e-06,
+      "loss": 4.1061,
+      "step": 3431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8396649810366626e-06,
+      "loss": 4.1806,
+      "step": 3432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839617572692794e-06,
+      "loss": 3.844,
+      "step": 3433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8395701643489256e-06,
+      "loss": 4.0128,
+      "step": 3434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8395227560050573e-06,
+      "loss": 4.0485,
+      "step": 3435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8394753476611885e-06,
+      "loss": 4.1208,
+      "step": 3436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.83942793931732e-06,
+      "loss": 3.8922,
+      "step": 3437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8393805309734515e-06,
+      "loss": 4.1176,
+      "step": 3438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839333122629583e-06,
+      "loss": 4.013,
+      "step": 3439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839285714285714e-06,
+      "loss": 3.8656,
+      "step": 3440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839238305941846e-06,
+      "loss": 3.9401,
+      "step": 3441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8391908975979775e-06,
+      "loss": 4.2168,
+      "step": 3442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8391434892541088e-06,
+      "loss": 3.9425,
+      "step": 3443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8390960809102405e-06,
+      "loss": 4.1865,
+      "step": 3444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8390486725663718e-06,
+      "loss": 3.4707,
+      "step": 3445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.839001264222503e-06,
+      "loss": 4.1574,
+      "step": 3446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8389538558786347e-06,
+      "loss": 3.9453,
+      "step": 3447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8389064475347664e-06,
+      "loss": 3.8269,
+      "step": 3448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8388590391908977e-06,
+      "loss": 3.9071,
+      "step": 3449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.838811630847029e-06,
+      "loss": 3.8512,
+      "step": 3450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.8387642225031607e-06,
+      "loss": 4.1724,
+      "step": 3451
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838716814159292e-06,
+      "loss": 3.6811,
+      "step": 3452
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8386694058154233e-06,
+      "loss": 4.0229,
+      "step": 3453
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838621997471555e-06,
+      "loss": 3.9839,
+      "step": 3454
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8385745891276867e-06,
+      "loss": 4.2506,
+      "step": 3455
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838527180783818e-06,
+      "loss": 4.282,
+      "step": 3456
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8384797724399497e-06,
+      "loss": 4.083,
+      "step": 3457
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838432364096081e-06,
+      "loss": 4.1752,
+      "step": 3458
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8383849557522122e-06,
+      "loss": 4.0262,
+      "step": 3459
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838337547408344e-06,
+      "loss": 4.0313,
+      "step": 3460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8382901390644756e-06,
+      "loss": 4.2727,
+      "step": 3461
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838242730720607e-06,
+      "loss": 3.7938,
+      "step": 3462
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8381953223767386e-06,
+      "loss": 4.0337,
+      "step": 3463
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.83814791403287e-06,
+      "loss": 4.0418,
+      "step": 3464
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838100505689001e-06,
+      "loss": 3.3498,
+      "step": 3465
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838053097345133e-06,
+      "loss": 4.1244,
+      "step": 3466
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.838005689001264e-06,
+      "loss": 3.9626,
+      "step": 3467
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837958280657396e-06,
+      "loss": 3.836,
+      "step": 3468
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8379108723135276e-06,
+      "loss": 4.0314,
+      "step": 3469
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837863463969659e-06,
+      "loss": 4.1698,
+      "step": 3470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.83781605562579e-06,
+      "loss": 3.9175,
+      "step": 3471
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8377686472819214e-06,
+      "loss": 4.2193,
+      "step": 3472
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837721238938053e-06,
+      "loss": 3.7701,
+      "step": 3473
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837673830594185e-06,
+      "loss": 4.0349,
+      "step": 3474
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837626422250316e-06,
+      "loss": 4.1449,
+      "step": 3475
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837579013906448e-06,
+      "loss": 3.8098,
+      "step": 3476
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837531605562579e-06,
+      "loss": 4.1362,
+      "step": 3477
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8374841972187104e-06,
+      "loss": 4.1968,
+      "step": 3478
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837436788874842e-06,
+      "loss": 3.8455,
+      "step": 3479
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8373893805309733e-06,
+      "loss": 3.873,
+      "step": 3480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837341972187105e-06,
+      "loss": 3.6977,
+      "step": 3481
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8372945638432368e-06,
+      "loss": 3.8767,
+      "step": 3482
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837247155499368e-06,
+      "loss": 4.0132,
+      "step": 3483
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8371997471554993e-06,
+      "loss": 3.6138,
+      "step": 3484
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837152338811631e-06,
+      "loss": 3.9224,
+      "step": 3485
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8371049304677623e-06,
+      "loss": 3.7778,
+      "step": 3486
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.837057522123894e-06,
+      "loss": 4.1474,
+      "step": 3487
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8370101137800257e-06,
+      "loss": 4.2146,
+      "step": 3488
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836962705436157e-06,
+      "loss": 3.3345,
+      "step": 3489
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8369152970922883e-06,
+      "loss": 3.9384,
+      "step": 3490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.83686788874842e-06,
+      "loss": 3.6592,
+      "step": 3491
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8368204804045513e-06,
+      "loss": 4.1021,
+      "step": 3492
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8367730720606825e-06,
+      "loss": 4.174,
+      "step": 3493
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8367256637168142e-06,
+      "loss": 3.8622,
+      "step": 3494
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836678255372946e-06,
+      "loss": 3.924,
+      "step": 3495
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8366308470290772e-06,
+      "loss": 4.0024,
+      "step": 3496
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8365834386852085e-06,
+      "loss": 4.3008,
+      "step": 3497
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.83653603034134e-06,
+      "loss": 3.8643,
+      "step": 3498
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8364886219974715e-06,
+      "loss": 3.8072,
+      "step": 3499
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836441213653603e-06,
+      "loss": 4.1964,
+      "step": 3500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836393805309735e-06,
+      "loss": 4.2174,
+      "step": 3501
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836346396965866e-06,
+      "loss": 3.9189,
+      "step": 3502
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8362989886219975e-06,
+      "loss": 4.0785,
+      "step": 3503
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836251580278129e-06,
+      "loss": 3.9647,
+      "step": 3504
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8362041719342604e-06,
+      "loss": 3.4989,
+      "step": 3505
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8361567635903917e-06,
+      "loss": 4.1475,
+      "step": 3506
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8361093552465234e-06,
+      "loss": 4.0742,
+      "step": 3507
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.836061946902655e-06,
+      "loss": 4.0846,
+      "step": 3508
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8360145385587864e-06,
+      "loss": 3.8149,
+      "step": 3509
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.835967130214918e-06,
+      "loss": 4.193,
+      "step": 3510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8359197218710494e-06,
+      "loss": 3.8085,
+      "step": 3511
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8358723135271807e-06,
+      "loss": 3.8431,
+      "step": 3512
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.8358249051833124e-06,
+      "loss": 4.1322,
+      "step": 3513
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.835777496839444e-06,
+      "loss": 4.0056,
+      "step": 3514
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8357300884955754e-06,
+      "loss": 4.0075,
+      "step": 3515
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8356826801517066e-06,
+      "loss": 3.9127,
+      "step": 3516
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8356352718078383e-06,
+      "loss": 4.1374,
+      "step": 3517
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8355878634639696e-06,
+      "loss": 3.7458,
+      "step": 3518
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.835540455120101e-06,
+      "loss": 3.9845,
+      "step": 3519
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8354930467762326e-06,
+      "loss": 3.841,
+      "step": 3520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8354456384323643e-06,
+      "loss": 4.0221,
+      "step": 3521
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8353982300884956e-06,
+      "loss": 3.8753,
+      "step": 3522
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8353508217446273e-06,
+      "loss": 3.9227,
+      "step": 3523
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8353034134007586e-06,
+      "loss": 3.7812,
+      "step": 3524
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.83525600505689e-06,
+      "loss": 4.2146,
+      "step": 3525
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8352085967130216e-06,
+      "loss": 4.21,
+      "step": 3526
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.835161188369153e-06,
+      "loss": 4.2212,
+      "step": 3527
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8351137800252845e-06,
+      "loss": 3.6452,
+      "step": 3528
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8350663716814163e-06,
+      "loss": 4.2154,
+      "step": 3529
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8350189633375475e-06,
+      "loss": 4.1273,
+      "step": 3530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834971554993679e-06,
+      "loss": 3.972,
+      "step": 3531
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8349241466498105e-06,
+      "loss": 3.9319,
+      "step": 3532
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834876738305942e-06,
+      "loss": 4.0366,
+      "step": 3533
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8348293299620735e-06,
+      "loss": 3.8577,
+      "step": 3534
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834781921618205e-06,
+      "loss": 4.1486,
+      "step": 3535
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8347345132743365e-06,
+      "loss": 3.6554,
+      "step": 3536
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8346871049304678e-06,
+      "loss": 3.8526,
+      "step": 3537
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834639696586599e-06,
+      "loss": 3.6221,
+      "step": 3538
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8345922882427308e-06,
+      "loss": 3.9363,
+      "step": 3539
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834544879898862e-06,
+      "loss": 3.7978,
+      "step": 3540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8344974715549937e-06,
+      "loss": 3.9598,
+      "step": 3541
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8344500632111254e-06,
+      "loss": 4.222,
+      "step": 3542
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8344026548672567e-06,
+      "loss": 4.1659,
+      "step": 3543
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834355246523388e-06,
+      "loss": 3.8724,
+      "step": 3544
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8343078381795197e-06,
+      "loss": 4.1221,
+      "step": 3545
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834260429835651e-06,
+      "loss": 3.7379,
+      "step": 3546
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8342130214917827e-06,
+      "loss": 4.0098,
+      "step": 3547
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8341656131479144e-06,
+      "loss": 3.9579,
+      "step": 3548
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8341182048040457e-06,
+      "loss": 4.1165,
+      "step": 3549
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.834070796460177e-06,
+      "loss": 3.859,
+      "step": 3550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8340233881163087e-06,
+      "loss": 4.0143,
+      "step": 3551
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.83397597977244e-06,
+      "loss": 4.1133,
+      "step": 3552
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8339285714285712e-06,
+      "loss": 3.9335,
+      "step": 3553
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833881163084703e-06,
+      "loss": 4.1607,
+      "step": 3554
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8338337547408346e-06,
+      "loss": 4.2192,
+      "step": 3555
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833786346396966e-06,
+      "loss": 3.8359,
+      "step": 3556
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8337389380530976e-06,
+      "loss": 3.8213,
+      "step": 3557
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833691529709229e-06,
+      "loss": 4.0112,
+      "step": 3558
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.83364412136536e-06,
+      "loss": 3.9163,
+      "step": 3559
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833596713021492e-06,
+      "loss": 4.0393,
+      "step": 3560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8335493046776236e-06,
+      "loss": 3.7816,
+      "step": 3561
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833501896333755e-06,
+      "loss": 4.094,
+      "step": 3562
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833454487989886e-06,
+      "loss": 4.023,
+      "step": 3563
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833407079646018e-06,
+      "loss": 3.5304,
+      "step": 3564
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833359671302149e-06,
+      "loss": 3.971,
+      "step": 3565
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8333122629582804e-06,
+      "loss": 3.9354,
+      "step": 3566
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833264854614412e-06,
+      "loss": 3.937,
+      "step": 3567
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833217446270544e-06,
+      "loss": 3.8235,
+      "step": 3568
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833170037926675e-06,
+      "loss": 4.0662,
+      "step": 3569
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833122629582807e-06,
+      "loss": 4.1433,
+      "step": 3570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.833075221238938e-06,
+      "loss": 3.88,
+      "step": 3571
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8330278128950694e-06,
+      "loss": 4.2829,
+      "step": 3572
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.832980404551201e-06,
+      "loss": 3.7391,
+      "step": 3573
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8329329962073328e-06,
+      "loss": 4.0723,
+      "step": 3574
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.832885587863464e-06,
+      "loss": 4.1344,
+      "step": 3575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8328381795195958e-06,
+      "loss": 3.7499,
+      "step": 3576
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.832790771175727e-06,
+      "loss": 3.6985,
+      "step": 3577
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.8327433628318583e-06,
+      "loss": 3.7548,
+      "step": 3578
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8326959544879896e-06,
+      "loss": 4.235,
+      "step": 3579
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8326485461441213e-06,
+      "loss": 3.7661,
+      "step": 3580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.832601137800253e-06,
+      "loss": 3.9527,
+      "step": 3581
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8325537294563843e-06,
+      "loss": 4.1362,
+      "step": 3582
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.832506321112516e-06,
+      "loss": 4.1772,
+      "step": 3583
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8324589127686473e-06,
+      "loss": 4.13,
+      "step": 3584
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8324115044247785e-06,
+      "loss": 3.7872,
+      "step": 3585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8323640960809102e-06,
+      "loss": 3.8142,
+      "step": 3586
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.832316687737042e-06,
+      "loss": 3.8717,
+      "step": 3587
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8322692793931732e-06,
+      "loss": 3.9671,
+      "step": 3588
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.832221871049305e-06,
+      "loss": 3.6861,
+      "step": 3589
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8321744627054362e-06,
+      "loss": 3.9969,
+      "step": 3590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8321270543615675e-06,
+      "loss": 3.6873,
+      "step": 3591
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.832079646017699e-06,
+      "loss": 3.8846,
+      "step": 3592
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8320322376738305e-06,
+      "loss": 3.6036,
+      "step": 3593
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831984829329962e-06,
+      "loss": 3.7084,
+      "step": 3594
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831937420986094e-06,
+      "loss": 3.8878,
+      "step": 3595
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831890012642225e-06,
+      "loss": 3.8776,
+      "step": 3596
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8318426042983565e-06,
+      "loss": 3.7264,
+      "step": 3597
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831795195954488e-06,
+      "loss": 3.8343,
+      "step": 3598
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8317477876106194e-06,
+      "loss": 3.8593,
+      "step": 3599
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831700379266751e-06,
+      "loss": 4.144,
+      "step": 3600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8316529709228824e-06,
+      "loss": 3.6951,
+      "step": 3601
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831605562579014e-06,
+      "loss": 3.7764,
+      "step": 3602
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8315581542351454e-06,
+      "loss": 3.7194,
+      "step": 3603
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8315107458912767e-06,
+      "loss": 4.0668,
+      "step": 3604
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8314633375474084e-06,
+      "loss": 4.0326,
+      "step": 3605
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8314159292035397e-06,
+      "loss": 3.8806,
+      "step": 3606
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8313685208596714e-06,
+      "loss": 3.795,
+      "step": 3607
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831321112515803e-06,
+      "loss": 4.2278,
+      "step": 3608
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8312737041719344e-06,
+      "loss": 4.2399,
+      "step": 3609
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8312262958280656e-06,
+      "loss": 4.1051,
+      "step": 3610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8311788874841973e-06,
+      "loss": 4.1573,
+      "step": 3611
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8311314791403286e-06,
+      "loss": 4.2793,
+      "step": 3612
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8310840707964603e-06,
+      "loss": 4.176,
+      "step": 3613
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.831036662452592e-06,
+      "loss": 3.9055,
+      "step": 3614
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8309892541087233e-06,
+      "loss": 3.8448,
+      "step": 3615
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8309418457648546e-06,
+      "loss": 3.9395,
+      "step": 3616
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8308944374209863e-06,
+      "loss": 3.6318,
+      "step": 3617
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8308470290771176e-06,
+      "loss": 3.5726,
+      "step": 3618
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.830799620733249e-06,
+      "loss": 4.2832,
+      "step": 3619
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8307522123893806e-06,
+      "loss": 3.9203,
+      "step": 3620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8307048040455123e-06,
+      "loss": 3.9242,
+      "step": 3621
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8306573957016435e-06,
+      "loss": 4.1784,
+      "step": 3622
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.830609987357775e-06,
+      "loss": 4.0664,
+      "step": 3623
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8305625790139065e-06,
+      "loss": 4.1483,
+      "step": 3624
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.830515170670038e-06,
+      "loss": 3.8382,
+      "step": 3625
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8304677623261695e-06,
+      "loss": 3.7899,
+      "step": 3626
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8304203539823012e-06,
+      "loss": 3.9264,
+      "step": 3627
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8303729456384325e-06,
+      "loss": 3.9851,
+      "step": 3628
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8303255372945638e-06,
+      "loss": 4.0252,
+      "step": 3629
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8302781289506955e-06,
+      "loss": 4.1737,
+      "step": 3630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8302307206068268e-06,
+      "loss": 3.932,
+      "step": 3631
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.830183312262958e-06,
+      "loss": 3.8661,
+      "step": 3632
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8301359039190897e-06,
+      "loss": 3.8146,
+      "step": 3633
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8300884955752215e-06,
+      "loss": 3.7875,
+      "step": 3634
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8300410872313527e-06,
+      "loss": 3.6858,
+      "step": 3635
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8299936788874844e-06,
+      "loss": 4.2086,
+      "step": 3636
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8299462705436157e-06,
+      "loss": 3.8457,
+      "step": 3637
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.829898862199747e-06,
+      "loss": 4.0507,
+      "step": 3638
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8298514538558787e-06,
+      "loss": 3.9343,
+      "step": 3639
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8298040455120104e-06,
+      "loss": 4.0525,
+      "step": 3640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.8297566371681417e-06,
+      "loss": 3.8268,
+      "step": 3641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8297092288242734e-06,
+      "loss": 4.1471,
+      "step": 3642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8296618204804047e-06,
+      "loss": 4.0294,
+      "step": 3643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829614412136536e-06,
+      "loss": 4.0263,
+      "step": 3644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8295670037926672e-06,
+      "loss": 3.6708,
+      "step": 3645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829519595448799e-06,
+      "loss": 4.0468,
+      "step": 3646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8294721871049306e-06,
+      "loss": 4.1384,
+      "step": 3647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829424778761062e-06,
+      "loss": 3.8188,
+      "step": 3648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8293773704171936e-06,
+      "loss": 3.7078,
+      "step": 3649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829329962073325e-06,
+      "loss": 4.1079,
+      "step": 3650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829282553729456e-06,
+      "loss": 3.8471,
+      "step": 3651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829235145385588e-06,
+      "loss": 3.8211,
+      "step": 3652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8291877370417196e-06,
+      "loss": 4.1784,
+      "step": 3653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829140328697851e-06,
+      "loss": 4.1116,
+      "step": 3654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8290929203539826e-06,
+      "loss": 4.1432,
+      "step": 3655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.829045512010114e-06,
+      "loss": 4.2904,
+      "step": 3656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828998103666245e-06,
+      "loss": 3.7886,
+      "step": 3657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828950695322377e-06,
+      "loss": 4.003,
+      "step": 3658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828903286978508e-06,
+      "loss": 4.1487,
+      "step": 3659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.82885587863464e-06,
+      "loss": 3.7709,
+      "step": 3660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8288084702907715e-06,
+      "loss": 4.1269,
+      "step": 3661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828761061946903e-06,
+      "loss": 3.9264,
+      "step": 3662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828713653603034e-06,
+      "loss": 3.9911,
+      "step": 3663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828666245259166e-06,
+      "loss": 3.8999,
+      "step": 3664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828618836915297e-06,
+      "loss": 4.1172,
+      "step": 3665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8285714285714288e-06,
+      "loss": 3.9893,
+      "step": 3666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.82852402022756e-06,
+      "loss": 3.8053,
+      "step": 3667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8284766118836918e-06,
+      "loss": 4.2677,
+      "step": 3668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828429203539823e-06,
+      "loss": 3.8916,
+      "step": 3669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8283817951959543e-06,
+      "loss": 4.0248,
+      "step": 3670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828334386852086e-06,
+      "loss": 4.0831,
+      "step": 3671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8282869785082173e-06,
+      "loss": 3.9534,
+      "step": 3672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828239570164349e-06,
+      "loss": 3.8027,
+      "step": 3673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8281921618204807e-06,
+      "loss": 3.746,
+      "step": 3674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828144753476612e-06,
+      "loss": 3.8802,
+      "step": 3675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8280973451327433e-06,
+      "loss": 4.0684,
+      "step": 3676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.828049936788875e-06,
+      "loss": 3.6349,
+      "step": 3677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8280025284450063e-06,
+      "loss": 3.5736,
+      "step": 3678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827955120101138e-06,
+      "loss": 4.0344,
+      "step": 3679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8279077117572697e-06,
+      "loss": 4.1746,
+      "step": 3680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827860303413401e-06,
+      "loss": 4.1237,
+      "step": 3681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8278128950695322e-06,
+      "loss": 4.1426,
+      "step": 3682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827765486725664e-06,
+      "loss": 3.7231,
+      "step": 3683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8277180783817952e-06,
+      "loss": 4.1278,
+      "step": 3684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8276706700379265e-06,
+      "loss": 4.2125,
+      "step": 3685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827623261694058e-06,
+      "loss": 3.9561,
+      "step": 3686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.82757585335019e-06,
+      "loss": 4.1017,
+      "step": 3687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827528445006321e-06,
+      "loss": 4.1872,
+      "step": 3688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8274810366624525e-06,
+      "loss": 4.3823,
+      "step": 3689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827433628318584e-06,
+      "loss": 3.9678,
+      "step": 3690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8273862199747154e-06,
+      "loss": 4.0669,
+      "step": 3691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827338811630847e-06,
+      "loss": 4.0943,
+      "step": 3692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827291403286979e-06,
+      "loss": 3.9183,
+      "step": 3693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.82724399494311e-06,
+      "loss": 4.1064,
+      "step": 3694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8271965865992414e-06,
+      "loss": 3.555,
+      "step": 3695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.827149178255373e-06,
+      "loss": 3.8711,
+      "step": 3696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8271017699115044e-06,
+      "loss": 3.8674,
+      "step": 3697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8270543615676357e-06,
+      "loss": 3.9763,
+      "step": 3698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8270069532237674e-06,
+      "loss": 3.9666,
+      "step": 3699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.826959544879899e-06,
+      "loss": 4.06,
+      "step": 3700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8269121365360304e-06,
+      "loss": 3.9358,
+      "step": 3701
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.826864728192162e-06,
+      "loss": 3.9167,
+      "step": 3702
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8268173198482934e-06,
+      "loss": 4.0725,
+      "step": 3703
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.8267699115044246e-06,
+      "loss": 4.1387,
+      "step": 3704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8267225031605563e-06,
+      "loss": 4.0027,
+      "step": 3705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.826675094816688e-06,
+      "loss": 4.1107,
+      "step": 3706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8266276864728193e-06,
+      "loss": 3.7142,
+      "step": 3707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.826580278128951e-06,
+      "loss": 3.4977,
+      "step": 3708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8265328697850823e-06,
+      "loss": 3.4584,
+      "step": 3709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8264854614412136e-06,
+      "loss": 3.6927,
+      "step": 3710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.826438053097345e-06,
+      "loss": 4.2128,
+      "step": 3711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8263906447534766e-06,
+      "loss": 3.9652,
+      "step": 3712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8263432364096083e-06,
+      "loss": 3.8211,
+      "step": 3713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8262958280657396e-06,
+      "loss": 3.9477,
+      "step": 3714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8262484197218713e-06,
+      "loss": 3.8251,
+      "step": 3715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8262010113780025e-06,
+      "loss": 3.9753,
+      "step": 3716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.826153603034134e-06,
+      "loss": 4.2439,
+      "step": 3717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8261061946902655e-06,
+      "loss": 4.0114,
+      "step": 3718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8260587863463972e-06,
+      "loss": 4.0683,
+      "step": 3719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8260113780025285e-06,
+      "loss": 4.1422,
+      "step": 3720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.82596396965866e-06,
+      "loss": 3.6089,
+      "step": 3721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8259165613147915e-06,
+      "loss": 4.1464,
+      "step": 3722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8258691529709228e-06,
+      "loss": 3.8865,
+      "step": 3723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8258217446270545e-06,
+      "loss": 4.3156,
+      "step": 3724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8257743362831858e-06,
+      "loss": 4.2179,
+      "step": 3725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8257269279393175e-06,
+      "loss": 3.9561,
+      "step": 3726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.825679519595449e-06,
+      "loss": 4.1375,
+      "step": 3727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8256321112515804e-06,
+      "loss": 3.9423,
+      "step": 3728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8255847029077117e-06,
+      "loss": 3.8773,
+      "step": 3729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8255372945638434e-06,
+      "loss": 3.7718,
+      "step": 3730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8254898862199747e-06,
+      "loss": 3.7184,
+      "step": 3731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8254424778761064e-06,
+      "loss": 4.1844,
+      "step": 3732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8253950695322377e-06,
+      "loss": 3.542,
+      "step": 3733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8253476611883694e-06,
+      "loss": 3.7422,
+      "step": 3734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8253002528445007e-06,
+      "loss": 4.039,
+      "step": 3735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.825252844500632e-06,
+      "loss": 3.7038,
+      "step": 3736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8252054361567637e-06,
+      "loss": 3.7835,
+      "step": 3737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.825158027812895e-06,
+      "loss": 4.0275,
+      "step": 3738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8251106194690266e-06,
+      "loss": 4.1085,
+      "step": 3739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8250632111251584e-06,
+      "loss": 3.7309,
+      "step": 3740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8250158027812896e-06,
+      "loss": 4.2165,
+      "step": 3741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824968394437421e-06,
+      "loss": 4.0035,
+      "step": 3742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8249209860935526e-06,
+      "loss": 3.9388,
+      "step": 3743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824873577749684e-06,
+      "loss": 4.1442,
+      "step": 3744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8248261694058156e-06,
+      "loss": 4.0502,
+      "step": 3745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8247787610619473e-06,
+      "loss": 3.6168,
+      "step": 3746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8247313527180786e-06,
+      "loss": 4.0217,
+      "step": 3747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.82468394437421e-06,
+      "loss": 4.0216,
+      "step": 3748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8246365360303416e-06,
+      "loss": 3.981,
+      "step": 3749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824589127686473e-06,
+      "loss": 3.8487,
+      "step": 3750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824541719342604e-06,
+      "loss": 3.8668,
+      "step": 3751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824494310998736e-06,
+      "loss": 3.992,
+      "step": 3752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8244469026548675e-06,
+      "loss": 3.8297,
+      "step": 3753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824399494310999e-06,
+      "loss": 3.8976,
+      "step": 3754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.82435208596713e-06,
+      "loss": 4.0109,
+      "step": 3755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824304677623262e-06,
+      "loss": 3.984,
+      "step": 3756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824257269279393e-06,
+      "loss": 4.0274,
+      "step": 3757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8242098609355248e-06,
+      "loss": 3.9717,
+      "step": 3758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8241624525916565e-06,
+      "loss": 4.0168,
+      "step": 3759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8241150442477878e-06,
+      "loss": 3.7284,
+      "step": 3760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.824067635903919e-06,
+      "loss": 3.6803,
+      "step": 3761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8240202275600508e-06,
+      "loss": 3.6673,
+      "step": 3762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.823972819216182e-06,
+      "loss": 3.9934,
+      "step": 3763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8239254108723133e-06,
+      "loss": 3.8149,
+      "step": 3764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.823878002528445e-06,
+      "loss": 3.8301,
+      "step": 3765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8238305941845767e-06,
+      "loss": 4.0959,
+      "step": 3766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.823783185840708e-06,
+      "loss": 3.7381,
+      "step": 3767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2.8237357774968397e-06,
+      "loss": 3.8009,
+      "step": 3768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823688369152971e-06,
+      "loss": 3.9449,
+      "step": 3769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8236409608091023e-06,
+      "loss": 4.0984,
+      "step": 3770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823593552465234e-06,
+      "loss": 3.8451,
+      "step": 3771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8235461441213657e-06,
+      "loss": 3.7158,
+      "step": 3772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823498735777497e-06,
+      "loss": 4.0231,
+      "step": 3773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8234513274336287e-06,
+      "loss": 3.8152,
+      "step": 3774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.82340391908976e-06,
+      "loss": 3.9403,
+      "step": 3775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8233565107458912e-06,
+      "loss": 3.7202,
+      "step": 3776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8233091024020225e-06,
+      "loss": 4.1743,
+      "step": 3777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823261694058154e-06,
+      "loss": 3.6637,
+      "step": 3778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823214285714286e-06,
+      "loss": 3.4872,
+      "step": 3779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823166877370417e-06,
+      "loss": 4.0626,
+      "step": 3780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.823119469026549e-06,
+      "loss": 3.8245,
+      "step": 3781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.82307206068268e-06,
+      "loss": 4.1223,
+      "step": 3782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8230246523388115e-06,
+      "loss": 3.7129,
+      "step": 3783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822977243994943e-06,
+      "loss": 4.1173,
+      "step": 3784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822929835651075e-06,
+      "loss": 3.7826,
+      "step": 3785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822882427307206e-06,
+      "loss": 3.9241,
+      "step": 3786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822835018963338e-06,
+      "loss": 4.1579,
+      "step": 3787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822787610619469e-06,
+      "loss": 4.109,
+      "step": 3788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8227402022756004e-06,
+      "loss": 3.8728,
+      "step": 3789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822692793931732e-06,
+      "loss": 4.2247,
+      "step": 3790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8226453855878634e-06,
+      "loss": 4.132,
+      "step": 3791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822597977243995e-06,
+      "loss": 4.0041,
+      "step": 3792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822550568900127e-06,
+      "loss": 4.0496,
+      "step": 3793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822503160556258e-06,
+      "loss": 3.6942,
+      "step": 3794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8224557522123894e-06,
+      "loss": 3.8315,
+      "step": 3795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822408343868521e-06,
+      "loss": 3.8838,
+      "step": 3796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8223609355246523e-06,
+      "loss": 3.8193,
+      "step": 3797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8223135271807836e-06,
+      "loss": 3.7496,
+      "step": 3798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8222661188369153e-06,
+      "loss": 3.6188,
+      "step": 3799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.822218710493047e-06,
+      "loss": 4.4304,
+      "step": 3800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8221713021491783e-06,
+      "loss": 3.7436,
+      "step": 3801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8221238938053096e-06,
+      "loss": 3.8376,
+      "step": 3802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8220764854614413e-06,
+      "loss": 3.8529,
+      "step": 3803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8220290771175726e-06,
+      "loss": 4.0401,
+      "step": 3804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8219816687737043e-06,
+      "loss": 3.7383,
+      "step": 3805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.821934260429836e-06,
+      "loss": 4.0783,
+      "step": 3806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8218868520859673e-06,
+      "loss": 3.8375,
+      "step": 3807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8218394437420985e-06,
+      "loss": 4.0679,
+      "step": 3808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8217920353982303e-06,
+      "loss": 4.1363,
+      "step": 3809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8217446270543615e-06,
+      "loss": 4.046,
+      "step": 3810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.821697218710493e-06,
+      "loss": 3.6446,
+      "step": 3811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8216498103666245e-06,
+      "loss": 4.0487,
+      "step": 3812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8216024020227562e-06,
+      "loss": 4.0809,
+      "step": 3813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8215549936788875e-06,
+      "loss": 3.9083,
+      "step": 3814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.821507585335019e-06,
+      "loss": 3.8927,
+      "step": 3815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8214601769911505e-06,
+      "loss": 3.7858,
+      "step": 3816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8214127686472818e-06,
+      "loss": 4.07,
+      "step": 3817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8213653603034135e-06,
+      "loss": 3.8705,
+      "step": 3818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.821317951959545e-06,
+      "loss": 3.9523,
+      "step": 3819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8212705436156765e-06,
+      "loss": 4.0025,
+      "step": 3820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8212231352718077e-06,
+      "loss": 4.2913,
+      "step": 3821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8211757269279394e-06,
+      "loss": 3.9826,
+      "step": 3822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8211283185840707e-06,
+      "loss": 4.0268,
+      "step": 3823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.821080910240202e-06,
+      "loss": 4.0431,
+      "step": 3824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8210335018963337e-06,
+      "loss": 3.9196,
+      "step": 3825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8209860935524654e-06,
+      "loss": 3.9267,
+      "step": 3826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8209386852085967e-06,
+      "loss": 3.688,
+      "step": 3827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8208912768647284e-06,
+      "loss": 3.6138,
+      "step": 3828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8208438685208597e-06,
+      "loss": 4.0462,
+      "step": 3829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.820796460176991e-06,
+      "loss": 3.9535,
+      "step": 3830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2.8207490518331227e-06,
+      "loss": 3.9108,
+      "step": 3831
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8207016434892544e-06,
+      "loss": 3.8493,
+      "step": 3832
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8206542351453856e-06,
+      "loss": 4.1706,
+      "step": 3833
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8206068268015173e-06,
+      "loss": 4.0106,
+      "step": 3834
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8205594184576486e-06,
+      "loss": 3.7954,
+      "step": 3835
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.82051201011378e-06,
+      "loss": 3.9368,
+      "step": 3836
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8204646017699116e-06,
+      "loss": 4.134,
+      "step": 3837
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.820417193426043e-06,
+      "loss": 3.976,
+      "step": 3838
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8203697850821746e-06,
+      "loss": 4.0672,
+      "step": 3839
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8203223767383063e-06,
+      "loss": 3.3968,
+      "step": 3840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8202749683944376e-06,
+      "loss": 3.6027,
+      "step": 3841
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.820227560050569e-06,
+      "loss": 3.7578,
+      "step": 3842
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8201801517067e-06,
+      "loss": 3.8396,
+      "step": 3843
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.820132743362832e-06,
+      "loss": 4.2794,
+      "step": 3844
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8200853350189635e-06,
+      "loss": 4.0612,
+      "step": 3845
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.820037926675095e-06,
+      "loss": 3.6949,
+      "step": 3846
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8199905183312265e-06,
+      "loss": 3.7631,
+      "step": 3847
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819943109987358e-06,
+      "loss": 4.2295,
+      "step": 3848
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819895701643489e-06,
+      "loss": 3.8403,
+      "step": 3849
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819848293299621e-06,
+      "loss": 3.5853,
+      "step": 3850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819800884955752e-06,
+      "loss": 4.2174,
+      "step": 3851
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8197534766118838e-06,
+      "loss": 3.6055,
+      "step": 3852
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8197060682680155e-06,
+      "loss": 3.7199,
+      "step": 3853
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8196586599241468e-06,
+      "loss": 3.9195,
+      "step": 3854
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819611251580278e-06,
+      "loss": 3.9695,
+      "step": 3855
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8195638432364097e-06,
+      "loss": 4.4103,
+      "step": 3856
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819516434892541e-06,
+      "loss": 4.1695,
+      "step": 3857
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8194690265486727e-06,
+      "loss": 4.038,
+      "step": 3858
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8194216182048044e-06,
+      "loss": 3.8075,
+      "step": 3859
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8193742098609357e-06,
+      "loss": 3.7085,
+      "step": 3860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819326801517067e-06,
+      "loss": 3.6408,
+      "step": 3861
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8192793931731987e-06,
+      "loss": 4.119,
+      "step": 3862
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.81923198482933e-06,
+      "loss": 4.1203,
+      "step": 3863
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8191845764854613e-06,
+      "loss": 3.7377,
+      "step": 3864
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819137168141593e-06,
+      "loss": 4.1225,
+      "step": 3865
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8190897597977247e-06,
+      "loss": 3.9104,
+      "step": 3866
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.819042351453856e-06,
+      "loss": 3.747,
+      "step": 3867
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8189949431099872e-06,
+      "loss": 4.3755,
+      "step": 3868
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818947534766119e-06,
+      "loss": 3.7521,
+      "step": 3869
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8189001264222502e-06,
+      "loss": 4.0499,
+      "step": 3870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818852718078382e-06,
+      "loss": 3.3514,
+      "step": 3871
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8188053097345136e-06,
+      "loss": 3.9498,
+      "step": 3872
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818757901390645e-06,
+      "loss": 3.9737,
+      "step": 3873
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818710493046776e-06,
+      "loss": 3.9898,
+      "step": 3874
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818663084702908e-06,
+      "loss": 3.7803,
+      "step": 3875
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818615676359039e-06,
+      "loss": 3.9394,
+      "step": 3876
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8185682680151704e-06,
+      "loss": 3.9928,
+      "step": 3877
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818520859671302e-06,
+      "loss": 4.0014,
+      "step": 3878
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818473451327434e-06,
+      "loss": 3.8423,
+      "step": 3879
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818426042983565e-06,
+      "loss": 4.1322,
+      "step": 3880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818378634639697e-06,
+      "loss": 3.4799,
+      "step": 3881
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818331226295828e-06,
+      "loss": 4.2279,
+      "step": 3882
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8182838179519594e-06,
+      "loss": 4.0478,
+      "step": 3883
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818236409608091e-06,
+      "loss": 3.9541,
+      "step": 3884
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818189001264223e-06,
+      "loss": 4.2333,
+      "step": 3885
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818141592920354e-06,
+      "loss": 3.9568,
+      "step": 3886
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8180941845764854e-06,
+      "loss": 4.0625,
+      "step": 3887
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.818046776232617e-06,
+      "loss": 3.8935,
+      "step": 3888
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8179993678887484e-06,
+      "loss": 3.6024,
+      "step": 3889
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8179519595448796e-06,
+      "loss": 3.9961,
+      "step": 3890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8179045512010113e-06,
+      "loss": 4.0742,
+      "step": 3891
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.817857142857143e-06,
+      "loss": 4.0218,
+      "step": 3892
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.8178097345132743e-06,
+      "loss": 3.937,
+      "step": 3893
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.817762326169406e-06,
+      "loss": 3.4956,
+      "step": 3894
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8177149178255373e-06,
+      "loss": 3.9546,
+      "step": 3895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8176675094816686e-06,
+      "loss": 3.9109,
+      "step": 3896
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8176201011378003e-06,
+      "loss": 4.1603,
+      "step": 3897
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.817572692793932e-06,
+      "loss": 3.5476,
+      "step": 3898
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8175252844500633e-06,
+      "loss": 3.9238,
+      "step": 3899
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.817477876106195e-06,
+      "loss": 4.2686,
+      "step": 3900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8174304677623263e-06,
+      "loss": 3.5842,
+      "step": 3901
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8173830594184575e-06,
+      "loss": 4.0605,
+      "step": 3902
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8173356510745892e-06,
+      "loss": 3.7026,
+      "step": 3903
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8172882427307205e-06,
+      "loss": 4.0904,
+      "step": 3904
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8172408343868522e-06,
+      "loss": 4.1766,
+      "step": 3905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.817193426042984e-06,
+      "loss": 4.0134,
+      "step": 3906
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8171460176991152e-06,
+      "loss": 3.6405,
+      "step": 3907
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8170986093552465e-06,
+      "loss": 3.9362,
+      "step": 3908
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8170512010113778e-06,
+      "loss": 3.8501,
+      "step": 3909
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8170037926675095e-06,
+      "loss": 3.4747,
+      "step": 3910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.816956384323641e-06,
+      "loss": 4.1305,
+      "step": 3911
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8169089759797725e-06,
+      "loss": 3.7824,
+      "step": 3912
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.816861567635904e-06,
+      "loss": 3.9252,
+      "step": 3913
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8168141592920354e-06,
+      "loss": 3.6506,
+      "step": 3914
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8167667509481667e-06,
+      "loss": 4.0923,
+      "step": 3915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8167193426042984e-06,
+      "loss": 4.2182,
+      "step": 3916
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8166719342604297e-06,
+      "loss": 3.9761,
+      "step": 3917
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8166245259165614e-06,
+      "loss": 3.5534,
+      "step": 3918
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.816577117572693e-06,
+      "loss": 3.8535,
+      "step": 3919
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8165297092288244e-06,
+      "loss": 3.8015,
+      "step": 3920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8164823008849557e-06,
+      "loss": 4.2051,
+      "step": 3921
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8164348925410874e-06,
+      "loss": 3.6977,
+      "step": 3922
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8163874841972187e-06,
+      "loss": 4.1563,
+      "step": 3923
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8163400758533504e-06,
+      "loss": 3.9906,
+      "step": 3924
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.816292667509482e-06,
+      "loss": 3.955,
+      "step": 3925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8162452591656134e-06,
+      "loss": 4.1209,
+      "step": 3926
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8161978508217446e-06,
+      "loss": 3.6741,
+      "step": 3927
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8161504424778763e-06,
+      "loss": 4.1443,
+      "step": 3928
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8161030341340076e-06,
+      "loss": 4.0403,
+      "step": 3929
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.816055625790139e-06,
+      "loss": 4.0426,
+      "step": 3930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8160082174462706e-06,
+      "loss": 3.8643,
+      "step": 3931
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8159608091024023e-06,
+      "loss": 3.8282,
+      "step": 3932
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8159134007585336e-06,
+      "loss": 3.7742,
+      "step": 3933
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.815865992414665e-06,
+      "loss": 4.0285,
+      "step": 3934
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8158185840707966e-06,
+      "loss": 3.7678,
+      "step": 3935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.815771175726928e-06,
+      "loss": 3.8219,
+      "step": 3936
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8157237673830596e-06,
+      "loss": 3.899,
+      "step": 3937
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8156763590391913e-06,
+      "loss": 4.3162,
+      "step": 3938
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8156289506953225e-06,
+      "loss": 4.0597,
+      "step": 3939
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.815581542351454e-06,
+      "loss": 3.7873,
+      "step": 3940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8155341340075855e-06,
+      "loss": 4.2831,
+      "step": 3941
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.815486725663717e-06,
+      "loss": 3.7975,
+      "step": 3942
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.815439317319848e-06,
+      "loss": 4.0296,
+      "step": 3943
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.81539190897598e-06,
+      "loss": 3.7706,
+      "step": 3944
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8153445006321115e-06,
+      "loss": 3.7667,
+      "step": 3945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8152970922882428e-06,
+      "loss": 3.8698,
+      "step": 3946
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8152496839443745e-06,
+      "loss": 3.7661,
+      "step": 3947
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8152022756005058e-06,
+      "loss": 3.7947,
+      "step": 3948
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.815154867256637e-06,
+      "loss": 3.9362,
+      "step": 3949
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8151074589127687e-06,
+      "loss": 4.1173,
+      "step": 3950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8150600505689004e-06,
+      "loss": 3.7898,
+      "step": 3951
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8150126422250317e-06,
+      "loss": 3.9204,
+      "step": 3952
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.814965233881163e-06,
+      "loss": 4.0478,
+      "step": 3953
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8149178255372947e-06,
+      "loss": 4.3033,
+      "step": 3954
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.814870417193426e-06,
+      "loss": 4.222,
+      "step": 3955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8148230088495573e-06,
+      "loss": 4.1841,
+      "step": 3956
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.814775600505689e-06,
+      "loss": 3.7731,
+      "step": 3957
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.8147281921618207e-06,
+      "loss": 3.5855,
+      "step": 3958
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814680783817952e-06,
+      "loss": 4.007,
+      "step": 3959
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8146333754740837e-06,
+      "loss": 3.6787,
+      "step": 3960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814585967130215e-06,
+      "loss": 4.0391,
+      "step": 3961
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8145385587863462e-06,
+      "loss": 3.841,
+      "step": 3962
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814491150442478e-06,
+      "loss": 4.1091,
+      "step": 3963
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8144437420986096e-06,
+      "loss": 3.8848,
+      "step": 3964
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814396333754741e-06,
+      "loss": 3.8279,
+      "step": 3965
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8143489254108726e-06,
+      "loss": 4.2027,
+      "step": 3966
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814301517067004e-06,
+      "loss": 4.0071,
+      "step": 3967
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814254108723135e-06,
+      "loss": 4.2117,
+      "step": 3968
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814206700379267e-06,
+      "loss": 3.9232,
+      "step": 3969
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814159292035398e-06,
+      "loss": 4.0293,
+      "step": 3970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.81411188369153e-06,
+      "loss": 4.2756,
+      "step": 3971
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8140644753476616e-06,
+      "loss": 4.1518,
+      "step": 3972
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.814017067003793e-06,
+      "loss": 3.6226,
+      "step": 3973
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813969658659924e-06,
+      "loss": 3.6505,
+      "step": 3974
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8139222503160554e-06,
+      "loss": 4.2813,
+      "step": 3975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813874841972187e-06,
+      "loss": 3.8919,
+      "step": 3976
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813827433628319e-06,
+      "loss": 4.0338,
+      "step": 3977
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.81378002528445e-06,
+      "loss": 4.0676,
+      "step": 3978
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813732616940582e-06,
+      "loss": 4.0621,
+      "step": 3979
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813685208596713e-06,
+      "loss": 4.1699,
+      "step": 3980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8136378002528444e-06,
+      "loss": 4.0833,
+      "step": 3981
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813590391908976e-06,
+      "loss": 3.7441,
+      "step": 3982
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8135429835651073e-06,
+      "loss": 4.0575,
+      "step": 3983
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813495575221239e-06,
+      "loss": 3.8104,
+      "step": 3984
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8134481668773708e-06,
+      "loss": 4.3451,
+      "step": 3985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813400758533502e-06,
+      "loss": 4.1524,
+      "step": 3986
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8133533501896333e-06,
+      "loss": 3.9897,
+      "step": 3987
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813305941845765e-06,
+      "loss": 4.0988,
+      "step": 3988
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8132585335018963e-06,
+      "loss": 3.7111,
+      "step": 3989
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813211125158028e-06,
+      "loss": 3.8676,
+      "step": 3990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8131637168141597e-06,
+      "loss": 3.867,
+      "step": 3991
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813116308470291e-06,
+      "loss": 3.7967,
+      "step": 3992
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8130689001264223e-06,
+      "loss": 4.206,
+      "step": 3993
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.813021491782554e-06,
+      "loss": 3.9656,
+      "step": 3994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8129740834386853e-06,
+      "loss": 4.0943,
+      "step": 3995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8129266750948165e-06,
+      "loss": 4.028,
+      "step": 3996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8128792667509482e-06,
+      "loss": 3.8793,
+      "step": 3997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.81283185840708e-06,
+      "loss": 3.8931,
+      "step": 3998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8127844500632112e-06,
+      "loss": 4.102,
+      "step": 3999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8127370417193425e-06,
+      "loss": 3.6335,
+      "step": 4000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.812689633375474e-06,
+      "loss": 3.8111,
+      "step": 4001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8126422250316055e-06,
+      "loss": 4.0268,
+      "step": 4002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.812594816687737e-06,
+      "loss": 3.797,
+      "step": 4003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.812547408343869e-06,
+      "loss": 4.0242,
+      "step": 4004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8125e-06,
+      "loss": 3.6653,
+      "step": 4005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8124525916561315e-06,
+      "loss": 4.1457,
+      "step": 4006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.812405183312263e-06,
+      "loss": 4.093,
+      "step": 4007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8123577749683944e-06,
+      "loss": 3.8305,
+      "step": 4008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8123103666245257e-06,
+      "loss": 4.1358,
+      "step": 4009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8122629582806574e-06,
+      "loss": 4.0611,
+      "step": 4010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.812215549936789e-06,
+      "loss": 3.9341,
+      "step": 4011
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8121681415929204e-06,
+      "loss": 4.1354,
+      "step": 4012
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.812120733249052e-06,
+      "loss": 4.0223,
+      "step": 4013
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8120733249051834e-06,
+      "loss": 3.9891,
+      "step": 4014
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8120259165613147e-06,
+      "loss": 3.9352,
+      "step": 4015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8119785082174464e-06,
+      "loss": 3.8194,
+      "step": 4016
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.811931099873578e-06,
+      "loss": 3.7627,
+      "step": 4017
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8118836915297094e-06,
+      "loss": 4.1192,
+      "step": 4018
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8118362831858406e-06,
+      "loss": 4.0113,
+      "step": 4019
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8117888748419723e-06,
+      "loss": 3.8016,
+      "step": 4020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.8117414664981036e-06,
+      "loss": 3.8811,
+      "step": 4021
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.811694058154235e-06,
+      "loss": 3.8393,
+      "step": 4022
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8116466498103666e-06,
+      "loss": 3.9708,
+      "step": 4023
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8115992414664983e-06,
+      "loss": 4.1059,
+      "step": 4024
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8115518331226296e-06,
+      "loss": 4.2157,
+      "step": 4025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8115044247787613e-06,
+      "loss": 3.7794,
+      "step": 4026
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8114570164348926e-06,
+      "loss": 3.5894,
+      "step": 4027
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.811409608091024e-06,
+      "loss": 3.6493,
+      "step": 4028
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8113621997471556e-06,
+      "loss": 3.6366,
+      "step": 4029
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8113147914032873e-06,
+      "loss": 4.1229,
+      "step": 4030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8112673830594186e-06,
+      "loss": 3.8133,
+      "step": 4031
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8112199747155503e-06,
+      "loss": 4.1094,
+      "step": 4032
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8111725663716815e-06,
+      "loss": 4.0941,
+      "step": 4033
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.811125158027813e-06,
+      "loss": 3.9622,
+      "step": 4034
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8110777496839445e-06,
+      "loss": 4.0502,
+      "step": 4035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.811030341340076e-06,
+      "loss": 3.8812,
+      "step": 4036
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8109829329962075e-06,
+      "loss": 4.092,
+      "step": 4037
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8109355246523388e-06,
+      "loss": 4.3447,
+      "step": 4038
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8108881163084705e-06,
+      "loss": 4.0535,
+      "step": 4039
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8108407079646018e-06,
+      "loss": 3.7133,
+      "step": 4040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.810793299620733e-06,
+      "loss": 3.671,
+      "step": 4041
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8107458912768648e-06,
+      "loss": 3.9409,
+      "step": 4042
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8106984829329965e-06,
+      "loss": 4.2501,
+      "step": 4043
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8106510745891277e-06,
+      "loss": 3.9494,
+      "step": 4044
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8106036662452594e-06,
+      "loss": 3.9136,
+      "step": 4045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8105562579013907e-06,
+      "loss": 3.6726,
+      "step": 4046
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.810508849557522e-06,
+      "loss": 3.5,
+      "step": 4047
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8104614412136537e-06,
+      "loss": 3.822,
+      "step": 4048
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.810414032869785e-06,
+      "loss": 3.9503,
+      "step": 4049
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8103666245259167e-06,
+      "loss": 4.0509,
+      "step": 4050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8103192161820484e-06,
+      "loss": 4.3353,
+      "step": 4051
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8102718078381797e-06,
+      "loss": 4.0584,
+      "step": 4052
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.810224399494311e-06,
+      "loss": 3.8591,
+      "step": 4053
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8101769911504427e-06,
+      "loss": 4.0055,
+      "step": 4054
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.810129582806574e-06,
+      "loss": 3.7686,
+      "step": 4055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8100821744627056e-06,
+      "loss": 3.7841,
+      "step": 4056
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8100347661188373e-06,
+      "loss": 3.8745,
+      "step": 4057
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8099873577749686e-06,
+      "loss": 4.2718,
+      "step": 4058
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8099399494311e-06,
+      "loss": 3.8207,
+      "step": 4059
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809892541087231e-06,
+      "loss": 3.8572,
+      "step": 4060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809845132743363e-06,
+      "loss": 4.0627,
+      "step": 4061
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809797724399494e-06,
+      "loss": 3.9515,
+      "step": 4062
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809750316055626e-06,
+      "loss": 3.8527,
+      "step": 4063
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8097029077117576e-06,
+      "loss": 4.1182,
+      "step": 4064
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809655499367889e-06,
+      "loss": 3.7921,
+      "step": 4065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.80960809102402e-06,
+      "loss": 4.0975,
+      "step": 4066
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809560682680152e-06,
+      "loss": 3.4197,
+      "step": 4067
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809513274336283e-06,
+      "loss": 4.0099,
+      "step": 4068
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8094658659924144e-06,
+      "loss": 4.0312,
+      "step": 4069
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8094184576485465e-06,
+      "loss": 3.7695,
+      "step": 4070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809371049304678e-06,
+      "loss": 4.3457,
+      "step": 4071
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809323640960809e-06,
+      "loss": 3.7917,
+      "step": 4072
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809276232616941e-06,
+      "loss": 3.8306,
+      "step": 4073
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809228824273072e-06,
+      "loss": 4.0736,
+      "step": 4074
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8091814159292034e-06,
+      "loss": 3.9,
+      "step": 4075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809134007585335e-06,
+      "loss": 3.5741,
+      "step": 4076
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8090865992414668e-06,
+      "loss": 3.8937,
+      "step": 4077
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809039190897598e-06,
+      "loss": 4.1666,
+      "step": 4078
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8089917825537298e-06,
+      "loss": 3.8848,
+      "step": 4079
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.808944374209861e-06,
+      "loss": 4.0339,
+      "step": 4080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8088969658659923e-06,
+      "loss": 3.6649,
+      "step": 4081
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8088495575221236e-06,
+      "loss": 4.3592,
+      "step": 4082
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8088021491782553e-06,
+      "loss": 3.9603,
+      "step": 4083
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.808754740834387e-06,
+      "loss": 4.0403,
+      "step": 4084
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8087073324905183e-06,
+      "loss": 4.0693,
+      "step": 4085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.80865992414665e-06,
+      "loss": 4.0279,
+      "step": 4086
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8086125158027813e-06,
+      "loss": 3.8784,
+      "step": 4087
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8085651074589125e-06,
+      "loss": 3.8397,
+      "step": 4088
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8085176991150442e-06,
+      "loss": 4.1212,
+      "step": 4089
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.808470290771176e-06,
+      "loss": 4.0278,
+      "step": 4090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8084228824273072e-06,
+      "loss": 4.3642,
+      "step": 4091
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.808375474083439e-06,
+      "loss": 3.3956,
+      "step": 4092
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8083280657395702e-06,
+      "loss": 4.036,
+      "step": 4093
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8082806573957015e-06,
+      "loss": 4.0967,
+      "step": 4094
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.808233249051833e-06,
+      "loss": 3.9534,
+      "step": 4095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8081858407079645e-06,
+      "loss": 3.8088,
+      "step": 4096
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.808138432364096e-06,
+      "loss": 3.9702,
+      "step": 4097
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.808091024020228e-06,
+      "loss": 3.7149,
+      "step": 4098
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.808043615676359e-06,
+      "loss": 4.0989,
+      "step": 4099
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8079962073324905e-06,
+      "loss": 3.8671,
+      "step": 4100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.807948798988622e-06,
+      "loss": 4.1493,
+      "step": 4101
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8079013906447534e-06,
+      "loss": 3.9213,
+      "step": 4102
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.807853982300885e-06,
+      "loss": 4.1783,
+      "step": 4103
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8078065739570164e-06,
+      "loss": 3.9682,
+      "step": 4104
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.807759165613148e-06,
+      "loss": 3.7379,
+      "step": 4105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8077117572692794e-06,
+      "loss": 4.0375,
+      "step": 4106
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8076643489254107e-06,
+      "loss": 4.0119,
+      "step": 4107
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8076169405815424e-06,
+      "loss": 3.7611,
+      "step": 4108
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8075695322376737e-06,
+      "loss": 4.4551,
+      "step": 4109
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8075221238938054e-06,
+      "loss": 4.0122,
+      "step": 4110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.807474715549937e-06,
+      "loss": 3.9625,
+      "step": 4111
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8074273072060684e-06,
+      "loss": 4.1001,
+      "step": 4112
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8073798988621996e-06,
+      "loss": 4.0693,
+      "step": 4113
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8073324905183313e-06,
+      "loss": 4.2402,
+      "step": 4114
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8072850821744626e-06,
+      "loss": 3.8598,
+      "step": 4115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8072376738305943e-06,
+      "loss": 4.4321,
+      "step": 4116
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.807190265486726e-06,
+      "loss": 4.0096,
+      "step": 4117
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8071428571428573e-06,
+      "loss": 4.0579,
+      "step": 4118
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8070954487989886e-06,
+      "loss": 4.0395,
+      "step": 4119
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8070480404551203e-06,
+      "loss": 4.0402,
+      "step": 4120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8070006321112516e-06,
+      "loss": 3.9292,
+      "step": 4121
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.806953223767383e-06,
+      "loss": 3.8997,
+      "step": 4122
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8069058154235146e-06,
+      "loss": 3.5618,
+      "step": 4123
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8068584070796463e-06,
+      "loss": 4.2056,
+      "step": 4124
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8068109987357775e-06,
+      "loss": 3.8079,
+      "step": 4125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.806763590391909e-06,
+      "loss": 4.0937,
+      "step": 4126
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8067161820480405e-06,
+      "loss": 4.1455,
+      "step": 4127
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.806668773704172e-06,
+      "loss": 4.2783,
+      "step": 4128
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8066213653603035e-06,
+      "loss": 3.6809,
+      "step": 4129
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8065739570164352e-06,
+      "loss": 3.6947,
+      "step": 4130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8065265486725665e-06,
+      "loss": 3.9876,
+      "step": 4131
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8064791403286978e-06,
+      "loss": 4.2558,
+      "step": 4132
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8064317319848295e-06,
+      "loss": 4.1948,
+      "step": 4133
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8063843236409608e-06,
+      "loss": 3.6077,
+      "step": 4134
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.806336915297092e-06,
+      "loss": 3.8305,
+      "step": 4135
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8062895069532237e-06,
+      "loss": 3.8331,
+      "step": 4136
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8062420986093555e-06,
+      "loss": 3.9289,
+      "step": 4137
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8061946902654867e-06,
+      "loss": 3.8458,
+      "step": 4138
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8061472819216184e-06,
+      "loss": 3.9116,
+      "step": 4139
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8060998735777497e-06,
+      "loss": 3.8362,
+      "step": 4140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.806052465233881e-06,
+      "loss": 4.1118,
+      "step": 4141
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8060050568900127e-06,
+      "loss": 4.1926,
+      "step": 4142
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8059576485461444e-06,
+      "loss": 3.7281,
+      "step": 4143
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8059102402022757e-06,
+      "loss": 3.9222,
+      "step": 4144
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8058628318584074e-06,
+      "loss": 4.2545,
+      "step": 4145
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8058154235145387e-06,
+      "loss": 4.1374,
+      "step": 4146
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.80576801517067e-06,
+      "loss": 3.6887,
+      "step": 4147
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8057206068268012e-06,
+      "loss": 3.795,
+      "step": 4148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805673198482933e-06,
+      "loss": 4.3316,
+      "step": 4149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8056257901390646e-06,
+      "loss": 3.9686,
+      "step": 4150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805578381795196e-06,
+      "loss": 4.0722,
+      "step": 4151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8055309734513276e-06,
+      "loss": 3.871,
+      "step": 4152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805483565107459e-06,
+      "loss": 3.6737,
+      "step": 4153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80543615676359e-06,
+      "loss": 4.1287,
+      "step": 4154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805388748419722e-06,
+      "loss": 4.2246,
+      "step": 4155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8053413400758536e-06,
+      "loss": 3.5229,
+      "step": 4156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805293931731985e-06,
+      "loss": 3.6227,
+      "step": 4157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8052465233881166e-06,
+      "loss": 4.4418,
+      "step": 4158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805199115044248e-06,
+      "loss": 3.6083,
+      "step": 4159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805151706700379e-06,
+      "loss": 3.8651,
+      "step": 4160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805104298356511e-06,
+      "loss": 3.5717,
+      "step": 4161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805056890012642e-06,
+      "loss": 3.9257,
+      "step": 4162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.805009481668774e-06,
+      "loss": 3.7993,
+      "step": 4163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8049620733249055e-06,
+      "loss": 3.974,
+      "step": 4164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804914664981037e-06,
+      "loss": 4.3142,
+      "step": 4165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804867256637168e-06,
+      "loss": 3.8319,
+      "step": 4166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8048198482933e-06,
+      "loss": 4.2638,
+      "step": 4167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804772439949431e-06,
+      "loss": 4.0199,
+      "step": 4168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8047250316055628e-06,
+      "loss": 3.6006,
+      "step": 4169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804677623261694e-06,
+      "loss": 4.1126,
+      "step": 4170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8046302149178258e-06,
+      "loss": 3.9863,
+      "step": 4171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804582806573957e-06,
+      "loss": 3.6507,
+      "step": 4172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8045353982300883e-06,
+      "loss": 3.905,
+      "step": 4173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.80448798988622e-06,
+      "loss": 4.0945,
+      "step": 4174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8044405815423513e-06,
+      "loss": 4.1273,
+      "step": 4175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804393173198483e-06,
+      "loss": 4.0201,
+      "step": 4176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8043457648546147e-06,
+      "loss": 3.9218,
+      "step": 4177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804298356510746e-06,
+      "loss": 3.6996,
+      "step": 4178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8042509481668773e-06,
+      "loss": 3.8744,
+      "step": 4179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804203539823009e-06,
+      "loss": 3.8507,
+      "step": 4180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8041561314791403e-06,
+      "loss": 3.8648,
+      "step": 4181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804108723135272e-06,
+      "loss": 3.8593,
+      "step": 4182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8040613147914037e-06,
+      "loss": 4.0048,
+      "step": 4183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.804013906447535e-06,
+      "loss": 3.9312,
+      "step": 4184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8039664981036662e-06,
+      "loss": 4.2459,
+      "step": 4185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803919089759798e-06,
+      "loss": 4.035,
+      "step": 4186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8038716814159292e-06,
+      "loss": 3.8811,
+      "step": 4187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8038242730720605e-06,
+      "loss": 4.0791,
+      "step": 4188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803776864728192e-06,
+      "loss": 3.6723,
+      "step": 4189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803729456384324e-06,
+      "loss": 3.8903,
+      "step": 4190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803682048040455e-06,
+      "loss": 4.1019,
+      "step": 4191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8036346396965865e-06,
+      "loss": 3.9571,
+      "step": 4192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803587231352718e-06,
+      "loss": 3.5625,
+      "step": 4193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8035398230088494e-06,
+      "loss": 3.8473,
+      "step": 4194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803492414664981e-06,
+      "loss": 3.6489,
+      "step": 4195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803445006321113e-06,
+      "loss": 3.9288,
+      "step": 4196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803397597977244e-06,
+      "loss": 3.7598,
+      "step": 4197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8033501896333754e-06,
+      "loss": 3.9981,
+      "step": 4198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803302781289507e-06,
+      "loss": 4.1874,
+      "step": 4199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8032553729456384e-06,
+      "loss": 4.0401,
+      "step": 4200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8032079646017697e-06,
+      "loss": 3.8777,
+      "step": 4201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8031605562579014e-06,
+      "loss": 3.8997,
+      "step": 4202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803113147914033e-06,
+      "loss": 4.1734,
+      "step": 4203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8030657395701644e-06,
+      "loss": 4.2208,
+      "step": 4204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.803018331226296e-06,
+      "loss": 3.727,
+      "step": 4205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8029709228824274e-06,
+      "loss": 4.0348,
+      "step": 4206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8029235145385586e-06,
+      "loss": 4.0629,
+      "step": 4207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8028761061946903e-06,
+      "loss": 3.9055,
+      "step": 4208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.802828697850822e-06,
+      "loss": 4.2173,
+      "step": 4209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8027812895069533e-06,
+      "loss": 4.1684,
+      "step": 4210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.802733881163085e-06,
+      "loss": 3.9981,
+      "step": 4211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8026864728192163e-06,
+      "loss": 4.2789,
+      "step": 4212
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8026390644753476e-06,
+      "loss": 3.6474,
+      "step": 4213
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.802591656131479e-06,
+      "loss": 4.0615,
+      "step": 4214
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8025442477876106e-06,
+      "loss": 4.2196,
+      "step": 4215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8024968394437423e-06,
+      "loss": 3.9817,
+      "step": 4216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8024494310998736e-06,
+      "loss": 4.0946,
+      "step": 4217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8024020227560053e-06,
+      "loss": 3.4934,
+      "step": 4218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8023546144121365e-06,
+      "loss": 3.7399,
+      "step": 4219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.802307206068268e-06,
+      "loss": 4.2989,
+      "step": 4220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8022597977243995e-06,
+      "loss": 4.0745,
+      "step": 4221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8022123893805312e-06,
+      "loss": 3.6997,
+      "step": 4222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8021649810366625e-06,
+      "loss": 3.8458,
+      "step": 4223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.802117572692794e-06,
+      "loss": 3.6531,
+      "step": 4224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8020701643489255e-06,
+      "loss": 4.0549,
+      "step": 4225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8020227560050568e-06,
+      "loss": 3.7444,
+      "step": 4226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8019753476611885e-06,
+      "loss": 3.7789,
+      "step": 4227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8019279393173198e-06,
+      "loss": 3.8701,
+      "step": 4228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8018805309734515e-06,
+      "loss": 4.0069,
+      "step": 4229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.801833122629583e-06,
+      "loss": 3.9311,
+      "step": 4230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8017857142857144e-06,
+      "loss": 3.9973,
+      "step": 4231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8017383059418457e-06,
+      "loss": 4.1574,
+      "step": 4232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8016908975979774e-06,
+      "loss": 4.0413,
+      "step": 4233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8016434892541087e-06,
+      "loss": 4.2367,
+      "step": 4234
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8015960809102404e-06,
+      "loss": 3.861,
+      "step": 4235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8015486725663717e-06,
+      "loss": 3.8307,
+      "step": 4236
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8015012642225034e-06,
+      "loss": 4.1215,
+      "step": 4237
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8014538558786347e-06,
+      "loss": 4.1297,
+      "step": 4238
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.801406447534766e-06,
+      "loss": 3.8956,
+      "step": 4239
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8013590391908977e-06,
+      "loss": 4.0798,
+      "step": 4240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.801311630847029e-06,
+      "loss": 4.0188,
+      "step": 4241
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8012642225031606e-06,
+      "loss": 4.0186,
+      "step": 4242
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8012168141592924e-06,
+      "loss": 3.7286,
+      "step": 4243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8011694058154236e-06,
+      "loss": 3.8604,
+      "step": 4244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.801121997471555e-06,
+      "loss": 3.8326,
+      "step": 4245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8010745891276866e-06,
+      "loss": 3.9179,
+      "step": 4246
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.801027180783818e-06,
+      "loss": 3.8643,
+      "step": 4247
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8009797724399496e-06,
+      "loss": 3.8183,
+      "step": 4248
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8009323640960813e-06,
+      "loss": 3.8847,
+      "step": 4249
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8008849557522126e-06,
+      "loss": 3.4738,
+      "step": 4250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800837547408344e-06,
+      "loss": 4.087,
+      "step": 4251
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8007901390644756e-06,
+      "loss": 3.8782,
+      "step": 4252
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800742730720607e-06,
+      "loss": 4.0497,
+      "step": 4253
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800695322376738e-06,
+      "loss": 3.9399,
+      "step": 4254
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.80064791403287e-06,
+      "loss": 4.0437,
+      "step": 4255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8006005056890015e-06,
+      "loss": 3.6529,
+      "step": 4256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800553097345133e-06,
+      "loss": 3.9993,
+      "step": 4257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800505689001264e-06,
+      "loss": 3.9424,
+      "step": 4258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800458280657396e-06,
+      "loss": 4.097,
+      "step": 4259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800410872313527e-06,
+      "loss": 4.1474,
+      "step": 4260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8003634639696588e-06,
+      "loss": 4.1329,
+      "step": 4261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8003160556257905e-06,
+      "loss": 3.7241,
+      "step": 4262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8002686472819218e-06,
+      "loss": 3.7249,
+      "step": 4263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800221238938053e-06,
+      "loss": 4.1973,
+      "step": 4264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8001738305941848e-06,
+      "loss": 3.7074,
+      "step": 4265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800126422250316e-06,
+      "loss": 3.9826,
+      "step": 4266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8000790139064473e-06,
+      "loss": 3.9869,
+      "step": 4267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.800031605562579e-06,
+      "loss": 4.1752,
+      "step": 4268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7999841972187107e-06,
+      "loss": 3.8314,
+      "step": 4269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.799936788874842e-06,
+      "loss": 3.916,
+      "step": 4270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7998893805309737e-06,
+      "loss": 4.1278,
+      "step": 4271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.799841972187105e-06,
+      "loss": 3.9577,
+      "step": 4272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7997945638432363e-06,
+      "loss": 3.4298,
+      "step": 4273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.799747155499368e-06,
+      "loss": 4.0681,
+      "step": 4274
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7996997471554997e-06,
+      "loss": 4.1514,
+      "step": 4275
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799652338811631e-06,
+      "loss": 3.6716,
+      "step": 4276
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7996049304677627e-06,
+      "loss": 4.0083,
+      "step": 4277
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799557522123894e-06,
+      "loss": 3.7317,
+      "step": 4278
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7995101137800252e-06,
+      "loss": 3.9689,
+      "step": 4279
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7994627054361565e-06,
+      "loss": 3.7963,
+      "step": 4280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799415297092288e-06,
+      "loss": 3.9829,
+      "step": 4281
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.79936788874842e-06,
+      "loss": 3.6015,
+      "step": 4282
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799320480404551e-06,
+      "loss": 3.309,
+      "step": 4283
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799273072060683e-06,
+      "loss": 3.8853,
+      "step": 4284
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799225663716814e-06,
+      "loss": 3.9955,
+      "step": 4285
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7991782553729455e-06,
+      "loss": 3.9386,
+      "step": 4286
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799130847029077e-06,
+      "loss": 3.9318,
+      "step": 4287
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.799083438685209e-06,
+      "loss": 3.8915,
+      "step": 4288
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.79903603034134e-06,
+      "loss": 4.0946,
+      "step": 4289
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798988621997472e-06,
+      "loss": 4.0497,
+      "step": 4290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798941213653603e-06,
+      "loss": 4.2324,
+      "step": 4291
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7988938053097344e-06,
+      "loss": 3.9838,
+      "step": 4292
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798846396965866e-06,
+      "loss": 3.8346,
+      "step": 4293
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7987989886219974e-06,
+      "loss": 3.7906,
+      "step": 4294
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798751580278129e-06,
+      "loss": 4.1787,
+      "step": 4295
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798704171934261e-06,
+      "loss": 3.992,
+      "step": 4296
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798656763590392e-06,
+      "loss": 4.2111,
+      "step": 4297
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7986093552465234e-06,
+      "loss": 4.179,
+      "step": 4298
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798561946902655e-06,
+      "loss": 4.0349,
+      "step": 4299
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7985145385587863e-06,
+      "loss": 3.5136,
+      "step": 4300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798467130214918e-06,
+      "loss": 4.0145,
+      "step": 4301
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7984197218710493e-06,
+      "loss": 3.8047,
+      "step": 4302
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.798372313527181e-06,
+      "loss": 4.1379,
+      "step": 4303
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7983249051833123e-06,
+      "loss": 3.9918,
+      "step": 4304
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7982774968394436e-06,
+      "loss": 3.5768,
+      "step": 4305
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7982300884955753e-06,
+      "loss": 3.9621,
+      "step": 4306
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7981826801517066e-06,
+      "loss": 3.6976,
+      "step": 4307
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7981352718078383e-06,
+      "loss": 4.0288,
+      "step": 4308
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.79808786346397e-06,
+      "loss": 3.6307,
+      "step": 4309
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7980404551201013e-06,
+      "loss": 3.6401,
+      "step": 4310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7979930467762325e-06,
+      "loss": 3.9226,
+      "step": 4311
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7979456384323643e-06,
+      "loss": 4.0012,
+      "step": 4312
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7978982300884955e-06,
+      "loss": 4.1611,
+      "step": 4313
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7978508217446272e-06,
+      "loss": 3.5858,
+      "step": 4314
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.797803413400759e-06,
+      "loss": 4.1069,
+      "step": 4315
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7977560050568902e-06,
+      "loss": 3.6189,
+      "step": 4316
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7977085967130215e-06,
+      "loss": 3.8096,
+      "step": 4317
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.797661188369153e-06,
+      "loss": 3.9888,
+      "step": 4318
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7976137800252845e-06,
+      "loss": 3.8878,
+      "step": 4319
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7975663716814158e-06,
+      "loss": 3.879,
+      "step": 4320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7975189633375475e-06,
+      "loss": 3.698,
+      "step": 4321
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.797471554993679e-06,
+      "loss": 3.7343,
+      "step": 4322
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7974241466498105e-06,
+      "loss": 4.1292,
+      "step": 4323
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7973767383059417e-06,
+      "loss": 3.7045,
+      "step": 4324
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7973293299620734e-06,
+      "loss": 4.0849,
+      "step": 4325
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7972819216182047e-06,
+      "loss": 4.0249,
+      "step": 4326
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7972345132743364e-06,
+      "loss": 4.3227,
+      "step": 4327
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.797187104930468e-06,
+      "loss": 4.3479,
+      "step": 4328
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7971396965865994e-06,
+      "loss": 4.1077,
+      "step": 4329
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7970922882427307e-06,
+      "loss": 3.8101,
+      "step": 4330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7970448798988624e-06,
+      "loss": 4.1556,
+      "step": 4331
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7969974715549937e-06,
+      "loss": 3.8404,
+      "step": 4332
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.796950063211125e-06,
+      "loss": 4.0695,
+      "step": 4333
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7969026548672567e-06,
+      "loss": 3.8629,
+      "step": 4334
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7968552465233884e-06,
+      "loss": 4.1508,
+      "step": 4335
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7968078381795196e-06,
+      "loss": 3.8772,
+      "step": 4336
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7967604298356513e-06,
+      "loss": 4.3285,
+      "step": 4337
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7967130214917826e-06,
+      "loss": 3.8698,
+      "step": 4338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796665613147914e-06,
+      "loss": 4.2908,
+      "step": 4339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7966182048040456e-06,
+      "loss": 3.8441,
+      "step": 4340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7965707964601773e-06,
+      "loss": 4.2006,
+      "step": 4341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7965233881163086e-06,
+      "loss": 3.6485,
+      "step": 4342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7964759797724403e-06,
+      "loss": 3.358,
+      "step": 4343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7964285714285716e-06,
+      "loss": 3.8518,
+      "step": 4344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796381163084703e-06,
+      "loss": 3.9492,
+      "step": 4345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796333754740834e-06,
+      "loss": 3.875,
+      "step": 4346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796286346396966e-06,
+      "loss": 3.934,
+      "step": 4347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7962389380530975e-06,
+      "loss": 4.0563,
+      "step": 4348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796191529709229e-06,
+      "loss": 3.6998,
+      "step": 4349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7961441213653605e-06,
+      "loss": 4.3159,
+      "step": 4350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796096713021492e-06,
+      "loss": 4.0483,
+      "step": 4351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796049304677623e-06,
+      "loss": 4.0268,
+      "step": 4352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.796001896333755e-06,
+      "loss": 4.0168,
+      "step": 4353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795954487989886e-06,
+      "loss": 3.9978,
+      "step": 4354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7959070796460178e-06,
+      "loss": 3.8197,
+      "step": 4355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7958596713021495e-06,
+      "loss": 4.1228,
+      "step": 4356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7958122629582808e-06,
+      "loss": 4.0641,
+      "step": 4357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795764854614412e-06,
+      "loss": 3.8415,
+      "step": 4358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7957174462705437e-06,
+      "loss": 4.0513,
+      "step": 4359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795670037926675e-06,
+      "loss": 3.4345,
+      "step": 4360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7956226295828067e-06,
+      "loss": 4.0902,
+      "step": 4361
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7955752212389384e-06,
+      "loss": 3.5723,
+      "step": 4362
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7955278128950697e-06,
+      "loss": 3.6651,
+      "step": 4363
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795480404551201e-06,
+      "loss": 4.0139,
+      "step": 4364
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7954329962073327e-06,
+      "loss": 4.2166,
+      "step": 4365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795385587863464e-06,
+      "loss": 3.829,
+      "step": 4366
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7953381795195953e-06,
+      "loss": 3.9882,
+      "step": 4367
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795290771175727e-06,
+      "loss": 3.9239,
+      "step": 4368
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7952433628318587e-06,
+      "loss": 3.9769,
+      "step": 4369
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.79519595448799e-06,
+      "loss": 4.0225,
+      "step": 4370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7951485461441212e-06,
+      "loss": 3.8956,
+      "step": 4371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795101137800253e-06,
+      "loss": 3.7391,
+      "step": 4372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7950537294563842e-06,
+      "loss": 3.8193,
+      "step": 4373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.795006321112516e-06,
+      "loss": 3.7945,
+      "step": 4374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7949589127686476e-06,
+      "loss": 3.9585,
+      "step": 4375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794911504424779e-06,
+      "loss": 4.1902,
+      "step": 4376
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.79486409608091e-06,
+      "loss": 3.8385,
+      "step": 4377
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794816687737042e-06,
+      "loss": 4.0168,
+      "step": 4378
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794769279393173e-06,
+      "loss": 4.2717,
+      "step": 4379
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7947218710493044e-06,
+      "loss": 4.0373,
+      "step": 4380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794674462705436e-06,
+      "loss": 3.5867,
+      "step": 4381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794627054361568e-06,
+      "loss": 3.7412,
+      "step": 4382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794579646017699e-06,
+      "loss": 4.0412,
+      "step": 4383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794532237673831e-06,
+      "loss": 4.1102,
+      "step": 4384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794484829329962e-06,
+      "loss": 3.4561,
+      "step": 4385
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7944374209860934e-06,
+      "loss": 3.6294,
+      "step": 4386
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794390012642225e-06,
+      "loss": 4.088,
+      "step": 4387
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794342604298357e-06,
+      "loss": 3.8415,
+      "step": 4388
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794295195954488e-06,
+      "loss": 3.9674,
+      "step": 4389
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7942477876106194e-06,
+      "loss": 3.776,
+      "step": 4390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794200379266751e-06,
+      "loss": 3.7552,
+      "step": 4391
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7941529709228824e-06,
+      "loss": 3.6101,
+      "step": 4392
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7941055625790136e-06,
+      "loss": 3.3936,
+      "step": 4393
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7940581542351453e-06,
+      "loss": 3.5191,
+      "step": 4394
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.794010745891277e-06,
+      "loss": 4.157,
+      "step": 4395
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7939633375474083e-06,
+      "loss": 4.0866,
+      "step": 4396
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.79391592920354e-06,
+      "loss": 3.795,
+      "step": 4397
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7938685208596713e-06,
+      "loss": 3.9505,
+      "step": 4398
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7938211125158026e-06,
+      "loss": 3.986,
+      "step": 4399
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7937737041719343e-06,
+      "loss": 3.754,
+      "step": 4400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.793726295828066e-06,
+      "loss": 3.9538,
+      "step": 4401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7936788874841973e-06,
+      "loss": 4.0934,
+      "step": 4402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.793631479140329e-06,
+      "loss": 4.1119,
+      "step": 4403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7935840707964603e-06,
+      "loss": 3.8205,
+      "step": 4404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7935366624525915e-06,
+      "loss": 3.7034,
+      "step": 4405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7934892541087232e-06,
+      "loss": 3.8234,
+      "step": 4406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7934418457648545e-06,
+      "loss": 3.7132,
+      "step": 4407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7933944374209862e-06,
+      "loss": 3.4723,
+      "step": 4408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.793347029077118e-06,
+      "loss": 3.9677,
+      "step": 4409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7932996207332492e-06,
+      "loss": 4.024,
+      "step": 4410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7932522123893805e-06,
+      "loss": 4.2619,
+      "step": 4411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7932048040455118e-06,
+      "loss": 4.2307,
+      "step": 4412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7931573957016435e-06,
+      "loss": 3.6786,
+      "step": 4413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.793109987357775e-06,
+      "loss": 3.9988,
+      "step": 4414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7930625790139065e-06,
+      "loss": 3.3422,
+      "step": 4415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.793015170670038e-06,
+      "loss": 3.7808,
+      "step": 4416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7929677623261694e-06,
+      "loss": 4.0668,
+      "step": 4417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7929203539823007e-06,
+      "loss": 3.9658,
+      "step": 4418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7928729456384324e-06,
+      "loss": 4.031,
+      "step": 4419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7928255372945637e-06,
+      "loss": 3.949,
+      "step": 4420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7927781289506954e-06,
+      "loss": 3.9902,
+      "step": 4421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.792730720606827e-06,
+      "loss": 4.2865,
+      "step": 4422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7926833122629584e-06,
+      "loss": 3.7198,
+      "step": 4423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7926359039190897e-06,
+      "loss": 4.2462,
+      "step": 4424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7925884955752214e-06,
+      "loss": 3.6421,
+      "step": 4425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7925410872313527e-06,
+      "loss": 4.2211,
+      "step": 4426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7924936788874844e-06,
+      "loss": 4.1818,
+      "step": 4427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.792446270543616e-06,
+      "loss": 3.7781,
+      "step": 4428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7923988621997474e-06,
+      "loss": 4.0445,
+      "step": 4429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7923514538558786e-06,
+      "loss": 3.5273,
+      "step": 4430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7923040455120103e-06,
+      "loss": 3.6701,
+      "step": 4431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7922566371681416e-06,
+      "loss": 4.3762,
+      "step": 4432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.792209228824273e-06,
+      "loss": 3.9419,
+      "step": 4433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7921618204804046e-06,
+      "loss": 3.6438,
+      "step": 4434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7921144121365363e-06,
+      "loss": 4.197,
+      "step": 4435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7920670037926676e-06,
+      "loss": 4.0082,
+      "step": 4436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.792019595448799e-06,
+      "loss": 3.8216,
+      "step": 4437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7919721871049306e-06,
+      "loss": 4.1267,
+      "step": 4438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791924778761062e-06,
+      "loss": 3.893,
+      "step": 4439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7918773704171936e-06,
+      "loss": 3.9107,
+      "step": 4440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7918299620733253e-06,
+      "loss": 3.9016,
+      "step": 4441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7917825537294565e-06,
+      "loss": 3.4417,
+      "step": 4442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791735145385588e-06,
+      "loss": 3.6831,
+      "step": 4443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7916877370417195e-06,
+      "loss": 3.9445,
+      "step": 4444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791640328697851e-06,
+      "loss": 3.7461,
+      "step": 4445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791592920353982e-06,
+      "loss": 3.9718,
+      "step": 4446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791545512010114e-06,
+      "loss": 3.9866,
+      "step": 4447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7914981036662455e-06,
+      "loss": 3.5879,
+      "step": 4448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7914506953223768e-06,
+      "loss": 3.7806,
+      "step": 4449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7914032869785085e-06,
+      "loss": 4.0197,
+      "step": 4450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7913558786346398e-06,
+      "loss": 3.9724,
+      "step": 4451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791308470290771e-06,
+      "loss": 3.8849,
+      "step": 4452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7912610619469027e-06,
+      "loss": 3.9821,
+      "step": 4453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7912136536030344e-06,
+      "loss": 3.9298,
+      "step": 4454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7911662452591657e-06,
+      "loss": 3.7626,
+      "step": 4455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.791118836915297e-06,
+      "loss": 3.8157,
+      "step": 4456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7910714285714287e-06,
+      "loss": 4.1894,
+      "step": 4457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.79102402022756e-06,
+      "loss": 3.8969,
+      "step": 4458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7909766118836913e-06,
+      "loss": 4.0182,
+      "step": 4459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.790929203539823e-06,
+      "loss": 4.0116,
+      "step": 4460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7908817951959547e-06,
+      "loss": 4.1147,
+      "step": 4461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.790834386852086e-06,
+      "loss": 4.0434,
+      "step": 4462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7907869785082177e-06,
+      "loss": 4.0481,
+      "step": 4463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.790739570164349e-06,
+      "loss": 4.0255,
+      "step": 4464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7906921618204802e-06,
+      "loss": 4.0891,
+      "step": 4465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790644753476612e-06,
+      "loss": 4.0627,
+      "step": 4466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7905973451327436e-06,
+      "loss": 4.188,
+      "step": 4467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790549936788875e-06,
+      "loss": 3.9822,
+      "step": 4468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7905025284450066e-06,
+      "loss": 4.1688,
+      "step": 4469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790455120101138e-06,
+      "loss": 3.5793,
+      "step": 4470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790407711757269e-06,
+      "loss": 3.9355,
+      "step": 4471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790360303413401e-06,
+      "loss": 3.6511,
+      "step": 4472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790312895069532e-06,
+      "loss": 4.0225,
+      "step": 4473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790265486725664e-06,
+      "loss": 3.943,
+      "step": 4474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790218078381795e-06,
+      "loss": 3.6309,
+      "step": 4475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790170670037927e-06,
+      "loss": 3.9286,
+      "step": 4476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790123261694058e-06,
+      "loss": 3.8915,
+      "step": 4477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7900758533501894e-06,
+      "loss": 4.0338,
+      "step": 4478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.790028445006321e-06,
+      "loss": 4.1912,
+      "step": 4479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789981036662453e-06,
+      "loss": 4.0385,
+      "step": 4480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789933628318584e-06,
+      "loss": 4.2787,
+      "step": 4481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789886219974716e-06,
+      "loss": 3.4923,
+      "step": 4482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789838811630847e-06,
+      "loss": 3.8811,
+      "step": 4483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7897914032869784e-06,
+      "loss": 3.8295,
+      "step": 4484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.78974399494311e-06,
+      "loss": 3.865,
+      "step": 4485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7896965865992413e-06,
+      "loss": 3.9551,
+      "step": 4486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789649178255373e-06,
+      "loss": 4.097,
+      "step": 4487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7896017699115048e-06,
+      "loss": 3.9745,
+      "step": 4488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789554361567636e-06,
+      "loss": 4.1884,
+      "step": 4489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7895069532237673e-06,
+      "loss": 3.5973,
+      "step": 4490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789459544879899e-06,
+      "loss": 4.1247,
+      "step": 4491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7894121365360303e-06,
+      "loss": 4.0388,
+      "step": 4492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789364728192162e-06,
+      "loss": 3.7757,
+      "step": 4493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7893173198482937e-06,
+      "loss": 3.5682,
+      "step": 4494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.789269911504425e-06,
+      "loss": 3.6744,
+      "step": 4495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7892225031605563e-06,
+      "loss": 3.9661,
+      "step": 4496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7891750948166876e-06,
+      "loss": 3.9772,
+      "step": 4497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7891276864728193e-06,
+      "loss": 3.745,
+      "step": 4498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7890802781289505e-06,
+      "loss": 3.7834,
+      "step": 4499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7890328697850822e-06,
+      "loss": 3.7265,
+      "step": 4500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788985461441214e-06,
+      "loss": 3.953,
+      "step": 4501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7889380530973452e-06,
+      "loss": 3.4632,
+      "step": 4502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7888906447534765e-06,
+      "loss": 3.5113,
+      "step": 4503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788843236409608e-06,
+      "loss": 3.7253,
+      "step": 4504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7887958280657395e-06,
+      "loss": 3.6952,
+      "step": 4505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788748419721871e-06,
+      "loss": 3.8055,
+      "step": 4506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788701011378003e-06,
+      "loss": 3.9947,
+      "step": 4507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788653603034134e-06,
+      "loss": 3.9086,
+      "step": 4508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7886061946902655e-06,
+      "loss": 3.4447,
+      "step": 4509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788558786346397e-06,
+      "loss": 3.4683,
+      "step": 4510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7885113780025284e-06,
+      "loss": 3.9338,
+      "step": 4511
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7884639696586597e-06,
+      "loss": 4.0318,
+      "step": 4512
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7884165613147914e-06,
+      "loss": 4.0148,
+      "step": 4513
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788369152970923e-06,
+      "loss": 4.1216,
+      "step": 4514
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7883217446270544e-06,
+      "loss": 4.18,
+      "step": 4515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788274336283186e-06,
+      "loss": 3.8545,
+      "step": 4516
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7882269279393174e-06,
+      "loss": 3.8749,
+      "step": 4517
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7881795195954487e-06,
+      "loss": 4.3152,
+      "step": 4518
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7881321112515804e-06,
+      "loss": 3.7452,
+      "step": 4519
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.788084702907712e-06,
+      "loss": 3.9955,
+      "step": 4520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7880372945638434e-06,
+      "loss": 4.0514,
+      "step": 4521
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7879898862199746e-06,
+      "loss": 3.8109,
+      "step": 4522
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7879424778761063e-06,
+      "loss": 3.7443,
+      "step": 4523
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7878950695322376e-06,
+      "loss": 3.7083,
+      "step": 4524
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.787847661188369e-06,
+      "loss": 3.8392,
+      "step": 4525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7878002528445006e-06,
+      "loss": 3.8328,
+      "step": 4526
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7877528445006323e-06,
+      "loss": 4.1585,
+      "step": 4527
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.7877054361567636e-06,
+      "loss": 4.1372,
+      "step": 4528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7876580278128953e-06,
+      "loss": 3.8979,
+      "step": 4529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7876106194690266e-06,
+      "loss": 4.1542,
+      "step": 4530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.787563211125158e-06,
+      "loss": 3.9855,
+      "step": 4531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7875158027812896e-06,
+      "loss": 3.9334,
+      "step": 4532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7874683944374213e-06,
+      "loss": 4.1466,
+      "step": 4533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7874209860935526e-06,
+      "loss": 3.7975,
+      "step": 4534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7873735777496843e-06,
+      "loss": 4.0969,
+      "step": 4535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7873261694058155e-06,
+      "loss": 4.0447,
+      "step": 4536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.787278761061947e-06,
+      "loss": 3.9884,
+      "step": 4537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7872313527180785e-06,
+      "loss": 3.6652,
+      "step": 4538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.78718394437421e-06,
+      "loss": 3.911,
+      "step": 4539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7871365360303415e-06,
+      "loss": 4.0735,
+      "step": 4540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7870891276864728e-06,
+      "loss": 4.2326,
+      "step": 4541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7870417193426045e-06,
+      "loss": 4.1993,
+      "step": 4542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7869943109987358e-06,
+      "loss": 3.8614,
+      "step": 4543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786946902654867e-06,
+      "loss": 3.7774,
+      "step": 4544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7868994943109988e-06,
+      "loss": 3.6949,
+      "step": 4545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7868520859671305e-06,
+      "loss": 3.9649,
+      "step": 4546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7868046776232617e-06,
+      "loss": 3.925,
+      "step": 4547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7867572692793934e-06,
+      "loss": 3.5542,
+      "step": 4548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7867098609355247e-06,
+      "loss": 3.8343,
+      "step": 4549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786662452591656e-06,
+      "loss": 4.0647,
+      "step": 4550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7866150442477877e-06,
+      "loss": 4.1635,
+      "step": 4551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786567635903919e-06,
+      "loss": 3.8609,
+      "step": 4552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7865202275600507e-06,
+      "loss": 4.0653,
+      "step": 4553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7864728192161824e-06,
+      "loss": 3.7536,
+      "step": 4554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7864254108723137e-06,
+      "loss": 4.146,
+      "step": 4555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786378002528445e-06,
+      "loss": 3.9573,
+      "step": 4556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7863305941845767e-06,
+      "loss": 3.6107,
+      "step": 4557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786283185840708e-06,
+      "loss": 3.9891,
+      "step": 4558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7862357774968396e-06,
+      "loss": 3.9524,
+      "step": 4559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7861883691529713e-06,
+      "loss": 3.9531,
+      "step": 4560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7861409608091026e-06,
+      "loss": 3.7152,
+      "step": 4561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786093552465234e-06,
+      "loss": 3.8944,
+      "step": 4562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.786046144121365e-06,
+      "loss": 3.7424,
+      "step": 4563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785998735777497e-06,
+      "loss": 3.991,
+      "step": 4564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785951327433628e-06,
+      "loss": 4.2052,
+      "step": 4565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.78590391908976e-06,
+      "loss": 3.8204,
+      "step": 4566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7858565107458916e-06,
+      "loss": 3.9851,
+      "step": 4567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785809102402023e-06,
+      "loss": 4.2111,
+      "step": 4568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785761694058154e-06,
+      "loss": 3.6899,
+      "step": 4569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785714285714286e-06,
+      "loss": 3.8954,
+      "step": 4570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785666877370417e-06,
+      "loss": 4.1666,
+      "step": 4571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785619469026549e-06,
+      "loss": 4.2184,
+      "step": 4572
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7855720606826805e-06,
+      "loss": 4.2688,
+      "step": 4573
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785524652338812e-06,
+      "loss": 3.7898,
+      "step": 4574
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785477243994943e-06,
+      "loss": 3.485,
+      "step": 4575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785429835651075e-06,
+      "loss": 4.0581,
+      "step": 4576
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785382427307206e-06,
+      "loss": 4.0928,
+      "step": 4577
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7853350189633374e-06,
+      "loss": 3.9986,
+      "step": 4578
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785287610619469e-06,
+      "loss": 3.708,
+      "step": 4579
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7852402022756008e-06,
+      "loss": 4.0124,
+      "step": 4580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785192793931732e-06,
+      "loss": 4.1208,
+      "step": 4581
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7851453855878638e-06,
+      "loss": 3.8716,
+      "step": 4582
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785097977243995e-06,
+      "loss": 3.7873,
+      "step": 4583
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7850505689001263e-06,
+      "loss": 3.7589,
+      "step": 4584
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.785003160556258e-06,
+      "loss": 4.3211,
+      "step": 4585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7849557522123897e-06,
+      "loss": 4.2281,
+      "step": 4586
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.784908343868521e-06,
+      "loss": 3.6635,
+      "step": 4587
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7848609355246523e-06,
+      "loss": 3.7057,
+      "step": 4588
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.784813527180784e-06,
+      "loss": 3.7452,
+      "step": 4589
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7847661188369153e-06,
+      "loss": 4.1808,
+      "step": 4590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.7847187104930465e-06,
+      "loss": 3.7227,
+      "step": 4591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7846713021491782e-06,
+      "loss": 4.2578,
+      "step": 4592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.78462389380531e-06,
+      "loss": 4.0777,
+      "step": 4593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7845764854614412e-06,
+      "loss": 3.7175,
+      "step": 4594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784529077117573e-06,
+      "loss": 3.8872,
+      "step": 4595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7844816687737042e-06,
+      "loss": 4.3924,
+      "step": 4596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7844342604298355e-06,
+      "loss": 3.853,
+      "step": 4597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784386852085967e-06,
+      "loss": 4.1999,
+      "step": 4598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784339443742099e-06,
+      "loss": 3.9593,
+      "step": 4599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.78429203539823e-06,
+      "loss": 4.1957,
+      "step": 4600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784244627054362e-06,
+      "loss": 3.3759,
+      "step": 4601
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784197218710493e-06,
+      "loss": 3.8047,
+      "step": 4602
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7841498103666245e-06,
+      "loss": 3.2496,
+      "step": 4603
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784102402022756e-06,
+      "loss": 3.8247,
+      "step": 4604
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7840549936788874e-06,
+      "loss": 3.8268,
+      "step": 4605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.784007585335019e-06,
+      "loss": 3.9586,
+      "step": 4606
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7839601769911504e-06,
+      "loss": 4.1222,
+      "step": 4607
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.783912768647282e-06,
+      "loss": 3.8065,
+      "step": 4608
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7838653603034134e-06,
+      "loss": 3.8205,
+      "step": 4609
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7838179519595447e-06,
+      "loss": 3.9747,
+      "step": 4610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7837705436156764e-06,
+      "loss": 3.966,
+      "step": 4611
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.783723135271808e-06,
+      "loss": 4.0191,
+      "step": 4612
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7836757269279394e-06,
+      "loss": 4.1518,
+      "step": 4613
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.783628318584071e-06,
+      "loss": 3.5818,
+      "step": 4614
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7835809102402024e-06,
+      "loss": 3.9335,
+      "step": 4615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7835335018963336e-06,
+      "loss": 4.2423,
+      "step": 4616
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7834860935524653e-06,
+      "loss": 3.9084,
+      "step": 4617
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7834386852085966e-06,
+      "loss": 3.454,
+      "step": 4618
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7833912768647283e-06,
+      "loss": 3.6681,
+      "step": 4619
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.78334386852086e-06,
+      "loss": 4.1083,
+      "step": 4620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7832964601769913e-06,
+      "loss": 3.9751,
+      "step": 4621
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7832490518331226e-06,
+      "loss": 3.6769,
+      "step": 4622
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7832016434892543e-06,
+      "loss": 3.7492,
+      "step": 4623
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7831542351453856e-06,
+      "loss": 4.0278,
+      "step": 4624
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.783106826801517e-06,
+      "loss": 3.4374,
+      "step": 4625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.783059418457649e-06,
+      "loss": 3.7588,
+      "step": 4626
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7830120101137803e-06,
+      "loss": 3.4626,
+      "step": 4627
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7829646017699115e-06,
+      "loss": 4.1636,
+      "step": 4628
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.782917193426043e-06,
+      "loss": 3.7189,
+      "step": 4629
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7828697850821745e-06,
+      "loss": 3.7298,
+      "step": 4630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.782822376738306e-06,
+      "loss": 4.1473,
+      "step": 4631
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7827749683944375e-06,
+      "loss": 3.71,
+      "step": 4632
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7827275600505692e-06,
+      "loss": 3.8493,
+      "step": 4633
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7826801517067005e-06,
+      "loss": 4.0856,
+      "step": 4634
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7826327433628318e-06,
+      "loss": 3.8996,
+      "step": 4635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7825853350189635e-06,
+      "loss": 3.9882,
+      "step": 4636
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7825379266750948e-06,
+      "loss": 3.6381,
+      "step": 4637
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.782490518331226e-06,
+      "loss": 3.4023,
+      "step": 4638
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7824431099873577e-06,
+      "loss": 3.6568,
+      "step": 4639
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7823957016434895e-06,
+      "loss": 3.4678,
+      "step": 4640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7823482932996207e-06,
+      "loss": 3.8073,
+      "step": 4641
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7823008849557524e-06,
+      "loss": 3.8893,
+      "step": 4642
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7822534766118837e-06,
+      "loss": 4.0288,
+      "step": 4643
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.782206068268015e-06,
+      "loss": 4.3437,
+      "step": 4644
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7821586599241467e-06,
+      "loss": 3.8442,
+      "step": 4645
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7821112515802784e-06,
+      "loss": 3.6695,
+      "step": 4646
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7820638432364097e-06,
+      "loss": 4.2814,
+      "step": 4647
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7820164348925414e-06,
+      "loss": 3.4952,
+      "step": 4648
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7819690265486727e-06,
+      "loss": 4.2314,
+      "step": 4649
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.781921618204804e-06,
+      "loss": 4.0637,
+      "step": 4650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7818742098609352e-06,
+      "loss": 3.5802,
+      "step": 4651
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.781826801517067e-06,
+      "loss": 3.9953,
+      "step": 4652
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.7817793931731986e-06,
+      "loss": 4.1505,
+      "step": 4653
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.78173198482933e-06,
+      "loss": 3.7698,
+      "step": 4654
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7816845764854616e-06,
+      "loss": 3.8807,
+      "step": 4655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781637168141593e-06,
+      "loss": 3.9138,
+      "step": 4656
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781589759797724e-06,
+      "loss": 3.7816,
+      "step": 4657
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781542351453856e-06,
+      "loss": 3.9548,
+      "step": 4658
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7814949431099876e-06,
+      "loss": 4.2384,
+      "step": 4659
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781447534766119e-06,
+      "loss": 3.8107,
+      "step": 4660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7814001264222506e-06,
+      "loss": 3.8574,
+      "step": 4661
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781352718078382e-06,
+      "loss": 3.6591,
+      "step": 4662
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781305309734513e-06,
+      "loss": 4.0862,
+      "step": 4663
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781257901390645e-06,
+      "loss": 3.9981,
+      "step": 4664
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781210493046776e-06,
+      "loss": 3.9483,
+      "step": 4665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781163084702908e-06,
+      "loss": 3.9937,
+      "step": 4666
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7811156763590395e-06,
+      "loss": 4.2388,
+      "step": 4667
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781068268015171e-06,
+      "loss": 4.0121,
+      "step": 4668
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.781020859671302e-06,
+      "loss": 3.6583,
+      "step": 4669
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780973451327434e-06,
+      "loss": 3.6042,
+      "step": 4670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780926042983565e-06,
+      "loss": 4.1838,
+      "step": 4671
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7808786346396968e-06,
+      "loss": 3.7833,
+      "step": 4672
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780831226295828e-06,
+      "loss": 3.3074,
+      "step": 4673
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7807838179519598e-06,
+      "loss": 4.0317,
+      "step": 4674
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780736409608091e-06,
+      "loss": 4.1231,
+      "step": 4675
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7806890012642223e-06,
+      "loss": 3.9622,
+      "step": 4676
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780641592920354e-06,
+      "loss": 3.8329,
+      "step": 4677
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7805941845764853e-06,
+      "loss": 3.8896,
+      "step": 4678
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780546776232617e-06,
+      "loss": 3.949,
+      "step": 4679
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7804993678887487e-06,
+      "loss": 3.7203,
+      "step": 4680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.78045195954488e-06,
+      "loss": 4.1887,
+      "step": 4681
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7804045512010113e-06,
+      "loss": 4.0633,
+      "step": 4682
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780357142857143e-06,
+      "loss": 3.9217,
+      "step": 4683
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7803097345132743e-06,
+      "loss": 4.0288,
+      "step": 4684
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780262326169406e-06,
+      "loss": 3.6733,
+      "step": 4685
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7802149178255377e-06,
+      "loss": 3.5726,
+      "step": 4686
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780167509481669e-06,
+      "loss": 3.6438,
+      "step": 4687
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7801201011378002e-06,
+      "loss": 4.0578,
+      "step": 4688
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.780072692793932e-06,
+      "loss": 4.1266,
+      "step": 4689
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7800252844500632e-06,
+      "loss": 3.5311,
+      "step": 4690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7799778761061945e-06,
+      "loss": 3.9347,
+      "step": 4691
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779930467762326e-06,
+      "loss": 3.4873,
+      "step": 4692
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779883059418458e-06,
+      "loss": 4.0285,
+      "step": 4693
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779835651074589e-06,
+      "loss": 3.2513,
+      "step": 4694
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7797882427307205e-06,
+      "loss": 3.842,
+      "step": 4695
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779740834386852e-06,
+      "loss": 3.5283,
+      "step": 4696
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7796934260429834e-06,
+      "loss": 3.8445,
+      "step": 4697
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779646017699115e-06,
+      "loss": 4.0256,
+      "step": 4698
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779598609355247e-06,
+      "loss": 3.9533,
+      "step": 4699
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779551201011378e-06,
+      "loss": 4.0956,
+      "step": 4700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7795037926675094e-06,
+      "loss": 3.7058,
+      "step": 4701
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779456384323641e-06,
+      "loss": 3.4683,
+      "step": 4702
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7794089759797724e-06,
+      "loss": 4.1739,
+      "step": 4703
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7793615676359037e-06,
+      "loss": 3.5965,
+      "step": 4704
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7793141592920354e-06,
+      "loss": 4.239,
+      "step": 4705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.779266750948167e-06,
+      "loss": 3.9912,
+      "step": 4706
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7792193426042984e-06,
+      "loss": 3.9265,
+      "step": 4707
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.77917193426043e-06,
+      "loss": 4.0493,
+      "step": 4708
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7791245259165614e-06,
+      "loss": 3.6519,
+      "step": 4709
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7790771175726926e-06,
+      "loss": 4.0943,
+      "step": 4710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7790297092288243e-06,
+      "loss": 3.9177,
+      "step": 4711
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.778982300884956e-06,
+      "loss": 3.6798,
+      "step": 4712
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7789348925410873e-06,
+      "loss": 3.8035,
+      "step": 4713
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.778887484197219e-06,
+      "loss": 3.9552,
+      "step": 4714
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7788400758533503e-06,
+      "loss": 3.4042,
+      "step": 4715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7787926675094816e-06,
+      "loss": 4.0123,
+      "step": 4716
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.778745259165613e-06,
+      "loss": 3.4805,
+      "step": 4717
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.7786978508217446e-06,
+      "loss": 3.7227,
+      "step": 4718
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7786504424778763e-06,
+      "loss": 3.6682,
+      "step": 4719
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7786030341340076e-06,
+      "loss": 4.0007,
+      "step": 4720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7785556257901393e-06,
+      "loss": 3.645,
+      "step": 4721
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7785082174462705e-06,
+      "loss": 3.8487,
+      "step": 4722
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.778460809102402e-06,
+      "loss": 3.2826,
+      "step": 4723
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7784134007585335e-06,
+      "loss": 3.7025,
+      "step": 4724
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7783659924146652e-06,
+      "loss": 4.0728,
+      "step": 4725
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7783185840707965e-06,
+      "loss": 4.1399,
+      "step": 4726
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.778271175726928e-06,
+      "loss": 4.1479,
+      "step": 4727
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7782237673830595e-06,
+      "loss": 3.6451,
+      "step": 4728
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7781763590391908e-06,
+      "loss": 3.6472,
+      "step": 4729
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7781289506953225e-06,
+      "loss": 4.0603,
+      "step": 4730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7780815423514538e-06,
+      "loss": 4.1718,
+      "step": 4731
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7780341340075855e-06,
+      "loss": 3.8619,
+      "step": 4732
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.777986725663717e-06,
+      "loss": 4.0086,
+      "step": 4733
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7779393173198484e-06,
+      "loss": 4.1903,
+      "step": 4734
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7778919089759797e-06,
+      "loss": 4.0106,
+      "step": 4735
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7778445006321114e-06,
+      "loss": 3.844,
+      "step": 4736
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7777970922882427e-06,
+      "loss": 3.7797,
+      "step": 4737
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7777496839443744e-06,
+      "loss": 3.485,
+      "step": 4738
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7777022756005057e-06,
+      "loss": 4.0924,
+      "step": 4739
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7776548672566374e-06,
+      "loss": 3.9816,
+      "step": 4740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7776074589127687e-06,
+      "loss": 3.4135,
+      "step": 4741
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7775600505689e-06,
+      "loss": 4.0882,
+      "step": 4742
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7775126422250317e-06,
+      "loss": 3.8946,
+      "step": 4743
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.777465233881163e-06,
+      "loss": 4.1916,
+      "step": 4744
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7774178255372946e-06,
+      "loss": 3.7949,
+      "step": 4745
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7773704171934264e-06,
+      "loss": 4.2737,
+      "step": 4746
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7773230088495576e-06,
+      "loss": 3.8965,
+      "step": 4747
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.777275600505689e-06,
+      "loss": 4.229,
+      "step": 4748
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7772281921618206e-06,
+      "loss": 3.8984,
+      "step": 4749
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.777180783817952e-06,
+      "loss": 4.0186,
+      "step": 4750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7771333754740836e-06,
+      "loss": 4.0681,
+      "step": 4751
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7770859671302153e-06,
+      "loss": 3.6184,
+      "step": 4752
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7770385587863466e-06,
+      "loss": 3.6974,
+      "step": 4753
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776991150442478e-06,
+      "loss": 3.8453,
+      "step": 4754
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7769437420986096e-06,
+      "loss": 3.8183,
+      "step": 4755
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776896333754741e-06,
+      "loss": 3.9434,
+      "step": 4756
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776848925410872e-06,
+      "loss": 3.5648,
+      "step": 4757
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776801517067004e-06,
+      "loss": 3.7261,
+      "step": 4758
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7767541087231355e-06,
+      "loss": 3.7342,
+      "step": 4759
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776706700379267e-06,
+      "loss": 3.8822,
+      "step": 4760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776659292035398e-06,
+      "loss": 3.748,
+      "step": 4761
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.77661188369153e-06,
+      "loss": 4.0804,
+      "step": 4762
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776564475347661e-06,
+      "loss": 3.8304,
+      "step": 4763
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7765170670037928e-06,
+      "loss": 3.6992,
+      "step": 4764
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7764696586599245e-06,
+      "loss": 3.4947,
+      "step": 4765
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7764222503160558e-06,
+      "loss": 3.7412,
+      "step": 4766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776374841972187e-06,
+      "loss": 4.087,
+      "step": 4767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7763274336283188e-06,
+      "loss": 3.964,
+      "step": 4768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.77628002528445e-06,
+      "loss": 3.654,
+      "step": 4769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7762326169405813e-06,
+      "loss": 4.1393,
+      "step": 4770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776185208596713e-06,
+      "loss": 3.7803,
+      "step": 4771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7761378002528447e-06,
+      "loss": 3.5126,
+      "step": 4772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.776090391908976e-06,
+      "loss": 3.7296,
+      "step": 4773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7760429835651077e-06,
+      "loss": 3.6883,
+      "step": 4774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.775995575221239e-06,
+      "loss": 3.8875,
+      "step": 4775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7759481668773703e-06,
+      "loss": 4.0231,
+      "step": 4776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.775900758533502e-06,
+      "loss": 4.1044,
+      "step": 4777
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7758533501896337e-06,
+      "loss": 3.839,
+      "step": 4778
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.775805941845765e-06,
+      "loss": 3.9728,
+      "step": 4779
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.7757585335018967e-06,
+      "loss": 3.7165,
+      "step": 4780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.775711125158028e-06,
+      "loss": 3.8233,
+      "step": 4781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7756637168141592e-06,
+      "loss": 4.0518,
+      "step": 4782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7756163084702905e-06,
+      "loss": 4.0948,
+      "step": 4783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775568900126422e-06,
+      "loss": 4.0093,
+      "step": 4784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775521491782554e-06,
+      "loss": 3.6354,
+      "step": 4785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775474083438685e-06,
+      "loss": 4.1195,
+      "step": 4786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775426675094817e-06,
+      "loss": 4.1968,
+      "step": 4787
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775379266750948e-06,
+      "loss": 3.8244,
+      "step": 4788
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7753318584070795e-06,
+      "loss": 4.3871,
+      "step": 4789
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775284450063211e-06,
+      "loss": 4.0847,
+      "step": 4790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775237041719343e-06,
+      "loss": 3.6921,
+      "step": 4791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775189633375474e-06,
+      "loss": 4.1705,
+      "step": 4792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775142225031606e-06,
+      "loss": 3.7469,
+      "step": 4793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775094816687737e-06,
+      "loss": 3.8166,
+      "step": 4794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7750474083438684e-06,
+      "loss": 4.0644,
+      "step": 4795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775e-06,
+      "loss": 3.939,
+      "step": 4796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7749525916561314e-06,
+      "loss": 3.928,
+      "step": 4797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774905183312263e-06,
+      "loss": 3.8019,
+      "step": 4798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774857774968395e-06,
+      "loss": 4.2635,
+      "step": 4799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774810366624526e-06,
+      "loss": 3.946,
+      "step": 4800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7747629582806574e-06,
+      "loss": 4.0792,
+      "step": 4801
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774715549936789e-06,
+      "loss": 3.8941,
+      "step": 4802
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7746681415929203e-06,
+      "loss": 3.9815,
+      "step": 4803
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774620733249052e-06,
+      "loss": 3.7527,
+      "step": 4804
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7745733249051833e-06,
+      "loss": 4.075,
+      "step": 4805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774525916561315e-06,
+      "loss": 4.0303,
+      "step": 4806
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7744785082174463e-06,
+      "loss": 4.3179,
+      "step": 4807
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7744310998735776e-06,
+      "loss": 3.6618,
+      "step": 4808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7743836915297093e-06,
+      "loss": 3.7759,
+      "step": 4809
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7743362831858406e-06,
+      "loss": 3.7317,
+      "step": 4810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7742888748419723e-06,
+      "loss": 3.9298,
+      "step": 4811
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774241466498104e-06,
+      "loss": 4.0604,
+      "step": 4812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7741940581542353e-06,
+      "loss": 3.8795,
+      "step": 4813
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7741466498103665e-06,
+      "loss": 3.8202,
+      "step": 4814
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7740992414664983e-06,
+      "loss": 3.8273,
+      "step": 4815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7740518331226295e-06,
+      "loss": 3.274,
+      "step": 4816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7740044247787612e-06,
+      "loss": 3.5245,
+      "step": 4817
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773957016434893e-06,
+      "loss": 3.7495,
+      "step": 4818
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7739096080910242e-06,
+      "loss": 4.0377,
+      "step": 4819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7738621997471555e-06,
+      "loss": 4.1172,
+      "step": 4820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773814791403287e-06,
+      "loss": 4.1243,
+      "step": 4821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7737673830594185e-06,
+      "loss": 3.9954,
+      "step": 4822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7737199747155498e-06,
+      "loss": 3.71,
+      "step": 4823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7736725663716815e-06,
+      "loss": 3.946,
+      "step": 4824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773625158027813e-06,
+      "loss": 3.8525,
+      "step": 4825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7735777496839445e-06,
+      "loss": 4.194,
+      "step": 4826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7735303413400757e-06,
+      "loss": 3.8839,
+      "step": 4827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7734829329962074e-06,
+      "loss": 3.9184,
+      "step": 4828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7734355246523387e-06,
+      "loss": 3.7063,
+      "step": 4829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7733881163084704e-06,
+      "loss": 3.7327,
+      "step": 4830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773340707964602e-06,
+      "loss": 3.6015,
+      "step": 4831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7732932996207334e-06,
+      "loss": 3.8463,
+      "step": 4832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7732458912768647e-06,
+      "loss": 4.1521,
+      "step": 4833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7731984829329964e-06,
+      "loss": 3.83,
+      "step": 4834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7731510745891277e-06,
+      "loss": 3.4106,
+      "step": 4835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773103666245259e-06,
+      "loss": 3.7489,
+      "step": 4836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7730562579013907e-06,
+      "loss": 3.971,
+      "step": 4837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7730088495575224e-06,
+      "loss": 4.1608,
+      "step": 4838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7729614412136536e-06,
+      "loss": 3.7164,
+      "step": 4839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7729140328697853e-06,
+      "loss": 3.3332,
+      "step": 4840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7728666245259166e-06,
+      "loss": 3.7637,
+      "step": 4841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.772819216182048e-06,
+      "loss": 3.8911,
+      "step": 4842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7727718078381796e-06,
+      "loss": 3.8419,
+      "step": 4843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7727243994943113e-06,
+      "loss": 4.1351,
+      "step": 4844
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7726769911504426e-06,
+      "loss": 3.7104,
+      "step": 4845
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7726295828065743e-06,
+      "loss": 3.6978,
+      "step": 4846
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7725821744627056e-06,
+      "loss": 3.7676,
+      "step": 4847
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772534766118837e-06,
+      "loss": 3.9781,
+      "step": 4848
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772487357774968e-06,
+      "loss": 3.6833,
+      "step": 4849
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7724399494311e-06,
+      "loss": 3.8734,
+      "step": 4850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7723925410872315e-06,
+      "loss": 3.9863,
+      "step": 4851
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772345132743363e-06,
+      "loss": 3.9125,
+      "step": 4852
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7722977243994945e-06,
+      "loss": 3.8411,
+      "step": 4853
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772250316055626e-06,
+      "loss": 4.2065,
+      "step": 4854
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772202907711757e-06,
+      "loss": 3.8965,
+      "step": 4855
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772155499367889e-06,
+      "loss": 3.8331,
+      "step": 4856
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7721080910240205e-06,
+      "loss": 3.8241,
+      "step": 4857
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7720606826801518e-06,
+      "loss": 3.9407,
+      "step": 4858
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7720132743362835e-06,
+      "loss": 3.8696,
+      "step": 4859
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7719658659924148e-06,
+      "loss": 4.0067,
+      "step": 4860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771918457648546e-06,
+      "loss": 3.6553,
+      "step": 4861
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7718710493046777e-06,
+      "loss": 3.7805,
+      "step": 4862
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771823640960809e-06,
+      "loss": 4.3224,
+      "step": 4863
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7717762326169407e-06,
+      "loss": 4.1281,
+      "step": 4864
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7717288242730724e-06,
+      "loss": 3.9287,
+      "step": 4865
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7716814159292037e-06,
+      "loss": 3.4987,
+      "step": 4866
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771634007585335e-06,
+      "loss": 4.3669,
+      "step": 4867
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7715865992414667e-06,
+      "loss": 4.1149,
+      "step": 4868
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771539190897598e-06,
+      "loss": 3.7462,
+      "step": 4869
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7714917825537297e-06,
+      "loss": 3.5562,
+      "step": 4870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771444374209861e-06,
+      "loss": 3.9194,
+      "step": 4871
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7713969658659927e-06,
+      "loss": 4.1098,
+      "step": 4872
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771349557522124e-06,
+      "loss": 4.2586,
+      "step": 4873
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7713021491782552e-06,
+      "loss": 3.6782,
+      "step": 4874
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771254740834387e-06,
+      "loss": 3.9017,
+      "step": 4875
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7712073324905182e-06,
+      "loss": 3.4615,
+      "step": 4876
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.77115992414665e-06,
+      "loss": 3.8546,
+      "step": 4877
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7711125158027816e-06,
+      "loss": 3.5098,
+      "step": 4878
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771065107458913e-06,
+      "loss": 3.5878,
+      "step": 4879
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.771017699115044e-06,
+      "loss": 4.1924,
+      "step": 4880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770970290771176e-06,
+      "loss": 4.0004,
+      "step": 4881
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770922882427307e-06,
+      "loss": 3.64,
+      "step": 4882
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770875474083439e-06,
+      "loss": 4.0184,
+      "step": 4883
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7708280657395706e-06,
+      "loss": 3.8626,
+      "step": 4884
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770780657395702e-06,
+      "loss": 3.6748,
+      "step": 4885
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770733249051833e-06,
+      "loss": 3.4716,
+      "step": 4886
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770685840707965e-06,
+      "loss": 4.1224,
+      "step": 4887
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770638432364096e-06,
+      "loss": 4.0387,
+      "step": 4888
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7705910240202274e-06,
+      "loss": 4.1018,
+      "step": 4889
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770543615676359e-06,
+      "loss": 4.1269,
+      "step": 4890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770496207332491e-06,
+      "loss": 3.815,
+      "step": 4891
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770448798988622e-06,
+      "loss": 3.9802,
+      "step": 4892
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7704013906447534e-06,
+      "loss": 3.6496,
+      "step": 4893
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770353982300885e-06,
+      "loss": 3.985,
+      "step": 4894
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7703065739570164e-06,
+      "loss": 3.4863,
+      "step": 4895
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770259165613148e-06,
+      "loss": 3.553,
+      "step": 4896
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7702117572692798e-06,
+      "loss": 3.9056,
+      "step": 4897
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770164348925411e-06,
+      "loss": 3.6156,
+      "step": 4898
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7701169405815423e-06,
+      "loss": 3.9682,
+      "step": 4899
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.770069532237674e-06,
+      "loss": 4.129,
+      "step": 4900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7700221238938053e-06,
+      "loss": 3.6074,
+      "step": 4901
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7699747155499366e-06,
+      "loss": 3.7178,
+      "step": 4902
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7699273072060683e-06,
+      "loss": 3.7867,
+      "step": 4903
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7698798988622e-06,
+      "loss": 3.9203,
+      "step": 4904
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7698324905183313e-06,
+      "loss": 3.8327,
+      "step": 4905
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.769785082174463e-06,
+      "loss": 3.8678,
+      "step": 4906
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7697376738305943e-06,
+      "loss": 3.9654,
+      "step": 4907
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7696902654867255e-06,
+      "loss": 4.2157,
+      "step": 4908
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7696428571428572e-06,
+      "loss": 4.2209,
+      "step": 4909
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.769595448798989e-06,
+      "loss": 3.9315,
+      "step": 4910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7695480404551202e-06,
+      "loss": 4.1794,
+      "step": 4911
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7695006321112515e-06,
+      "loss": 3.7759,
+      "step": 4912
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7694532237673832e-06,
+      "loss": 4.0255,
+      "step": 4913
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7694058154235145e-06,
+      "loss": 3.9564,
+      "step": 4914
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7693584070796458e-06,
+      "loss": 3.275,
+      "step": 4915
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7693109987357775e-06,
+      "loss": 3.7737,
+      "step": 4916
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.769263590391909e-06,
+      "loss": 3.7485,
+      "step": 4917
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7692161820480405e-06,
+      "loss": 3.3672,
+      "step": 4918
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.769168773704172e-06,
+      "loss": 4.1923,
+      "step": 4919
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7691213653603034e-06,
+      "loss": 4.2483,
+      "step": 4920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7690739570164347e-06,
+      "loss": 3.6486,
+      "step": 4921
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7690265486725664e-06,
+      "loss": 3.9229,
+      "step": 4922
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7689791403286977e-06,
+      "loss": 4.1072,
+      "step": 4923
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7689317319848294e-06,
+      "loss": 3.7987,
+      "step": 4924
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.768884323640961e-06,
+      "loss": 3.8509,
+      "step": 4925
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7688369152970924e-06,
+      "loss": 3.7979,
+      "step": 4926
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7687895069532237e-06,
+      "loss": 3.8929,
+      "step": 4927
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7687420986093554e-06,
+      "loss": 4.2035,
+      "step": 4928
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7686946902654867e-06,
+      "loss": 3.7636,
+      "step": 4929
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7686472819216184e-06,
+      "loss": 3.9981,
+      "step": 4930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.76859987357775e-06,
+      "loss": 3.7531,
+      "step": 4931
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7685524652338814e-06,
+      "loss": 3.2443,
+      "step": 4932
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7685050568900126e-06,
+      "loss": 4.058,
+      "step": 4933
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.768457648546144e-06,
+      "loss": 4.1507,
+      "step": 4934
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7684102402022756e-06,
+      "loss": 4.0945,
+      "step": 4935
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.768362831858407e-06,
+      "loss": 3.9076,
+      "step": 4936
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7683154235145386e-06,
+      "loss": 3.9386,
+      "step": 4937
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7682680151706703e-06,
+      "loss": 3.8018,
+      "step": 4938
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7682206068268016e-06,
+      "loss": 3.9444,
+      "step": 4939
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.768173198482933e-06,
+      "loss": 4.2173,
+      "step": 4940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7681257901390646e-06,
+      "loss": 3.6849,
+      "step": 4941
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.768078381795196e-06,
+      "loss": 3.9386,
+      "step": 4942
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7680309734513276e-06,
+      "loss": 3.913,
+      "step": 4943
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7679835651074593e-06,
+      "loss": 3.8044,
+      "step": 4944
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7679361567635905e-06,
+      "loss": 3.9726,
+      "step": 4945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767888748419722e-06,
+      "loss": 3.7731,
+      "step": 4946
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7678413400758535e-06,
+      "loss": 3.6029,
+      "step": 4947
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767793931731985e-06,
+      "loss": 3.5457,
+      "step": 4948
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767746523388116e-06,
+      "loss": 3.5037,
+      "step": 4949
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767699115044248e-06,
+      "loss": 3.7663,
+      "step": 4950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7676517067003795e-06,
+      "loss": 3.9385,
+      "step": 4951
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7676042983565108e-06,
+      "loss": 3.793,
+      "step": 4952
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7675568900126425e-06,
+      "loss": 3.9338,
+      "step": 4953
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7675094816687738e-06,
+      "loss": 3.9748,
+      "step": 4954
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767462073324905e-06,
+      "loss": 3.9013,
+      "step": 4955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7674146649810367e-06,
+      "loss": 3.8119,
+      "step": 4956
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7673672566371684e-06,
+      "loss": 3.8836,
+      "step": 4957
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7673198482932997e-06,
+      "loss": 3.4791,
+      "step": 4958
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767272439949431e-06,
+      "loss": 3.8141,
+      "step": 4959
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7672250316055627e-06,
+      "loss": 3.7052,
+      "step": 4960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767177623261694e-06,
+      "loss": 3.7205,
+      "step": 4961
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7671302149178253e-06,
+      "loss": 4.0038,
+      "step": 4962
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.767082806573957e-06,
+      "loss": 3.8754,
+      "step": 4963
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7670353982300887e-06,
+      "loss": 4.0007,
+      "step": 4964
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.76698798988622e-06,
+      "loss": 3.5394,
+      "step": 4965
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7669405815423517e-06,
+      "loss": 4.0315,
+      "step": 4966
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.766893173198483e-06,
+      "loss": 3.7246,
+      "step": 4967
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7668457648546142e-06,
+      "loss": 3.6284,
+      "step": 4968
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.766798356510746e-06,
+      "loss": 3.7708,
+      "step": 4969
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.7667509481668776e-06,
+      "loss": 4.2199,
+      "step": 4970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.766703539823009e-06,
+      "loss": 4.0333,
+      "step": 4971
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7666561314791406e-06,
+      "loss": 4.4732,
+      "step": 4972
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766608723135272e-06,
+      "loss": 3.647,
+      "step": 4973
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766561314791403e-06,
+      "loss": 3.9864,
+      "step": 4974
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766513906447535e-06,
+      "loss": 4.1705,
+      "step": 4975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766466498103666e-06,
+      "loss": 4.0281,
+      "step": 4976
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766419089759798e-06,
+      "loss": 3.877,
+      "step": 4977
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766371681415929e-06,
+      "loss": 4.0571,
+      "step": 4978
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766324273072061e-06,
+      "loss": 4.0119,
+      "step": 4979
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766276864728192e-06,
+      "loss": 3.7027,
+      "step": 4980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7662294563843234e-06,
+      "loss": 4.2928,
+      "step": 4981
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766182048040455e-06,
+      "loss": 3.968,
+      "step": 4982
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766134639696587e-06,
+      "loss": 3.6755,
+      "step": 4983
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.766087231352718e-06,
+      "loss": 3.847,
+      "step": 4984
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.76603982300885e-06,
+      "loss": 4.0692,
+      "step": 4985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765992414664981e-06,
+      "loss": 3.478,
+      "step": 4986
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7659450063211124e-06,
+      "loss": 3.7109,
+      "step": 4987
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765897597977244e-06,
+      "loss": 3.6928,
+      "step": 4988
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7658501896333753e-06,
+      "loss": 4.0739,
+      "step": 4989
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765802781289507e-06,
+      "loss": 3.7565,
+      "step": 4990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7657553729456388e-06,
+      "loss": 3.9355,
+      "step": 4991
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.76570796460177e-06,
+      "loss": 3.4547,
+      "step": 4992
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7656605562579013e-06,
+      "loss": 3.8244,
+      "step": 4993
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765613147914033e-06,
+      "loss": 3.6122,
+      "step": 4994
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7655657395701643e-06,
+      "loss": 3.7626,
+      "step": 4995
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765518331226296e-06,
+      "loss": 4.1753,
+      "step": 4996
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7654709228824277e-06,
+      "loss": 4.5251,
+      "step": 4997
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765423514538559e-06,
+      "loss": 3.8475,
+      "step": 4998
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7653761061946903e-06,
+      "loss": 3.9925,
+      "step": 4999
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7653286978508216e-06,
+      "loss": 3.5123,
+      "step": 5000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7652812895069533e-06,
+      "loss": 3.9215,
+      "step": 5001
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7652338811630845e-06,
+      "loss": 3.6091,
+      "step": 5002
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7651864728192162e-06,
+      "loss": 3.521,
+      "step": 5003
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.765139064475348e-06,
+      "loss": 3.8482,
+      "step": 5004
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7650916561314792e-06,
+      "loss": 3.9534,
+      "step": 5005
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7650442477876105e-06,
+      "loss": 4.0761,
+      "step": 5006
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764996839443742e-06,
+      "loss": 3.933,
+      "step": 5007
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7649494310998735e-06,
+      "loss": 3.7779,
+      "step": 5008
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764902022756005e-06,
+      "loss": 3.9277,
+      "step": 5009
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764854614412137e-06,
+      "loss": 3.9767,
+      "step": 5010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764807206068268e-06,
+      "loss": 3.8432,
+      "step": 5011
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7647597977243995e-06,
+      "loss": 3.8045,
+      "step": 5012
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764712389380531e-06,
+      "loss": 3.6664,
+      "step": 5013
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7646649810366624e-06,
+      "loss": 3.7958,
+      "step": 5014
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7646175726927937e-06,
+      "loss": 3.9034,
+      "step": 5015
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7645701643489254e-06,
+      "loss": 4.0061,
+      "step": 5016
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764522756005057e-06,
+      "loss": 3.9801,
+      "step": 5017
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7644753476611884e-06,
+      "loss": 4.0232,
+      "step": 5018
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.76442793931732e-06,
+      "loss": 4.2445,
+      "step": 5019
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7643805309734514e-06,
+      "loss": 3.9356,
+      "step": 5020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7643331226295827e-06,
+      "loss": 4.2783,
+      "step": 5021
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7642857142857144e-06,
+      "loss": 3.8146,
+      "step": 5022
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764238305941846e-06,
+      "loss": 3.7908,
+      "step": 5023
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7641908975979774e-06,
+      "loss": 4.0304,
+      "step": 5024
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7641434892541086e-06,
+      "loss": 3.4245,
+      "step": 5025
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7640960809102403e-06,
+      "loss": 3.7949,
+      "step": 5026
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7640486725663716e-06,
+      "loss": 3.8615,
+      "step": 5027
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.764001264222503e-06,
+      "loss": 4.1121,
+      "step": 5028
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7639538558786346e-06,
+      "loss": 3.4986,
+      "step": 5029
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7639064475347663e-06,
+      "loss": 3.6226,
+      "step": 5030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7638590391908976e-06,
+      "loss": 3.8273,
+      "step": 5031
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7638116308470293e-06,
+      "loss": 3.9827,
+      "step": 5032
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.7637642225031606e-06,
+      "loss": 4.2908,
+      "step": 5033
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.763716814159292e-06,
+      "loss": 3.5348,
+      "step": 5034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7636694058154236e-06,
+      "loss": 3.8792,
+      "step": 5035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7636219974715553e-06,
+      "loss": 3.5785,
+      "step": 5036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7635745891276866e-06,
+      "loss": 3.3155,
+      "step": 5037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7635271807838183e-06,
+      "loss": 3.8995,
+      "step": 5038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7634797724399495e-06,
+      "loss": 3.54,
+      "step": 5039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.763432364096081e-06,
+      "loss": 3.6759,
+      "step": 5040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7633849557522125e-06,
+      "loss": 3.8582,
+      "step": 5041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.763337547408344e-06,
+      "loss": 4.0952,
+      "step": 5042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7632901390644755e-06,
+      "loss": 4.0649,
+      "step": 5043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7632427307206068e-06,
+      "loss": 4.1809,
+      "step": 5044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7631953223767385e-06,
+      "loss": 3.9493,
+      "step": 5045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7631479140328698e-06,
+      "loss": 4.0856,
+      "step": 5046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.763100505689001e-06,
+      "loss": 3.6631,
+      "step": 5047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7630530973451328e-06,
+      "loss": 3.8802,
+      "step": 5048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7630056890012645e-06,
+      "loss": 4.0432,
+      "step": 5049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7629582806573957e-06,
+      "loss": 3.6321,
+      "step": 5050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7629108723135274e-06,
+      "loss": 4.0806,
+      "step": 5051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7628634639696587e-06,
+      "loss": 3.6885,
+      "step": 5052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.76281605562579e-06,
+      "loss": 3.5687,
+      "step": 5053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7627686472819217e-06,
+      "loss": 3.7826,
+      "step": 5054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762721238938053e-06,
+      "loss": 3.8397,
+      "step": 5055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7626738305941847e-06,
+      "loss": 4.1275,
+      "step": 5056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7626264222503164e-06,
+      "loss": 4.0911,
+      "step": 5057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7625790139064477e-06,
+      "loss": 4.0562,
+      "step": 5058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762531605562579e-06,
+      "loss": 3.9157,
+      "step": 5059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7624841972187107e-06,
+      "loss": 4.225,
+      "step": 5060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762436788874842e-06,
+      "loss": 3.4949,
+      "step": 5061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7623893805309736e-06,
+      "loss": 4.3535,
+      "step": 5062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7623419721871053e-06,
+      "loss": 4.2831,
+      "step": 5063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7622945638432366e-06,
+      "loss": 3.6499,
+      "step": 5064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762247155499368e-06,
+      "loss": 3.708,
+      "step": 5065
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762199747155499e-06,
+      "loss": 4.0624,
+      "step": 5066
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762152338811631e-06,
+      "loss": 4.0803,
+      "step": 5067
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762104930467762e-06,
+      "loss": 3.7624,
+      "step": 5068
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762057522123894e-06,
+      "loss": 3.7801,
+      "step": 5069
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7620101137800256e-06,
+      "loss": 3.8621,
+      "step": 5070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761962705436157e-06,
+      "loss": 3.6114,
+      "step": 5071
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761915297092288e-06,
+      "loss": 4.0122,
+      "step": 5072
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.76186788874842e-06,
+      "loss": 3.5876,
+      "step": 5073
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761820480404551e-06,
+      "loss": 3.9992,
+      "step": 5074
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761773072060683e-06,
+      "loss": 4.141,
+      "step": 5075
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7617256637168145e-06,
+      "loss": 3.9819,
+      "step": 5076
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761678255372946e-06,
+      "loss": 3.4779,
+      "step": 5077
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761630847029077e-06,
+      "loss": 3.8287,
+      "step": 5078
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761583438685209e-06,
+      "loss": 4.3045,
+      "step": 5079
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.76153603034134e-06,
+      "loss": 3.9538,
+      "step": 5080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7614886219974714e-06,
+      "loss": 4.0286,
+      "step": 5081
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761441213653603e-06,
+      "loss": 3.7819,
+      "step": 5082
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7613938053097348e-06,
+      "loss": 3.9838,
+      "step": 5083
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761346396965866e-06,
+      "loss": 4.0181,
+      "step": 5084
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7612989886219978e-06,
+      "loss": 4.3531,
+      "step": 5085
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761251580278129e-06,
+      "loss": 4.1571,
+      "step": 5086
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7612041719342603e-06,
+      "loss": 3.618,
+      "step": 5087
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761156763590392e-06,
+      "loss": 3.9011,
+      "step": 5088
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7611093552465237e-06,
+      "loss": 4.3207,
+      "step": 5089
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761061946902655e-06,
+      "loss": 3.8825,
+      "step": 5090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7610145385587863e-06,
+      "loss": 3.5754,
+      "step": 5091
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.760967130214918e-06,
+      "loss": 4.0628,
+      "step": 5092
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7609197218710493e-06,
+      "loss": 3.2954,
+      "step": 5093
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7608723135271805e-06,
+      "loss": 3.6703,
+      "step": 5094
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7608249051833123e-06,
+      "loss": 3.9522,
+      "step": 5095
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.760777496839444e-06,
+      "loss": 3.7933,
+      "step": 5096
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7607300884955752e-06,
+      "loss": 3.9709,
+      "step": 5097
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.760682680151707e-06,
+      "loss": 3.5225,
+      "step": 5098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7606352718078382e-06,
+      "loss": 3.6374,
+      "step": 5099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7605878634639695e-06,
+      "loss": 3.9162,
+      "step": 5100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760540455120101e-06,
+      "loss": 3.8605,
+      "step": 5101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760493046776233e-06,
+      "loss": 4.0981,
+      "step": 5102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760445638432364e-06,
+      "loss": 3.936,
+      "step": 5103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760398230088496e-06,
+      "loss": 3.9241,
+      "step": 5104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760350821744627e-06,
+      "loss": 4.1307,
+      "step": 5105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7603034134007585e-06,
+      "loss": 3.3874,
+      "step": 5106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.76025600505689e-06,
+      "loss": 3.4506,
+      "step": 5107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7602085967130214e-06,
+      "loss": 3.9682,
+      "step": 5108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760161188369153e-06,
+      "loss": 4.1223,
+      "step": 5109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7601137800252844e-06,
+      "loss": 4.0679,
+      "step": 5110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.760066371681416e-06,
+      "loss": 4.1233,
+      "step": 5111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7600189633375474e-06,
+      "loss": 4.0409,
+      "step": 5112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7599715549936787e-06,
+      "loss": 3.6717,
+      "step": 5113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7599241466498104e-06,
+      "loss": 3.7948,
+      "step": 5114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.759876738305942e-06,
+      "loss": 4.015,
+      "step": 5115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7598293299620734e-06,
+      "loss": 3.7699,
+      "step": 5116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.759781921618205e-06,
+      "loss": 4.334,
+      "step": 5117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7597345132743364e-06,
+      "loss": 4.0065,
+      "step": 5118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7596871049304676e-06,
+      "loss": 3.8065,
+      "step": 5119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7596396965865993e-06,
+      "loss": 3.4417,
+      "step": 5120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7595922882427306e-06,
+      "loss": 3.5058,
+      "step": 5121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7595448798988623e-06,
+      "loss": 3.7243,
+      "step": 5122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.759497471554994e-06,
+      "loss": 3.6772,
+      "step": 5123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7594500632111253e-06,
+      "loss": 3.6464,
+      "step": 5124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7594026548672566e-06,
+      "loss": 4.0502,
+      "step": 5125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7593552465233883e-06,
+      "loss": 3.7743,
+      "step": 5126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7593078381795196e-06,
+      "loss": 3.5742,
+      "step": 5127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7592604298356513e-06,
+      "loss": 4.1867,
+      "step": 5128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.759213021491783e-06,
+      "loss": 3.8719,
+      "step": 5129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7591656131479143e-06,
+      "loss": 3.9555,
+      "step": 5130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7591182048040455e-06,
+      "loss": 4.1557,
+      "step": 5131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.759070796460177e-06,
+      "loss": 4.389,
+      "step": 5132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7590233881163085e-06,
+      "loss": 4.2808,
+      "step": 5133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.75897597977244e-06,
+      "loss": 3.5599,
+      "step": 5134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7589285714285715e-06,
+      "loss": 3.9416,
+      "step": 5135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7588811630847032e-06,
+      "loss": 3.9959,
+      "step": 5136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7588337547408345e-06,
+      "loss": 3.7118,
+      "step": 5137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7587863463969658e-06,
+      "loss": 3.9252,
+      "step": 5138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7587389380530975e-06,
+      "loss": 4.0396,
+      "step": 5139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7586915297092288e-06,
+      "loss": 3.5273,
+      "step": 5140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7586441213653605e-06,
+      "loss": 3.4209,
+      "step": 5141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.758596713021492e-06,
+      "loss": 3.9258,
+      "step": 5142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7585493046776235e-06,
+      "loss": 3.933,
+      "step": 5143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7585018963337547e-06,
+      "loss": 3.6555,
+      "step": 5144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7584544879898864e-06,
+      "loss": 3.9262,
+      "step": 5145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7584070796460177e-06,
+      "loss": 3.7641,
+      "step": 5146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.758359671302149e-06,
+      "loss": 3.8704,
+      "step": 5147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7583122629582807e-06,
+      "loss": 4.1091,
+      "step": 5148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7582648546144124e-06,
+      "loss": 3.8912,
+      "step": 5149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7582174462705437e-06,
+      "loss": 4.2653,
+      "step": 5150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7581700379266754e-06,
+      "loss": 3.7791,
+      "step": 5151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7581226295828067e-06,
+      "loss": 3.7955,
+      "step": 5152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.758075221238938e-06,
+      "loss": 3.6838,
+      "step": 5153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7580278128950697e-06,
+      "loss": 3.8281,
+      "step": 5154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7579804045512014e-06,
+      "loss": 3.5386,
+      "step": 5155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7579329962073326e-06,
+      "loss": 3.9224,
+      "step": 5156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.757885587863464e-06,
+      "loss": 4.1753,
+      "step": 5157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7578381795195956e-06,
+      "loss": 3.8596,
+      "step": 5158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.757790771175727e-06,
+      "loss": 4.1155,
+      "step": 5159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.757743362831858e-06,
+      "loss": 3.6424,
+      "step": 5160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.75769595448799e-06,
+      "loss": 3.881,
+      "step": 5161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7576485461441216e-06,
+      "loss": 3.8609,
+      "step": 5162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757601137800253e-06,
+      "loss": 3.4805,
+      "step": 5163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7575537294563846e-06,
+      "loss": 4.0588,
+      "step": 5164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757506321112516e-06,
+      "loss": 3.783,
+      "step": 5165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757458912768647e-06,
+      "loss": 3.7482,
+      "step": 5166
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757411504424779e-06,
+      "loss": 3.6603,
+      "step": 5167
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7573640960809105e-06,
+      "loss": 3.8148,
+      "step": 5168
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757316687737042e-06,
+      "loss": 4.2153,
+      "step": 5169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7572692793931735e-06,
+      "loss": 3.9772,
+      "step": 5170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757221871049305e-06,
+      "loss": 3.9727,
+      "step": 5171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757174462705436e-06,
+      "loss": 3.8904,
+      "step": 5172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757127054361568e-06,
+      "loss": 3.5933,
+      "step": 5173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.757079646017699e-06,
+      "loss": 3.6111,
+      "step": 5174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7570322376738308e-06,
+      "loss": 3.92,
+      "step": 5175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756984829329962e-06,
+      "loss": 3.7745,
+      "step": 5176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7569374209860938e-06,
+      "loss": 3.399,
+      "step": 5177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756890012642225e-06,
+      "loss": 4.065,
+      "step": 5178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7568426042983563e-06,
+      "loss": 4.0082,
+      "step": 5179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756795195954488e-06,
+      "loss": 4.1293,
+      "step": 5180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7567477876106197e-06,
+      "loss": 3.6608,
+      "step": 5181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756700379266751e-06,
+      "loss": 3.693,
+      "step": 5182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7566529709228827e-06,
+      "loss": 3.8456,
+      "step": 5183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756605562579014e-06,
+      "loss": 3.6179,
+      "step": 5184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7565581542351453e-06,
+      "loss": 3.8154,
+      "step": 5185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756510745891277e-06,
+      "loss": 3.7849,
+      "step": 5186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7564633375474083e-06,
+      "loss": 3.7722,
+      "step": 5187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.75641592920354e-06,
+      "loss": 3.749,
+      "step": 5188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7563685208596717e-06,
+      "loss": 3.6168,
+      "step": 5189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756321112515803e-06,
+      "loss": 3.6446,
+      "step": 5190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7562737041719342e-06,
+      "loss": 4.0572,
+      "step": 5191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756226295828066e-06,
+      "loss": 3.7218,
+      "step": 5192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7561788874841972e-06,
+      "loss": 4.049,
+      "step": 5193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7561314791403285e-06,
+      "loss": 3.645,
+      "step": 5194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7560840707964606e-06,
+      "loss": 3.8924,
+      "step": 5195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.756036662452592e-06,
+      "loss": 3.7101,
+      "step": 5196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755989254108723e-06,
+      "loss": 3.7173,
+      "step": 5197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7559418457648545e-06,
+      "loss": 3.7392,
+      "step": 5198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755894437420986e-06,
+      "loss": 4.2499,
+      "step": 5199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7558470290771174e-06,
+      "loss": 3.7476,
+      "step": 5200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755799620733249e-06,
+      "loss": 4.1794,
+      "step": 5201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755752212389381e-06,
+      "loss": 3.607,
+      "step": 5202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755704804045512e-06,
+      "loss": 3.8216,
+      "step": 5203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7556573957016434e-06,
+      "loss": 3.984,
+      "step": 5204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755609987357775e-06,
+      "loss": 3.7686,
+      "step": 5205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7555625790139064e-06,
+      "loss": 3.4352,
+      "step": 5206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7555151706700377e-06,
+      "loss": 4.013,
+      "step": 5207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7554677623261694e-06,
+      "loss": 3.7343,
+      "step": 5208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755420353982301e-06,
+      "loss": 3.7478,
+      "step": 5209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7553729456384324e-06,
+      "loss": 3.9366,
+      "step": 5210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755325537294564e-06,
+      "loss": 4.1512,
+      "step": 5211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7552781289506954e-06,
+      "loss": 4.2064,
+      "step": 5212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7552307206068266e-06,
+      "loss": 3.7248,
+      "step": 5213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7551833122629583e-06,
+      "loss": 3.1819,
+      "step": 5214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.75513590391909e-06,
+      "loss": 3.6296,
+      "step": 5215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7550884955752213e-06,
+      "loss": 3.4455,
+      "step": 5216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.755041087231353e-06,
+      "loss": 3.6192,
+      "step": 5217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7549936788874843e-06,
+      "loss": 4.2627,
+      "step": 5218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7549462705436156e-06,
+      "loss": 4.0232,
+      "step": 5219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.754898862199747e-06,
+      "loss": 3.8962,
+      "step": 5220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7548514538558786e-06,
+      "loss": 3.9946,
+      "step": 5221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7548040455120103e-06,
+      "loss": 4.2449,
+      "step": 5222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7547566371681416e-06,
+      "loss": 3.6664,
+      "step": 5223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.7547092288242733e-06,
+      "loss": 3.7684,
+      "step": 5224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7546618204804045e-06,
+      "loss": 3.7607,
+      "step": 5225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.754614412136536e-06,
+      "loss": 3.9055,
+      "step": 5226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7545670037926675e-06,
+      "loss": 4.0017,
+      "step": 5227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7545195954487992e-06,
+      "loss": 3.6831,
+      "step": 5228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7544721871049305e-06,
+      "loss": 3.9264,
+      "step": 5229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.754424778761062e-06,
+      "loss": 3.2887,
+      "step": 5230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7543773704171935e-06,
+      "loss": 3.9134,
+      "step": 5231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7543299620733248e-06,
+      "loss": 3.91,
+      "step": 5232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7542825537294565e-06,
+      "loss": 3.9342,
+      "step": 5233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7542351453855878e-06,
+      "loss": 4.1477,
+      "step": 5234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7541877370417195e-06,
+      "loss": 3.5948,
+      "step": 5235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.754140328697851e-06,
+      "loss": 4.0584,
+      "step": 5236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7540929203539824e-06,
+      "loss": 3.8281,
+      "step": 5237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7540455120101137e-06,
+      "loss": 3.639,
+      "step": 5238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7539981036662454e-06,
+      "loss": 4.0673,
+      "step": 5239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7539506953223767e-06,
+      "loss": 3.9184,
+      "step": 5240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7539032869785084e-06,
+      "loss": 3.5842,
+      "step": 5241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7538558786346397e-06,
+      "loss": 3.5018,
+      "step": 5242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7538084702907714e-06,
+      "loss": 4.0775,
+      "step": 5243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7537610619469027e-06,
+      "loss": 4.195,
+      "step": 5244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753713653603034e-06,
+      "loss": 3.962,
+      "step": 5245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7536662452591657e-06,
+      "loss": 4.1513,
+      "step": 5246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753618836915297e-06,
+      "loss": 3.684,
+      "step": 5247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7535714285714286e-06,
+      "loss": 3.5313,
+      "step": 5248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7535240202275604e-06,
+      "loss": 3.9355,
+      "step": 5249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7534766118836916e-06,
+      "loss": 3.4543,
+      "step": 5250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753429203539823e-06,
+      "loss": 3.7372,
+      "step": 5251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7533817951959546e-06,
+      "loss": 3.6852,
+      "step": 5252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753334386852086e-06,
+      "loss": 4.1126,
+      "step": 5253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7532869785082176e-06,
+      "loss": 4.1665,
+      "step": 5254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7532395701643493e-06,
+      "loss": 4.1902,
+      "step": 5255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7531921618204806e-06,
+      "loss": 3.8939,
+      "step": 5256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753144753476612e-06,
+      "loss": 3.8729,
+      "step": 5257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7530973451327436e-06,
+      "loss": 3.2331,
+      "step": 5258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753049936788875e-06,
+      "loss": 3.7747,
+      "step": 5259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.753002528445006e-06,
+      "loss": 3.364,
+      "step": 5260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752955120101138e-06,
+      "loss": 3.6645,
+      "step": 5261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7529077117572695e-06,
+      "loss": 4.1386,
+      "step": 5262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752860303413401e-06,
+      "loss": 3.6446,
+      "step": 5263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752812895069532e-06,
+      "loss": 4.1604,
+      "step": 5264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752765486725664e-06,
+      "loss": 4.1109,
+      "step": 5265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752718078381795e-06,
+      "loss": 3.8487,
+      "step": 5266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7526706700379268e-06,
+      "loss": 4.0218,
+      "step": 5267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7526232616940585e-06,
+      "loss": 3.5054,
+      "step": 5268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7525758533501898e-06,
+      "loss": 3.8485,
+      "step": 5269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752528445006321e-06,
+      "loss": 3.8917,
+      "step": 5270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7524810366624528e-06,
+      "loss": 3.5371,
+      "step": 5271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752433628318584e-06,
+      "loss": 3.7526,
+      "step": 5272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7523862199747153e-06,
+      "loss": 3.7918,
+      "step": 5273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752338811630847e-06,
+      "loss": 3.4915,
+      "step": 5274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7522914032869787e-06,
+      "loss": 3.7166,
+      "step": 5275
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.75224399494311e-06,
+      "loss": 3.724,
+      "step": 5276
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7521965865992417e-06,
+      "loss": 3.8577,
+      "step": 5277
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752149178255373e-06,
+      "loss": 3.7649,
+      "step": 5278
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7521017699115043e-06,
+      "loss": 4.3014,
+      "step": 5279
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.752054361567636e-06,
+      "loss": 4.1177,
+      "step": 5280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7520069532237677e-06,
+      "loss": 3.444,
+      "step": 5281
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.751959544879899e-06,
+      "loss": 4.2278,
+      "step": 5282
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7519121365360307e-06,
+      "loss": 3.4957,
+      "step": 5283
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.751864728192162e-06,
+      "loss": 3.8329,
+      "step": 5284
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7518173198482932e-06,
+      "loss": 3.4934,
+      "step": 5285
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.7517699115044245e-06,
+      "loss": 3.8577,
+      "step": 5286
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.751722503160556e-06,
+      "loss": 3.647,
+      "step": 5287
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.751675094816688e-06,
+      "loss": 4.2109,
+      "step": 5288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751627686472819e-06,
+      "loss": 3.782,
+      "step": 5289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751580278128951e-06,
+      "loss": 3.3092,
+      "step": 5290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751532869785082e-06,
+      "loss": 3.3107,
+      "step": 5291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7514854614412135e-06,
+      "loss": 3.4186,
+      "step": 5292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751438053097345e-06,
+      "loss": 3.7471,
+      "step": 5293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751390644753477e-06,
+      "loss": 4.0466,
+      "step": 5294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751343236409608e-06,
+      "loss": 3.9676,
+      "step": 5295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.75129582806574e-06,
+      "loss": 3.848,
+      "step": 5296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751248419721871e-06,
+      "loss": 3.7726,
+      "step": 5297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7512010113780024e-06,
+      "loss": 4.0338,
+      "step": 5298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751153603034134e-06,
+      "loss": 3.6192,
+      "step": 5299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7511061946902654e-06,
+      "loss": 3.2267,
+      "step": 5300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751058786346397e-06,
+      "loss": 3.7091,
+      "step": 5301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.751011378002529e-06,
+      "loss": 3.8568,
+      "step": 5302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.75096396965866e-06,
+      "loss": 3.9647,
+      "step": 5303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7509165613147914e-06,
+      "loss": 3.7598,
+      "step": 5304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.750869152970923e-06,
+      "loss": 3.5724,
+      "step": 5305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7508217446270543e-06,
+      "loss": 3.8929,
+      "step": 5306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.750774336283186e-06,
+      "loss": 3.8123,
+      "step": 5307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7507269279393173e-06,
+      "loss": 3.7277,
+      "step": 5308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.750679519595449e-06,
+      "loss": 3.6679,
+      "step": 5309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7506321112515803e-06,
+      "loss": 3.9918,
+      "step": 5310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7505847029077116e-06,
+      "loss": 4.0705,
+      "step": 5311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7505372945638433e-06,
+      "loss": 3.8341,
+      "step": 5312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7504898862199746e-06,
+      "loss": 4.1089,
+      "step": 5313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7504424778761063e-06,
+      "loss": 3.9377,
+      "step": 5314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.750395069532238e-06,
+      "loss": 3.3426,
+      "step": 5315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7503476611883693e-06,
+      "loss": 4.0647,
+      "step": 5316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7503002528445005e-06,
+      "loss": 3.6791,
+      "step": 5317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7502528445006323e-06,
+      "loss": 3.5705,
+      "step": 5318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7502054361567635e-06,
+      "loss": 4.0937,
+      "step": 5319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7501580278128952e-06,
+      "loss": 3.8665,
+      "step": 5320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.750110619469027e-06,
+      "loss": 3.7793,
+      "step": 5321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7500632111251582e-06,
+      "loss": 3.809,
+      "step": 5322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7500158027812895e-06,
+      "loss": 3.6823,
+      "step": 5323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.749968394437421e-06,
+      "loss": 4.0678,
+      "step": 5324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7499209860935525e-06,
+      "loss": 2.9999,
+      "step": 5325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7498735777496838e-06,
+      "loss": 3.5158,
+      "step": 5326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7498261694058155e-06,
+      "loss": 3.9519,
+      "step": 5327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.749778761061947e-06,
+      "loss": 3.628,
+      "step": 5328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7497313527180785e-06,
+      "loss": 4.1948,
+      "step": 5329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7496839443742097e-06,
+      "loss": 3.8222,
+      "step": 5330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7496365360303414e-06,
+      "loss": 3.301,
+      "step": 5331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7495891276864727e-06,
+      "loss": 3.8378,
+      "step": 5332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7495417193426044e-06,
+      "loss": 4.3019,
+      "step": 5333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.749494310998736e-06,
+      "loss": 4.1026,
+      "step": 5334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7494469026548674e-06,
+      "loss": 3.9076,
+      "step": 5335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7493994943109987e-06,
+      "loss": 3.6126,
+      "step": 5336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7493520859671304e-06,
+      "loss": 4.0743,
+      "step": 5337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7493046776232617e-06,
+      "loss": 4.0075,
+      "step": 5338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.749257269279393e-06,
+      "loss": 4.0649,
+      "step": 5339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7492098609355247e-06,
+      "loss": 3.8805,
+      "step": 5340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7491624525916564e-06,
+      "loss": 4.0785,
+      "step": 5341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7491150442477876e-06,
+      "loss": 3.6953,
+      "step": 5342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7490676359039193e-06,
+      "loss": 4.1617,
+      "step": 5343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7490202275600506e-06,
+      "loss": 4.0474,
+      "step": 5344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.748972819216182e-06,
+      "loss": 4.1491,
+      "step": 5345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7489254108723136e-06,
+      "loss": 3.5278,
+      "step": 5346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7488780025284453e-06,
+      "loss": 4.1338,
+      "step": 5347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7488305941845766e-06,
+      "loss": 3.8368,
+      "step": 5348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.748783185840708e-06,
+      "loss": 4.014,
+      "step": 5349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.7487357774968396e-06,
+      "loss": 3.6144,
+      "step": 5350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.748688369152971e-06,
+      "loss": 3.6266,
+      "step": 5351
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.748640960809102e-06,
+      "loss": 3.3171,
+      "step": 5352
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.748593552465234e-06,
+      "loss": 4.1047,
+      "step": 5353
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7485461441213655e-06,
+      "loss": 3.4974,
+      "step": 5354
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.748498735777497e-06,
+      "loss": 3.8295,
+      "step": 5355
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7484513274336285e-06,
+      "loss": 3.9057,
+      "step": 5356
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.74840391908976e-06,
+      "loss": 4.0383,
+      "step": 5357
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.748356510745891e-06,
+      "loss": 3.8299,
+      "step": 5358
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.748309102402023e-06,
+      "loss": 3.5115,
+      "step": 5359
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7482616940581545e-06,
+      "loss": 4.2335,
+      "step": 5360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7482142857142858e-06,
+      "loss": 3.7334,
+      "step": 5361
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7481668773704175e-06,
+      "loss": 3.8835,
+      "step": 5362
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7481194690265488e-06,
+      "loss": 4.0781,
+      "step": 5363
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.74807206068268e-06,
+      "loss": 3.92,
+      "step": 5364
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7480246523388117e-06,
+      "loss": 4.0072,
+      "step": 5365
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747977243994943e-06,
+      "loss": 3.8089,
+      "step": 5366
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7479298356510747e-06,
+      "loss": 3.7719,
+      "step": 5367
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7478824273072064e-06,
+      "loss": 3.9671,
+      "step": 5368
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7478350189633377e-06,
+      "loss": 4.002,
+      "step": 5369
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747787610619469e-06,
+      "loss": 3.9288,
+      "step": 5370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7477402022756003e-06,
+      "loss": 4.4297,
+      "step": 5371
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747692793931732e-06,
+      "loss": 3.8652,
+      "step": 5372
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7476453855878637e-06,
+      "loss": 3.8267,
+      "step": 5373
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747597977243995e-06,
+      "loss": 3.7736,
+      "step": 5374
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7475505689001267e-06,
+      "loss": 3.8309,
+      "step": 5375
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747503160556258e-06,
+      "loss": 4.0531,
+      "step": 5376
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7474557522123892e-06,
+      "loss": 3.8465,
+      "step": 5377
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747408343868521e-06,
+      "loss": 4.0243,
+      "step": 5378
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7473609355246522e-06,
+      "loss": 3.7007,
+      "step": 5379
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747313527180784e-06,
+      "loss": 4.0718,
+      "step": 5380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7472661188369156e-06,
+      "loss": 3.585,
+      "step": 5381
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747218710493047e-06,
+      "loss": 3.9905,
+      "step": 5382
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747171302149178e-06,
+      "loss": 4.1751,
+      "step": 5383
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.74712389380531e-06,
+      "loss": 3.5498,
+      "step": 5384
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747076485461441e-06,
+      "loss": 3.9798,
+      "step": 5385
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.747029077117573e-06,
+      "loss": 4.0033,
+      "step": 5386
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7469816687737046e-06,
+      "loss": 3.9169,
+      "step": 5387
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746934260429836e-06,
+      "loss": 3.8996,
+      "step": 5388
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746886852085967e-06,
+      "loss": 4.3685,
+      "step": 5389
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746839443742099e-06,
+      "loss": 3.4584,
+      "step": 5390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.74679203539823e-06,
+      "loss": 4.134,
+      "step": 5391
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7467446270543614e-06,
+      "loss": 3.9758,
+      "step": 5392
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746697218710493e-06,
+      "loss": 3.8337,
+      "step": 5393
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746649810366625e-06,
+      "loss": 3.6037,
+      "step": 5394
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746602402022756e-06,
+      "loss": 4.386,
+      "step": 5395
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7465549936788874e-06,
+      "loss": 3.6452,
+      "step": 5396
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746507585335019e-06,
+      "loss": 3.6665,
+      "step": 5397
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7464601769911504e-06,
+      "loss": 3.5825,
+      "step": 5398
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746412768647282e-06,
+      "loss": 3.3021,
+      "step": 5399
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7463653603034138e-06,
+      "loss": 3.6195,
+      "step": 5400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746317951959545e-06,
+      "loss": 4.0112,
+      "step": 5401
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7462705436156763e-06,
+      "loss": 3.7149,
+      "step": 5402
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746223135271808e-06,
+      "loss": 3.9844,
+      "step": 5403
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7461757269279393e-06,
+      "loss": 3.8237,
+      "step": 5404
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7461283185840706e-06,
+      "loss": 3.5581,
+      "step": 5405
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7460809102402023e-06,
+      "loss": 3.7582,
+      "step": 5406
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.746033501896334e-06,
+      "loss": 3.6976,
+      "step": 5407
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7459860935524653e-06,
+      "loss": 4.1703,
+      "step": 5408
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.745938685208597e-06,
+      "loss": 3.21,
+      "step": 5409
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7458912768647283e-06,
+      "loss": 4.1725,
+      "step": 5410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7458438685208595e-06,
+      "loss": 3.3745,
+      "step": 5411
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7457964601769912e-06,
+      "loss": 3.8136,
+      "step": 5412
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.745749051833123e-06,
+      "loss": 4.3114,
+      "step": 5413
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7457016434892542e-06,
+      "loss": 3.5691,
+      "step": 5414
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7456542351453855e-06,
+      "loss": 3.9229,
+      "step": 5415
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7456068268015172e-06,
+      "loss": 4.0446,
+      "step": 5416
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7455594184576485e-06,
+      "loss": 3.7161,
+      "step": 5417
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7455120101137798e-06,
+      "loss": 3.7238,
+      "step": 5418
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7454646017699115e-06,
+      "loss": 4.0293,
+      "step": 5419
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.745417193426043e-06,
+      "loss": 3.8745,
+      "step": 5420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7453697850821745e-06,
+      "loss": 3.4926,
+      "step": 5421
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.745322376738306e-06,
+      "loss": 3.6072,
+      "step": 5422
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7452749683944374e-06,
+      "loss": 3.9075,
+      "step": 5423
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7452275600505687e-06,
+      "loss": 4.0847,
+      "step": 5424
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7451801517067004e-06,
+      "loss": 4.0095,
+      "step": 5425
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.745132743362832e-06,
+      "loss": 3.6568,
+      "step": 5426
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7450853350189634e-06,
+      "loss": 4.0546,
+      "step": 5427
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.745037926675095e-06,
+      "loss": 3.9685,
+      "step": 5428
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7449905183312264e-06,
+      "loss": 3.6091,
+      "step": 5429
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7449431099873577e-06,
+      "loss": 4.083,
+      "step": 5430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7448957016434894e-06,
+      "loss": 4.2175,
+      "step": 5431
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7448482932996207e-06,
+      "loss": 3.6966,
+      "step": 5432
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7448008849557524e-06,
+      "loss": 3.5394,
+      "step": 5433
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.744753476611884e-06,
+      "loss": 4.1334,
+      "step": 5434
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7447060682680154e-06,
+      "loss": 4.2289,
+      "step": 5435
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7446586599241466e-06,
+      "loss": 3.6813,
+      "step": 5436
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.744611251580278e-06,
+      "loss": 3.7754,
+      "step": 5437
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7445638432364096e-06,
+      "loss": 4.1553,
+      "step": 5438
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7445164348925413e-06,
+      "loss": 3.6812,
+      "step": 5439
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7444690265486726e-06,
+      "loss": 3.8466,
+      "step": 5440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7444216182048043e-06,
+      "loss": 3.7738,
+      "step": 5441
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7443742098609356e-06,
+      "loss": 3.9224,
+      "step": 5442
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.744326801517067e-06,
+      "loss": 3.246,
+      "step": 5443
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7442793931731986e-06,
+      "loss": 3.4359,
+      "step": 5444
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.74423198482933e-06,
+      "loss": 3.9275,
+      "step": 5445
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7441845764854616e-06,
+      "loss": 3.9374,
+      "step": 5446
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7441371681415933e-06,
+      "loss": 3.9803,
+      "step": 5447
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7440897597977245e-06,
+      "loss": 3.7358,
+      "step": 5448
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.744042351453856e-06,
+      "loss": 3.9143,
+      "step": 5449
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7439949431099875e-06,
+      "loss": 3.5727,
+      "step": 5450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.743947534766119e-06,
+      "loss": 4.0592,
+      "step": 5451
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7439001264222505e-06,
+      "loss": 3.8985,
+      "step": 5452
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7438527180783822e-06,
+      "loss": 4.0572,
+      "step": 5453
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7438053097345135e-06,
+      "loss": 4.0137,
+      "step": 5454
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7437579013906448e-06,
+      "loss": 4.1755,
+      "step": 5455
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7437104930467765e-06,
+      "loss": 3.606,
+      "step": 5456
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7436630847029078e-06,
+      "loss": 3.6713,
+      "step": 5457
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.743615676359039e-06,
+      "loss": 3.8074,
+      "step": 5458
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7435682680151707e-06,
+      "loss": 4.0015,
+      "step": 5459
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7435208596713024e-06,
+      "loss": 3.4016,
+      "step": 5460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7434734513274337e-06,
+      "loss": 3.5956,
+      "step": 5461
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.743426042983565e-06,
+      "loss": 4.213,
+      "step": 5462
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7433786346396967e-06,
+      "loss": 4.2211,
+      "step": 5463
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.743331226295828e-06,
+      "loss": 3.9253,
+      "step": 5464
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7432838179519593e-06,
+      "loss": 3.6624,
+      "step": 5465
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7432364096080914e-06,
+      "loss": 3.829,
+      "step": 5466
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7431890012642227e-06,
+      "loss": 3.657,
+      "step": 5467
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.743141592920354e-06,
+      "loss": 3.8727,
+      "step": 5468
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7430941845764857e-06,
+      "loss": 3.8431,
+      "step": 5469
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.743046776232617e-06,
+      "loss": 3.8209,
+      "step": 5470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7429993678887482e-06,
+      "loss": 3.9417,
+      "step": 5471
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.74295195954488e-06,
+      "loss": 3.9029,
+      "step": 5472
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7429045512010116e-06,
+      "loss": 3.7076,
+      "step": 5473
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.742857142857143e-06,
+      "loss": 3.642,
+      "step": 5474
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.7428097345132746e-06,
+      "loss": 3.8846,
+      "step": 5475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.742762326169406e-06,
+      "loss": 3.6819,
+      "step": 5476
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.742714917825537e-06,
+      "loss": 3.5169,
+      "step": 5477
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.742667509481669e-06,
+      "loss": 4.3221,
+      "step": 5478
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7426201011378e-06,
+      "loss": 3.6426,
+      "step": 5479
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742572692793932e-06,
+      "loss": 4.1178,
+      "step": 5480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742525284450063e-06,
+      "loss": 4.0327,
+      "step": 5481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742477876106195e-06,
+      "loss": 3.9859,
+      "step": 5482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742430467762326e-06,
+      "loss": 4.1252,
+      "step": 5483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7423830594184574e-06,
+      "loss": 3.9865,
+      "step": 5484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742335651074589e-06,
+      "loss": 3.6908,
+      "step": 5485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742288242730721e-06,
+      "loss": 3.5323,
+      "step": 5486
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742240834386852e-06,
+      "loss": 3.7107,
+      "step": 5487
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742193426042984e-06,
+      "loss": 3.8252,
+      "step": 5488
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742146017699115e-06,
+      "loss": 4.0656,
+      "step": 5489
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7420986093552464e-06,
+      "loss": 3.4899,
+      "step": 5490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.742051201011378e-06,
+      "loss": 3.7915,
+      "step": 5491
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7420037926675094e-06,
+      "loss": 3.9498,
+      "step": 5492
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741956384323641e-06,
+      "loss": 4.0522,
+      "step": 5493
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7419089759797728e-06,
+      "loss": 4.0545,
+      "step": 5494
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741861567635904e-06,
+      "loss": 3.7362,
+      "step": 5495
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7418141592920353e-06,
+      "loss": 3.9872,
+      "step": 5496
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741766750948167e-06,
+      "loss": 3.5901,
+      "step": 5497
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7417193426042983e-06,
+      "loss": 3.8402,
+      "step": 5498
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.74167193426043e-06,
+      "loss": 3.6831,
+      "step": 5499
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7416245259165617e-06,
+      "loss": 3.3855,
+      "step": 5500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741577117572693e-06,
+      "loss": 3.8999,
+      "step": 5501
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7415297092288243e-06,
+      "loss": 3.7029,
+      "step": 5502
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7414823008849556e-06,
+      "loss": 3.7524,
+      "step": 5503
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7414348925410873e-06,
+      "loss": 3.7928,
+      "step": 5504
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7413874841972185e-06,
+      "loss": 3.2197,
+      "step": 5505
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7413400758533502e-06,
+      "loss": 3.7681,
+      "step": 5506
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741292667509482e-06,
+      "loss": 4.105,
+      "step": 5507
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7412452591656132e-06,
+      "loss": 3.869,
+      "step": 5508
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7411978508217445e-06,
+      "loss": 3.3439,
+      "step": 5509
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741150442477876e-06,
+      "loss": 4.2343,
+      "step": 5510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7411030341340075e-06,
+      "loss": 3.1755,
+      "step": 5511
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741055625790139e-06,
+      "loss": 3.759,
+      "step": 5512
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.741008217446271e-06,
+      "loss": 3.7436,
+      "step": 5513
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.740960809102402e-06,
+      "loss": 3.9341,
+      "step": 5514
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7409134007585335e-06,
+      "loss": 3.43,
+      "step": 5515
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.740865992414665e-06,
+      "loss": 3.948,
+      "step": 5516
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7408185840707964e-06,
+      "loss": 3.6648,
+      "step": 5517
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7407711757269277e-06,
+      "loss": 3.4589,
+      "step": 5518
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7407237673830594e-06,
+      "loss": 4.4697,
+      "step": 5519
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.740676359039191e-06,
+      "loss": 3.8095,
+      "step": 5520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7406289506953224e-06,
+      "loss": 3.8635,
+      "step": 5521
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.740581542351454e-06,
+      "loss": 3.5257,
+      "step": 5522
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7405341340075854e-06,
+      "loss": 4.2245,
+      "step": 5523
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7404867256637167e-06,
+      "loss": 3.5168,
+      "step": 5524
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7404393173198484e-06,
+      "loss": 3.6754,
+      "step": 5525
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.74039190897598e-06,
+      "loss": 3.8756,
+      "step": 5526
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7403445006321114e-06,
+      "loss": 3.997,
+      "step": 5527
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7402970922882426e-06,
+      "loss": 4.0555,
+      "step": 5528
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7402496839443743e-06,
+      "loss": 4.1277,
+      "step": 5529
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7402022756005056e-06,
+      "loss": 3.3698,
+      "step": 5530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.740154867256637e-06,
+      "loss": 3.6601,
+      "step": 5531
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7401074589127686e-06,
+      "loss": 3.8729,
+      "step": 5532
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7400600505689003e-06,
+      "loss": 3.8089,
+      "step": 5533
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7400126422250316e-06,
+      "loss": 3.3086,
+      "step": 5534
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7399652338811633e-06,
+      "loss": 4.0605,
+      "step": 5535
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7399178255372946e-06,
+      "loss": 3.54,
+      "step": 5536
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.739870417193426e-06,
+      "loss": 3.7702,
+      "step": 5537
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7398230088495576e-06,
+      "loss": 4.1028,
+      "step": 5538
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7397756005056893e-06,
+      "loss": 4.2023,
+      "step": 5539
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7397281921618206e-06,
+      "loss": 4.142,
+      "step": 5540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.7396807838179523e-06,
+      "loss": 4.1108,
+      "step": 5541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7396333754740835e-06,
+      "loss": 4.0404,
+      "step": 5542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.739585967130215e-06,
+      "loss": 4.2297,
+      "step": 5543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7395385587863465e-06,
+      "loss": 3.7112,
+      "step": 5544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.739491150442478e-06,
+      "loss": 3.4175,
+      "step": 5545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7394437420986095e-06,
+      "loss": 3.8399,
+      "step": 5546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7393963337547408e-06,
+      "loss": 3.6604,
+      "step": 5547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7393489254108725e-06,
+      "loss": 4.2596,
+      "step": 5548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7393015170670038e-06,
+      "loss": 3.1943,
+      "step": 5549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.739254108723135e-06,
+      "loss": 4.0246,
+      "step": 5550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7392067003792668e-06,
+      "loss": 3.0864,
+      "step": 5551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7391592920353985e-06,
+      "loss": 3.2371,
+      "step": 5552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7391118836915297e-06,
+      "loss": 3.249,
+      "step": 5553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7390644753476614e-06,
+      "loss": 3.9567,
+      "step": 5554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7390170670037927e-06,
+      "loss": 3.8674,
+      "step": 5555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738969658659924e-06,
+      "loss": 3.6141,
+      "step": 5556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7389222503160557e-06,
+      "loss": 3.7449,
+      "step": 5557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738874841972187e-06,
+      "loss": 3.6683,
+      "step": 5558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7388274336283187e-06,
+      "loss": 3.4429,
+      "step": 5559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7387800252844504e-06,
+      "loss": 3.794,
+      "step": 5560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7387326169405817e-06,
+      "loss": 3.4142,
+      "step": 5561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738685208596713e-06,
+      "loss": 3.6638,
+      "step": 5562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7386378002528447e-06,
+      "loss": 3.7391,
+      "step": 5563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738590391908976e-06,
+      "loss": 3.5936,
+      "step": 5564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7385429835651076e-06,
+      "loss": 4.0141,
+      "step": 5565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7384955752212393e-06,
+      "loss": 3.1016,
+      "step": 5566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7384481668773706e-06,
+      "loss": 4.0198,
+      "step": 5567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738400758533502e-06,
+      "loss": 3.6981,
+      "step": 5568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738353350189633e-06,
+      "loss": 4.3162,
+      "step": 5569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738305941845765e-06,
+      "loss": 4.1612,
+      "step": 5570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738258533501896e-06,
+      "loss": 3.8575,
+      "step": 5571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738211125158028e-06,
+      "loss": 3.6272,
+      "step": 5572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7381637168141596e-06,
+      "loss": 4.3851,
+      "step": 5573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738116308470291e-06,
+      "loss": 4.0323,
+      "step": 5574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738068900126422e-06,
+      "loss": 4.5054,
+      "step": 5575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.738021491782554e-06,
+      "loss": 3.3907,
+      "step": 5576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737974083438685e-06,
+      "loss": 3.9466,
+      "step": 5577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737926675094817e-06,
+      "loss": 4.0271,
+      "step": 5578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7378792667509485e-06,
+      "loss": 4.0135,
+      "step": 5579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.73783185840708e-06,
+      "loss": 3.5844,
+      "step": 5580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737784450063211e-06,
+      "loss": 3.8788,
+      "step": 5581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737737041719343e-06,
+      "loss": 4.3279,
+      "step": 5582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737689633375474e-06,
+      "loss": 3.4805,
+      "step": 5583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7376422250316054e-06,
+      "loss": 4.3092,
+      "step": 5584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737594816687737e-06,
+      "loss": 3.4333,
+      "step": 5585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7375474083438688e-06,
+      "loss": 3.7552,
+      "step": 5586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7375e-06,
+      "loss": 3.708,
+      "step": 5587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7374525916561318e-06,
+      "loss": 3.893,
+      "step": 5588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737405183312263e-06,
+      "loss": 3.8786,
+      "step": 5589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7373577749683943e-06,
+      "loss": 3.9142,
+      "step": 5590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737310366624526e-06,
+      "loss": 3.7545,
+      "step": 5591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7372629582806577e-06,
+      "loss": 4.0745,
+      "step": 5592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737215549936789e-06,
+      "loss": 3.7638,
+      "step": 5593
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7371681415929203e-06,
+      "loss": 3.7867,
+      "step": 5594
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.737120733249052e-06,
+      "loss": 3.9871,
+      "step": 5595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7370733249051833e-06,
+      "loss": 4.1412,
+      "step": 5596
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7370259165613145e-06,
+      "loss": 4.2233,
+      "step": 5597
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7369785082174463e-06,
+      "loss": 3.8006,
+      "step": 5598
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.736931099873578e-06,
+      "loss": 3.5919,
+      "step": 5599
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7368836915297092e-06,
+      "loss": 4.3203,
+      "step": 5600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.736836283185841e-06,
+      "loss": 3.4877,
+      "step": 5601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7367888748419722e-06,
+      "loss": 4.006,
+      "step": 5602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.7367414664981035e-06,
+      "loss": 3.7258,
+      "step": 5603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.736694058154235e-06,
+      "loss": 4.0371,
+      "step": 5604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.736646649810367e-06,
+      "loss": 3.6185,
+      "step": 5605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.736599241466498e-06,
+      "loss": 3.6056,
+      "step": 5606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.73655183312263e-06,
+      "loss": 4.1337,
+      "step": 5607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.736504424778761e-06,
+      "loss": 3.7626,
+      "step": 5608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7364570164348925e-06,
+      "loss": 4.0834,
+      "step": 5609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.736409608091024e-06,
+      "loss": 4.1222,
+      "step": 5610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7363621997471554e-06,
+      "loss": 3.9089,
+      "step": 5611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.736314791403287e-06,
+      "loss": 3.5868,
+      "step": 5612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7362673830594184e-06,
+      "loss": 3.6393,
+      "step": 5613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.73621997471555e-06,
+      "loss": 3.4617,
+      "step": 5614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7361725663716814e-06,
+      "loss": 3.6225,
+      "step": 5615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7361251580278127e-06,
+      "loss": 4.284,
+      "step": 5616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7360777496839444e-06,
+      "loss": 4.0466,
+      "step": 5617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.736030341340076e-06,
+      "loss": 3.5374,
+      "step": 5618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7359829329962074e-06,
+      "loss": 4.0766,
+      "step": 5619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.735935524652339e-06,
+      "loss": 3.5895,
+      "step": 5620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7358881163084704e-06,
+      "loss": 4.1168,
+      "step": 5621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7358407079646016e-06,
+      "loss": 4.1093,
+      "step": 5622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7357932996207333e-06,
+      "loss": 3.756,
+      "step": 5623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7357458912768646e-06,
+      "loss": 3.8734,
+      "step": 5624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7356984829329963e-06,
+      "loss": 3.7112,
+      "step": 5625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.735651074589128e-06,
+      "loss": 3.9709,
+      "step": 5626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7356036662452593e-06,
+      "loss": 3.5512,
+      "step": 5627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7355562579013906e-06,
+      "loss": 3.3931,
+      "step": 5628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7355088495575223e-06,
+      "loss": 3.8005,
+      "step": 5629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7354614412136536e-06,
+      "loss": 3.2069,
+      "step": 5630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7354140328697853e-06,
+      "loss": 3.7825,
+      "step": 5631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.735366624525917e-06,
+      "loss": 3.9333,
+      "step": 5632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7353192161820483e-06,
+      "loss": 3.9642,
+      "step": 5633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7352718078381795e-06,
+      "loss": 3.7107,
+      "step": 5634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.735224399494311e-06,
+      "loss": 4.2611,
+      "step": 5635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7351769911504425e-06,
+      "loss": 3.7595,
+      "step": 5636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.735129582806574e-06,
+      "loss": 3.9921,
+      "step": 5637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7350821744627055e-06,
+      "loss": 4.0877,
+      "step": 5638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7350347661188372e-06,
+      "loss": 4.0245,
+      "step": 5639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7349873577749685e-06,
+      "loss": 3.6669,
+      "step": 5640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7349399494310998e-06,
+      "loss": 4.03,
+      "step": 5641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7348925410872315e-06,
+      "loss": 3.9914,
+      "step": 5642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7348451327433628e-06,
+      "loss": 3.8092,
+      "step": 5643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7347977243994945e-06,
+      "loss": 4.1387,
+      "step": 5644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.734750316055626e-06,
+      "loss": 4.0683,
+      "step": 5645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7347029077117575e-06,
+      "loss": 3.9843,
+      "step": 5646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7346554993678887e-06,
+      "loss": 3.7812,
+      "step": 5647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7346080910240204e-06,
+      "loss": 4.1915,
+      "step": 5648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7345606826801517e-06,
+      "loss": 3.9051,
+      "step": 5649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.734513274336283e-06,
+      "loss": 4.0365,
+      "step": 5650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7344658659924147e-06,
+      "loss": 3.4998,
+      "step": 5651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7344184576485464e-06,
+      "loss": 3.6309,
+      "step": 5652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7343710493046777e-06,
+      "loss": 4.0447,
+      "step": 5653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7343236409608094e-06,
+      "loss": 3.6909,
+      "step": 5654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7342762326169407e-06,
+      "loss": 3.8077,
+      "step": 5655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.734228824273072e-06,
+      "loss": 3.6895,
+      "step": 5656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7341814159292037e-06,
+      "loss": 3.8487,
+      "step": 5657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7341340075853354e-06,
+      "loss": 3.7695,
+      "step": 5658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7340865992414666e-06,
+      "loss": 3.7921,
+      "step": 5659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.734039190897598e-06,
+      "loss": 3.9123,
+      "step": 5660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7339917825537296e-06,
+      "loss": 3.8485,
+      "step": 5661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.733944374209861e-06,
+      "loss": 3.3329,
+      "step": 5662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.733896965865992e-06,
+      "loss": 3.9552,
+      "step": 5663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.733849557522124e-06,
+      "loss": 3.5173,
+      "step": 5664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7338021491782556e-06,
+      "loss": 3.8816,
+      "step": 5665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.733754740834387e-06,
+      "loss": 3.8783,
+      "step": 5666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.7337073324905186e-06,
+      "loss": 3.2961,
+      "step": 5667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.73365992414665e-06,
+      "loss": 3.1875,
+      "step": 5668
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733612515802781e-06,
+      "loss": 4.0661,
+      "step": 5669
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733565107458913e-06,
+      "loss": 4.164,
+      "step": 5670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7335176991150445e-06,
+      "loss": 3.7811,
+      "step": 5671
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733470290771176e-06,
+      "loss": 3.8891,
+      "step": 5672
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7334228824273075e-06,
+      "loss": 3.7759,
+      "step": 5673
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733375474083439e-06,
+      "loss": 3.7744,
+      "step": 5674
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73332806573957e-06,
+      "loss": 4.3181,
+      "step": 5675
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733280657395702e-06,
+      "loss": 3.6358,
+      "step": 5676
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733233249051833e-06,
+      "loss": 3.6432,
+      "step": 5677
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7331858407079648e-06,
+      "loss": 3.9822,
+      "step": 5678
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733138432364096e-06,
+      "loss": 4.3111,
+      "step": 5679
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7330910240202278e-06,
+      "loss": 3.9419,
+      "step": 5680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.733043615676359e-06,
+      "loss": 3.542,
+      "step": 5681
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7329962073324903e-06,
+      "loss": 3.5871,
+      "step": 5682
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732948798988622e-06,
+      "loss": 3.3511,
+      "step": 5683
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7329013906447537e-06,
+      "loss": 3.8691,
+      "step": 5684
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732853982300885e-06,
+      "loss": 3.6659,
+      "step": 5685
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7328065739570167e-06,
+      "loss": 3.7108,
+      "step": 5686
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732759165613148e-06,
+      "loss": 3.5008,
+      "step": 5687
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7327117572692793e-06,
+      "loss": 3.7189,
+      "step": 5688
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732664348925411e-06,
+      "loss": 3.7908,
+      "step": 5689
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7326169405815423e-06,
+      "loss": 3.6753,
+      "step": 5690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732569532237674e-06,
+      "loss": 3.871,
+      "step": 5691
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7325221238938057e-06,
+      "loss": 3.8113,
+      "step": 5692
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732474715549937e-06,
+      "loss": 3.6968,
+      "step": 5693
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7324273072060682e-06,
+      "loss": 3.8683,
+      "step": 5694
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7323798988622e-06,
+      "loss": 3.679,
+      "step": 5695
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7323324905183312e-06,
+      "loss": 3.2113,
+      "step": 5696
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732285082174463e-06,
+      "loss": 3.59,
+      "step": 5697
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7322376738305946e-06,
+      "loss": 4.0883,
+      "step": 5698
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732190265486726e-06,
+      "loss": 4.0608,
+      "step": 5699
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.732142857142857e-06,
+      "loss": 3.8527,
+      "step": 5700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7320954487989885e-06,
+      "loss": 4.126,
+      "step": 5701
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73204804045512e-06,
+      "loss": 4.2553,
+      "step": 5702
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7320006321112514e-06,
+      "loss": 3.8547,
+      "step": 5703
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731953223767383e-06,
+      "loss": 3.0147,
+      "step": 5704
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731905815423515e-06,
+      "loss": 3.8954,
+      "step": 5705
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731858407079646e-06,
+      "loss": 4.0184,
+      "step": 5706
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7318109987357774e-06,
+      "loss": 3.6992,
+      "step": 5707
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731763590391909e-06,
+      "loss": 3.8729,
+      "step": 5708
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7317161820480404e-06,
+      "loss": 3.6938,
+      "step": 5709
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731668773704172e-06,
+      "loss": 4.1153,
+      "step": 5710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731621365360304e-06,
+      "loss": 3.5566,
+      "step": 5711
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731573957016435e-06,
+      "loss": 3.3575,
+      "step": 5712
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7315265486725664e-06,
+      "loss": 3.77,
+      "step": 5713
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731479140328698e-06,
+      "loss": 3.8951,
+      "step": 5714
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7314317319848294e-06,
+      "loss": 3.4504,
+      "step": 5715
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7313843236409606e-06,
+      "loss": 3.3087,
+      "step": 5716
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7313369152970923e-06,
+      "loss": 3.5077,
+      "step": 5717
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731289506953224e-06,
+      "loss": 3.8967,
+      "step": 5718
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7312420986093553e-06,
+      "loss": 3.8835,
+      "step": 5719
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731194690265487e-06,
+      "loss": 4.2437,
+      "step": 5720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7311472819216183e-06,
+      "loss": 3.5958,
+      "step": 5721
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7310998735777496e-06,
+      "loss": 3.9999,
+      "step": 5722
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7310524652338813e-06,
+      "loss": 3.613,
+      "step": 5723
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.731005056890013e-06,
+      "loss": 3.3016,
+      "step": 5724
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7309576485461443e-06,
+      "loss": 3.7811,
+      "step": 5725
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7309102402022756e-06,
+      "loss": 3.7671,
+      "step": 5726
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7308628318584073e-06,
+      "loss": 3.4581,
+      "step": 5727
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7308154235145385e-06,
+      "loss": 4.106,
+      "step": 5728
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.73076801517067e-06,
+      "loss": 3.8898,
+      "step": 5729
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7307206068268015e-06,
+      "loss": 3.7321,
+      "step": 5730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.7306731984829332e-06,
+      "loss": 4.0308,
+      "step": 5731
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7306257901390645e-06,
+      "loss": 4.0623,
+      "step": 5732
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7305783817951962e-06,
+      "loss": 4.383,
+      "step": 5733
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7305309734513275e-06,
+      "loss": 4.2249,
+      "step": 5734
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7304835651074588e-06,
+      "loss": 3.956,
+      "step": 5735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7304361567635905e-06,
+      "loss": 3.8691,
+      "step": 5736
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.730388748419722e-06,
+      "loss": 3.3833,
+      "step": 5737
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7303413400758535e-06,
+      "loss": 3.5739,
+      "step": 5738
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.730293931731985e-06,
+      "loss": 3.7074,
+      "step": 5739
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7302465233881164e-06,
+      "loss": 3.8563,
+      "step": 5740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7301991150442477e-06,
+      "loss": 4.345,
+      "step": 5741
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7301517067003794e-06,
+      "loss": 3.6024,
+      "step": 5742
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7301042983565107e-06,
+      "loss": 4.0981,
+      "step": 5743
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7300568900126424e-06,
+      "loss": 3.3361,
+      "step": 5744
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7300094816687737e-06,
+      "loss": 3.8267,
+      "step": 5745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7299620733249054e-06,
+      "loss": 3.9664,
+      "step": 5746
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7299146649810367e-06,
+      "loss": 3.8385,
+      "step": 5747
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729867256637168e-06,
+      "loss": 3.7232,
+      "step": 5748
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7298198482932997e-06,
+      "loss": 3.7521,
+      "step": 5749
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729772439949431e-06,
+      "loss": 3.7845,
+      "step": 5750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7297250316055626e-06,
+      "loss": 4.2057,
+      "step": 5751
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7296776232616944e-06,
+      "loss": 3.9477,
+      "step": 5752
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7296302149178256e-06,
+      "loss": 3.7301,
+      "step": 5753
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729582806573957e-06,
+      "loss": 3.754,
+      "step": 5754
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7295353982300886e-06,
+      "loss": 3.9194,
+      "step": 5755
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.72948798988622e-06,
+      "loss": 3.3139,
+      "step": 5756
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7294405815423516e-06,
+      "loss": 4.0736,
+      "step": 5757
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7293931731984833e-06,
+      "loss": 4.1007,
+      "step": 5758
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7293457648546146e-06,
+      "loss": 4.0334,
+      "step": 5759
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729298356510746e-06,
+      "loss": 4.0916,
+      "step": 5760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7292509481668776e-06,
+      "loss": 4.162,
+      "step": 5761
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729203539823009e-06,
+      "loss": 3.5097,
+      "step": 5762
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.72915613147914e-06,
+      "loss": 3.8172,
+      "step": 5763
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729108723135272e-06,
+      "loss": 3.6626,
+      "step": 5764
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7290613147914035e-06,
+      "loss": 4.1312,
+      "step": 5765
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.729013906447535e-06,
+      "loss": 3.8106,
+      "step": 5766
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728966498103666e-06,
+      "loss": 3.7051,
+      "step": 5767
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728919089759798e-06,
+      "loss": 4.2509,
+      "step": 5768
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728871681415929e-06,
+      "loss": 3.3331,
+      "step": 5769
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7288242730720608e-06,
+      "loss": 3.5723,
+      "step": 5770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7287768647281925e-06,
+      "loss": 3.7841,
+      "step": 5771
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7287294563843238e-06,
+      "loss": 3.6238,
+      "step": 5772
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728682048040455e-06,
+      "loss": 3.9403,
+      "step": 5773
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7286346396965868e-06,
+      "loss": 3.9769,
+      "step": 5774
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728587231352718e-06,
+      "loss": 3.51,
+      "step": 5775
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7285398230088493e-06,
+      "loss": 3.3491,
+      "step": 5776
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728492414664981e-06,
+      "loss": 3.7644,
+      "step": 5777
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7284450063211127e-06,
+      "loss": 3.7635,
+      "step": 5778
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728397597977244e-06,
+      "loss": 3.7306,
+      "step": 5779
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7283501896333757e-06,
+      "loss": 4.0192,
+      "step": 5780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728302781289507e-06,
+      "loss": 4.2507,
+      "step": 5781
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7282553729456383e-06,
+      "loss": 3.6766,
+      "step": 5782
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.72820796460177e-06,
+      "loss": 3.8461,
+      "step": 5783
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7281605562579017e-06,
+      "loss": 3.639,
+      "step": 5784
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728113147914033e-06,
+      "loss": 3.6766,
+      "step": 5785
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7280657395701642e-06,
+      "loss": 4.0211,
+      "step": 5786
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.728018331226296e-06,
+      "loss": 3.8615,
+      "step": 5787
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7279709228824272e-06,
+      "loss": 3.9187,
+      "step": 5788
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.7279235145385585e-06,
+      "loss": 3.6146,
+      "step": 5789
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.72787610619469e-06,
+      "loss": 3.9555,
+      "step": 5790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.727828697850822e-06,
+      "loss": 3.7135,
+      "step": 5791
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.727781289506953e-06,
+      "loss": 4.1323,
+      "step": 5792
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.727733881163085e-06,
+      "loss": 3.3884,
+      "step": 5793
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.727686472819216e-06,
+      "loss": 3.6023,
+      "step": 5794
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7276390644753475e-06,
+      "loss": 3.8336,
+      "step": 5795
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727591656131479e-06,
+      "loss": 3.8815,
+      "step": 5796
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727544247787611e-06,
+      "loss": 3.8649,
+      "step": 5797
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727496839443742e-06,
+      "loss": 3.316,
+      "step": 5798
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727449431099874e-06,
+      "loss": 4.2389,
+      "step": 5799
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727402022756005e-06,
+      "loss": 3.4862,
+      "step": 5800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7273546144121364e-06,
+      "loss": 3.6866,
+      "step": 5801
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727307206068268e-06,
+      "loss": 3.9171,
+      "step": 5802
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7272597977243994e-06,
+      "loss": 3.6384,
+      "step": 5803
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727212389380531e-06,
+      "loss": 3.7991,
+      "step": 5804
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727164981036663e-06,
+      "loss": 3.662,
+      "step": 5805
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.727117572692794e-06,
+      "loss": 3.9565,
+      "step": 5806
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7270701643489254e-06,
+      "loss": 3.9845,
+      "step": 5807
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7270227560050566e-06,
+      "loss": 3.9372,
+      "step": 5808
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7269753476611883e-06,
+      "loss": 3.6479,
+      "step": 5809
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.72692793931732e-06,
+      "loss": 3.1249,
+      "step": 5810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7268805309734513e-06,
+      "loss": 3.4404,
+      "step": 5811
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.726833122629583e-06,
+      "loss": 3.2178,
+      "step": 5812
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7267857142857143e-06,
+      "loss": 3.4679,
+      "step": 5813
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7267383059418456e-06,
+      "loss": 3.8622,
+      "step": 5814
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7266908975979773e-06,
+      "loss": 4.0177,
+      "step": 5815
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7266434892541086e-06,
+      "loss": 3.1181,
+      "step": 5816
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7265960809102403e-06,
+      "loss": 3.7012,
+      "step": 5817
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.726548672566372e-06,
+      "loss": 3.7904,
+      "step": 5818
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7265012642225033e-06,
+      "loss": 4.2522,
+      "step": 5819
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7264538558786345e-06,
+      "loss": 4.1339,
+      "step": 5820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7264064475347663e-06,
+      "loss": 3.9048,
+      "step": 5821
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7263590391908975e-06,
+      "loss": 3.7741,
+      "step": 5822
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7263116308470292e-06,
+      "loss": 3.5992,
+      "step": 5823
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.726264222503161e-06,
+      "loss": 3.9501,
+      "step": 5824
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7262168141592922e-06,
+      "loss": 3.3105,
+      "step": 5825
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7261694058154235e-06,
+      "loss": 3.5525,
+      "step": 5826
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.726121997471555e-06,
+      "loss": 4.2314,
+      "step": 5827
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7260745891276865e-06,
+      "loss": 3.5187,
+      "step": 5828
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7260271807838178e-06,
+      "loss": 4.1918,
+      "step": 5829
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7259797724399495e-06,
+      "loss": 4.3229,
+      "step": 5830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.725932364096081e-06,
+      "loss": 3.8861,
+      "step": 5831
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7258849557522125e-06,
+      "loss": 3.6055,
+      "step": 5832
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7258375474083437e-06,
+      "loss": 3.7647,
+      "step": 5833
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7257901390644754e-06,
+      "loss": 4.2812,
+      "step": 5834
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7257427307206067e-06,
+      "loss": 3.8111,
+      "step": 5835
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7256953223767384e-06,
+      "loss": 3.5725,
+      "step": 5836
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.72564791403287e-06,
+      "loss": 3.2878,
+      "step": 5837
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7256005056890014e-06,
+      "loss": 4.1116,
+      "step": 5838
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7255530973451327e-06,
+      "loss": 3.8854,
+      "step": 5839
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7255056890012644e-06,
+      "loss": 3.8919,
+      "step": 5840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7254582806573957e-06,
+      "loss": 4.417,
+      "step": 5841
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.725410872313527e-06,
+      "loss": 3.9487,
+      "step": 5842
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7253634639696587e-06,
+      "loss": 3.9874,
+      "step": 5843
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7253160556257904e-06,
+      "loss": 4.0327,
+      "step": 5844
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7252686472819216e-06,
+      "loss": 3.8901,
+      "step": 5845
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7252212389380533e-06,
+      "loss": 3.9048,
+      "step": 5846
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7251738305941846e-06,
+      "loss": 3.8557,
+      "step": 5847
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.725126422250316e-06,
+      "loss": 4.0954,
+      "step": 5848
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7250790139064476e-06,
+      "loss": 4.2908,
+      "step": 5849
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7250316055625793e-06,
+      "loss": 3.7559,
+      "step": 5850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7249841972187106e-06,
+      "loss": 4.0163,
+      "step": 5851
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.724936788874842e-06,
+      "loss": 3.7391,
+      "step": 5852
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7248893805309736e-06,
+      "loss": 4.0817,
+      "step": 5853
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.724841972187105e-06,
+      "loss": 3.9887,
+      "step": 5854
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.724794563843236e-06,
+      "loss": 3.8102,
+      "step": 5855
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.724747155499368e-06,
+      "loss": 4.0045,
+      "step": 5856
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.7246997471554995e-06,
+      "loss": 4.3558,
+      "step": 5857
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.724652338811631e-06,
+      "loss": 4.3272,
+      "step": 5858
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7246049304677625e-06,
+      "loss": 4.4278,
+      "step": 5859
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.724557522123894e-06,
+      "loss": 4.1126,
+      "step": 5860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.724510113780025e-06,
+      "loss": 3.9509,
+      "step": 5861
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.724462705436157e-06,
+      "loss": 3.5459,
+      "step": 5862
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7244152970922885e-06,
+      "loss": 3.7801,
+      "step": 5863
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7243678887484198e-06,
+      "loss": 3.8948,
+      "step": 5864
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7243204804045515e-06,
+      "loss": 4.132,
+      "step": 5865
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7242730720606828e-06,
+      "loss": 4.1603,
+      "step": 5866
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.724225663716814e-06,
+      "loss": 3.4293,
+      "step": 5867
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7241782553729457e-06,
+      "loss": 4.0657,
+      "step": 5868
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.724130847029077e-06,
+      "loss": 3.6031,
+      "step": 5869
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7240834386852087e-06,
+      "loss": 4.2146,
+      "step": 5870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7240360303413404e-06,
+      "loss": 4.0613,
+      "step": 5871
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7239886219974717e-06,
+      "loss": 3.8864,
+      "step": 5872
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723941213653603e-06,
+      "loss": 4.0011,
+      "step": 5873
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7238938053097343e-06,
+      "loss": 3.7886,
+      "step": 5874
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723846396965866e-06,
+      "loss": 3.7661,
+      "step": 5875
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7237989886219977e-06,
+      "loss": 3.6796,
+      "step": 5876
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723751580278129e-06,
+      "loss": 3.7113,
+      "step": 5877
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7237041719342607e-06,
+      "loss": 3.6466,
+      "step": 5878
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723656763590392e-06,
+      "loss": 3.4492,
+      "step": 5879
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7236093552465232e-06,
+      "loss": 3.6601,
+      "step": 5880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723561946902655e-06,
+      "loss": 4.1324,
+      "step": 5881
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7235145385587862e-06,
+      "loss": 3.6515,
+      "step": 5882
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723467130214918e-06,
+      "loss": 4.0389,
+      "step": 5883
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7234197218710496e-06,
+      "loss": 3.8755,
+      "step": 5884
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723372313527181e-06,
+      "loss": 3.7322,
+      "step": 5885
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723324905183312e-06,
+      "loss": 3.7764,
+      "step": 5886
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723277496839444e-06,
+      "loss": 4.3157,
+      "step": 5887
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723230088495575e-06,
+      "loss": 3.7372,
+      "step": 5888
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723182680151707e-06,
+      "loss": 3.6994,
+      "step": 5889
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7231352718078386e-06,
+      "loss": 3.9245,
+      "step": 5890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.72308786346397e-06,
+      "loss": 3.6364,
+      "step": 5891
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723040455120101e-06,
+      "loss": 3.7397,
+      "step": 5892
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722993046776233e-06,
+      "loss": 3.8627,
+      "step": 5893
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722945638432364e-06,
+      "loss": 4.2519,
+      "step": 5894
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7228982300884954e-06,
+      "loss": 3.9311,
+      "step": 5895
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722850821744627e-06,
+      "loss": 3.3829,
+      "step": 5896
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722803413400759e-06,
+      "loss": 3.8356,
+      "step": 5897
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.72275600505689e-06,
+      "loss": 3.627,
+      "step": 5898
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7227085967130214e-06,
+      "loss": 3.5454,
+      "step": 5899
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722661188369153e-06,
+      "loss": 3.7691,
+      "step": 5900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7226137800252844e-06,
+      "loss": 3.7129,
+      "step": 5901
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722566371681416e-06,
+      "loss": 3.8082,
+      "step": 5902
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7225189633375478e-06,
+      "loss": 4.14,
+      "step": 5903
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722471554993679e-06,
+      "loss": 3.4986,
+      "step": 5904
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7224241466498103e-06,
+      "loss": 3.899,
+      "step": 5905
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722376738305942e-06,
+      "loss": 3.9995,
+      "step": 5906
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7223293299620733e-06,
+      "loss": 3.5346,
+      "step": 5907
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7222819216182046e-06,
+      "loss": 3.586,
+      "step": 5908
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7222345132743363e-06,
+      "loss": 3.5687,
+      "step": 5909
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722187104930468e-06,
+      "loss": 3.9503,
+      "step": 5910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7221396965865993e-06,
+      "loss": 3.8354,
+      "step": 5911
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.722092288242731e-06,
+      "loss": 4.1098,
+      "step": 5912
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7220448798988623e-06,
+      "loss": 3.705,
+      "step": 5913
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7219974715549935e-06,
+      "loss": 3.89,
+      "step": 5914
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7219500632111252e-06,
+      "loss": 3.8217,
+      "step": 5915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.721902654867257e-06,
+      "loss": 3.9153,
+      "step": 5916
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7218552465233882e-06,
+      "loss": 3.534,
+      "step": 5917
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7218078381795195e-06,
+      "loss": 4.0434,
+      "step": 5918
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7217604298356512e-06,
+      "loss": 3.7303,
+      "step": 5919
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7217130214917825e-06,
+      "loss": 3.7665,
+      "step": 5920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7216656131479138e-06,
+      "loss": 3.5771,
+      "step": 5921
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7216182048040455e-06,
+      "loss": 3.9492,
+      "step": 5922
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.721570796460177e-06,
+      "loss": 3.9528,
+      "step": 5923
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7215233881163085e-06,
+      "loss": 3.7165,
+      "step": 5924
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.72147597977244e-06,
+      "loss": 3.9312,
+      "step": 5925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7214285714285714e-06,
+      "loss": 3.8096,
+      "step": 5926
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7213811630847027e-06,
+      "loss": 4.0128,
+      "step": 5927
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7213337547408344e-06,
+      "loss": 3.8827,
+      "step": 5928
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.721286346396966e-06,
+      "loss": 3.9878,
+      "step": 5929
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7212389380530974e-06,
+      "loss": 3.881,
+      "step": 5930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.721191529709229e-06,
+      "loss": 4.1435,
+      "step": 5931
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7211441213653604e-06,
+      "loss": 3.878,
+      "step": 5932
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7210967130214917e-06,
+      "loss": 3.4748,
+      "step": 5933
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7210493046776234e-06,
+      "loss": 3.7815,
+      "step": 5934
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7210018963337547e-06,
+      "loss": 4.2746,
+      "step": 5935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7209544879898864e-06,
+      "loss": 3.8618,
+      "step": 5936
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.720907079646018e-06,
+      "loss": 3.3146,
+      "step": 5937
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7208596713021494e-06,
+      "loss": 3.8685,
+      "step": 5938
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7208122629582806e-06,
+      "loss": 3.614,
+      "step": 5939
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.720764854614412e-06,
+      "loss": 3.746,
+      "step": 5940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7207174462705436e-06,
+      "loss": 4.0831,
+      "step": 5941
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7206700379266753e-06,
+      "loss": 3.8251,
+      "step": 5942
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7206226295828066e-06,
+      "loss": 3.4173,
+      "step": 5943
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7205752212389383e-06,
+      "loss": 3.7556,
+      "step": 5944
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7205278128950696e-06,
+      "loss": 3.4595,
+      "step": 5945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.720480404551201e-06,
+      "loss": 3.4653,
+      "step": 5946
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7204329962073326e-06,
+      "loss": 4.1166,
+      "step": 5947
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.720385587863464e-06,
+      "loss": 3.3062,
+      "step": 5948
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7203381795195956e-06,
+      "loss": 3.7341,
+      "step": 5949
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7202907711757273e-06,
+      "loss": 4.011,
+      "step": 5950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7202433628318585e-06,
+      "loss": 3.7628,
+      "step": 5951
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.72019595448799e-06,
+      "loss": 3.0818,
+      "step": 5952
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7201485461441215e-06,
+      "loss": 3.3064,
+      "step": 5953
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.720101137800253e-06,
+      "loss": 3.5897,
+      "step": 5954
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7200537294563845e-06,
+      "loss": 2.9495,
+      "step": 5955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7200063211125162e-06,
+      "loss": 3.5712,
+      "step": 5956
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7199589127686475e-06,
+      "loss": 3.8288,
+      "step": 5957
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7199115044247788e-06,
+      "loss": 3.5259,
+      "step": 5958
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7198640960809105e-06,
+      "loss": 3.5345,
+      "step": 5959
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7198166877370418e-06,
+      "loss": 3.709,
+      "step": 5960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719769279393173e-06,
+      "loss": 3.349,
+      "step": 5961
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7197218710493047e-06,
+      "loss": 3.9864,
+      "step": 5962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7196744627054364e-06,
+      "loss": 3.815,
+      "step": 5963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7196270543615677e-06,
+      "loss": 4.0359,
+      "step": 5964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719579646017699e-06,
+      "loss": 3.5827,
+      "step": 5965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7195322376738307e-06,
+      "loss": 3.3426,
+      "step": 5966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719484829329962e-06,
+      "loss": 4.1318,
+      "step": 5967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7194374209860937e-06,
+      "loss": 3.7168,
+      "step": 5968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7193900126422254e-06,
+      "loss": 3.6285,
+      "step": 5969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7193426042983567e-06,
+      "loss": 3.1847,
+      "step": 5970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719295195954488e-06,
+      "loss": 3.993,
+      "step": 5971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7192477876106197e-06,
+      "loss": 3.8772,
+      "step": 5972
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719200379266751e-06,
+      "loss": 3.801,
+      "step": 5973
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7191529709228822e-06,
+      "loss": 3.9109,
+      "step": 5974
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719105562579014e-06,
+      "loss": 3.9045,
+      "step": 5975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7190581542351456e-06,
+      "loss": 3.9655,
+      "step": 5976
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719010745891277e-06,
+      "loss": 3.9817,
+      "step": 5977
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7189633375474086e-06,
+      "loss": 3.4333,
+      "step": 5978
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.71891592920354e-06,
+      "loss": 3.9523,
+      "step": 5979
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.718868520859671e-06,
+      "loss": 3.6991,
+      "step": 5980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.718821112515803e-06,
+      "loss": 4.1929,
+      "step": 5981
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7187737041719346e-06,
+      "loss": 3.5378,
+      "step": 5982
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.718726295828066e-06,
+      "loss": 3.9984,
+      "step": 5983
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.718678887484197e-06,
+      "loss": 3.8924,
+      "step": 5984
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718631479140329e-06,
+      "loss": 3.5647,
+      "step": 5985
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.71858407079646e-06,
+      "loss": 3.2314,
+      "step": 5986
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7185366624525914e-06,
+      "loss": 3.6357,
+      "step": 5987
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718489254108723e-06,
+      "loss": 3.7247,
+      "step": 5988
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718441845764855e-06,
+      "loss": 3.9295,
+      "step": 5989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718394437420986e-06,
+      "loss": 3.8391,
+      "step": 5990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718347029077118e-06,
+      "loss": 3.8898,
+      "step": 5991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718299620733249e-06,
+      "loss": 3.9583,
+      "step": 5992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7182522123893804e-06,
+      "loss": 3.7783,
+      "step": 5993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718204804045512e-06,
+      "loss": 3.3223,
+      "step": 5994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7181573957016438e-06,
+      "loss": 3.478,
+      "step": 5995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718109987357775e-06,
+      "loss": 3.7793,
+      "step": 5996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7180625790139068e-06,
+      "loss": 3.8436,
+      "step": 5997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.718015170670038e-06,
+      "loss": 4.1743,
+      "step": 5998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7179677623261693e-06,
+      "loss": 3.8189,
+      "step": 5999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717920353982301e-06,
+      "loss": 3.8637,
+      "step": 6000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7178729456384323e-06,
+      "loss": 3.9189,
+      "step": 6001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717825537294564e-06,
+      "loss": 4.2005,
+      "step": 6002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7177781289506957e-06,
+      "loss": 3.8818,
+      "step": 6003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717730720606827e-06,
+      "loss": 3.9062,
+      "step": 6004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7176833122629583e-06,
+      "loss": 3.9504,
+      "step": 6005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7176359039190896e-06,
+      "loss": 3.6211,
+      "step": 6006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7175884955752213e-06,
+      "loss": 3.8756,
+      "step": 6007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717541087231353e-06,
+      "loss": 3.8331,
+      "step": 6008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7174936788874842e-06,
+      "loss": 4.0769,
+      "step": 6009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717446270543616e-06,
+      "loss": 3.0925,
+      "step": 6010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7173988621997472e-06,
+      "loss": 4.1826,
+      "step": 6011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7173514538558785e-06,
+      "loss": 3.7698,
+      "step": 6012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.71730404551201e-06,
+      "loss": 3.7077,
+      "step": 6013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7172566371681415e-06,
+      "loss": 4.0025,
+      "step": 6014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717209228824273e-06,
+      "loss": 3.8917,
+      "step": 6015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717161820480405e-06,
+      "loss": 3.7589,
+      "step": 6016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717114412136536e-06,
+      "loss": 3.9501,
+      "step": 6017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7170670037926675e-06,
+      "loss": 3.908,
+      "step": 6018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717019595448799e-06,
+      "loss": 3.4887,
+      "step": 6019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7169721871049304e-06,
+      "loss": 4.0385,
+      "step": 6020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7169247787610617e-06,
+      "loss": 3.6675,
+      "step": 6021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.716877370417194e-06,
+      "loss": 4.0303,
+      "step": 6022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.716829962073325e-06,
+      "loss": 3.8454,
+      "step": 6023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7167825537294564e-06,
+      "loss": 3.6783,
+      "step": 6024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.716735145385588e-06,
+      "loss": 4.0768,
+      "step": 6025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7166877370417194e-06,
+      "loss": 3.6035,
+      "step": 6026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7166403286978507e-06,
+      "loss": 3.3961,
+      "step": 6027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7165929203539824e-06,
+      "loss": 4.2449,
+      "step": 6028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.716545512010114e-06,
+      "loss": 3.9897,
+      "step": 6029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7164981036662454e-06,
+      "loss": 3.9433,
+      "step": 6030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7164506953223766e-06,
+      "loss": 4.0161,
+      "step": 6031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7164032869785083e-06,
+      "loss": 3.3442,
+      "step": 6032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7163558786346396e-06,
+      "loss": 4.1538,
+      "step": 6033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.716308470290771e-06,
+      "loss": 3.564,
+      "step": 6034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7162610619469026e-06,
+      "loss": 3.7192,
+      "step": 6035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7162136536030343e-06,
+      "loss": 3.5124,
+      "step": 6036
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7161662452591656e-06,
+      "loss": 3.8528,
+      "step": 6037
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7161188369152973e-06,
+      "loss": 4.0324,
+      "step": 6038
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7160714285714286e-06,
+      "loss": 4.198,
+      "step": 6039
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.71602402022756e-06,
+      "loss": 4.0028,
+      "step": 6040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7159766118836916e-06,
+      "loss": 3.9392,
+      "step": 6041
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7159292035398233e-06,
+      "loss": 3.3184,
+      "step": 6042
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7158817951959546e-06,
+      "loss": 3.7523,
+      "step": 6043
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7158343868520863e-06,
+      "loss": 3.987,
+      "step": 6044
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7157869785082175e-06,
+      "loss": 3.8617,
+      "step": 6045
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.715739570164349e-06,
+      "loss": 3.719,
+      "step": 6046
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7156921618204805e-06,
+      "loss": 3.53,
+      "step": 6047
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.715644753476612e-06,
+      "loss": 3.6125,
+      "step": 6048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7155973451327435e-06,
+      "loss": 3.6852,
+      "step": 6049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7155499367888748e-06,
+      "loss": 3.5403,
+      "step": 6050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7155025284450065e-06,
+      "loss": 3.7598,
+      "step": 6051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7154551201011378e-06,
+      "loss": 4.0731,
+      "step": 6052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.715407711757269e-06,
+      "loss": 3.8996,
+      "step": 6053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7153603034134008e-06,
+      "loss": 3.839,
+      "step": 6054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7153128950695325e-06,
+      "loss": 3.5114,
+      "step": 6055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7152654867256637e-06,
+      "loss": 3.9563,
+      "step": 6056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7152180783817954e-06,
+      "loss": 3.7706,
+      "step": 6057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7151706700379267e-06,
+      "loss": 3.7701,
+      "step": 6058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.715123261694058e-06,
+      "loss": 3.7411,
+      "step": 6059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7150758533501897e-06,
+      "loss": 3.7783,
+      "step": 6060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.715028445006321e-06,
+      "loss": 3.9706,
+      "step": 6061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7149810366624527e-06,
+      "loss": 4.0276,
+      "step": 6062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7149336283185844e-06,
+      "loss": 3.5523,
+      "step": 6063
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7148862199747157e-06,
+      "loss": 4.0321,
+      "step": 6064
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714838811630847e-06,
+      "loss": 3.7525,
+      "step": 6065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7147914032869787e-06,
+      "loss": 3.7354,
+      "step": 6066
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.71474399494311e-06,
+      "loss": 3.7636,
+      "step": 6067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7146965865992416e-06,
+      "loss": 4.0348,
+      "step": 6068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7146491782553733e-06,
+      "loss": 4.0209,
+      "step": 6069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7146017699115046e-06,
+      "loss": 3.7154,
+      "step": 6070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714554361567636e-06,
+      "loss": 4.2326,
+      "step": 6071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714506953223767e-06,
+      "loss": 4.1299,
+      "step": 6072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714459544879899e-06,
+      "loss": 3.8525,
+      "step": 6073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.71441213653603e-06,
+      "loss": 3.3124,
+      "step": 6074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714364728192162e-06,
+      "loss": 3.561,
+      "step": 6075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7143173198482936e-06,
+      "loss": 4.0084,
+      "step": 6076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714269911504425e-06,
+      "loss": 3.6732,
+      "step": 6077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714222503160556e-06,
+      "loss": 4.0624,
+      "step": 6078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714175094816688e-06,
+      "loss": 3.7575,
+      "step": 6079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714127686472819e-06,
+      "loss": 3.9694,
+      "step": 6080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.714080278128951e-06,
+      "loss": 3.9762,
+      "step": 6081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7140328697850825e-06,
+      "loss": 3.9386,
+      "step": 6082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713985461441214e-06,
+      "loss": 3.773,
+      "step": 6083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713938053097345e-06,
+      "loss": 3.7633,
+      "step": 6084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713890644753477e-06,
+      "loss": 4.3192,
+      "step": 6085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713843236409608e-06,
+      "loss": 3.887,
+      "step": 6086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7137958280657394e-06,
+      "loss": 3.8095,
+      "step": 6087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713748419721871e-06,
+      "loss": 4.1921,
+      "step": 6088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7137010113780028e-06,
+      "loss": 3.9789,
+      "step": 6089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713653603034134e-06,
+      "loss": 3.4147,
+      "step": 6090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7136061946902658e-06,
+      "loss": 3.8347,
+      "step": 6091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713558786346397e-06,
+      "loss": 3.7431,
+      "step": 6092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7135113780025283e-06,
+      "loss": 4.1294,
+      "step": 6093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.71346396965866e-06,
+      "loss": 3.2951,
+      "step": 6094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7134165613147917e-06,
+      "loss": 3.4638,
+      "step": 6095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713369152970923e-06,
+      "loss": 3.4492,
+      "step": 6096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7133217446270543e-06,
+      "loss": 3.4727,
+      "step": 6097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713274336283186e-06,
+      "loss": 3.5823,
+      "step": 6098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7132269279393173e-06,
+      "loss": 3.962,
+      "step": 6099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7131795195954485e-06,
+      "loss": 4.1627,
+      "step": 6100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7131321112515803e-06,
+      "loss": 4.0612,
+      "step": 6101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.713084702907712e-06,
+      "loss": 4.0407,
+      "step": 6102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7130372945638432e-06,
+      "loss": 3.8865,
+      "step": 6103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.712989886219975e-06,
+      "loss": 3.7115,
+      "step": 6104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7129424778761062e-06,
+      "loss": 3.9879,
+      "step": 6105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.7128950695322375e-06,
+      "loss": 3.7303,
+      "step": 6106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.712847661188369e-06,
+      "loss": 4.0781,
+      "step": 6107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.712800252844501e-06,
+      "loss": 3.2979,
+      "step": 6108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.712752844500632e-06,
+      "loss": 3.8447,
+      "step": 6109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.712705436156764e-06,
+      "loss": 3.9297,
+      "step": 6110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.712658027812895e-06,
+      "loss": 3.9029,
+      "step": 6111
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7126106194690265e-06,
+      "loss": 3.6346,
+      "step": 6112
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.712563211125158e-06,
+      "loss": 3.6526,
+      "step": 6113
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7125158027812894e-06,
+      "loss": 3.7169,
+      "step": 6114
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.712468394437421e-06,
+      "loss": 3.7002,
+      "step": 6115
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7124209860935524e-06,
+      "loss": 3.7301,
+      "step": 6116
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.712373577749684e-06,
+      "loss": 3.7194,
+      "step": 6117
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7123261694058154e-06,
+      "loss": 3.5165,
+      "step": 6118
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7122787610619467e-06,
+      "loss": 3.2492,
+      "step": 6119
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7122313527180784e-06,
+      "loss": 3.585,
+      "step": 6120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.71218394437421e-06,
+      "loss": 3.7535,
+      "step": 6121
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7121365360303414e-06,
+      "loss": 3.4492,
+      "step": 6122
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.712089127686473e-06,
+      "loss": 3.2259,
+      "step": 6123
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7120417193426044e-06,
+      "loss": 4.2663,
+      "step": 6124
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7119943109987356e-06,
+      "loss": 3.9361,
+      "step": 6125
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7119469026548673e-06,
+      "loss": 4.1829,
+      "step": 6126
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7118994943109986e-06,
+      "loss": 3.6088,
+      "step": 6127
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7118520859671303e-06,
+      "loss": 4.1258,
+      "step": 6128
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.711804677623262e-06,
+      "loss": 3.8379,
+      "step": 6129
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7117572692793933e-06,
+      "loss": 3.7488,
+      "step": 6130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7117098609355246e-06,
+      "loss": 3.5362,
+      "step": 6131
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7116624525916563e-06,
+      "loss": 3.9577,
+      "step": 6132
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7116150442477876e-06,
+      "loss": 3.6452,
+      "step": 6133
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7115676359039193e-06,
+      "loss": 3.0657,
+      "step": 6134
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.711520227560051e-06,
+      "loss": 3.3502,
+      "step": 6135
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7114728192161823e-06,
+      "loss": 3.7108,
+      "step": 6136
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7114254108723135e-06,
+      "loss": 3.6768,
+      "step": 6137
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.711378002528445e-06,
+      "loss": 3.7006,
+      "step": 6138
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7113305941845765e-06,
+      "loss": 3.7542,
+      "step": 6139
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.711283185840708e-06,
+      "loss": 4.3629,
+      "step": 6140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7112357774968395e-06,
+      "loss": 3.5251,
+      "step": 6141
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7111883691529712e-06,
+      "loss": 4.1054,
+      "step": 6142
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7111409608091025e-06,
+      "loss": 3.5678,
+      "step": 6143
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7110935524652338e-06,
+      "loss": 4.1657,
+      "step": 6144
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7110461441213655e-06,
+      "loss": 4.084,
+      "step": 6145
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7109987357774968e-06,
+      "loss": 3.695,
+      "step": 6146
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7109513274336285e-06,
+      "loss": 3.176,
+      "step": 6147
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.71090391908976e-06,
+      "loss": 3.5744,
+      "step": 6148
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7108565107458915e-06,
+      "loss": 3.6257,
+      "step": 6149
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7108091024020227e-06,
+      "loss": 3.9422,
+      "step": 6150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7107616940581544e-06,
+      "loss": 3.7905,
+      "step": 6151
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7107142857142857e-06,
+      "loss": 3.8326,
+      "step": 6152
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.710666877370417e-06,
+      "loss": 3.5618,
+      "step": 6153
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7106194690265487e-06,
+      "loss": 3.9805,
+      "step": 6154
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7105720606826804e-06,
+      "loss": 3.8127,
+      "step": 6155
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7105246523388117e-06,
+      "loss": 3.6635,
+      "step": 6156
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7104772439949434e-06,
+      "loss": 4.2537,
+      "step": 6157
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7104298356510747e-06,
+      "loss": 4.2704,
+      "step": 6158
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.710382427307206e-06,
+      "loss": 3.7103,
+      "step": 6159
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7103350189633377e-06,
+      "loss": 3.9587,
+      "step": 6160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7102876106194694e-06,
+      "loss": 3.9615,
+      "step": 6161
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7102402022756006e-06,
+      "loss": 4.077,
+      "step": 6162
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.710192793931732e-06,
+      "loss": 3.7764,
+      "step": 6163
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7101453855878636e-06,
+      "loss": 3.7447,
+      "step": 6164
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.710097977243995e-06,
+      "loss": 3.8691,
+      "step": 6165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.710050568900126e-06,
+      "loss": 3.7368,
+      "step": 6166
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.710003160556258e-06,
+      "loss": 4.2474,
+      "step": 6167
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7099557522123896e-06,
+      "loss": 3.8642,
+      "step": 6168
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.709908343868521e-06,
+      "loss": 4.3391,
+      "step": 6169
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7098609355246526e-06,
+      "loss": 3.4013,
+      "step": 6170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.709813527180784e-06,
+      "loss": 3.5452,
+      "step": 6171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.709766118836915e-06,
+      "loss": 3.7704,
+      "step": 6172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.709718710493047e-06,
+      "loss": 3.6173,
+      "step": 6173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2.7096713021491785e-06,
+      "loss": 3.5891,
+      "step": 6174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.70962389380531e-06,
+      "loss": 3.5913,
+      "step": 6175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7095764854614415e-06,
+      "loss": 3.9185,
+      "step": 6176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709529077117573e-06,
+      "loss": 2.9793,
+      "step": 6177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709481668773704e-06,
+      "loss": 3.8717,
+      "step": 6178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709434260429836e-06,
+      "loss": 3.9201,
+      "step": 6179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709386852085967e-06,
+      "loss": 4.0593,
+      "step": 6180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7093394437420988e-06,
+      "loss": 3.8683,
+      "step": 6181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.70929203539823e-06,
+      "loss": 4.0256,
+      "step": 6182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7092446270543618e-06,
+      "loss": 4.0841,
+      "step": 6183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709197218710493e-06,
+      "loss": 4.2461,
+      "step": 6184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7091498103666243e-06,
+      "loss": 3.7932,
+      "step": 6185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709102402022756e-06,
+      "loss": 4.1115,
+      "step": 6186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7090549936788877e-06,
+      "loss": 3.5623,
+      "step": 6187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.709007585335019e-06,
+      "loss": 3.9902,
+      "step": 6188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7089601769911507e-06,
+      "loss": 3.5706,
+      "step": 6189
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708912768647282e-06,
+      "loss": 3.2392,
+      "step": 6190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7088653603034133e-06,
+      "loss": 3.8721,
+      "step": 6191
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708817951959545e-06,
+      "loss": 3.7416,
+      "step": 6192
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7087705436156763e-06,
+      "loss": 3.6574,
+      "step": 6193
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708723135271808e-06,
+      "loss": 4.1288,
+      "step": 6194
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7086757269279397e-06,
+      "loss": 4.0588,
+      "step": 6195
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708628318584071e-06,
+      "loss": 3.4213,
+      "step": 6196
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7085809102402022e-06,
+      "loss": 3.3366,
+      "step": 6197
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708533501896334e-06,
+      "loss": 3.4603,
+      "step": 6198
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7084860935524652e-06,
+      "loss": 3.7211,
+      "step": 6199
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708438685208597e-06,
+      "loss": 3.9504,
+      "step": 6200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7083912768647286e-06,
+      "loss": 3.7506,
+      "step": 6201
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.70834386852086e-06,
+      "loss": 4.1342,
+      "step": 6202
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708296460176991e-06,
+      "loss": 3.4447,
+      "step": 6203
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7082490518331225e-06,
+      "loss": 4.1022,
+      "step": 6204
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708201643489254e-06,
+      "loss": 3.7958,
+      "step": 6205
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7081542351453854e-06,
+      "loss": 4.0293,
+      "step": 6206
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708106826801517e-06,
+      "loss": 3.4739,
+      "step": 6207
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708059418457649e-06,
+      "loss": 3.7444,
+      "step": 6208
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.70801201011378e-06,
+      "loss": 3.7954,
+      "step": 6209
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7079646017699114e-06,
+      "loss": 3.9317,
+      "step": 6210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707917193426043e-06,
+      "loss": 3.6946,
+      "step": 6211
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7078697850821744e-06,
+      "loss": 3.6653,
+      "step": 6212
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707822376738306e-06,
+      "loss": 3.4014,
+      "step": 6213
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707774968394438e-06,
+      "loss": 3.6637,
+      "step": 6214
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707727560050569e-06,
+      "loss": 3.6084,
+      "step": 6215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7076801517067004e-06,
+      "loss": 4.154,
+      "step": 6216
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707632743362832e-06,
+      "loss": 3.7981,
+      "step": 6217
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7075853350189634e-06,
+      "loss": 3.6122,
+      "step": 6218
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7075379266750946e-06,
+      "loss": 3.9746,
+      "step": 6219
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7074905183312263e-06,
+      "loss": 3.6458,
+      "step": 6220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707443109987358e-06,
+      "loss": 2.8506,
+      "step": 6221
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7073957016434893e-06,
+      "loss": 3.8209,
+      "step": 6222
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7073482932996206e-06,
+      "loss": 3.6425,
+      "step": 6223
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7073008849557523e-06,
+      "loss": 3.9258,
+      "step": 6224
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7072534766118836e-06,
+      "loss": 3.4998,
+      "step": 6225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7072060682680153e-06,
+      "loss": 3.7015,
+      "step": 6226
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.707158659924147e-06,
+      "loss": 4.1121,
+      "step": 6227
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7071112515802783e-06,
+      "loss": 3.5852,
+      "step": 6228
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7070638432364096e-06,
+      "loss": 3.4448,
+      "step": 6229
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7070164348925413e-06,
+      "loss": 4.079,
+      "step": 6230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7069690265486725e-06,
+      "loss": 4.1855,
+      "step": 6231
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.706921618204804e-06,
+      "loss": 4.3432,
+      "step": 6232
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7068742098609355e-06,
+      "loss": 4.1226,
+      "step": 6233
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7068268015170672e-06,
+      "loss": 3.4724,
+      "step": 6234
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7067793931731985e-06,
+      "loss": 3.9299,
+      "step": 6235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7067319848293302e-06,
+      "loss": 3.9316,
+      "step": 6236
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7066845764854615e-06,
+      "loss": 3.7931,
+      "step": 6237
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7066371681415928e-06,
+      "loss": 3.6923,
+      "step": 6238
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7065897597977245e-06,
+      "loss": 3.7965,
+      "step": 6239
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.706542351453856e-06,
+      "loss": 3.9609,
+      "step": 6240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7064949431099875e-06,
+      "loss": 3.8271,
+      "step": 6241
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.706447534766119e-06,
+      "loss": 4.1853,
+      "step": 6242
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7064001264222504e-06,
+      "loss": 3.6969,
+      "step": 6243
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7063527180783817e-06,
+      "loss": 4.1772,
+      "step": 6244
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.706305309734513e-06,
+      "loss": 4.2387,
+      "step": 6245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7062579013906447e-06,
+      "loss": 3.3773,
+      "step": 6246
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7062104930467764e-06,
+      "loss": 3.5879,
+      "step": 6247
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7061630847029077e-06,
+      "loss": 4.1895,
+      "step": 6248
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7061156763590394e-06,
+      "loss": 3.8115,
+      "step": 6249
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7060682680151707e-06,
+      "loss": 3.6032,
+      "step": 6250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.706020859671302e-06,
+      "loss": 3.6629,
+      "step": 6251
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7059734513274337e-06,
+      "loss": 3.5053,
+      "step": 6252
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7059260429835654e-06,
+      "loss": 3.3349,
+      "step": 6253
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7058786346396966e-06,
+      "loss": 3.2509,
+      "step": 6254
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7058312262958284e-06,
+      "loss": 3.7964,
+      "step": 6255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7057838179519596e-06,
+      "loss": 3.8647,
+      "step": 6256
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705736409608091e-06,
+      "loss": 3.8832,
+      "step": 6257
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7056890012642226e-06,
+      "loss": 4.241,
+      "step": 6258
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705641592920354e-06,
+      "loss": 4.2358,
+      "step": 6259
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7055941845764856e-06,
+      "loss": 3.7663,
+      "step": 6260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7055467762326173e-06,
+      "loss": 3.8914,
+      "step": 6261
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7054993678887486e-06,
+      "loss": 3.7877,
+      "step": 6262
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.70545195954488e-06,
+      "loss": 4.0982,
+      "step": 6263
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7054045512010116e-06,
+      "loss": 3.8852,
+      "step": 6264
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705357142857143e-06,
+      "loss": 3.9731,
+      "step": 6265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7053097345132746e-06,
+      "loss": 3.7204,
+      "step": 6266
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705262326169406e-06,
+      "loss": 4.0251,
+      "step": 6267
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7052149178255375e-06,
+      "loss": 3.7833,
+      "step": 6268
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705167509481669e-06,
+      "loss": 3.8794,
+      "step": 6269
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7051201011378e-06,
+      "loss": 3.5346,
+      "step": 6270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705072692793932e-06,
+      "loss": 3.8638,
+      "step": 6271
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.705025284450063e-06,
+      "loss": 3.8251,
+      "step": 6272
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7049778761061948e-06,
+      "loss": 3.4136,
+      "step": 6273
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7049304677623265e-06,
+      "loss": 3.4837,
+      "step": 6274
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7048830594184578e-06,
+      "loss": 3.8537,
+      "step": 6275
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.704835651074589e-06,
+      "loss": 3.8374,
+      "step": 6276
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7047882427307208e-06,
+      "loss": 3.8406,
+      "step": 6277
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.704740834386852e-06,
+      "loss": 4.1917,
+      "step": 6278
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7046934260429837e-06,
+      "loss": 4.0842,
+      "step": 6279
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7046460176991154e-06,
+      "loss": 3.642,
+      "step": 6280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7045986093552467e-06,
+      "loss": 3.8057,
+      "step": 6281
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.704551201011378e-06,
+      "loss": 3.335,
+      "step": 6282
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7045037926675097e-06,
+      "loss": 4.0618,
+      "step": 6283
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.704456384323641e-06,
+      "loss": 4.1778,
+      "step": 6284
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7044089759797723e-06,
+      "loss": 3.5248,
+      "step": 6285
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.704361567635904e-06,
+      "loss": 3.6284,
+      "step": 6286
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7043141592920357e-06,
+      "loss": 3.6638,
+      "step": 6287
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.704266750948167e-06,
+      "loss": 3.8279,
+      "step": 6288
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7042193426042982e-06,
+      "loss": 3.8963,
+      "step": 6289
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.70417193426043e-06,
+      "loss": 3.8243,
+      "step": 6290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7041245259165612e-06,
+      "loss": 3.8863,
+      "step": 6291
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7040771175726925e-06,
+      "loss": 3.6676,
+      "step": 6292
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7040297092288246e-06,
+      "loss": 3.7791,
+      "step": 6293
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.703982300884956e-06,
+      "loss": 3.4627,
+      "step": 6294
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.703934892541087e-06,
+      "loss": 3.4243,
+      "step": 6295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.703887484197219e-06,
+      "loss": 3.6384,
+      "step": 6296
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.70384007585335e-06,
+      "loss": 4.0332,
+      "step": 6297
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7037926675094815e-06,
+      "loss": 4.0858,
+      "step": 6298
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.703745259165613e-06,
+      "loss": 3.7914,
+      "step": 6299
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.703697850821745e-06,
+      "loss": 4.1678,
+      "step": 6300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.703650442477876e-06,
+      "loss": 3.0917,
+      "step": 6301
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703603034134008e-06,
+      "loss": 3.3649,
+      "step": 6302
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703555625790139e-06,
+      "loss": 3.7192,
+      "step": 6303
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7035082174462704e-06,
+      "loss": 3.481,
+      "step": 6304
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703460809102402e-06,
+      "loss": 3.8008,
+      "step": 6305
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7034134007585334e-06,
+      "loss": 3.7361,
+      "step": 6306
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703365992414665e-06,
+      "loss": 4.0495,
+      "step": 6307
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703318584070797e-06,
+      "loss": 3.9494,
+      "step": 6308
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703271175726928e-06,
+      "loss": 3.141,
+      "step": 6309
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7032237673830594e-06,
+      "loss": 3.5294,
+      "step": 6310
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7031763590391906e-06,
+      "loss": 3.559,
+      "step": 6311
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7031289506953223e-06,
+      "loss": 3.5326,
+      "step": 6312
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.703081542351454e-06,
+      "loss": 3.8947,
+      "step": 6313
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7030341340075853e-06,
+      "loss": 3.9723,
+      "step": 6314
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.702986725663717e-06,
+      "loss": 4.0313,
+      "step": 6315
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7029393173198483e-06,
+      "loss": 3.9322,
+      "step": 6316
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7028919089759796e-06,
+      "loss": 3.6937,
+      "step": 6317
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7028445006321113e-06,
+      "loss": 3.102,
+      "step": 6318
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7027970922882426e-06,
+      "loss": 3.6221,
+      "step": 6319
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7027496839443743e-06,
+      "loss": 3.7255,
+      "step": 6320
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.702702275600506e-06,
+      "loss": 3.3472,
+      "step": 6321
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7026548672566373e-06,
+      "loss": 3.7344,
+      "step": 6322
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7026074589127685e-06,
+      "loss": 3.7907,
+      "step": 6323
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7025600505689003e-06,
+      "loss": 3.7566,
+      "step": 6324
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7025126422250315e-06,
+      "loss": 4.1162,
+      "step": 6325
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7024652338811632e-06,
+      "loss": 3.8727,
+      "step": 6326
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.702417825537295e-06,
+      "loss": 3.5441,
+      "step": 6327
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7023704171934262e-06,
+      "loss": 3.6772,
+      "step": 6328
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7023230088495575e-06,
+      "loss": 3.7066,
+      "step": 6329
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.702275600505689e-06,
+      "loss": 4.084,
+      "step": 6330
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7022281921618205e-06,
+      "loss": 3.6159,
+      "step": 6331
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7021807838179518e-06,
+      "loss": 3.4325,
+      "step": 6332
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7021333754740835e-06,
+      "loss": 4.805,
+      "step": 6333
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.16554174067495558,
+      "eval_loss": 3.6870830059051514,
+      "eval_runtime": 152.8439,
+      "eval_samples_per_second": 18.417,
+      "eval_steps_per_second": 4.606,
+      "step": 6333
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.702085967130215e-06,
+      "loss": 3.8474,
+      "step": 6334
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7020385587863465e-06,
+      "loss": 3.8004,
+      "step": 6335
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7019911504424777e-06,
+      "loss": 3.2923,
+      "step": 6336
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7019437420986094e-06,
+      "loss": 3.5296,
+      "step": 6337
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7018963337547407e-06,
+      "loss": 3.7302,
+      "step": 6338
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7018489254108724e-06,
+      "loss": 3.7888,
+      "step": 6339
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.701801517067004e-06,
+      "loss": 3.6139,
+      "step": 6340
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7017541087231354e-06,
+      "loss": 3.7312,
+      "step": 6341
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7017067003792667e-06,
+      "loss": 3.7834,
+      "step": 6342
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7016592920353984e-06,
+      "loss": 4.0486,
+      "step": 6343
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7016118836915297e-06,
+      "loss": 4.0534,
+      "step": 6344
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.701564475347661e-06,
+      "loss": 3.3324,
+      "step": 6345
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7015170670037927e-06,
+      "loss": 3.8201,
+      "step": 6346
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7014696586599244e-06,
+      "loss": 3.5619,
+      "step": 6347
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7014222503160556e-06,
+      "loss": 3.1394,
+      "step": 6348
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7013748419721873e-06,
+      "loss": 3.8522,
+      "step": 6349
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7013274336283186e-06,
+      "loss": 3.4706,
+      "step": 6350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.70128002528445e-06,
+      "loss": 3.7581,
+      "step": 6351
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7012326169405816e-06,
+      "loss": 4.544,
+      "step": 6352
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7011852085967133e-06,
+      "loss": 3.7902,
+      "step": 6353
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7011378002528446e-06,
+      "loss": 3.3339,
+      "step": 6354
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.701090391908976e-06,
+      "loss": 4.338,
+      "step": 6355
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7010429835651076e-06,
+      "loss": 3.3558,
+      "step": 6356
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.700995575221239e-06,
+      "loss": 4.1473,
+      "step": 6357
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.70094816687737e-06,
+      "loss": 3.6462,
+      "step": 6358
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.700900758533502e-06,
+      "loss": 3.9864,
+      "step": 6359
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7008533501896335e-06,
+      "loss": 3.8196,
+      "step": 6360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.700805941845765e-06,
+      "loss": 3.9812,
+      "step": 6361
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7007585335018965e-06,
+      "loss": 4.1715,
+      "step": 6362
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.700711125158028e-06,
+      "loss": 3.6991,
+      "step": 6363
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.700663716814159e-06,
+      "loss": 4.0327,
+      "step": 6364
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.700616308470291e-06,
+      "loss": 4.1829,
+      "step": 6365
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7005689001264225e-06,
+      "loss": 3.8431,
+      "step": 6366
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7005214917825538e-06,
+      "loss": 3.6968,
+      "step": 6367
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7004740834386855e-06,
+      "loss": 4.1326,
+      "step": 6368
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7004266750948168e-06,
+      "loss": 3.8709,
+      "step": 6369
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.700379266750948e-06,
+      "loss": 3.4842,
+      "step": 6370
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7003318584070798e-06,
+      "loss": 4.2053,
+      "step": 6371
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.700284450063211e-06,
+      "loss": 3.7023,
+      "step": 6372
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7002370417193427e-06,
+      "loss": 3.4517,
+      "step": 6373
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7001896333754744e-06,
+      "loss": 4.1742,
+      "step": 6374
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7001422250316057e-06,
+      "loss": 3.9876,
+      "step": 6375
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.700094816687737e-06,
+      "loss": 3.909,
+      "step": 6376
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7000474083438683e-06,
+      "loss": 3.4036,
+      "step": 6377
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.7e-06,
+      "loss": 3.4045,
+      "step": 6378
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6999525916561317e-06,
+      "loss": 4.1866,
+      "step": 6379
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699905183312263e-06,
+      "loss": 4.0486,
+      "step": 6380
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6998577749683947e-06,
+      "loss": 3.2579,
+      "step": 6381
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699810366624526e-06,
+      "loss": 3.88,
+      "step": 6382
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6997629582806572e-06,
+      "loss": 3.9879,
+      "step": 6383
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699715549936789e-06,
+      "loss": 3.8387,
+      "step": 6384
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6996681415929202e-06,
+      "loss": 4.1119,
+      "step": 6385
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699620733249052e-06,
+      "loss": 3.5418,
+      "step": 6386
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6995733249051836e-06,
+      "loss": 3.6874,
+      "step": 6387
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699525916561315e-06,
+      "loss": 3.8642,
+      "step": 6388
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699478508217446e-06,
+      "loss": 4.0047,
+      "step": 6389
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699431099873578e-06,
+      "loss": 3.5757,
+      "step": 6390
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699383691529709e-06,
+      "loss": 4.2153,
+      "step": 6391
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699336283185841e-06,
+      "loss": 3.7316,
+      "step": 6392
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6992888748419726e-06,
+      "loss": 3.9361,
+      "step": 6393
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699241466498104e-06,
+      "loss": 3.9315,
+      "step": 6394
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699194058154235e-06,
+      "loss": 3.8061,
+      "step": 6395
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699146649810367e-06,
+      "loss": 3.9519,
+      "step": 6396
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699099241466498e-06,
+      "loss": 3.5062,
+      "step": 6397
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6990518331226294e-06,
+      "loss": 4.1646,
+      "step": 6398
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.699004424778761e-06,
+      "loss": 3.4083,
+      "step": 6399
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698957016434893e-06,
+      "loss": 3.9385,
+      "step": 6400
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698909608091024e-06,
+      "loss": 3.7382,
+      "step": 6401
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6988621997471554e-06,
+      "loss": 3.5805,
+      "step": 6402
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698814791403287e-06,
+      "loss": 3.6089,
+      "step": 6403
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6987673830594184e-06,
+      "loss": 4.3301,
+      "step": 6404
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.69871997471555e-06,
+      "loss": 3.3812,
+      "step": 6405
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6986725663716818e-06,
+      "loss": 3.7599,
+      "step": 6406
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698625158027813e-06,
+      "loss": 4.0478,
+      "step": 6407
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6985777496839443e-06,
+      "loss": 3.7943,
+      "step": 6408
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698530341340076e-06,
+      "loss": 3.6417,
+      "step": 6409
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6984829329962073e-06,
+      "loss": 4.0298,
+      "step": 6410
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6984355246523386e-06,
+      "loss": 4.0037,
+      "step": 6411
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6983881163084703e-06,
+      "loss": 4.092,
+      "step": 6412
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698340707964602e-06,
+      "loss": 3.8556,
+      "step": 6413
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6982932996207333e-06,
+      "loss": 3.4379,
+      "step": 6414
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698245891276865e-06,
+      "loss": 3.895,
+      "step": 6415
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6981984829329963e-06,
+      "loss": 3.5125,
+      "step": 6416
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6981510745891275e-06,
+      "loss": 3.6608,
+      "step": 6417
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6981036662452592e-06,
+      "loss": 4.0443,
+      "step": 6418
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.698056257901391e-06,
+      "loss": 3.3029,
+      "step": 6419
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6980088495575222e-06,
+      "loss": 3.5342,
+      "step": 6420
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6979614412136535e-06,
+      "loss": 3.2356,
+      "step": 6421
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6979140328697852e-06,
+      "loss": 3.3545,
+      "step": 6422
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6978666245259165e-06,
+      "loss": 4.1295,
+      "step": 6423
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6978192161820478e-06,
+      "loss": 3.7551,
+      "step": 6424
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6977718078381795e-06,
+      "loss": 4.288,
+      "step": 6425
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.697724399494311e-06,
+      "loss": 3.9069,
+      "step": 6426
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2.6976769911504425e-06,
+      "loss": 4.2412,
+      "step": 6427
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.697629582806574e-06,
+      "loss": 3.6585,
+      "step": 6428
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6975821744627054e-06,
+      "loss": 3.7223,
+      "step": 6429
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6975347661188367e-06,
+      "loss": 3.8331,
+      "step": 6430
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6974873577749684e-06,
+      "loss": 3.5963,
+      "step": 6431
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6974399494311e-06,
+      "loss": 3.667,
+      "step": 6432
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6973925410872314e-06,
+      "loss": 4.06,
+      "step": 6433
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.697345132743363e-06,
+      "loss": 3.9456,
+      "step": 6434
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6972977243994944e-06,
+      "loss": 3.2438,
+      "step": 6435
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6972503160556257e-06,
+      "loss": 3.9632,
+      "step": 6436
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6972029077117574e-06,
+      "loss": 4.016,
+      "step": 6437
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6971554993678887e-06,
+      "loss": 3.4978,
+      "step": 6438
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6971080910240204e-06,
+      "loss": 4.1426,
+      "step": 6439
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.697060682680152e-06,
+      "loss": 3.7895,
+      "step": 6440
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6970132743362834e-06,
+      "loss": 3.4729,
+      "step": 6441
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6969658659924146e-06,
+      "loss": 3.4182,
+      "step": 6442
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.696918457648546e-06,
+      "loss": 3.7069,
+      "step": 6443
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6968710493046776e-06,
+      "loss": 3.6743,
+      "step": 6444
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6968236409608093e-06,
+      "loss": 4.1389,
+      "step": 6445
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6967762326169406e-06,
+      "loss": 3.8932,
+      "step": 6446
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6967288242730723e-06,
+      "loss": 3.9017,
+      "step": 6447
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6966814159292036e-06,
+      "loss": 3.7964,
+      "step": 6448
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.696634007585335e-06,
+      "loss": 3.3922,
+      "step": 6449
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6965865992414666e-06,
+      "loss": 3.5143,
+      "step": 6450
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.696539190897598e-06,
+      "loss": 3.0442,
+      "step": 6451
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6964917825537296e-06,
+      "loss": 3.5211,
+      "step": 6452
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6964443742098613e-06,
+      "loss": 3.7814,
+      "step": 6453
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6963969658659925e-06,
+      "loss": 3.8088,
+      "step": 6454
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.696349557522124e-06,
+      "loss": 3.6618,
+      "step": 6455
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6963021491782555e-06,
+      "loss": 3.8648,
+      "step": 6456
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.696254740834387e-06,
+      "loss": 3.5507,
+      "step": 6457
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6962073324905185e-06,
+      "loss": 3.501,
+      "step": 6458
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6961599241466502e-06,
+      "loss": 3.6395,
+      "step": 6459
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6961125158027815e-06,
+      "loss": 3.2463,
+      "step": 6460
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6960651074589128e-06,
+      "loss": 4.0205,
+      "step": 6461
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6960176991150445e-06,
+      "loss": 3.2124,
+      "step": 6462
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6959702907711758e-06,
+      "loss": 3.9658,
+      "step": 6463
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695922882427307e-06,
+      "loss": 3.9544,
+      "step": 6464
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6958754740834387e-06,
+      "loss": 3.647,
+      "step": 6465
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6958280657395704e-06,
+      "loss": 4.1096,
+      "step": 6466
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6957806573957017e-06,
+      "loss": 3.6392,
+      "step": 6467
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695733249051833e-06,
+      "loss": 3.9,
+      "step": 6468
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6956858407079647e-06,
+      "loss": 3.6699,
+      "step": 6469
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695638432364096e-06,
+      "loss": 4.3129,
+      "step": 6470
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6955910240202277e-06,
+      "loss": 3.5377,
+      "step": 6471
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6955436156763594e-06,
+      "loss": 4.2467,
+      "step": 6472
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6954962073324907e-06,
+      "loss": 3.5287,
+      "step": 6473
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695448798988622e-06,
+      "loss": 3.2124,
+      "step": 6474
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6954013906447537e-06,
+      "loss": 3.6593,
+      "step": 6475
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695353982300885e-06,
+      "loss": 4.1917,
+      "step": 6476
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6953065739570162e-06,
+      "loss": 3.8103,
+      "step": 6477
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695259165613148e-06,
+      "loss": 3.7299,
+      "step": 6478
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6952117572692796e-06,
+      "loss": 3.7982,
+      "step": 6479
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695164348925411e-06,
+      "loss": 3.7979,
+      "step": 6480
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6951169405815426e-06,
+      "loss": 4.1287,
+      "step": 6481
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695069532237674e-06,
+      "loss": 4.1184,
+      "step": 6482
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.695022123893805e-06,
+      "loss": 3.7918,
+      "step": 6483
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.694974715549937e-06,
+      "loss": 3.9921,
+      "step": 6484
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6949273072060686e-06,
+      "loss": 3.8936,
+      "step": 6485
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6948798988622e-06,
+      "loss": 3.9461,
+      "step": 6486
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.694832490518331e-06,
+      "loss": 4.1453,
+      "step": 6487
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.694785082174463e-06,
+      "loss": 3.8953,
+      "step": 6488
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.694737673830594e-06,
+      "loss": 4.0065,
+      "step": 6489
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.6946902654867254e-06,
+      "loss": 3.9882,
+      "step": 6490
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2.694642857142857e-06,
+      "loss": 3.7215,
+      "step": 6491
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694595448798989e-06,
+      "loss": 3.4531,
+      "step": 6492
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.69454804045512e-06,
+      "loss": 4.2067,
+      "step": 6493
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694500632111252e-06,
+      "loss": 3.815,
+      "step": 6494
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694453223767383e-06,
+      "loss": 3.8192,
+      "step": 6495
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6944058154235144e-06,
+      "loss": 4.291,
+      "step": 6496
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694358407079646e-06,
+      "loss": 3.6398,
+      "step": 6497
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6943109987357778e-06,
+      "loss": 4.3485,
+      "step": 6498
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694263590391909e-06,
+      "loss": 3.7044,
+      "step": 6499
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6942161820480408e-06,
+      "loss": 3.8826,
+      "step": 6500
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694168773704172e-06,
+      "loss": 3.9406,
+      "step": 6501
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6941213653603033e-06,
+      "loss": 3.3543,
+      "step": 6502
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.694073957016435e-06,
+      "loss": 3.7183,
+      "step": 6503
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6940265486725663e-06,
+      "loss": 3.6164,
+      "step": 6504
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693979140328698e-06,
+      "loss": 3.784,
+      "step": 6505
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6939317319848297e-06,
+      "loss": 3.6865,
+      "step": 6506
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693884323640961e-06,
+      "loss": 3.6627,
+      "step": 6507
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6938369152970923e-06,
+      "loss": 3.6802,
+      "step": 6508
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6937895069532236e-06,
+      "loss": 4.1551,
+      "step": 6509
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6937420986093553e-06,
+      "loss": 4.018,
+      "step": 6510
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693694690265487e-06,
+      "loss": 3.5054,
+      "step": 6511
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6936472819216182e-06,
+      "loss": 3.2995,
+      "step": 6512
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.69359987357775e-06,
+      "loss": 4.0718,
+      "step": 6513
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6935524652338812e-06,
+      "loss": 3.3366,
+      "step": 6514
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6935050568900125e-06,
+      "loss": 3.2989,
+      "step": 6515
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693457648546144e-06,
+      "loss": 3.7735,
+      "step": 6516
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6934102402022755e-06,
+      "loss": 3.5797,
+      "step": 6517
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693362831858407e-06,
+      "loss": 4.1121,
+      "step": 6518
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693315423514539e-06,
+      "loss": 3.6195,
+      "step": 6519
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.69326801517067e-06,
+      "loss": 3.406,
+      "step": 6520
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6932206068268015e-06,
+      "loss": 3.666,
+      "step": 6521
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693173198482933e-06,
+      "loss": 3.4714,
+      "step": 6522
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6931257901390644e-06,
+      "loss": 3.8249,
+      "step": 6523
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693078381795196e-06,
+      "loss": 3.3501,
+      "step": 6524
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.693030973451328e-06,
+      "loss": 3.2428,
+      "step": 6525
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.692983565107459e-06,
+      "loss": 3.5137,
+      "step": 6526
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6929361567635904e-06,
+      "loss": 3.9774,
+      "step": 6527
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.692888748419722e-06,
+      "loss": 4.1243,
+      "step": 6528
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6928413400758534e-06,
+      "loss": 3.8489,
+      "step": 6529
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6927939317319847e-06,
+      "loss": 3.6962,
+      "step": 6530
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6927465233881164e-06,
+      "loss": 3.7805,
+      "step": 6531
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.692699115044248e-06,
+      "loss": 3.8329,
+      "step": 6532
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6926517067003794e-06,
+      "loss": 3.8122,
+      "step": 6533
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6926042983565106e-06,
+      "loss": 3.7114,
+      "step": 6534
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6925568900126423e-06,
+      "loss": 4.1078,
+      "step": 6535
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6925094816687736e-06,
+      "loss": 3.6483,
+      "step": 6536
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6924620733249053e-06,
+      "loss": 4.1728,
+      "step": 6537
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.692414664981037e-06,
+      "loss": 3.3584,
+      "step": 6538
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6923672566371683e-06,
+      "loss": 3.5457,
+      "step": 6539
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6923198482932996e-06,
+      "loss": 3.9745,
+      "step": 6540
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6922724399494313e-06,
+      "loss": 4.2192,
+      "step": 6541
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6922250316055626e-06,
+      "loss": 3.9707,
+      "step": 6542
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.692177623261694e-06,
+      "loss": 3.6539,
+      "step": 6543
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6921302149178256e-06,
+      "loss": 3.9792,
+      "step": 6544
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6920828065739573e-06,
+      "loss": 3.9333,
+      "step": 6545
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6920353982300886e-06,
+      "loss": 3.7604,
+      "step": 6546
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6919879898862203e-06,
+      "loss": 3.7306,
+      "step": 6547
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6919405815423515e-06,
+      "loss": 4.0705,
+      "step": 6548
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.691893173198483e-06,
+      "loss": 3.9516,
+      "step": 6549
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6918457648546145e-06,
+      "loss": 3.9426,
+      "step": 6550
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6917983565107462e-06,
+      "loss": 3.3924,
+      "step": 6551
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6917509481668775e-06,
+      "loss": 3.7732,
+      "step": 6552
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6917035398230088e-06,
+      "loss": 3.5765,
+      "step": 6553
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2.6916561314791405e-06,
+      "loss": 3.3175,
+      "step": 6554
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6916087231352718e-06,
+      "loss": 3.6888,
+      "step": 6555
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.691561314791403e-06,
+      "loss": 3.5897,
+      "step": 6556
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6915139064475348e-06,
+      "loss": 4.078,
+      "step": 6557
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6914664981036665e-06,
+      "loss": 3.6218,
+      "step": 6558
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6914190897597977e-06,
+      "loss": 3.552,
+      "step": 6559
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6913716814159294e-06,
+      "loss": 3.4598,
+      "step": 6560
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6913242730720607e-06,
+      "loss": 3.7873,
+      "step": 6561
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.691276864728192e-06,
+      "loss": 3.6712,
+      "step": 6562
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6912294563843237e-06,
+      "loss": 3.1481,
+      "step": 6563
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6911820480404554e-06,
+      "loss": 4.038,
+      "step": 6564
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6911346396965867e-06,
+      "loss": 3.5729,
+      "step": 6565
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6910872313527184e-06,
+      "loss": 3.4647,
+      "step": 6566
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6910398230088497e-06,
+      "loss": 3.8109,
+      "step": 6567
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690992414664981e-06,
+      "loss": 3.3711,
+      "step": 6568
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6909450063211127e-06,
+      "loss": 3.3782,
+      "step": 6569
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690897597977244e-06,
+      "loss": 3.3708,
+      "step": 6570
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6908501896333756e-06,
+      "loss": 3.5904,
+      "step": 6571
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6908027812895073e-06,
+      "loss": 4.0192,
+      "step": 6572
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6907553729456386e-06,
+      "loss": 3.8819,
+      "step": 6573
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.69070796460177e-06,
+      "loss": 3.887,
+      "step": 6574
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690660556257901e-06,
+      "loss": 3.635,
+      "step": 6575
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690613147914033e-06,
+      "loss": 3.9664,
+      "step": 6576
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690565739570164e-06,
+      "loss": 3.8193,
+      "step": 6577
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690518331226296e-06,
+      "loss": 3.4522,
+      "step": 6578
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6904709228824276e-06,
+      "loss": 3.8046,
+      "step": 6579
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690423514538559e-06,
+      "loss": 3.3814,
+      "step": 6580
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.69037610619469e-06,
+      "loss": 3.8212,
+      "step": 6581
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690328697850822e-06,
+      "loss": 3.9314,
+      "step": 6582
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690281289506953e-06,
+      "loss": 3.4919,
+      "step": 6583
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690233881163085e-06,
+      "loss": 3.5253,
+      "step": 6584
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6901864728192165e-06,
+      "loss": 4.0879,
+      "step": 6585
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690139064475348e-06,
+      "loss": 3.3902,
+      "step": 6586
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690091656131479e-06,
+      "loss": 3.8481,
+      "step": 6587
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.690044247787611e-06,
+      "loss": 3.57,
+      "step": 6588
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689996839443742e-06,
+      "loss": 3.8919,
+      "step": 6589
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6899494310998734e-06,
+      "loss": 3.4538,
+      "step": 6590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689902022756005e-06,
+      "loss": 4.0375,
+      "step": 6591
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6898546144121368e-06,
+      "loss": 3.9558,
+      "step": 6592
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689807206068268e-06,
+      "loss": 4.1182,
+      "step": 6593
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6897597977243998e-06,
+      "loss": 3.9056,
+      "step": 6594
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689712389380531e-06,
+      "loss": 3.8114,
+      "step": 6595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6896649810366623e-06,
+      "loss": 3.6314,
+      "step": 6596
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689617572692794e-06,
+      "loss": 3.7324,
+      "step": 6597
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6895701643489257e-06,
+      "loss": 3.6632,
+      "step": 6598
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689522756005057e-06,
+      "loss": 3.4825,
+      "step": 6599
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6894753476611883e-06,
+      "loss": 3.5493,
+      "step": 6600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.68942793931732e-06,
+      "loss": 3.7411,
+      "step": 6601
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6893805309734513e-06,
+      "loss": 3.894,
+      "step": 6602
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6893331226295825e-06,
+      "loss": 3.6582,
+      "step": 6603
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6892857142857143e-06,
+      "loss": 4.085,
+      "step": 6604
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689238305941846e-06,
+      "loss": 3.8808,
+      "step": 6605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6891908975979772e-06,
+      "loss": 4.1299,
+      "step": 6606
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689143489254109e-06,
+      "loss": 4.1848,
+      "step": 6607
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6890960809102402e-06,
+      "loss": 3.7107,
+      "step": 6608
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6890486725663715e-06,
+      "loss": 3.2036,
+      "step": 6609
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.689001264222503e-06,
+      "loss": 3.366,
+      "step": 6610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.688953855878635e-06,
+      "loss": 4.134,
+      "step": 6611
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.688906447534766e-06,
+      "loss": 3.6002,
+      "step": 6612
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.688859039190898e-06,
+      "loss": 3.9114,
+      "step": 6613
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.688811630847029e-06,
+      "loss": 3.6418,
+      "step": 6614
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6887642225031605e-06,
+      "loss": 3.9654,
+      "step": 6615
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.688716814159292e-06,
+      "loss": 4.1393,
+      "step": 6616
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2.6886694058154234e-06,
+      "loss": 3.5628,
+      "step": 6617
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.688621997471555e-06,
+      "loss": 3.2316,
+      "step": 6618
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6885745891276864e-06,
+      "loss": 3.3878,
+      "step": 6619
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.688527180783818e-06,
+      "loss": 3.876,
+      "step": 6620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6884797724399494e-06,
+      "loss": 3.9112,
+      "step": 6621
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6884323640960807e-06,
+      "loss": 3.6415,
+      "step": 6622
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6883849557522124e-06,
+      "loss": 3.6551,
+      "step": 6623
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.688337547408344e-06,
+      "loss": 3.8386,
+      "step": 6624
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6882901390644754e-06,
+      "loss": 4.391,
+      "step": 6625
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.688242730720607e-06,
+      "loss": 3.2696,
+      "step": 6626
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6881953223767384e-06,
+      "loss": 3.3073,
+      "step": 6627
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6881479140328696e-06,
+      "loss": 3.5939,
+      "step": 6628
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6881005056890013e-06,
+      "loss": 3.2795,
+      "step": 6629
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6880530973451326e-06,
+      "loss": 4.1852,
+      "step": 6630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6880056890012643e-06,
+      "loss": 3.4802,
+      "step": 6631
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.687958280657396e-06,
+      "loss": 4.1375,
+      "step": 6632
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6879108723135273e-06,
+      "loss": 3.5418,
+      "step": 6633
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6878634639696586e-06,
+      "loss": 3.3968,
+      "step": 6634
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6878160556257903e-06,
+      "loss": 3.9343,
+      "step": 6635
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6877686472819216e-06,
+      "loss": 3.5965,
+      "step": 6636
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6877212389380533e-06,
+      "loss": 3.5636,
+      "step": 6637
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.687673830594185e-06,
+      "loss": 3.5326,
+      "step": 6638
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6876264222503163e-06,
+      "loss": 3.4606,
+      "step": 6639
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6875790139064475e-06,
+      "loss": 3.9988,
+      "step": 6640
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.687531605562579e-06,
+      "loss": 4.0209,
+      "step": 6641
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6874841972187105e-06,
+      "loss": 3.5444,
+      "step": 6642
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.687436788874842e-06,
+      "loss": 3.5436,
+      "step": 6643
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6873893805309735e-06,
+      "loss": 3.6091,
+      "step": 6644
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6873419721871052e-06,
+      "loss": 4.0119,
+      "step": 6645
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6872945638432365e-06,
+      "loss": 3.3578,
+      "step": 6646
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6872471554993678e-06,
+      "loss": 3.2073,
+      "step": 6647
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6871997471554995e-06,
+      "loss": 3.8572,
+      "step": 6648
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6871523388116308e-06,
+      "loss": 3.9856,
+      "step": 6649
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6871049304677625e-06,
+      "loss": 3.2721,
+      "step": 6650
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.687057522123894e-06,
+      "loss": 4.0237,
+      "step": 6651
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6870101137800255e-06,
+      "loss": 3.0783,
+      "step": 6652
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6869627054361567e-06,
+      "loss": 3.5969,
+      "step": 6653
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6869152970922884e-06,
+      "loss": 4.0592,
+      "step": 6654
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6868678887484197e-06,
+      "loss": 2.8928,
+      "step": 6655
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.686820480404551e-06,
+      "loss": 3.6943,
+      "step": 6656
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6867730720606827e-06,
+      "loss": 3.7918,
+      "step": 6657
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6867256637168144e-06,
+      "loss": 3.9468,
+      "step": 6658
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6866782553729457e-06,
+      "loss": 3.8742,
+      "step": 6659
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.686630847029077e-06,
+      "loss": 3.913,
+      "step": 6660
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6865834386852087e-06,
+      "loss": 3.8921,
+      "step": 6661
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.68653603034134e-06,
+      "loss": 3.7408,
+      "step": 6662
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6864886219974717e-06,
+      "loss": 3.4565,
+      "step": 6663
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6864412136536034e-06,
+      "loss": 3.4813,
+      "step": 6664
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6863938053097346e-06,
+      "loss": 3.9855,
+      "step": 6665
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.686346396965866e-06,
+      "loss": 3.5275,
+      "step": 6666
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6862989886219976e-06,
+      "loss": 3.9067,
+      "step": 6667
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.686251580278129e-06,
+      "loss": 4.169,
+      "step": 6668
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.68620417193426e-06,
+      "loss": 3.5494,
+      "step": 6669
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.686156763590392e-06,
+      "loss": 3.7044,
+      "step": 6670
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6861093552465236e-06,
+      "loss": 4.1067,
+      "step": 6671
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.686061946902655e-06,
+      "loss": 2.9708,
+      "step": 6672
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6860145385587866e-06,
+      "loss": 3.6635,
+      "step": 6673
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.685967130214918e-06,
+      "loss": 3.7308,
+      "step": 6674
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.685919721871049e-06,
+      "loss": 3.3119,
+      "step": 6675
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.685872313527181e-06,
+      "loss": 3.5158,
+      "step": 6676
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6858249051833125e-06,
+      "loss": 4.1004,
+      "step": 6677
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.685777496839444e-06,
+      "loss": 3.6835,
+      "step": 6678
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.6857300884955755e-06,
+      "loss": 3.3892,
+      "step": 6679
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.685682680151707e-06,
+      "loss": 3.9578,
+      "step": 6680
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2.685635271807838e-06,
+      "loss": 3.6705,
+      "step": 6681
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6855878634639694e-06,
+      "loss": 4.1254,
+      "step": 6682
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.685540455120101e-06,
+      "loss": 3.8861,
+      "step": 6683
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6854930467762328e-06,
+      "loss": 3.5365,
+      "step": 6684
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.685445638432364e-06,
+      "loss": 4.0215,
+      "step": 6685
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6853982300884958e-06,
+      "loss": 4.2623,
+      "step": 6686
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.685350821744627e-06,
+      "loss": 3.1654,
+      "step": 6687
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6853034134007583e-06,
+      "loss": 3.8478,
+      "step": 6688
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.68525600505689e-06,
+      "loss": 3.5304,
+      "step": 6689
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6852085967130217e-06,
+      "loss": 3.2797,
+      "step": 6690
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.685161188369153e-06,
+      "loss": 3.7442,
+      "step": 6691
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6851137800252847e-06,
+      "loss": 2.9725,
+      "step": 6692
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.685066371681416e-06,
+      "loss": 3.8482,
+      "step": 6693
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6850189633375473e-06,
+      "loss": 3.6924,
+      "step": 6694
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684971554993679e-06,
+      "loss": 3.8611,
+      "step": 6695
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6849241466498103e-06,
+      "loss": 3.7028,
+      "step": 6696
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684876738305942e-06,
+      "loss": 4.1286,
+      "step": 6697
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6848293299620737e-06,
+      "loss": 4.3509,
+      "step": 6698
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684781921618205e-06,
+      "loss": 4.1249,
+      "step": 6699
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6847345132743362e-06,
+      "loss": 3.8286,
+      "step": 6700
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684687104930468e-06,
+      "loss": 3.7561,
+      "step": 6701
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6846396965865992e-06,
+      "loss": 3.9127,
+      "step": 6702
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684592288242731e-06,
+      "loss": 3.7985,
+      "step": 6703
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684544879898862e-06,
+      "loss": 3.9211,
+      "step": 6704
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684497471554994e-06,
+      "loss": 3.6196,
+      "step": 6705
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684450063211125e-06,
+      "loss": 3.6471,
+      "step": 6706
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6844026548672565e-06,
+      "loss": 3.8595,
+      "step": 6707
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684355246523388e-06,
+      "loss": 3.4528,
+      "step": 6708
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6843078381795194e-06,
+      "loss": 3.4595,
+      "step": 6709
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684260429835651e-06,
+      "loss": 3.4263,
+      "step": 6710
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684213021491783e-06,
+      "loss": 3.9349,
+      "step": 6711
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684165613147914e-06,
+      "loss": 3.5392,
+      "step": 6712
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6841182048040454e-06,
+      "loss": 3.7794,
+      "step": 6713
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.684070796460177e-06,
+      "loss": 3.905,
+      "step": 6714
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6840233881163084e-06,
+      "loss": 3.2923,
+      "step": 6715
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.68397597977244e-06,
+      "loss": 3.9449,
+      "step": 6716
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.683928571428572e-06,
+      "loss": 3.4514,
+      "step": 6717
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.683881163084703e-06,
+      "loss": 3.4531,
+      "step": 6718
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6838337547408344e-06,
+      "loss": 4.1405,
+      "step": 6719
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.683786346396966e-06,
+      "loss": 3.3547,
+      "step": 6720
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6837389380530974e-06,
+      "loss": 3.4868,
+      "step": 6721
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6836915297092286e-06,
+      "loss": 3.2854,
+      "step": 6722
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6836441213653603e-06,
+      "loss": 3.867,
+      "step": 6723
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.683596713021492e-06,
+      "loss": 3.5514,
+      "step": 6724
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6835493046776233e-06,
+      "loss": 3.7014,
+      "step": 6725
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6835018963337546e-06,
+      "loss": 3.8436,
+      "step": 6726
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6834544879898863e-06,
+      "loss": 4.3396,
+      "step": 6727
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6834070796460176e-06,
+      "loss": 3.6979,
+      "step": 6728
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6833596713021493e-06,
+      "loss": 3.5489,
+      "step": 6729
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.683312262958281e-06,
+      "loss": 3.6954,
+      "step": 6730
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6832648546144123e-06,
+      "loss": 3.7336,
+      "step": 6731
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6832174462705436e-06,
+      "loss": 3.9149,
+      "step": 6732
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6831700379266753e-06,
+      "loss": 3.6964,
+      "step": 6733
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6831226295828065e-06,
+      "loss": 3.8438,
+      "step": 6734
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.683075221238938e-06,
+      "loss": 3.4804,
+      "step": 6735
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6830278128950695e-06,
+      "loss": 3.8,
+      "step": 6736
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6829804045512012e-06,
+      "loss": 3.652,
+      "step": 6737
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6829329962073325e-06,
+      "loss": 4.0472,
+      "step": 6738
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6828855878634642e-06,
+      "loss": 3.6118,
+      "step": 6739
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6828381795195955e-06,
+      "loss": 3.8301,
+      "step": 6740
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6827907711757268e-06,
+      "loss": 4.1462,
+      "step": 6741
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6827433628318585e-06,
+      "loss": 3.0986,
+      "step": 6742
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.68269595448799e-06,
+      "loss": 4.0986,
+      "step": 6743
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2.6826485461441215e-06,
+      "loss": 3.79,
+      "step": 6744
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.682601137800253e-06,
+      "loss": 3.1376,
+      "step": 6745
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6825537294563844e-06,
+      "loss": 3.6035,
+      "step": 6746
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6825063211125157e-06,
+      "loss": 3.6219,
+      "step": 6747
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.682458912768647e-06,
+      "loss": 3.7477,
+      "step": 6748
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6824115044247787e-06,
+      "loss": 3.8054,
+      "step": 6749
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6823640960809104e-06,
+      "loss": 3.2166,
+      "step": 6750
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6823166877370417e-06,
+      "loss": 4.2484,
+      "step": 6751
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6822692793931734e-06,
+      "loss": 3.6385,
+      "step": 6752
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6822218710493047e-06,
+      "loss": 3.6899,
+      "step": 6753
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.682174462705436e-06,
+      "loss": 3.7014,
+      "step": 6754
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6821270543615677e-06,
+      "loss": 3.6171,
+      "step": 6755
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6820796460176994e-06,
+      "loss": 3.9399,
+      "step": 6756
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6820322376738306e-06,
+      "loss": 4.1173,
+      "step": 6757
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6819848293299624e-06,
+      "loss": 4.1699,
+      "step": 6758
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6819374209860936e-06,
+      "loss": 3.8982,
+      "step": 6759
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681890012642225e-06,
+      "loss": 4.1465,
+      "step": 6760
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6818426042983566e-06,
+      "loss": 3.5019,
+      "step": 6761
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681795195954488e-06,
+      "loss": 3.8276,
+      "step": 6762
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6817477876106196e-06,
+      "loss": 3.4111,
+      "step": 6763
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6817003792667513e-06,
+      "loss": 3.3192,
+      "step": 6764
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6816529709228826e-06,
+      "loss": 3.9399,
+      "step": 6765
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681605562579014e-06,
+      "loss": 4.1668,
+      "step": 6766
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6815581542351456e-06,
+      "loss": 3.4101,
+      "step": 6767
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681510745891277e-06,
+      "loss": 3.8794,
+      "step": 6768
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6814633375474086e-06,
+      "loss": 3.8088,
+      "step": 6769
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.68141592920354e-06,
+      "loss": 4.0874,
+      "step": 6770
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6813685208596715e-06,
+      "loss": 3.5589,
+      "step": 6771
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681321112515803e-06,
+      "loss": 3.5742,
+      "step": 6772
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681273704171934e-06,
+      "loss": 3.9393,
+      "step": 6773
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681226295828066e-06,
+      "loss": 3.9866,
+      "step": 6774
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.681178887484197e-06,
+      "loss": 3.9555,
+      "step": 6775
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6811314791403288e-06,
+      "loss": 3.9165,
+      "step": 6776
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6810840707964605e-06,
+      "loss": 3.7995,
+      "step": 6777
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6810366624525918e-06,
+      "loss": 3.6872,
+      "step": 6778
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680989254108723e-06,
+      "loss": 4.191,
+      "step": 6779
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6809418457648548e-06,
+      "loss": 3.8416,
+      "step": 6780
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680894437420986e-06,
+      "loss": 3.9017,
+      "step": 6781
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6808470290771177e-06,
+      "loss": 3.8696,
+      "step": 6782
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6807996207332494e-06,
+      "loss": 3.46,
+      "step": 6783
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6807522123893807e-06,
+      "loss": 4.0029,
+      "step": 6784
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680704804045512e-06,
+      "loss": 3.4617,
+      "step": 6785
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6806573957016437e-06,
+      "loss": 4.3198,
+      "step": 6786
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680609987357775e-06,
+      "loss": 2.6921,
+      "step": 6787
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6805625790139063e-06,
+      "loss": 3.7044,
+      "step": 6788
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680515170670038e-06,
+      "loss": 3.9199,
+      "step": 6789
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6804677623261697e-06,
+      "loss": 3.4203,
+      "step": 6790
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680420353982301e-06,
+      "loss": 4.2678,
+      "step": 6791
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6803729456384322e-06,
+      "loss": 4.016,
+      "step": 6792
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680325537294564e-06,
+      "loss": 3.8142,
+      "step": 6793
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6802781289506952e-06,
+      "loss": 3.919,
+      "step": 6794
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680230720606827e-06,
+      "loss": 3.4563,
+      "step": 6795
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6801833122629586e-06,
+      "loss": 3.6462,
+      "step": 6796
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.68013590391909e-06,
+      "loss": 3.8299,
+      "step": 6797
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680088495575221e-06,
+      "loss": 3.8915,
+      "step": 6798
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.680041087231353e-06,
+      "loss": 4.0755,
+      "step": 6799
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.679993678887484e-06,
+      "loss": 3.3659,
+      "step": 6800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6799462705436155e-06,
+      "loss": 3.5971,
+      "step": 6801
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.679898862199747e-06,
+      "loss": 3.6214,
+      "step": 6802
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.679851453855879e-06,
+      "loss": 3.8397,
+      "step": 6803
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.67980404551201e-06,
+      "loss": 3.9277,
+      "step": 6804
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.679756637168142e-06,
+      "loss": 4.0454,
+      "step": 6805
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.679709228824273e-06,
+      "loss": 4.1069,
+      "step": 6806
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2.6796618204804044e-06,
+      "loss": 3.9159,
+      "step": 6807
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679614412136536e-06,
+      "loss": 3.5922,
+      "step": 6808
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679567003792668e-06,
+      "loss": 4.0239,
+      "step": 6809
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679519595448799e-06,
+      "loss": 4.1162,
+      "step": 6810
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679472187104931e-06,
+      "loss": 3.6044,
+      "step": 6811
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679424778761062e-06,
+      "loss": 4.0664,
+      "step": 6812
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6793773704171934e-06,
+      "loss": 3.5313,
+      "step": 6813
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6793299620733246e-06,
+      "loss": 3.662,
+      "step": 6814
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6792825537294563e-06,
+      "loss": 3.9973,
+      "step": 6815
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679235145385588e-06,
+      "loss": 3.6503,
+      "step": 6816
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6791877370417193e-06,
+      "loss": 3.9218,
+      "step": 6817
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.679140328697851e-06,
+      "loss": 4.1156,
+      "step": 6818
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6790929203539823e-06,
+      "loss": 3.8999,
+      "step": 6819
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6790455120101136e-06,
+      "loss": 3.6559,
+      "step": 6820
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6789981036662453e-06,
+      "loss": 3.7582,
+      "step": 6821
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.678950695322377e-06,
+      "loss": 4.2124,
+      "step": 6822
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6789032869785083e-06,
+      "loss": 3.7839,
+      "step": 6823
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.67885587863464e-06,
+      "loss": 3.7595,
+      "step": 6824
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6788084702907713e-06,
+      "loss": 3.6189,
+      "step": 6825
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6787610619469025e-06,
+      "loss": 3.9751,
+      "step": 6826
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6787136536030343e-06,
+      "loss": 3.6432,
+      "step": 6827
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6786662452591655e-06,
+      "loss": 3.4147,
+      "step": 6828
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6786188369152972e-06,
+      "loss": 4.1153,
+      "step": 6829
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.678571428571429e-06,
+      "loss": 3.6698,
+      "step": 6830
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6785240202275602e-06,
+      "loss": 3.7649,
+      "step": 6831
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6784766118836915e-06,
+      "loss": 3.9489,
+      "step": 6832
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.678429203539823e-06,
+      "loss": 3.551,
+      "step": 6833
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6783817951959545e-06,
+      "loss": 4.1306,
+      "step": 6834
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.678334386852086e-06,
+      "loss": 4.2042,
+      "step": 6835
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6782869785082175e-06,
+      "loss": 3.6075,
+      "step": 6836
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.678239570164349e-06,
+      "loss": 3.1497,
+      "step": 6837
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6781921618204805e-06,
+      "loss": 4.0315,
+      "step": 6838
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6781447534766117e-06,
+      "loss": 4.1027,
+      "step": 6839
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6780973451327434e-06,
+      "loss": 3.3581,
+      "step": 6840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6780499367888747e-06,
+      "loss": 3.8092,
+      "step": 6841
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6780025284450064e-06,
+      "loss": 2.8497,
+      "step": 6842
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677955120101138e-06,
+      "loss": 3.6737,
+      "step": 6843
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6779077117572694e-06,
+      "loss": 4.0174,
+      "step": 6844
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6778603034134007e-06,
+      "loss": 4.0906,
+      "step": 6845
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6778128950695324e-06,
+      "loss": 3.4044,
+      "step": 6846
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6777654867256637e-06,
+      "loss": 3.6921,
+      "step": 6847
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677718078381795e-06,
+      "loss": 3.9571,
+      "step": 6848
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677670670037927e-06,
+      "loss": 4.1703,
+      "step": 6849
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6776232616940584e-06,
+      "loss": 3.4625,
+      "step": 6850
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6775758533501896e-06,
+      "loss": 3.384,
+      "step": 6851
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6775284450063213e-06,
+      "loss": 3.5076,
+      "step": 6852
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6774810366624526e-06,
+      "loss": 3.61,
+      "step": 6853
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677433628318584e-06,
+      "loss": 3.63,
+      "step": 6854
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6773862199747156e-06,
+      "loss": 3.5996,
+      "step": 6855
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6773388116308473e-06,
+      "loss": 3.5542,
+      "step": 6856
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6772914032869786e-06,
+      "loss": 3.6841,
+      "step": 6857
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.67724399494311e-06,
+      "loss": 4.2536,
+      "step": 6858
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6771965865992416e-06,
+      "loss": 3.7953,
+      "step": 6859
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677149178255373e-06,
+      "loss": 3.569,
+      "step": 6860
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677101769911504e-06,
+      "loss": 3.6638,
+      "step": 6861
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.677054361567636e-06,
+      "loss": 4.3127,
+      "step": 6862
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6770069532237675e-06,
+      "loss": 3.6953,
+      "step": 6863
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.676959544879899e-06,
+      "loss": 4.09,
+      "step": 6864
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6769121365360305e-06,
+      "loss": 4.1503,
+      "step": 6865
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.676864728192162e-06,
+      "loss": 3.6371,
+      "step": 6866
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.676817319848293e-06,
+      "loss": 2.944,
+      "step": 6867
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.676769911504425e-06,
+      "loss": 3.9112,
+      "step": 6868
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6767225031605565e-06,
+      "loss": 3.7013,
+      "step": 6869
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6766750948166878e-06,
+      "loss": 4.0919,
+      "step": 6870
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2.6766276864728195e-06,
+      "loss": 4.0408,
+      "step": 6871
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6765802781289508e-06,
+      "loss": 3.9282,
+      "step": 6872
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.676532869785082e-06,
+      "loss": 3.9995,
+      "step": 6873
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6764854614412138e-06,
+      "loss": 3.8003,
+      "step": 6874
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.676438053097345e-06,
+      "loss": 3.708,
+      "step": 6875
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6763906447534767e-06,
+      "loss": 3.3947,
+      "step": 6876
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6763432364096084e-06,
+      "loss": 3.6584,
+      "step": 6877
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6762958280657397e-06,
+      "loss": 3.5041,
+      "step": 6878
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.676248419721871e-06,
+      "loss": 4.4456,
+      "step": 6879
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6762010113780023e-06,
+      "loss": 3.2319,
+      "step": 6880
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.676153603034134e-06,
+      "loss": 3.5777,
+      "step": 6881
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6761061946902657e-06,
+      "loss": 4.2399,
+      "step": 6882
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.676058786346397e-06,
+      "loss": 3.9619,
+      "step": 6883
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6760113780025287e-06,
+      "loss": 3.877,
+      "step": 6884
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.67596396965866e-06,
+      "loss": 3.62,
+      "step": 6885
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6759165613147912e-06,
+      "loss": 4.0399,
+      "step": 6886
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675869152970923e-06,
+      "loss": 3.6921,
+      "step": 6887
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6758217446270542e-06,
+      "loss": 3.8914,
+      "step": 6888
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675774336283186e-06,
+      "loss": 3.9918,
+      "step": 6889
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6757269279393176e-06,
+      "loss": 3.8168,
+      "step": 6890
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675679519595449e-06,
+      "loss": 3.6652,
+      "step": 6891
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.67563211125158e-06,
+      "loss": 4.0046,
+      "step": 6892
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675584702907712e-06,
+      "loss": 3.4266,
+      "step": 6893
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675537294563843e-06,
+      "loss": 3.8879,
+      "step": 6894
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675489886219975e-06,
+      "loss": 3.9302,
+      "step": 6895
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6754424778761066e-06,
+      "loss": 3.0915,
+      "step": 6896
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675395069532238e-06,
+      "loss": 3.8029,
+      "step": 6897
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675347661188369e-06,
+      "loss": 3.9003,
+      "step": 6898
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675300252844501e-06,
+      "loss": 3.7666,
+      "step": 6899
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675252844500632e-06,
+      "loss": 3.9325,
+      "step": 6900
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6752054361567634e-06,
+      "loss": 3.9406,
+      "step": 6901
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675158027812895e-06,
+      "loss": 4.1055,
+      "step": 6902
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675110619469027e-06,
+      "loss": 4.1117,
+      "step": 6903
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.675063211125158e-06,
+      "loss": 3.7515,
+      "step": 6904
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6750158027812894e-06,
+      "loss": 3.3803,
+      "step": 6905
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.674968394437421e-06,
+      "loss": 3.932,
+      "step": 6906
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6749209860935524e-06,
+      "loss": 3.5131,
+      "step": 6907
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.674873577749684e-06,
+      "loss": 3.6118,
+      "step": 6908
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6748261694058158e-06,
+      "loss": 3.3802,
+      "step": 6909
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.674778761061947e-06,
+      "loss": 3.9875,
+      "step": 6910
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6747313527180783e-06,
+      "loss": 3.8753,
+      "step": 6911
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.67468394437421e-06,
+      "loss": 3.4158,
+      "step": 6912
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6746365360303413e-06,
+      "loss": 3.7955,
+      "step": 6913
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6745891276864726e-06,
+      "loss": 3.9198,
+      "step": 6914
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6745417193426043e-06,
+      "loss": 3.0122,
+      "step": 6915
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.674494310998736e-06,
+      "loss": 3.5052,
+      "step": 6916
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6744469026548673e-06,
+      "loss": 3.8243,
+      "step": 6917
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.674399494310999e-06,
+      "loss": 3.5972,
+      "step": 6918
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6743520859671303e-06,
+      "loss": 3.771,
+      "step": 6919
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6743046776232615e-06,
+      "loss": 4.0618,
+      "step": 6920
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6742572692793932e-06,
+      "loss": 3.8196,
+      "step": 6921
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.674209860935525e-06,
+      "loss": 4.1161,
+      "step": 6922
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6741624525916562e-06,
+      "loss": 3.8215,
+      "step": 6923
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6741150442477875e-06,
+      "loss": 3.6607,
+      "step": 6924
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6740676359039192e-06,
+      "loss": 3.6884,
+      "step": 6925
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6740202275600505e-06,
+      "loss": 3.9649,
+      "step": 6926
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6739728192161818e-06,
+      "loss": 3.7721,
+      "step": 6927
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6739254108723135e-06,
+      "loss": 3.2198,
+      "step": 6928
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.673878002528445e-06,
+      "loss": 3.6823,
+      "step": 6929
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6738305941845765e-06,
+      "loss": 3.2594,
+      "step": 6930
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.673783185840708e-06,
+      "loss": 3.9199,
+      "step": 6931
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6737357774968394e-06,
+      "loss": 4.2366,
+      "step": 6932
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6736883691529707e-06,
+      "loss": 4.0108,
+      "step": 6933
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2.6736409608091024e-06,
+      "loss": 3.4569,
+      "step": 6934
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.673593552465234e-06,
+      "loss": 4.1979,
+      "step": 6935
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6735461441213654e-06,
+      "loss": 3.7034,
+      "step": 6936
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.673498735777497e-06,
+      "loss": 3.8947,
+      "step": 6937
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6734513274336284e-06,
+      "loss": 3.4798,
+      "step": 6938
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6734039190897597e-06,
+      "loss": 3.527,
+      "step": 6939
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6733565107458914e-06,
+      "loss": 3.7974,
+      "step": 6940
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6733091024020227e-06,
+      "loss": 4.1023,
+      "step": 6941
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6732616940581544e-06,
+      "loss": 4.072,
+      "step": 6942
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.673214285714286e-06,
+      "loss": 3.4905,
+      "step": 6943
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6731668773704174e-06,
+      "loss": 3.9654,
+      "step": 6944
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6731194690265486e-06,
+      "loss": 4.2008,
+      "step": 6945
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.67307206068268e-06,
+      "loss": 3.3514,
+      "step": 6946
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6730246523388116e-06,
+      "loss": 3.6982,
+      "step": 6947
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6729772439949433e-06,
+      "loss": 3.4637,
+      "step": 6948
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6729298356510746e-06,
+      "loss": 3.7692,
+      "step": 6949
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6728824273072063e-06,
+      "loss": 3.9464,
+      "step": 6950
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6728350189633376e-06,
+      "loss": 3.9016,
+      "step": 6951
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.672787610619469e-06,
+      "loss": 4.3033,
+      "step": 6952
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6727402022756006e-06,
+      "loss": 3.1719,
+      "step": 6953
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.672692793931732e-06,
+      "loss": 3.5028,
+      "step": 6954
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6726453855878636e-06,
+      "loss": 3.2191,
+      "step": 6955
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6725979772439953e-06,
+      "loss": 3.8149,
+      "step": 6956
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6725505689001265e-06,
+      "loss": 3.8019,
+      "step": 6957
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.672503160556258e-06,
+      "loss": 3.7683,
+      "step": 6958
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6724557522123895e-06,
+      "loss": 3.5075,
+      "step": 6959
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.672408343868521e-06,
+      "loss": 3.8207,
+      "step": 6960
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6723609355246525e-06,
+      "loss": 4.1388,
+      "step": 6961
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6723135271807842e-06,
+      "loss": 3.4954,
+      "step": 6962
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6722661188369155e-06,
+      "loss": 4.3022,
+      "step": 6963
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6722187104930468e-06,
+      "loss": 3.547,
+      "step": 6964
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6721713021491785e-06,
+      "loss": 4.0236,
+      "step": 6965
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6721238938053098e-06,
+      "loss": 3.664,
+      "step": 6966
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.672076485461441e-06,
+      "loss": 2.9909,
+      "step": 6967
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6720290771175727e-06,
+      "loss": 3.4177,
+      "step": 6968
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6719816687737044e-06,
+      "loss": 3.8911,
+      "step": 6969
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6719342604298357e-06,
+      "loss": 3.4381,
+      "step": 6970
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671886852085967e-06,
+      "loss": 3.8637,
+      "step": 6971
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6718394437420987e-06,
+      "loss": 3.0248,
+      "step": 6972
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.67179203539823e-06,
+      "loss": 3.7424,
+      "step": 6973
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6717446270543617e-06,
+      "loss": 3.5789,
+      "step": 6974
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6716972187104934e-06,
+      "loss": 3.4589,
+      "step": 6975
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6716498103666247e-06,
+      "loss": 3.3613,
+      "step": 6976
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671602402022756e-06,
+      "loss": 3.9982,
+      "step": 6977
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6715549936788877e-06,
+      "loss": 3.625,
+      "step": 6978
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671507585335019e-06,
+      "loss": 4.1265,
+      "step": 6979
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6714601769911502e-06,
+      "loss": 3.7303,
+      "step": 6980
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671412768647282e-06,
+      "loss": 3.6846,
+      "step": 6981
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6713653603034136e-06,
+      "loss": 3.809,
+      "step": 6982
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671317951959545e-06,
+      "loss": 3.853,
+      "step": 6983
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6712705436156766e-06,
+      "loss": 3.7791,
+      "step": 6984
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671223135271808e-06,
+      "loss": 3.3334,
+      "step": 6985
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671175726927939e-06,
+      "loss": 3.9982,
+      "step": 6986
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671128318584071e-06,
+      "loss": 3.5221,
+      "step": 6987
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6710809102402026e-06,
+      "loss": 3.2922,
+      "step": 6988
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.671033501896334e-06,
+      "loss": 3.768,
+      "step": 6989
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670986093552465e-06,
+      "loss": 3.9305,
+      "step": 6990
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670938685208597e-06,
+      "loss": 3.0468,
+      "step": 6991
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670891276864728e-06,
+      "loss": 3.3453,
+      "step": 6992
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.6708438685208594e-06,
+      "loss": 3.6699,
+      "step": 6993
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670796460176991e-06,
+      "loss": 3.2406,
+      "step": 6994
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670749051833123e-06,
+      "loss": 3.8282,
+      "step": 6995
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670701643489254e-06,
+      "loss": 3.6236,
+      "step": 6996
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2.670654235145386e-06,
+      "loss": 3.6592,
+      "step": 6997
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.670606826801517e-06,
+      "loss": 3.9921,
+      "step": 6998
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6705594184576484e-06,
+      "loss": 3.3442,
+      "step": 6999
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.67051201011378e-06,
+      "loss": 3.7834,
+      "step": 7000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6704646017699118e-06,
+      "loss": 4.0131,
+      "step": 7001
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.670417193426043e-06,
+      "loss": 3.9974,
+      "step": 7002
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6703697850821748e-06,
+      "loss": 3.5378,
+      "step": 7003
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.670322376738306e-06,
+      "loss": 3.9102,
+      "step": 7004
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6702749683944373e-06,
+      "loss": 3.8183,
+      "step": 7005
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.670227560050569e-06,
+      "loss": 3.6669,
+      "step": 7006
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6701801517067003e-06,
+      "loss": 3.4572,
+      "step": 7007
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.670132743362832e-06,
+      "loss": 3.8038,
+      "step": 7008
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6700853350189637e-06,
+      "loss": 3.7224,
+      "step": 7009
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.670037926675095e-06,
+      "loss": 3.3031,
+      "step": 7010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6699905183312263e-06,
+      "loss": 3.6876,
+      "step": 7011
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6699431099873576e-06,
+      "loss": 3.8396,
+      "step": 7012
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6698957016434893e-06,
+      "loss": 3.0581,
+      "step": 7013
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669848293299621e-06,
+      "loss": 4.0329,
+      "step": 7014
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6698008849557522e-06,
+      "loss": 3.4764,
+      "step": 7015
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669753476611884e-06,
+      "loss": 3.557,
+      "step": 7016
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6697060682680152e-06,
+      "loss": 3.6192,
+      "step": 7017
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6696586599241465e-06,
+      "loss": 3.8484,
+      "step": 7018
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669611251580278e-06,
+      "loss": 3.9156,
+      "step": 7019
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6695638432364095e-06,
+      "loss": 3.3326,
+      "step": 7020
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669516434892541e-06,
+      "loss": 3.7713,
+      "step": 7021
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669469026548673e-06,
+      "loss": 3.7275,
+      "step": 7022
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669421618204804e-06,
+      "loss": 3.6975,
+      "step": 7023
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6693742098609355e-06,
+      "loss": 3.2109,
+      "step": 7024
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669326801517067e-06,
+      "loss": 3.3758,
+      "step": 7025
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6692793931731984e-06,
+      "loss": 3.6289,
+      "step": 7026
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.66923198482933e-06,
+      "loss": 3.5744,
+      "step": 7027
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669184576485462e-06,
+      "loss": 3.526,
+      "step": 7028
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669137168141593e-06,
+      "loss": 3.6348,
+      "step": 7029
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6690897597977244e-06,
+      "loss": 3.6918,
+      "step": 7030
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.669042351453856e-06,
+      "loss": 3.669,
+      "step": 7031
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6689949431099874e-06,
+      "loss": 3.9153,
+      "step": 7032
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6689475347661187e-06,
+      "loss": 4.0556,
+      "step": 7033
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6689001264222504e-06,
+      "loss": 3.7568,
+      "step": 7034
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.668852718078382e-06,
+      "loss": 3.8746,
+      "step": 7035
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6688053097345134e-06,
+      "loss": 3.7816,
+      "step": 7036
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6687579013906446e-06,
+      "loss": 3.7582,
+      "step": 7037
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6687104930467763e-06,
+      "loss": 3.2788,
+      "step": 7038
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6686630847029076e-06,
+      "loss": 3.5061,
+      "step": 7039
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6686156763590393e-06,
+      "loss": 3.3573,
+      "step": 7040
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.668568268015171e-06,
+      "loss": 3.5118,
+      "step": 7041
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6685208596713023e-06,
+      "loss": 3.8798,
+      "step": 7042
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6684734513274336e-06,
+      "loss": 4.1256,
+      "step": 7043
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6684260429835653e-06,
+      "loss": 3.4501,
+      "step": 7044
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6683786346396966e-06,
+      "loss": 3.8871,
+      "step": 7045
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.668331226295828e-06,
+      "loss": 3.5584,
+      "step": 7046
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6682838179519596e-06,
+      "loss": 3.5889,
+      "step": 7047
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6682364096080913e-06,
+      "loss": 3.855,
+      "step": 7048
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6681890012642226e-06,
+      "loss": 3.716,
+      "step": 7049
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6681415929203543e-06,
+      "loss": 3.9838,
+      "step": 7050
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6680941845764855e-06,
+      "loss": 2.5118,
+      "step": 7051
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.668046776232617e-06,
+      "loss": 3.7902,
+      "step": 7052
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6679993678887485e-06,
+      "loss": 3.6851,
+      "step": 7053
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6679519595448802e-06,
+      "loss": 3.8459,
+      "step": 7054
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6679045512010115e-06,
+      "loss": 3.679,
+      "step": 7055
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6678571428571428e-06,
+      "loss": 3.314,
+      "step": 7056
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6678097345132745e-06,
+      "loss": 2.6831,
+      "step": 7057
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6677623261694058e-06,
+      "loss": 3.4114,
+      "step": 7058
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.667714917825537e-06,
+      "loss": 4.1804,
+      "step": 7059
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6676675094816688e-06,
+      "loss": 3.3098,
+      "step": 7060
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2.6676201011378005e-06,
+      "loss": 3.4412,
+      "step": 7061
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6675726927939317e-06,
+      "loss": 4.086,
+      "step": 7062
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6675252844500634e-06,
+      "loss": 3.4484,
+      "step": 7063
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6674778761061947e-06,
+      "loss": 4.1215,
+      "step": 7064
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.667430467762326e-06,
+      "loss": 3.6589,
+      "step": 7065
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6673830594184577e-06,
+      "loss": 3.7947,
+      "step": 7066
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6673356510745894e-06,
+      "loss": 3.4186,
+      "step": 7067
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6672882427307207e-06,
+      "loss": 3.6003,
+      "step": 7068
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6672408343868524e-06,
+      "loss": 3.4762,
+      "step": 7069
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6671934260429837e-06,
+      "loss": 4.0938,
+      "step": 7070
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.667146017699115e-06,
+      "loss": 2.7937,
+      "step": 7071
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6670986093552467e-06,
+      "loss": 4.2166,
+      "step": 7072
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.667051201011378e-06,
+      "loss": 3.0614,
+      "step": 7073
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6670037926675096e-06,
+      "loss": 3.6217,
+      "step": 7074
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6669563843236413e-06,
+      "loss": 3.6943,
+      "step": 7075
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6669089759797726e-06,
+      "loss": 3.8892,
+      "step": 7076
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666861567635904e-06,
+      "loss": 3.5891,
+      "step": 7077
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666814159292035e-06,
+      "loss": 3.7453,
+      "step": 7078
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666766750948167e-06,
+      "loss": 3.7101,
+      "step": 7079
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6667193426042986e-06,
+      "loss": 2.9644,
+      "step": 7080
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.66667193426043e-06,
+      "loss": 3.8018,
+      "step": 7081
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6666245259165616e-06,
+      "loss": 3.8695,
+      "step": 7082
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666577117572693e-06,
+      "loss": 3.2914,
+      "step": 7083
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666529709228824e-06,
+      "loss": 3.5812,
+      "step": 7084
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666482300884956e-06,
+      "loss": 4.1647,
+      "step": 7085
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666434892541087e-06,
+      "loss": 3.3771,
+      "step": 7086
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666387484197219e-06,
+      "loss": 3.389,
+      "step": 7087
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6663400758533505e-06,
+      "loss": 3.7848,
+      "step": 7088
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666292667509482e-06,
+      "loss": 3.3113,
+      "step": 7089
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666245259165613e-06,
+      "loss": 3.8177,
+      "step": 7090
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666197850821745e-06,
+      "loss": 3.7971,
+      "step": 7091
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.666150442477876e-06,
+      "loss": 3.7475,
+      "step": 7092
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6661030341340078e-06,
+      "loss": 4.2465,
+      "step": 7093
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6660556257901395e-06,
+      "loss": 3.9303,
+      "step": 7094
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6660082174462708e-06,
+      "loss": 3.5461,
+      "step": 7095
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665960809102402e-06,
+      "loss": 3.5429,
+      "step": 7096
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6659134007585333e-06,
+      "loss": 4.0256,
+      "step": 7097
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665865992414665e-06,
+      "loss": 3.087,
+      "step": 7098
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6658185840707963e-06,
+      "loss": 3.9948,
+      "step": 7099
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665771175726928e-06,
+      "loss": 3.5266,
+      "step": 7100
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6657237673830597e-06,
+      "loss": 4.0298,
+      "step": 7101
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665676359039191e-06,
+      "loss": 3.7701,
+      "step": 7102
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6656289506953223e-06,
+      "loss": 3.8407,
+      "step": 7103
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665581542351454e-06,
+      "loss": 3.5876,
+      "step": 7104
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6655341340075853e-06,
+      "loss": 3.5347,
+      "step": 7105
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665486725663717e-06,
+      "loss": 3.4213,
+      "step": 7106
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6654393173198487e-06,
+      "loss": 3.6153,
+      "step": 7107
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.66539190897598e-06,
+      "loss": 3.5138,
+      "step": 7108
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6653445006321112e-06,
+      "loss": 4.0187,
+      "step": 7109
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665297092288243e-06,
+      "loss": 3.0893,
+      "step": 7110
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6652496839443742e-06,
+      "loss": 3.3127,
+      "step": 7111
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6652022756005055e-06,
+      "loss": 3.9725,
+      "step": 7112
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665154867256637e-06,
+      "loss": 3.2765,
+      "step": 7113
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665107458912769e-06,
+      "loss": 4.1041,
+      "step": 7114
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6650600505689e-06,
+      "loss": 3.7695,
+      "step": 7115
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.665012642225032e-06,
+      "loss": 3.5748,
+      "step": 7116
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.664965233881163e-06,
+      "loss": 3.5617,
+      "step": 7117
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6649178255372945e-06,
+      "loss": 4.0911,
+      "step": 7118
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6648704171934257e-06,
+      "loss": 3.4323,
+      "step": 7119
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.664823008849558e-06,
+      "loss": 3.552,
+      "step": 7120
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.664775600505689e-06,
+      "loss": 3.9305,
+      "step": 7121
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6647281921618204e-06,
+      "loss": 3.7881,
+      "step": 7122
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.664680783817952e-06,
+      "loss": 3.5845,
+      "step": 7123
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2.6646333754740834e-06,
+      "loss": 3.9166,
+      "step": 7124
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6645859671302147e-06,
+      "loss": 3.8396,
+      "step": 7125
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6645385587863464e-06,
+      "loss": 3.734,
+      "step": 7126
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.664491150442478e-06,
+      "loss": 4.2044,
+      "step": 7127
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6644437420986094e-06,
+      "loss": 4.0961,
+      "step": 7128
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.664396333754741e-06,
+      "loss": 3.9232,
+      "step": 7129
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6643489254108724e-06,
+      "loss": 3.6895,
+      "step": 7130
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6643015170670036e-06,
+      "loss": 4.1135,
+      "step": 7131
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6642541087231353e-06,
+      "loss": 3.8898,
+      "step": 7132
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6642067003792666e-06,
+      "loss": 4.0122,
+      "step": 7133
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6641592920353983e-06,
+      "loss": 3.3812,
+      "step": 7134
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.66411188369153e-06,
+      "loss": 4.276,
+      "step": 7135
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6640644753476613e-06,
+      "loss": 3.1913,
+      "step": 7136
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6640170670037926e-06,
+      "loss": 4.0556,
+      "step": 7137
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6639696586599243e-06,
+      "loss": 3.6385,
+      "step": 7138
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6639222503160556e-06,
+      "loss": 4.0012,
+      "step": 7139
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6638748419721873e-06,
+      "loss": 4.0714,
+      "step": 7140
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6638274336283186e-06,
+      "loss": 3.6962,
+      "step": 7141
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6637800252844503e-06,
+      "loss": 4.0556,
+      "step": 7142
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6637326169405815e-06,
+      "loss": 3.9708,
+      "step": 7143
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.663685208596713e-06,
+      "loss": 3.9239,
+      "step": 7144
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6636378002528445e-06,
+      "loss": 4.0268,
+      "step": 7145
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.663590391908976e-06,
+      "loss": 3.5299,
+      "step": 7146
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6635429835651075e-06,
+      "loss": 3.6408,
+      "step": 7147
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6634955752212392e-06,
+      "loss": 3.8963,
+      "step": 7148
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6634481668773705e-06,
+      "loss": 4.1787,
+      "step": 7149
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6634007585335018e-06,
+      "loss": 3.7647,
+      "step": 7150
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6633533501896335e-06,
+      "loss": 3.8449,
+      "step": 7151
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6633059418457648e-06,
+      "loss": 3.8325,
+      "step": 7152
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6632585335018965e-06,
+      "loss": 3.6661,
+      "step": 7153
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.663211125158028e-06,
+      "loss": 3.8933,
+      "step": 7154
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6631637168141595e-06,
+      "loss": 3.5216,
+      "step": 7155
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6631163084702907e-06,
+      "loss": 4.0014,
+      "step": 7156
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6630689001264224e-06,
+      "loss": 3.8342,
+      "step": 7157
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6630214917825537e-06,
+      "loss": 3.5696,
+      "step": 7158
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662974083438685e-06,
+      "loss": 3.4492,
+      "step": 7159
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6629266750948167e-06,
+      "loss": 3.7382,
+      "step": 7160
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6628792667509484e-06,
+      "loss": 3.2687,
+      "step": 7161
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6628318584070797e-06,
+      "loss": 3.6692,
+      "step": 7162
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662784450063211e-06,
+      "loss": 3.6747,
+      "step": 7163
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6627370417193427e-06,
+      "loss": 3.2155,
+      "step": 7164
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662689633375474e-06,
+      "loss": 3.5731,
+      "step": 7165
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6626422250316057e-06,
+      "loss": 3.5782,
+      "step": 7166
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6625948166877374e-06,
+      "loss": 3.419,
+      "step": 7167
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6625474083438686e-06,
+      "loss": 3.7718,
+      "step": 7168
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6625e-06,
+      "loss": 3.6759,
+      "step": 7169
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6624525916561316e-06,
+      "loss": 3.9435,
+      "step": 7170
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662405183312263e-06,
+      "loss": 2.8878,
+      "step": 7171
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662357774968394e-06,
+      "loss": 3.3644,
+      "step": 7172
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662310366624526e-06,
+      "loss": 4.0477,
+      "step": 7173
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6622629582806576e-06,
+      "loss": 3.3583,
+      "step": 7174
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662215549936789e-06,
+      "loss": 2.9589,
+      "step": 7175
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6621681415929206e-06,
+      "loss": 3.6585,
+      "step": 7176
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662120733249052e-06,
+      "loss": 3.9786,
+      "step": 7177
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662073324905183e-06,
+      "loss": 3.335,
+      "step": 7178
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.662025916561315e-06,
+      "loss": 3.7562,
+      "step": 7179
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6619785082174465e-06,
+      "loss": 3.2958,
+      "step": 7180
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.661931099873578e-06,
+      "loss": 4.1692,
+      "step": 7181
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6618836915297095e-06,
+      "loss": 3.2725,
+      "step": 7182
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.661836283185841e-06,
+      "loss": 3.8405,
+      "step": 7183
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.661788874841972e-06,
+      "loss": 3.8264,
+      "step": 7184
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6617414664981034e-06,
+      "loss": 4.1473,
+      "step": 7185
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.661694058154235e-06,
+      "loss": 4.12,
+      "step": 7186
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2.6616466498103668e-06,
+      "loss": 3.5816,
+      "step": 7187
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.661599241466498e-06,
+      "loss": 3.9099,
+      "step": 7188
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6615518331226298e-06,
+      "loss": 3.4443,
+      "step": 7189
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.661504424778761e-06,
+      "loss": 4.0163,
+      "step": 7190
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6614570164348923e-06,
+      "loss": 3.565,
+      "step": 7191
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.661409608091024e-06,
+      "loss": 3.9801,
+      "step": 7192
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6613621997471557e-06,
+      "loss": 3.9699,
+      "step": 7193
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.661314791403287e-06,
+      "loss": 3.6181,
+      "step": 7194
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6612673830594187e-06,
+      "loss": 3.3317,
+      "step": 7195
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.66121997471555e-06,
+      "loss": 3.7844,
+      "step": 7196
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6611725663716813e-06,
+      "loss": 3.5208,
+      "step": 7197
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.661125158027813e-06,
+      "loss": 3.289,
+      "step": 7198
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6610777496839443e-06,
+      "loss": 3.8267,
+      "step": 7199
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.661030341340076e-06,
+      "loss": 3.1306,
+      "step": 7200
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6609829329962077e-06,
+      "loss": 3.9896,
+      "step": 7201
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660935524652339e-06,
+      "loss": 3.9057,
+      "step": 7202
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6608881163084702e-06,
+      "loss": 3.9101,
+      "step": 7203
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660840707964602e-06,
+      "loss": 3.2721,
+      "step": 7204
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6607932996207332e-06,
+      "loss": 3.5895,
+      "step": 7205
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660745891276865e-06,
+      "loss": 3.955,
+      "step": 7206
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660698482932996e-06,
+      "loss": 3.2843,
+      "step": 7207
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660651074589128e-06,
+      "loss": 3.4522,
+      "step": 7208
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660603666245259e-06,
+      "loss": 3.8414,
+      "step": 7209
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6605562579013905e-06,
+      "loss": 3.2472,
+      "step": 7210
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660508849557522e-06,
+      "loss": 3.6027,
+      "step": 7211
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6604614412136534e-06,
+      "loss": 4.0086,
+      "step": 7212
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660414032869785e-06,
+      "loss": 3.8123,
+      "step": 7213
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660366624525917e-06,
+      "loss": 4.1786,
+      "step": 7214
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660319216182048e-06,
+      "loss": 4.3765,
+      "step": 7215
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6602718078381794e-06,
+      "loss": 3.7308,
+      "step": 7216
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660224399494311e-06,
+      "loss": 4.1826,
+      "step": 7217
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6601769911504424e-06,
+      "loss": 4.0004,
+      "step": 7218
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660129582806574e-06,
+      "loss": 3.5774,
+      "step": 7219
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660082174462706e-06,
+      "loss": 3.9413,
+      "step": 7220
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.660034766118837e-06,
+      "loss": 3.7823,
+      "step": 7221
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6599873577749684e-06,
+      "loss": 4.1281,
+      "step": 7222
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6599399494311e-06,
+      "loss": 3.7558,
+      "step": 7223
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6598925410872314e-06,
+      "loss": 3.3414,
+      "step": 7224
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6598451327433626e-06,
+      "loss": 2.7817,
+      "step": 7225
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6597977243994943e-06,
+      "loss": 3.686,
+      "step": 7226
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.659750316055626e-06,
+      "loss": 4.0598,
+      "step": 7227
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6597029077117573e-06,
+      "loss": 3.2816,
+      "step": 7228
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6596554993678886e-06,
+      "loss": 3.551,
+      "step": 7229
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6596080910240203e-06,
+      "loss": 3.9355,
+      "step": 7230
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6595606826801516e-06,
+      "loss": 3.843,
+      "step": 7231
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6595132743362833e-06,
+      "loss": 3.9004,
+      "step": 7232
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.659465865992415e-06,
+      "loss": 3.9452,
+      "step": 7233
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6594184576485463e-06,
+      "loss": 3.2724,
+      "step": 7234
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6593710493046776e-06,
+      "loss": 3.9213,
+      "step": 7235
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6593236409608093e-06,
+      "loss": 4.0703,
+      "step": 7236
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6592762326169405e-06,
+      "loss": 3.7097,
+      "step": 7237
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.659228824273072e-06,
+      "loss": 4.0098,
+      "step": 7238
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6591814159292035e-06,
+      "loss": 3.2486,
+      "step": 7239
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6591340075853352e-06,
+      "loss": 3.3135,
+      "step": 7240
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6590865992414665e-06,
+      "loss": 3.2596,
+      "step": 7241
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6590391908975982e-06,
+      "loss": 3.6624,
+      "step": 7242
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6589917825537295e-06,
+      "loss": 3.4546,
+      "step": 7243
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6589443742098608e-06,
+      "loss": 3.6634,
+      "step": 7244
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6588969658659925e-06,
+      "loss": 2.7604,
+      "step": 7245
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.658849557522124e-06,
+      "loss": 3.337,
+      "step": 7246
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6588021491782555e-06,
+      "loss": 4.1258,
+      "step": 7247
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.658754740834387e-06,
+      "loss": 4.2388,
+      "step": 7248
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6587073324905184e-06,
+      "loss": 3.4201,
+      "step": 7249
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.6586599241466497e-06,
+      "loss": 4.1468,
+      "step": 7250
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2.658612515802781e-06,
+      "loss": 3.4813,
+      "step": 7251
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6585651074589127e-06,
+      "loss": 3.9151,
+      "step": 7252
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6585176991150444e-06,
+      "loss": 2.8557,
+      "step": 7253
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6584702907711757e-06,
+      "loss": 3.977,
+      "step": 7254
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6584228824273074e-06,
+      "loss": 3.9687,
+      "step": 7255
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6583754740834387e-06,
+      "loss": 3.0763,
+      "step": 7256
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.65832806573957e-06,
+      "loss": 3.5374,
+      "step": 7257
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6582806573957017e-06,
+      "loss": 3.4593,
+      "step": 7258
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6582332490518334e-06,
+      "loss": 3.4302,
+      "step": 7259
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6581858407079646e-06,
+      "loss": 3.5269,
+      "step": 7260
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6581384323640964e-06,
+      "loss": 4.0333,
+      "step": 7261
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6580910240202276e-06,
+      "loss": 4.0058,
+      "step": 7262
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.658043615676359e-06,
+      "loss": 3.7198,
+      "step": 7263
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6579962073324906e-06,
+      "loss": 4.0872,
+      "step": 7264
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657948798988622e-06,
+      "loss": 3.479,
+      "step": 7265
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6579013906447536e-06,
+      "loss": 3.3539,
+      "step": 7266
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6578539823008853e-06,
+      "loss": 3.1935,
+      "step": 7267
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6578065739570166e-06,
+      "loss": 4.0169,
+      "step": 7268
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657759165613148e-06,
+      "loss": 3.3658,
+      "step": 7269
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6577117572692796e-06,
+      "loss": 3.8866,
+      "step": 7270
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657664348925411e-06,
+      "loss": 3.5671,
+      "step": 7271
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6576169405815426e-06,
+      "loss": 3.1706,
+      "step": 7272
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657569532237674e-06,
+      "loss": 3.6173,
+      "step": 7273
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6575221238938055e-06,
+      "loss": 3.6284,
+      "step": 7274
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657474715549937e-06,
+      "loss": 3.7479,
+      "step": 7275
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657427307206068e-06,
+      "loss": 4.0223,
+      "step": 7276
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6573798988622e-06,
+      "loss": 3.9277,
+      "step": 7277
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657332490518331e-06,
+      "loss": 3.7356,
+      "step": 7278
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6572850821744628e-06,
+      "loss": 3.5953,
+      "step": 7279
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6572376738305945e-06,
+      "loss": 3.7659,
+      "step": 7280
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6571902654867258e-06,
+      "loss": 3.7521,
+      "step": 7281
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.657142857142857e-06,
+      "loss": 3.7055,
+      "step": 7282
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6570954487989888e-06,
+      "loss": 3.4983,
+      "step": 7283
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.65704804045512e-06,
+      "loss": 3.8704,
+      "step": 7284
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6570006321112517e-06,
+      "loss": 3.6043,
+      "step": 7285
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6569532237673834e-06,
+      "loss": 4.0188,
+      "step": 7286
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6569058154235147e-06,
+      "loss": 3.6048,
+      "step": 7287
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656858407079646e-06,
+      "loss": 3.7776,
+      "step": 7288
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6568109987357777e-06,
+      "loss": 3.7313,
+      "step": 7289
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656763590391909e-06,
+      "loss": 3.8464,
+      "step": 7290
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6567161820480403e-06,
+      "loss": 4.2109,
+      "step": 7291
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656668773704172e-06,
+      "loss": 3.5483,
+      "step": 7292
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6566213653603037e-06,
+      "loss": 3.5381,
+      "step": 7293
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656573957016435e-06,
+      "loss": 3.7213,
+      "step": 7294
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6565265486725662e-06,
+      "loss": 3.4354,
+      "step": 7295
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656479140328698e-06,
+      "loss": 3.5586,
+      "step": 7296
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6564317319848292e-06,
+      "loss": 4.0596,
+      "step": 7297
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656384323640961e-06,
+      "loss": 3.7766,
+      "step": 7298
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6563369152970926e-06,
+      "loss": 3.5204,
+      "step": 7299
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656289506953224e-06,
+      "loss": 3.2559,
+      "step": 7300
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656242098609355e-06,
+      "loss": 3.6509,
+      "step": 7301
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656194690265487e-06,
+      "loss": 3.8243,
+      "step": 7302
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656147281921618e-06,
+      "loss": 4.0251,
+      "step": 7303
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6560998735777495e-06,
+      "loss": 4.3132,
+      "step": 7304
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656052465233881e-06,
+      "loss": 3.9708,
+      "step": 7305
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.656005056890013e-06,
+      "loss": 3.3131,
+      "step": 7306
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.655957648546144e-06,
+      "loss": 4.0329,
+      "step": 7307
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.655910240202276e-06,
+      "loss": 4.2045,
+      "step": 7308
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.655862831858407e-06,
+      "loss": 3.6215,
+      "step": 7309
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.6558154235145384e-06,
+      "loss": 3.5077,
+      "step": 7310
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.65576801517067e-06,
+      "loss": 3.7684,
+      "step": 7311
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.655720606826802e-06,
+      "loss": 4.108,
+      "step": 7312
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.655673198482933e-06,
+      "loss": 3.5484,
+      "step": 7313
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2.655625790139065e-06,
+      "loss": 4.0104,
+      "step": 7314
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.655578381795196e-06,
+      "loss": 4.1889,
+      "step": 7315
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6555309734513274e-06,
+      "loss": 3.5951,
+      "step": 7316
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6554835651074586e-06,
+      "loss": 3.1406,
+      "step": 7317
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6554361567635903e-06,
+      "loss": 3.7618,
+      "step": 7318
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.655388748419722e-06,
+      "loss": 2.8908,
+      "step": 7319
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6553413400758533e-06,
+      "loss": 3.4631,
+      "step": 7320
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.655293931731985e-06,
+      "loss": 3.5021,
+      "step": 7321
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6552465233881163e-06,
+      "loss": 3.8702,
+      "step": 7322
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6551991150442476e-06,
+      "loss": 3.2459,
+      "step": 7323
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6551517067003793e-06,
+      "loss": 3.4563,
+      "step": 7324
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.655104298356511e-06,
+      "loss": 3.092,
+      "step": 7325
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6550568900126423e-06,
+      "loss": 3.1879,
+      "step": 7326
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.655009481668774e-06,
+      "loss": 3.5175,
+      "step": 7327
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6549620733249053e-06,
+      "loss": 3.4784,
+      "step": 7328
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6549146649810365e-06,
+      "loss": 4.0996,
+      "step": 7329
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6548672566371683e-06,
+      "loss": 3.4336,
+      "step": 7330
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6548198482932995e-06,
+      "loss": 3.8053,
+      "step": 7331
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6547724399494312e-06,
+      "loss": 4.232,
+      "step": 7332
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.654725031605563e-06,
+      "loss": 4.203,
+      "step": 7333
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6546776232616942e-06,
+      "loss": 3.842,
+      "step": 7334
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6546302149178255e-06,
+      "loss": 3.3529,
+      "step": 7335
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.654582806573957e-06,
+      "loss": 3.9905,
+      "step": 7336
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6545353982300885e-06,
+      "loss": 4.3057,
+      "step": 7337
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.65448798988622e-06,
+      "loss": 3.9567,
+      "step": 7338
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6544405815423515e-06,
+      "loss": 4.2641,
+      "step": 7339
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.654393173198483e-06,
+      "loss": 3.7823,
+      "step": 7340
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6543457648546145e-06,
+      "loss": 3.6913,
+      "step": 7341
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6542983565107457e-06,
+      "loss": 3.6879,
+      "step": 7342
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6542509481668774e-06,
+      "loss": 3.1491,
+      "step": 7343
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6542035398230087e-06,
+      "loss": 3.5701,
+      "step": 7344
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6541561314791404e-06,
+      "loss": 3.6071,
+      "step": 7345
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.654108723135272e-06,
+      "loss": 3.673,
+      "step": 7346
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6540613147914034e-06,
+      "loss": 3.6919,
+      "step": 7347
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6540139064475347e-06,
+      "loss": 3.4739,
+      "step": 7348
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6539664981036664e-06,
+      "loss": 3.7006,
+      "step": 7349
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6539190897597977e-06,
+      "loss": 4.0786,
+      "step": 7350
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6538716814159294e-06,
+      "loss": 3.8757,
+      "step": 7351
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.653824273072061e-06,
+      "loss": 3.469,
+      "step": 7352
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6537768647281924e-06,
+      "loss": 3.6367,
+      "step": 7353
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6537294563843236e-06,
+      "loss": 3.5741,
+      "step": 7354
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6536820480404553e-06,
+      "loss": 3.4511,
+      "step": 7355
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6536346396965866e-06,
+      "loss": 3.733,
+      "step": 7356
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.653587231352718e-06,
+      "loss": 3.6237,
+      "step": 7357
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6535398230088496e-06,
+      "loss": 3.5143,
+      "step": 7358
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6534924146649813e-06,
+      "loss": 3.4516,
+      "step": 7359
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6534450063211126e-06,
+      "loss": 3.238,
+      "step": 7360
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.653397597977244e-06,
+      "loss": 3.6095,
+      "step": 7361
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6533501896333756e-06,
+      "loss": 3.6337,
+      "step": 7362
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.653302781289507e-06,
+      "loss": 3.8591,
+      "step": 7363
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6532553729456386e-06,
+      "loss": 3.4199,
+      "step": 7364
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6532079646017703e-06,
+      "loss": 4.0532,
+      "step": 7365
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6531605562579015e-06,
+      "loss": 3.7071,
+      "step": 7366
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.653113147914033e-06,
+      "loss": 3.5767,
+      "step": 7367
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6530657395701645e-06,
+      "loss": 4.1937,
+      "step": 7368
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.653018331226296e-06,
+      "loss": 3.6356,
+      "step": 7369
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.652970922882427e-06,
+      "loss": 2.8172,
+      "step": 7370
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.652923514538559e-06,
+      "loss": 3.4628,
+      "step": 7371
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6528761061946905e-06,
+      "loss": 3.2326,
+      "step": 7372
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6528286978508218e-06,
+      "loss": 4.1284,
+      "step": 7373
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6527812895069535e-06,
+      "loss": 3.2303,
+      "step": 7374
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6527338811630848e-06,
+      "loss": 3.506,
+      "step": 7375
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.652686472819216e-06,
+      "loss": 3.7275,
+      "step": 7376
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2.6526390644753478e-06,
+      "loss": 3.6307,
+      "step": 7377
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6525916561314795e-06,
+      "loss": 3.3153,
+      "step": 7378
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6525442477876107e-06,
+      "loss": 3.5872,
+      "step": 7379
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6524968394437424e-06,
+      "loss": 2.8774,
+      "step": 7380
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6524494310998737e-06,
+      "loss": 3.5729,
+      "step": 7381
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.652402022756005e-06,
+      "loss": 3.7208,
+      "step": 7382
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6523546144121363e-06,
+      "loss": 3.7597,
+      "step": 7383
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.652307206068268e-06,
+      "loss": 3.1684,
+      "step": 7384
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6522597977243997e-06,
+      "loss": 4.3166,
+      "step": 7385
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.652212389380531e-06,
+      "loss": 3.5437,
+      "step": 7386
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6521649810366627e-06,
+      "loss": 3.5576,
+      "step": 7387
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.652117572692794e-06,
+      "loss": 3.8685,
+      "step": 7388
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6520701643489252e-06,
+      "loss": 4.321,
+      "step": 7389
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.652022756005057e-06,
+      "loss": 4.0194,
+      "step": 7390
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6519753476611886e-06,
+      "loss": 3.4279,
+      "step": 7391
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.65192793931732e-06,
+      "loss": 3.3044,
+      "step": 7392
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6518805309734516e-06,
+      "loss": 2.8194,
+      "step": 7393
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651833122629583e-06,
+      "loss": 3.9144,
+      "step": 7394
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651785714285714e-06,
+      "loss": 3.4691,
+      "step": 7395
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651738305941846e-06,
+      "loss": 4.0009,
+      "step": 7396
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651690897597977e-06,
+      "loss": 3.1249,
+      "step": 7397
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651643489254109e-06,
+      "loss": 3.7677,
+      "step": 7398
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6515960809102406e-06,
+      "loss": 4.0737,
+      "step": 7399
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651548672566372e-06,
+      "loss": 3.8022,
+      "step": 7400
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651501264222503e-06,
+      "loss": 3.9147,
+      "step": 7401
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651453855878635e-06,
+      "loss": 3.5594,
+      "step": 7402
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651406447534766e-06,
+      "loss": 3.7713,
+      "step": 7403
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6513590391908974e-06,
+      "loss": 3.5129,
+      "step": 7404
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651311630847029e-06,
+      "loss": 4.0803,
+      "step": 7405
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651264222503161e-06,
+      "loss": 4.4949,
+      "step": 7406
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651216814159292e-06,
+      "loss": 3.34,
+      "step": 7407
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6511694058154234e-06,
+      "loss": 3.6543,
+      "step": 7408
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651121997471555e-06,
+      "loss": 3.8103,
+      "step": 7409
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6510745891276864e-06,
+      "loss": 3.4353,
+      "step": 7410
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.651027180783818e-06,
+      "loss": 3.7207,
+      "step": 7411
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6509797724399498e-06,
+      "loss": 3.0764,
+      "step": 7412
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.650932364096081e-06,
+      "loss": 3.0697,
+      "step": 7413
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6508849557522123e-06,
+      "loss": 3.815,
+      "step": 7414
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.650837547408344e-06,
+      "loss": 3.5753,
+      "step": 7415
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6507901390644753e-06,
+      "loss": 3.2835,
+      "step": 7416
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6507427307206066e-06,
+      "loss": 3.5875,
+      "step": 7417
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6506953223767383e-06,
+      "loss": 3.9778,
+      "step": 7418
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.65064791403287e-06,
+      "loss": 3.689,
+      "step": 7419
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6506005056890013e-06,
+      "loss": 3.5113,
+      "step": 7420
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.650553097345133e-06,
+      "loss": 3.3483,
+      "step": 7421
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6505056890012643e-06,
+      "loss": 3.6161,
+      "step": 7422
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6504582806573955e-06,
+      "loss": 3.571,
+      "step": 7423
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6504108723135272e-06,
+      "loss": 4.0847,
+      "step": 7424
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.650363463969659e-06,
+      "loss": 4.1911,
+      "step": 7425
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6503160556257902e-06,
+      "loss": 3.0361,
+      "step": 7426
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6502686472819215e-06,
+      "loss": 3.4006,
+      "step": 7427
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6502212389380532e-06,
+      "loss": 3.0789,
+      "step": 7428
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6501738305941845e-06,
+      "loss": 3.6538,
+      "step": 7429
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6501264222503158e-06,
+      "loss": 4.1483,
+      "step": 7430
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6500790139064475e-06,
+      "loss": 3.8873,
+      "step": 7431
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.650031605562579e-06,
+      "loss": 3.5314,
+      "step": 7432
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6499841972187105e-06,
+      "loss": 3.7673,
+      "step": 7433
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.649936788874842e-06,
+      "loss": 3.4608,
+      "step": 7434
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6498893805309735e-06,
+      "loss": 3.286,
+      "step": 7435
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6498419721871047e-06,
+      "loss": 3.2003,
+      "step": 7436
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6497945638432364e-06,
+      "loss": 3.438,
+      "step": 7437
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.649747155499368e-06,
+      "loss": 3.263,
+      "step": 7438
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6496997471554994e-06,
+      "loss": 3.8045,
+      "step": 7439
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.649652338811631e-06,
+      "loss": 4.0855,
+      "step": 7440
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2.6496049304677624e-06,
+      "loss": 2.916,
+      "step": 7441
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6495575221238937e-06,
+      "loss": 4.1953,
+      "step": 7442
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6495101137800254e-06,
+      "loss": 3.2256,
+      "step": 7443
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6494627054361567e-06,
+      "loss": 3.7418,
+      "step": 7444
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6494152970922884e-06,
+      "loss": 3.0436,
+      "step": 7445
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.64936788874842e-06,
+      "loss": 3.2554,
+      "step": 7446
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6493204804045514e-06,
+      "loss": 3.5459,
+      "step": 7447
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6492730720606826e-06,
+      "loss": 3.765,
+      "step": 7448
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.649225663716814e-06,
+      "loss": 4.2899,
+      "step": 7449
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6491782553729456e-06,
+      "loss": 4.1609,
+      "step": 7450
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6491308470290773e-06,
+      "loss": 3.6472,
+      "step": 7451
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6490834386852086e-06,
+      "loss": 4.0739,
+      "step": 7452
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6490360303413403e-06,
+      "loss": 3.8022,
+      "step": 7453
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6489886219974716e-06,
+      "loss": 3.8632,
+      "step": 7454
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.648941213653603e-06,
+      "loss": 3.4963,
+      "step": 7455
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6488938053097346e-06,
+      "loss": 3.8145,
+      "step": 7456
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.648846396965866e-06,
+      "loss": 2.8132,
+      "step": 7457
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6487989886219976e-06,
+      "loss": 3.749,
+      "step": 7458
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6487515802781293e-06,
+      "loss": 3.2791,
+      "step": 7459
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6487041719342605e-06,
+      "loss": 3.5998,
+      "step": 7460
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.648656763590392e-06,
+      "loss": 3.7923,
+      "step": 7461
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6486093552465235e-06,
+      "loss": 3.2077,
+      "step": 7462
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.648561946902655e-06,
+      "loss": 3.6232,
+      "step": 7463
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6485145385587865e-06,
+      "loss": 3.3487,
+      "step": 7464
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6484671302149182e-06,
+      "loss": 4.403,
+      "step": 7465
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6484197218710495e-06,
+      "loss": 3.9484,
+      "step": 7466
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6483723135271808e-06,
+      "loss": 3.9603,
+      "step": 7467
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6483249051833125e-06,
+      "loss": 3.2968,
+      "step": 7468
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6482774968394438e-06,
+      "loss": 3.4887,
+      "step": 7469
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.648230088495575e-06,
+      "loss": 3.5994,
+      "step": 7470
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6481826801517067e-06,
+      "loss": 4.1273,
+      "step": 7471
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6481352718078384e-06,
+      "loss": 4.0153,
+      "step": 7472
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6480878634639697e-06,
+      "loss": 3.8714,
+      "step": 7473
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.648040455120101e-06,
+      "loss": 3.2508,
+      "step": 7474
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6479930467762327e-06,
+      "loss": 3.5041,
+      "step": 7475
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647945638432364e-06,
+      "loss": 3.3375,
+      "step": 7476
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6478982300884957e-06,
+      "loss": 3.6089,
+      "step": 7477
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6478508217446274e-06,
+      "loss": 3.5482,
+      "step": 7478
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6478034134007587e-06,
+      "loss": 3.8688,
+      "step": 7479
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.64775600505689e-06,
+      "loss": 3.9053,
+      "step": 7480
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6477085967130217e-06,
+      "loss": 3.475,
+      "step": 7481
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647661188369153e-06,
+      "loss": 3.8516,
+      "step": 7482
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6476137800252842e-06,
+      "loss": 3.5922,
+      "step": 7483
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647566371681416e-06,
+      "loss": 3.3367,
+      "step": 7484
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6475189633375476e-06,
+      "loss": 3.0498,
+      "step": 7485
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647471554993679e-06,
+      "loss": 4.131,
+      "step": 7486
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6474241466498106e-06,
+      "loss": 4.0992,
+      "step": 7487
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647376738305942e-06,
+      "loss": 3.9812,
+      "step": 7488
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647329329962073e-06,
+      "loss": 3.0278,
+      "step": 7489
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647281921618205e-06,
+      "loss": 3.512,
+      "step": 7490
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6472345132743366e-06,
+      "loss": 3.697,
+      "step": 7491
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647187104930468e-06,
+      "loss": 3.0671,
+      "step": 7492
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647139696586599e-06,
+      "loss": 3.8965,
+      "step": 7493
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647092288242731e-06,
+      "loss": 3.4228,
+      "step": 7494
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.647044879898862e-06,
+      "loss": 3.6388,
+      "step": 7495
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6469974715549934e-06,
+      "loss": 3.4627,
+      "step": 7496
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.646950063211125e-06,
+      "loss": 3.406,
+      "step": 7497
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.646902654867257e-06,
+      "loss": 2.9841,
+      "step": 7498
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.646855246523388e-06,
+      "loss": 3.5508,
+      "step": 7499
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.64680783817952e-06,
+      "loss": 4.0658,
+      "step": 7500
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.646760429835651e-06,
+      "loss": 3.0716,
+      "step": 7501
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6467130214917824e-06,
+      "loss": 3.8196,
+      "step": 7502
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.646665613147914e-06,
+      "loss": 3.644,
+      "step": 7503
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2.6466182048040458e-06,
+      "loss": 3.272,
+      "step": 7504
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.646570796460177e-06,
+      "loss": 3.8285,
+      "step": 7505
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6465233881163088e-06,
+      "loss": 3.6312,
+      "step": 7506
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.64647597977244e-06,
+      "loss": 4.1477,
+      "step": 7507
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6464285714285713e-06,
+      "loss": 4.1702,
+      "step": 7508
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.646381163084703e-06,
+      "loss": 3.642,
+      "step": 7509
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6463337547408343e-06,
+      "loss": 3.9964,
+      "step": 7510
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.646286346396966e-06,
+      "loss": 3.0288,
+      "step": 7511
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6462389380530977e-06,
+      "loss": 3.8925,
+      "step": 7512
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.646191529709229e-06,
+      "loss": 3.4473,
+      "step": 7513
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6461441213653603e-06,
+      "loss": 3.0837,
+      "step": 7514
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6460967130214916e-06,
+      "loss": 3.8159,
+      "step": 7515
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6460493046776233e-06,
+      "loss": 3.4107,
+      "step": 7516
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.646001896333755e-06,
+      "loss": 3.1981,
+      "step": 7517
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6459544879898862e-06,
+      "loss": 3.9146,
+      "step": 7518
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645907079646018e-06,
+      "loss": 3.574,
+      "step": 7519
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6458596713021492e-06,
+      "loss": 3.3943,
+      "step": 7520
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6458122629582805e-06,
+      "loss": 3.5614,
+      "step": 7521
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645764854614412e-06,
+      "loss": 3.3866,
+      "step": 7522
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6457174462705435e-06,
+      "loss": 3.9855,
+      "step": 7523
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645670037926675e-06,
+      "loss": 4.0539,
+      "step": 7524
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645622629582807e-06,
+      "loss": 4.021,
+      "step": 7525
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645575221238938e-06,
+      "loss": 3.9575,
+      "step": 7526
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6455278128950695e-06,
+      "loss": 3.7138,
+      "step": 7527
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645480404551201e-06,
+      "loss": 3.9649,
+      "step": 7528
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6454329962073324e-06,
+      "loss": 4.0828,
+      "step": 7529
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645385587863464e-06,
+      "loss": 3.4634,
+      "step": 7530
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645338179519596e-06,
+      "loss": 3.8931,
+      "step": 7531
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645290771175727e-06,
+      "loss": 3.9454,
+      "step": 7532
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6452433628318584e-06,
+      "loss": 3.432,
+      "step": 7533
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6451959544879897e-06,
+      "loss": 3.6412,
+      "step": 7534
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6451485461441214e-06,
+      "loss": 3.5391,
+      "step": 7535
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6451011378002527e-06,
+      "loss": 4.1138,
+      "step": 7536
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6450537294563844e-06,
+      "loss": 4.0354,
+      "step": 7537
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.645006321112516e-06,
+      "loss": 3.619,
+      "step": 7538
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6449589127686474e-06,
+      "loss": 3.8532,
+      "step": 7539
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6449115044247786e-06,
+      "loss": 3.8948,
+      "step": 7540
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6448640960809104e-06,
+      "loss": 3.2583,
+      "step": 7541
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6448166877370416e-06,
+      "loss": 3.3557,
+      "step": 7542
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6447692793931733e-06,
+      "loss": 3.8657,
+      "step": 7543
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.644721871049305e-06,
+      "loss": 3.471,
+      "step": 7544
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6446744627054363e-06,
+      "loss": 4.0981,
+      "step": 7545
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6446270543615676e-06,
+      "loss": 4.0017,
+      "step": 7546
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6445796460176993e-06,
+      "loss": 3.5572,
+      "step": 7547
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6445322376738306e-06,
+      "loss": 3.1794,
+      "step": 7548
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.644484829329962e-06,
+      "loss": 4.242,
+      "step": 7549
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6444374209860936e-06,
+      "loss": 3.9139,
+      "step": 7550
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6443900126422253e-06,
+      "loss": 3.086,
+      "step": 7551
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6443426042983566e-06,
+      "loss": 4.337,
+      "step": 7552
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6442951959544883e-06,
+      "loss": 3.6173,
+      "step": 7553
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6442477876106195e-06,
+      "loss": 3.2465,
+      "step": 7554
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.644200379266751e-06,
+      "loss": 3.2975,
+      "step": 7555
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6441529709228825e-06,
+      "loss": 3.7854,
+      "step": 7556
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6441055625790142e-06,
+      "loss": 3.859,
+      "step": 7557
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6440581542351455e-06,
+      "loss": 3.1411,
+      "step": 7558
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6440107458912768e-06,
+      "loss": 4.0808,
+      "step": 7559
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6439633375474085e-06,
+      "loss": 3.1246,
+      "step": 7560
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6439159292035398e-06,
+      "loss": 3.481,
+      "step": 7561
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.643868520859671e-06,
+      "loss": 3.4353,
+      "step": 7562
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6438211125158028e-06,
+      "loss": 3.1343,
+      "step": 7563
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6437737041719345e-06,
+      "loss": 3.7367,
+      "step": 7564
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6437262958280657e-06,
+      "loss": 3.7072,
+      "step": 7565
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6436788874841974e-06,
+      "loss": 3.0161,
+      "step": 7566
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2.6436314791403287e-06,
+      "loss": 3.5491,
+      "step": 7567
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.64358407079646e-06,
+      "loss": 2.9745,
+      "step": 7568
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6435366624525917e-06,
+      "loss": 3.8936,
+      "step": 7569
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6434892541087234e-06,
+      "loss": 3.6188,
+      "step": 7570
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6434418457648547e-06,
+      "loss": 3.2027,
+      "step": 7571
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6433944374209864e-06,
+      "loss": 4.0832,
+      "step": 7572
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6433470290771177e-06,
+      "loss": 3.5958,
+      "step": 7573
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.643299620733249e-06,
+      "loss": 3.6871,
+      "step": 7574
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6432522123893807e-06,
+      "loss": 4.1088,
+      "step": 7575
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.643204804045512e-06,
+      "loss": 4.1355,
+      "step": 7576
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6431573957016436e-06,
+      "loss": 3.5565,
+      "step": 7577
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.643109987357775e-06,
+      "loss": 3.4472,
+      "step": 7578
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6430625790139066e-06,
+      "loss": 3.9754,
+      "step": 7579
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.643015170670038e-06,
+      "loss": 3.0357,
+      "step": 7580
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642967762326169e-06,
+      "loss": 3.5884,
+      "step": 7581
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642920353982301e-06,
+      "loss": 3.1081,
+      "step": 7582
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6428729456384326e-06,
+      "loss": 3.622,
+      "step": 7583
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642825537294564e-06,
+      "loss": 3.8488,
+      "step": 7584
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6427781289506956e-06,
+      "loss": 4.0777,
+      "step": 7585
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642730720606827e-06,
+      "loss": 3.715,
+      "step": 7586
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642683312262958e-06,
+      "loss": 3.7693,
+      "step": 7587
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.64263590391909e-06,
+      "loss": 3.9073,
+      "step": 7588
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642588495575221e-06,
+      "loss": 4.1026,
+      "step": 7589
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642541087231353e-06,
+      "loss": 3.3981,
+      "step": 7590
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6424936788874845e-06,
+      "loss": 3.996,
+      "step": 7591
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642446270543616e-06,
+      "loss": 3.8177,
+      "step": 7592
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642398862199747e-06,
+      "loss": 3.5991,
+      "step": 7593
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642351453855879e-06,
+      "loss": 3.1602,
+      "step": 7594
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.64230404551201e-06,
+      "loss": 3.958,
+      "step": 7595
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6422566371681418e-06,
+      "loss": 2.7726,
+      "step": 7596
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6422092288242735e-06,
+      "loss": 3.7998,
+      "step": 7597
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6421618204804048e-06,
+      "loss": 3.9922,
+      "step": 7598
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642114412136536e-06,
+      "loss": 3.6125,
+      "step": 7599
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6420670037926673e-06,
+      "loss": 3.816,
+      "step": 7600
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.642019595448799e-06,
+      "loss": 3.9647,
+      "step": 7601
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6419721871049303e-06,
+      "loss": 3.651,
+      "step": 7602
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641924778761062e-06,
+      "loss": 3.1069,
+      "step": 7603
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6418773704171937e-06,
+      "loss": 4.0557,
+      "step": 7604
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641829962073325e-06,
+      "loss": 3.7555,
+      "step": 7605
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6417825537294563e-06,
+      "loss": 3.3893,
+      "step": 7606
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641735145385588e-06,
+      "loss": 3.7275,
+      "step": 7607
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6416877370417193e-06,
+      "loss": 3.4343,
+      "step": 7608
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641640328697851e-06,
+      "loss": 3.5884,
+      "step": 7609
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6415929203539827e-06,
+      "loss": 4.1921,
+      "step": 7610
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641545512010114e-06,
+      "loss": 4.0153,
+      "step": 7611
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6414981036662452e-06,
+      "loss": 3.6203,
+      "step": 7612
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641450695322377e-06,
+      "loss": 3.5433,
+      "step": 7613
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6414032869785082e-06,
+      "loss": 4.0855,
+      "step": 7614
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6413558786346395e-06,
+      "loss": 3.6966,
+      "step": 7615
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641308470290771e-06,
+      "loss": 3.5886,
+      "step": 7616
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641261061946903e-06,
+      "loss": 2.9741,
+      "step": 7617
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641213653603034e-06,
+      "loss": 3.5235,
+      "step": 7618
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641166245259166e-06,
+      "loss": 4.2122,
+      "step": 7619
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.641118836915297e-06,
+      "loss": 3.9057,
+      "step": 7620
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6410714285714285e-06,
+      "loss": 3.9646,
+      "step": 7621
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.64102402022756e-06,
+      "loss": 3.3176,
+      "step": 7622
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.640976611883692e-06,
+      "loss": 3.3274,
+      "step": 7623
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.640929203539823e-06,
+      "loss": 3.2879,
+      "step": 7624
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6408817951959544e-06,
+      "loss": 3.7301,
+      "step": 7625
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.640834386852086e-06,
+      "loss": 4.0043,
+      "step": 7626
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6407869785082174e-06,
+      "loss": 3.9373,
+      "step": 7627
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6407395701643487e-06,
+      "loss": 3.4679,
+      "step": 7628
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6406921618204804e-06,
+      "loss": 3.5864,
+      "step": 7629
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.640644753476612e-06,
+      "loss": 3.7094,
+      "step": 7630
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2.6405973451327434e-06,
+      "loss": 3.2943,
+      "step": 7631
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.640549936788875e-06,
+      "loss": 3.246,
+      "step": 7632
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6405025284450064e-06,
+      "loss": 3.5863,
+      "step": 7633
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6404551201011376e-06,
+      "loss": 3.4252,
+      "step": 7634
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6404077117572693e-06,
+      "loss": 3.8556,
+      "step": 7635
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.640360303413401e-06,
+      "loss": 3.6704,
+      "step": 7636
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6403128950695323e-06,
+      "loss": 4.2491,
+      "step": 7637
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.640265486725664e-06,
+      "loss": 2.9717,
+      "step": 7638
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6402180783817953e-06,
+      "loss": 3.4595,
+      "step": 7639
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6401706700379266e-06,
+      "loss": 3.236,
+      "step": 7640
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6401232616940583e-06,
+      "loss": 4.0019,
+      "step": 7641
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6400758533501896e-06,
+      "loss": 3.7903,
+      "step": 7642
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6400284450063213e-06,
+      "loss": 3.4935,
+      "step": 7643
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6399810366624526e-06,
+      "loss": 3.6819,
+      "step": 7644
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6399336283185843e-06,
+      "loss": 4.0337,
+      "step": 7645
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6398862199747155e-06,
+      "loss": 3.7444,
+      "step": 7646
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.639838811630847e-06,
+      "loss": 3.2891,
+      "step": 7647
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6397914032869785e-06,
+      "loss": 3.6824,
+      "step": 7648
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6397439949431102e-06,
+      "loss": 3.6766,
+      "step": 7649
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6396965865992415e-06,
+      "loss": 3.3778,
+      "step": 7650
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6396491782553732e-06,
+      "loss": 3.0422,
+      "step": 7651
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6396017699115045e-06,
+      "loss": 3.4107,
+      "step": 7652
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6395543615676358e-06,
+      "loss": 3.6031,
+      "step": 7653
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6395069532237675e-06,
+      "loss": 3.5535,
+      "step": 7654
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6394595448798988e-06,
+      "loss": 3.6888,
+      "step": 7655
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6394121365360305e-06,
+      "loss": 3.3253,
+      "step": 7656
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.639364728192162e-06,
+      "loss": 3.064,
+      "step": 7657
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6393173198482935e-06,
+      "loss": 3.4012,
+      "step": 7658
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6392699115044247e-06,
+      "loss": 3.9954,
+      "step": 7659
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6392225031605564e-06,
+      "loss": 3.4647,
+      "step": 7660
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6391750948166877e-06,
+      "loss": 3.9534,
+      "step": 7661
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6391276864728194e-06,
+      "loss": 2.9063,
+      "step": 7662
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.639080278128951e-06,
+      "loss": 3.2412,
+      "step": 7663
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6390328697850824e-06,
+      "loss": 3.5399,
+      "step": 7664
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6389854614412137e-06,
+      "loss": 3.175,
+      "step": 7665
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638938053097345e-06,
+      "loss": 3.7792,
+      "step": 7666
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6388906447534767e-06,
+      "loss": 3.1741,
+      "step": 7667
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638843236409608e-06,
+      "loss": 3.7295,
+      "step": 7668
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6387958280657397e-06,
+      "loss": 3.7561,
+      "step": 7669
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6387484197218714e-06,
+      "loss": 3.0781,
+      "step": 7670
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6387010113780026e-06,
+      "loss": 3.0341,
+      "step": 7671
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638653603034134e-06,
+      "loss": 3.7454,
+      "step": 7672
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6386061946902656e-06,
+      "loss": 4.1513,
+      "step": 7673
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638558786346397e-06,
+      "loss": 3.7785,
+      "step": 7674
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638511378002528e-06,
+      "loss": 3.0078,
+      "step": 7675
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6384639696586603e-06,
+      "loss": 4.1686,
+      "step": 7676
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6384165613147916e-06,
+      "loss": 4.0148,
+      "step": 7677
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638369152970923e-06,
+      "loss": 4.0972,
+      "step": 7678
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6383217446270546e-06,
+      "loss": 4.1127,
+      "step": 7679
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638274336283186e-06,
+      "loss": 3.843,
+      "step": 7680
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638226927939317e-06,
+      "loss": 3.1204,
+      "step": 7681
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638179519595449e-06,
+      "loss": 3.8913,
+      "step": 7682
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6381321112515805e-06,
+      "loss": 4.35,
+      "step": 7683
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.638084702907712e-06,
+      "loss": 3.5833,
+      "step": 7684
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6380372945638435e-06,
+      "loss": 3.7165,
+      "step": 7685
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.637989886219975e-06,
+      "loss": 3.8733,
+      "step": 7686
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.637942477876106e-06,
+      "loss": 3.7192,
+      "step": 7687
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6378950695322374e-06,
+      "loss": 4.1867,
+      "step": 7688
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.637847661188369e-06,
+      "loss": 3.5911,
+      "step": 7689
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6378002528445008e-06,
+      "loss": 3.7871,
+      "step": 7690
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.637752844500632e-06,
+      "loss": 3.2547,
+      "step": 7691
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6377054361567638e-06,
+      "loss": 2.8213,
+      "step": 7692
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.637658027812895e-06,
+      "loss": 3.673,
+      "step": 7693
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2.6376106194690263e-06,
+      "loss": 3.7176,
+      "step": 7694
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.637563211125158e-06,
+      "loss": 3.8001,
+      "step": 7695
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6375158027812897e-06,
+      "loss": 3.5623,
+      "step": 7696
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.637468394437421e-06,
+      "loss": 4.0416,
+      "step": 7697
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6374209860935527e-06,
+      "loss": 3.372,
+      "step": 7698
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.637373577749684e-06,
+      "loss": 3.6768,
+      "step": 7699
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6373261694058153e-06,
+      "loss": 3.3454,
+      "step": 7700
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.637278761061947e-06,
+      "loss": 3.4849,
+      "step": 7701
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6372313527180783e-06,
+      "loss": 3.4179,
+      "step": 7702
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.63718394437421e-06,
+      "loss": 3.3656,
+      "step": 7703
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6371365360303417e-06,
+      "loss": 3.4299,
+      "step": 7704
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.637089127686473e-06,
+      "loss": 3.2278,
+      "step": 7705
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6370417193426042e-06,
+      "loss": 3.5024,
+      "step": 7706
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636994310998736e-06,
+      "loss": 3.825,
+      "step": 7707
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6369469026548672e-06,
+      "loss": 3.8935,
+      "step": 7708
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636899494310999e-06,
+      "loss": 3.4819,
+      "step": 7709
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.63685208596713e-06,
+      "loss": 4.122,
+      "step": 7710
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636804677623262e-06,
+      "loss": 3.8118,
+      "step": 7711
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636757269279393e-06,
+      "loss": 3.9318,
+      "step": 7712
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6367098609355245e-06,
+      "loss": 3.6828,
+      "step": 7713
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636662452591656e-06,
+      "loss": 3.556,
+      "step": 7714
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6366150442477874e-06,
+      "loss": 3.8995,
+      "step": 7715
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636567635903919e-06,
+      "loss": 3.5746,
+      "step": 7716
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636520227560051e-06,
+      "loss": 3.3172,
+      "step": 7717
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636472819216182e-06,
+      "loss": 3.7453,
+      "step": 7718
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6364254108723134e-06,
+      "loss": 3.567,
+      "step": 7719
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636378002528445e-06,
+      "loss": 3.7749,
+      "step": 7720
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6363305941845764e-06,
+      "loss": 3.2659,
+      "step": 7721
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636283185840708e-06,
+      "loss": 3.4936,
+      "step": 7722
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.63623577749684e-06,
+      "loss": 3.5975,
+      "step": 7723
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636188369152971e-06,
+      "loss": 3.6829,
+      "step": 7724
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6361409608091024e-06,
+      "loss": 3.0534,
+      "step": 7725
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.636093552465234e-06,
+      "loss": 3.2907,
+      "step": 7726
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6360461441213654e-06,
+      "loss": 3.7003,
+      "step": 7727
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6359987357774966e-06,
+      "loss": 3.2206,
+      "step": 7728
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6359513274336283e-06,
+      "loss": 3.5445,
+      "step": 7729
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.63590391908976e-06,
+      "loss": 3.6435,
+      "step": 7730
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6358565107458913e-06,
+      "loss": 2.927,
+      "step": 7731
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6358091024020226e-06,
+      "loss": 3.3616,
+      "step": 7732
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6357616940581543e-06,
+      "loss": 3.4742,
+      "step": 7733
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6357142857142856e-06,
+      "loss": 3.1168,
+      "step": 7734
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6356668773704173e-06,
+      "loss": 3.623,
+      "step": 7735
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.635619469026549e-06,
+      "loss": 3.7037,
+      "step": 7736
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6355720606826803e-06,
+      "loss": 3.3418,
+      "step": 7737
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6355246523388116e-06,
+      "loss": 3.7856,
+      "step": 7738
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6354772439949433e-06,
+      "loss": 3.3337,
+      "step": 7739
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6354298356510745e-06,
+      "loss": 3.5247,
+      "step": 7740
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.635382427307206e-06,
+      "loss": 3.8309,
+      "step": 7741
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6353350189633375e-06,
+      "loss": 3.3538,
+      "step": 7742
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6352876106194692e-06,
+      "loss": 3.1189,
+      "step": 7743
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6352402022756005e-06,
+      "loss": 3.0687,
+      "step": 7744
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6351927939317322e-06,
+      "loss": 3.5474,
+      "step": 7745
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6351453855878635e-06,
+      "loss": 3.3895,
+      "step": 7746
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6350979772439948e-06,
+      "loss": 3.9772,
+      "step": 7747
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6350505689001265e-06,
+      "loss": 4.3745,
+      "step": 7748
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.635003160556258e-06,
+      "loss": 3.2992,
+      "step": 7749
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6349557522123895e-06,
+      "loss": 3.8142,
+      "step": 7750
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.634908343868521e-06,
+      "loss": 2.9092,
+      "step": 7751
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6348609355246524e-06,
+      "loss": 3.3349,
+      "step": 7752
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6348135271807837e-06,
+      "loss": 3.5755,
+      "step": 7753
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.634766118836915e-06,
+      "loss": 3.6209,
+      "step": 7754
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6347187104930467e-06,
+      "loss": 3.4184,
+      "step": 7755
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6346713021491784e-06,
+      "loss": 3.0447,
+      "step": 7756
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.6346238938053097e-06,
+      "loss": 4.0379,
+      "step": 7757
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6345764854614414e-06,
+      "loss": 3.4218,
+      "step": 7758
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6345290771175727e-06,
+      "loss": 3.0074,
+      "step": 7759
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.634481668773704e-06,
+      "loss": 4.0252,
+      "step": 7760
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6344342604298357e-06,
+      "loss": 2.9218,
+      "step": 7761
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6343868520859674e-06,
+      "loss": 3.8144,
+      "step": 7762
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6343394437420986e-06,
+      "loss": 3.8316,
+      "step": 7763
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6342920353982304e-06,
+      "loss": 3.3826,
+      "step": 7764
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6342446270543616e-06,
+      "loss": 3.7015,
+      "step": 7765
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.634197218710493e-06,
+      "loss": 3.0469,
+      "step": 7766
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6341498103666246e-06,
+      "loss": 3.7255,
+      "step": 7767
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.634102402022756e-06,
+      "loss": 2.9868,
+      "step": 7768
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6340549936788876e-06,
+      "loss": 3.7191,
+      "step": 7769
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6340075853350193e-06,
+      "loss": 3.4138,
+      "step": 7770
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6339601769911506e-06,
+      "loss": 3.3538,
+      "step": 7771
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633912768647282e-06,
+      "loss": 3.3416,
+      "step": 7772
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6338653603034136e-06,
+      "loss": 3.6374,
+      "step": 7773
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633817951959545e-06,
+      "loss": 3.9394,
+      "step": 7774
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6337705436156766e-06,
+      "loss": 3.3607,
+      "step": 7775
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633723135271808e-06,
+      "loss": 3.8964,
+      "step": 7776
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6336757269279395e-06,
+      "loss": 3.6913,
+      "step": 7777
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633628318584071e-06,
+      "loss": 3.4856,
+      "step": 7778
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633580910240202e-06,
+      "loss": 3.615,
+      "step": 7779
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633533501896334e-06,
+      "loss": 3.4187,
+      "step": 7780
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633486093552465e-06,
+      "loss": 3.7705,
+      "step": 7781
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633438685208597e-06,
+      "loss": 4.2161,
+      "step": 7782
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6333912768647285e-06,
+      "loss": 3.808,
+      "step": 7783
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6333438685208598e-06,
+      "loss": 4.1439,
+      "step": 7784
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633296460176991e-06,
+      "loss": 3.4769,
+      "step": 7785
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6332490518331228e-06,
+      "loss": 3.6751,
+      "step": 7786
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.633201643489254e-06,
+      "loss": 3.6486,
+      "step": 7787
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6331542351453857e-06,
+      "loss": 4.1791,
+      "step": 7788
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6331068268015174e-06,
+      "loss": 3.3169,
+      "step": 7789
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6330594184576487e-06,
+      "loss": 3.6494,
+      "step": 7790
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.63301201011378e-06,
+      "loss": 4.1657,
+      "step": 7791
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6329646017699117e-06,
+      "loss": 3.8202,
+      "step": 7792
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632917193426043e-06,
+      "loss": 3.5415,
+      "step": 7793
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6328697850821743e-06,
+      "loss": 3.1079,
+      "step": 7794
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632822376738306e-06,
+      "loss": 4.0595,
+      "step": 7795
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6327749683944377e-06,
+      "loss": 3.6641,
+      "step": 7796
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632727560050569e-06,
+      "loss": 3.223,
+      "step": 7797
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6326801517067002e-06,
+      "loss": 4.1631,
+      "step": 7798
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632632743362832e-06,
+      "loss": 2.763,
+      "step": 7799
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6325853350189632e-06,
+      "loss": 3.6544,
+      "step": 7800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632537926675095e-06,
+      "loss": 3.8701,
+      "step": 7801
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6324905183312266e-06,
+      "loss": 3.4756,
+      "step": 7802
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632443109987358e-06,
+      "loss": 4.0005,
+      "step": 7803
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632395701643489e-06,
+      "loss": 3.3571,
+      "step": 7804
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632348293299621e-06,
+      "loss": 3.5905,
+      "step": 7805
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632300884955752e-06,
+      "loss": 3.2169,
+      "step": 7806
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6322534766118835e-06,
+      "loss": 3.3898,
+      "step": 7807
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632206068268015e-06,
+      "loss": 3.3093,
+      "step": 7808
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632158659924147e-06,
+      "loss": 3.4427,
+      "step": 7809
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632111251580278e-06,
+      "loss": 3.8348,
+      "step": 7810
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.63206384323641e-06,
+      "loss": 3.7992,
+      "step": 7811
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.632016434892541e-06,
+      "loss": 3.5427,
+      "step": 7812
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6319690265486724e-06,
+      "loss": 3.8387,
+      "step": 7813
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.631921618204804e-06,
+      "loss": 3.9936,
+      "step": 7814
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.631874209860936e-06,
+      "loss": 3.9796,
+      "step": 7815
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.631826801517067e-06,
+      "loss": 2.7681,
+      "step": 7816
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.631779393173199e-06,
+      "loss": 3.7072,
+      "step": 7817
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.63173198482933e-06,
+      "loss": 4.1857,
+      "step": 7818
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6316845764854614e-06,
+      "loss": 3.2209,
+      "step": 7819
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6316371681415926e-06,
+      "loss": 3.9735,
+      "step": 7820
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.6315897597977243e-06,
+      "loss": 3.9959,
+      "step": 7821
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.631542351453856e-06,
+      "loss": 3.5824,
+      "step": 7822
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6314949431099873e-06,
+      "loss": 3.8254,
+      "step": 7823
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.631447534766119e-06,
+      "loss": 2.484,
+      "step": 7824
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6314001264222503e-06,
+      "loss": 3.2278,
+      "step": 7825
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6313527180783816e-06,
+      "loss": 4.1383,
+      "step": 7826
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6313053097345133e-06,
+      "loss": 3.8888,
+      "step": 7827
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.631257901390645e-06,
+      "loss": 3.3323,
+      "step": 7828
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6312104930467763e-06,
+      "loss": 3.5829,
+      "step": 7829
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.631163084702908e-06,
+      "loss": 3.6963,
+      "step": 7830
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6311156763590393e-06,
+      "loss": 4.2393,
+      "step": 7831
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6310682680151706e-06,
+      "loss": 3.7647,
+      "step": 7832
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6310208596713023e-06,
+      "loss": 3.8426,
+      "step": 7833
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6309734513274335e-06,
+      "loss": 2.612,
+      "step": 7834
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6309260429835652e-06,
+      "loss": 4.1201,
+      "step": 7835
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.630878634639697e-06,
+      "loss": 3.1184,
+      "step": 7836
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6308312262958282e-06,
+      "loss": 3.7096,
+      "step": 7837
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6307838179519595e-06,
+      "loss": 3.3065,
+      "step": 7838
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.630736409608091e-06,
+      "loss": 2.84,
+      "step": 7839
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6306890012642225e-06,
+      "loss": 3.4715,
+      "step": 7840
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.630641592920354e-06,
+      "loss": 4.0461,
+      "step": 7841
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6305941845764855e-06,
+      "loss": 4.1329,
+      "step": 7842
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.630546776232617e-06,
+      "loss": 3.4367,
+      "step": 7843
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6304993678887485e-06,
+      "loss": 3.4328,
+      "step": 7844
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6304519595448797e-06,
+      "loss": 3.7246,
+      "step": 7845
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6304045512010114e-06,
+      "loss": 3.7666,
+      "step": 7846
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6303571428571427e-06,
+      "loss": 4.1222,
+      "step": 7847
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6303097345132744e-06,
+      "loss": 3.9874,
+      "step": 7848
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.630262326169406e-06,
+      "loss": 3.9746,
+      "step": 7849
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6302149178255374e-06,
+      "loss": 3.8761,
+      "step": 7850
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6301675094816687e-06,
+      "loss": 3.3189,
+      "step": 7851
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6301201011378004e-06,
+      "loss": 3.8486,
+      "step": 7852
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6300726927939317e-06,
+      "loss": 4.1011,
+      "step": 7853
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6300252844500634e-06,
+      "loss": 3.6922,
+      "step": 7854
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629977876106195e-06,
+      "loss": 4.0377,
+      "step": 7855
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6299304677623264e-06,
+      "loss": 3.5958,
+      "step": 7856
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6298830594184576e-06,
+      "loss": 3.8034,
+      "step": 7857
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6298356510745893e-06,
+      "loss": 3.4784,
+      "step": 7858
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6297882427307206e-06,
+      "loss": 3.039,
+      "step": 7859
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629740834386852e-06,
+      "loss": 3.7874,
+      "step": 7860
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6296934260429836e-06,
+      "loss": 3.929,
+      "step": 7861
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6296460176991153e-06,
+      "loss": 4.253,
+      "step": 7862
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6295986093552466e-06,
+      "loss": 3.6173,
+      "step": 7863
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629551201011378e-06,
+      "loss": 3.8563,
+      "step": 7864
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6295037926675096e-06,
+      "loss": 4.0781,
+      "step": 7865
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629456384323641e-06,
+      "loss": 3.8528,
+      "step": 7866
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6294089759797726e-06,
+      "loss": 3.8417,
+      "step": 7867
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6293615676359043e-06,
+      "loss": 3.4283,
+      "step": 7868
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6293141592920355e-06,
+      "loss": 4.0645,
+      "step": 7869
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629266750948167e-06,
+      "loss": 3.3411,
+      "step": 7870
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6292193426042985e-06,
+      "loss": 3.8214,
+      "step": 7871
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.62917193426043e-06,
+      "loss": 3.7184,
+      "step": 7872
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629124525916561e-06,
+      "loss": 3.0977,
+      "step": 7873
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.629077117572693e-06,
+      "loss": 4.0227,
+      "step": 7874
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6290297092288245e-06,
+      "loss": 2.9626,
+      "step": 7875
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6289823008849558e-06,
+      "loss": 3.9201,
+      "step": 7876
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6289348925410875e-06,
+      "loss": 3.8379,
+      "step": 7877
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6288874841972188e-06,
+      "loss": 3.4491,
+      "step": 7878
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.62884007585335e-06,
+      "loss": 3.3601,
+      "step": 7879
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6287926675094818e-06,
+      "loss": 3.4596,
+      "step": 7880
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6287452591656135e-06,
+      "loss": 3.4495,
+      "step": 7881
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6286978508217447e-06,
+      "loss": 3.9117,
+      "step": 7882
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6286504424778764e-06,
+      "loss": 3.8888,
+      "step": 7883
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2.6286030341340077e-06,
+      "loss": 4.3783,
+      "step": 7884
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.628555625790139e-06,
+      "loss": 3.7863,
+      "step": 7885
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6285082174462703e-06,
+      "loss": 4.1688,
+      "step": 7886
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.628460809102402e-06,
+      "loss": 3.8463,
+      "step": 7887
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6284134007585337e-06,
+      "loss": 3.7372,
+      "step": 7888
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.628365992414665e-06,
+      "loss": 3.4652,
+      "step": 7889
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6283185840707967e-06,
+      "loss": 4.0316,
+      "step": 7890
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.628271175726928e-06,
+      "loss": 3.9609,
+      "step": 7891
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6282237673830592e-06,
+      "loss": 3.7396,
+      "step": 7892
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.628176359039191e-06,
+      "loss": 3.5888,
+      "step": 7893
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6281289506953226e-06,
+      "loss": 3.6783,
+      "step": 7894
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.628081542351454e-06,
+      "loss": 3.4138,
+      "step": 7895
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6280341340075856e-06,
+      "loss": 3.6869,
+      "step": 7896
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627986725663717e-06,
+      "loss": 4.0272,
+      "step": 7897
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627939317319848e-06,
+      "loss": 4.1561,
+      "step": 7898
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.62789190897598e-06,
+      "loss": 3.1326,
+      "step": 7899
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627844500632111e-06,
+      "loss": 3.805,
+      "step": 7900
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627797092288243e-06,
+      "loss": 3.2096,
+      "step": 7901
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6277496839443746e-06,
+      "loss": 3.4012,
+      "step": 7902
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627702275600506e-06,
+      "loss": 3.5636,
+      "step": 7903
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627654867256637e-06,
+      "loss": 2.9336,
+      "step": 7904
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627607458912769e-06,
+      "loss": 3.2948,
+      "step": 7905
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6275600505689e-06,
+      "loss": 4.0026,
+      "step": 7906
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627512642225032e-06,
+      "loss": 4.3814,
+      "step": 7907
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627465233881163e-06,
+      "loss": 2.835,
+      "step": 7908
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627417825537295e-06,
+      "loss": 3.9385,
+      "step": 7909
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627370417193426e-06,
+      "loss": 2.7803,
+      "step": 7910
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6273230088495574e-06,
+      "loss": 3.9693,
+      "step": 7911
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627275600505689e-06,
+      "loss": 3.609,
+      "step": 7912
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6272281921618204e-06,
+      "loss": 4.3251,
+      "step": 7913
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627180783817952e-06,
+      "loss": 3.8238,
+      "step": 7914
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6271333754740838e-06,
+      "loss": 3.6959,
+      "step": 7915
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.627085967130215e-06,
+      "loss": 3.9632,
+      "step": 7916
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6270385587863463e-06,
+      "loss": 4.0699,
+      "step": 7917
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626991150442478e-06,
+      "loss": 2.9,
+      "step": 7918
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6269437420986093e-06,
+      "loss": 3.378,
+      "step": 7919
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626896333754741e-06,
+      "loss": 3.1347,
+      "step": 7920
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6268489254108727e-06,
+      "loss": 3.2246,
+      "step": 7921
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626801517067004e-06,
+      "loss": 3.174,
+      "step": 7922
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6267541087231353e-06,
+      "loss": 3.6827,
+      "step": 7923
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626706700379267e-06,
+      "loss": 3.8774,
+      "step": 7924
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6266592920353983e-06,
+      "loss": 3.7792,
+      "step": 7925
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6266118836915295e-06,
+      "loss": 3.3984,
+      "step": 7926
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6265644753476612e-06,
+      "loss": 3.5266,
+      "step": 7927
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626517067003793e-06,
+      "loss": 3.1153,
+      "step": 7928
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6264696586599242e-06,
+      "loss": 3.8727,
+      "step": 7929
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6264222503160555e-06,
+      "loss": 3.4093,
+      "step": 7930
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6263748419721872e-06,
+      "loss": 3.8844,
+      "step": 7931
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6263274336283185e-06,
+      "loss": 3.3531,
+      "step": 7932
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.62628002528445e-06,
+      "loss": 3.5071,
+      "step": 7933
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626232616940582e-06,
+      "loss": 3.5601,
+      "step": 7934
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626185208596713e-06,
+      "loss": 3.7129,
+      "step": 7935
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6261378002528445e-06,
+      "loss": 4.0151,
+      "step": 7936
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.626090391908976e-06,
+      "loss": 4.1979,
+      "step": 7937
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6260429835651075e-06,
+      "loss": 3.3069,
+      "step": 7938
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6259955752212387e-06,
+      "loss": 3.5926,
+      "step": 7939
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6259481668773704e-06,
+      "loss": 3.6645,
+      "step": 7940
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.625900758533502e-06,
+      "loss": 3.5309,
+      "step": 7941
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6258533501896334e-06,
+      "loss": 3.8514,
+      "step": 7942
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.625805941845765e-06,
+      "loss": 3.7838,
+      "step": 7943
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6257585335018964e-06,
+      "loss": 4.1391,
+      "step": 7944
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6257111251580277e-06,
+      "loss": 3.7903,
+      "step": 7945
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.6256637168141594e-06,
+      "loss": 3.666,
+      "step": 7946
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.625616308470291e-06,
+      "loss": 3.7808,
+      "step": 7947
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6255689001264224e-06,
+      "loss": 3.2262,
+      "step": 7948
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.625521491782554e-06,
+      "loss": 3.0226,
+      "step": 7949
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6254740834386854e-06,
+      "loss": 3.3737,
+      "step": 7950
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6254266750948166e-06,
+      "loss": 3.2891,
+      "step": 7951
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.625379266750948e-06,
+      "loss": 3.1923,
+      "step": 7952
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6253318584070796e-06,
+      "loss": 3.5831,
+      "step": 7953
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6252844500632113e-06,
+      "loss": 3.6436,
+      "step": 7954
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6252370417193426e-06,
+      "loss": 3.5343,
+      "step": 7955
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6251896333754743e-06,
+      "loss": 3.9327,
+      "step": 7956
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6251422250316056e-06,
+      "loss": 3.6683,
+      "step": 7957
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.625094816687737e-06,
+      "loss": 3.9906,
+      "step": 7958
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6250474083438686e-06,
+      "loss": 3.9926,
+      "step": 7959
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6250000000000003e-06,
+      "loss": 3.5309,
+      "step": 7960
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6249525916561316e-06,
+      "loss": 3.5191,
+      "step": 7961
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6249051833122633e-06,
+      "loss": 4.0616,
+      "step": 7962
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6248577749683945e-06,
+      "loss": 3.3043,
+      "step": 7963
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.624810366624526e-06,
+      "loss": 3.6602,
+      "step": 7964
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6247629582806575e-06,
+      "loss": 3.4118,
+      "step": 7965
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.624715549936789e-06,
+      "loss": 3.67,
+      "step": 7966
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6246681415929205e-06,
+      "loss": 3.3113,
+      "step": 7967
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6246207332490522e-06,
+      "loss": 3.5686,
+      "step": 7968
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6245733249051835e-06,
+      "loss": 3.8586,
+      "step": 7969
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6245259165613148e-06,
+      "loss": 3.5649,
+      "step": 7970
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.624478508217446e-06,
+      "loss": 3.0121,
+      "step": 7971
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6244310998735778e-06,
+      "loss": 3.6387,
+      "step": 7972
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.624383691529709e-06,
+      "loss": 2.9092,
+      "step": 7973
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6243362831858407e-06,
+      "loss": 3.4874,
+      "step": 7974
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6242888748419724e-06,
+      "loss": 3.6248,
+      "step": 7975
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6242414664981037e-06,
+      "loss": 3.1014,
+      "step": 7976
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.624194058154235e-06,
+      "loss": 3.3994,
+      "step": 7977
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6241466498103667e-06,
+      "loss": 3.3717,
+      "step": 7978
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.624099241466498e-06,
+      "loss": 3.8361,
+      "step": 7979
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6240518331226297e-06,
+      "loss": 3.3323,
+      "step": 7980
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6240044247787614e-06,
+      "loss": 3.7524,
+      "step": 7981
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6239570164348927e-06,
+      "loss": 4.1517,
+      "step": 7982
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623909608091024e-06,
+      "loss": 3.1089,
+      "step": 7983
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6238621997471557e-06,
+      "loss": 3.6824,
+      "step": 7984
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623814791403287e-06,
+      "loss": 3.6084,
+      "step": 7985
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6237673830594182e-06,
+      "loss": 3.8905,
+      "step": 7986
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.62371997471555e-06,
+      "loss": 3.2065,
+      "step": 7987
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6236725663716816e-06,
+      "loss": 3.8074,
+      "step": 7988
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623625158027813e-06,
+      "loss": 3.5008,
+      "step": 7989
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6235777496839446e-06,
+      "loss": 3.912,
+      "step": 7990
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623530341340076e-06,
+      "loss": 3.5273,
+      "step": 7991
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623482932996207e-06,
+      "loss": 3.5157,
+      "step": 7992
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623435524652339e-06,
+      "loss": 3.3066,
+      "step": 7993
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6233881163084706e-06,
+      "loss": 3.0615,
+      "step": 7994
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623340707964602e-06,
+      "loss": 3.8746,
+      "step": 7995
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623293299620733e-06,
+      "loss": 3.6638,
+      "step": 7996
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623245891276865e-06,
+      "loss": 3.7536,
+      "step": 7997
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623198482932996e-06,
+      "loss": 3.8951,
+      "step": 7998
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6231510745891274e-06,
+      "loss": 3.5475,
+      "step": 7999
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623103666245259e-06,
+      "loss": 3.5722,
+      "step": 8000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623056257901391e-06,
+      "loss": 3.0004,
+      "step": 8001
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.623008849557522e-06,
+      "loss": 3.152,
+      "step": 8002
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.622961441213654e-06,
+      "loss": 3.6865,
+      "step": 8003
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.622914032869785e-06,
+      "loss": 3.6676,
+      "step": 8004
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6228666245259164e-06,
+      "loss": 3.9322,
+      "step": 8005
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.622819216182048e-06,
+      "loss": 3.796,
+      "step": 8006
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6227718078381798e-06,
+      "loss": 3.447,
+      "step": 8007
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.622724399494311e-06,
+      "loss": 3.7935,
+      "step": 8008
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6226769911504428e-06,
+      "loss": 3.7392,
+      "step": 8009
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.622629582806574e-06,
+      "loss": 3.3558,
+      "step": 8010
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2.6225821744627053e-06,
+      "loss": 3.3705,
+      "step": 8011
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.622534766118837e-06,
+      "loss": 3.9435,
+      "step": 8012
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6224873577749683e-06,
+      "loss": 3.8761,
+      "step": 8013
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6224399494311e-06,
+      "loss": 3.7281,
+      "step": 8014
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6223925410872313e-06,
+      "loss": 3.5363,
+      "step": 8015
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.622345132743363e-06,
+      "loss": 3.6831,
+      "step": 8016
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6222977243994943e-06,
+      "loss": 4.2054,
+      "step": 8017
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6222503160556256e-06,
+      "loss": 3.739,
+      "step": 8018
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6222029077117573e-06,
+      "loss": 3.5866,
+      "step": 8019
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.622155499367889e-06,
+      "loss": 3.2784,
+      "step": 8020
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6221080910240202e-06,
+      "loss": 3.5656,
+      "step": 8021
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.622060682680152e-06,
+      "loss": 3.9452,
+      "step": 8022
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6220132743362832e-06,
+      "loss": 3.6694,
+      "step": 8023
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6219658659924145e-06,
+      "loss": 3.436,
+      "step": 8024
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621918457648546e-06,
+      "loss": 4.1676,
+      "step": 8025
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6218710493046775e-06,
+      "loss": 3.8666,
+      "step": 8026
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621823640960809e-06,
+      "loss": 3.4816,
+      "step": 8027
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621776232616941e-06,
+      "loss": 3.8541,
+      "step": 8028
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621728824273072e-06,
+      "loss": 3.7031,
+      "step": 8029
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6216814159292035e-06,
+      "loss": 3.2558,
+      "step": 8030
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621634007585335e-06,
+      "loss": 3.5229,
+      "step": 8031
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6215865992414664e-06,
+      "loss": 4.0162,
+      "step": 8032
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621539190897598e-06,
+      "loss": 3.2747,
+      "step": 8033
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.62149178255373e-06,
+      "loss": 3.8875,
+      "step": 8034
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.621444374209861e-06,
+      "loss": 3.6336,
+      "step": 8035
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6213969658659924e-06,
+      "loss": 3.8014,
+      "step": 8036
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6213495575221237e-06,
+      "loss": 2.7137,
+      "step": 8037
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6213021491782554e-06,
+      "loss": 3.5328,
+      "step": 8038
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6212547408343867e-06,
+      "loss": 3.4963,
+      "step": 8039
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6212073324905184e-06,
+      "loss": 3.9313,
+      "step": 8040
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.62115992414665e-06,
+      "loss": 3.7748,
+      "step": 8041
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6211125158027814e-06,
+      "loss": 3.3967,
+      "step": 8042
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6210651074589126e-06,
+      "loss": 4.0456,
+      "step": 8043
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6210176991150444e-06,
+      "loss": 3.3364,
+      "step": 8044
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6209702907711756e-06,
+      "loss": 3.5747,
+      "step": 8045
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6209228824273073e-06,
+      "loss": 3.1982,
+      "step": 8046
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.620875474083439e-06,
+      "loss": 3.3069,
+      "step": 8047
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6208280657395703e-06,
+      "loss": 4.059,
+      "step": 8048
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6207806573957016e-06,
+      "loss": 3.1711,
+      "step": 8049
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6207332490518333e-06,
+      "loss": 4.0847,
+      "step": 8050
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6206858407079646e-06,
+      "loss": 3.3451,
+      "step": 8051
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.620638432364096e-06,
+      "loss": 3.8929,
+      "step": 8052
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6205910240202276e-06,
+      "loss": 3.1972,
+      "step": 8053
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6205436156763593e-06,
+      "loss": 3.9085,
+      "step": 8054
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6204962073324906e-06,
+      "loss": 3.5493,
+      "step": 8055
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6204487989886223e-06,
+      "loss": 3.8557,
+      "step": 8056
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6204013906447535e-06,
+      "loss": 4.0709,
+      "step": 8057
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.620353982300885e-06,
+      "loss": 3.5003,
+      "step": 8058
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6203065739570165e-06,
+      "loss": 3.9751,
+      "step": 8059
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6202591656131482e-06,
+      "loss": 3.7621,
+      "step": 8060
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6202117572692795e-06,
+      "loss": 3.8486,
+      "step": 8061
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6201643489254108e-06,
+      "loss": 4.2274,
+      "step": 8062
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6201169405815425e-06,
+      "loss": 2.8123,
+      "step": 8063
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6200695322376738e-06,
+      "loss": 3.8739,
+      "step": 8064
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.620022123893805e-06,
+      "loss": 3.6034,
+      "step": 8065
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6199747155499368e-06,
+      "loss": 3.627,
+      "step": 8066
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6199273072060685e-06,
+      "loss": 3.6183,
+      "step": 8067
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6198798988621997e-06,
+      "loss": 3.73,
+      "step": 8068
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6198324905183314e-06,
+      "loss": 3.7825,
+      "step": 8069
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6197850821744627e-06,
+      "loss": 3.7387,
+      "step": 8070
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.619737673830594e-06,
+      "loss": 3.9603,
+      "step": 8071
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6196902654867257e-06,
+      "loss": 3.8751,
+      "step": 8072
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6196428571428574e-06,
+      "loss": 3.467,
+      "step": 8073
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.6195954487989887e-06,
+      "loss": 3.9593,
+      "step": 8074
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6195480404551204e-06,
+      "loss": 4.0799,
+      "step": 8075
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6195006321112517e-06,
+      "loss": 4.4602,
+      "step": 8076
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.619453223767383e-06,
+      "loss": 3.2443,
+      "step": 8077
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6194058154235147e-06,
+      "loss": 3.8165,
+      "step": 8078
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.619358407079646e-06,
+      "loss": 3.6309,
+      "step": 8079
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6193109987357776e-06,
+      "loss": 3.8103,
+      "step": 8080
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.619263590391909e-06,
+      "loss": 3.0438,
+      "step": 8081
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6192161820480406e-06,
+      "loss": 3.5027,
+      "step": 8082
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.619168773704172e-06,
+      "loss": 3.3464,
+      "step": 8083
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.619121365360303e-06,
+      "loss": 3.1376,
+      "step": 8084
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.619073957016435e-06,
+      "loss": 3.2731,
+      "step": 8085
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6190265486725666e-06,
+      "loss": 4.0843,
+      "step": 8086
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618979140328698e-06,
+      "loss": 3.7818,
+      "step": 8087
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6189317319848296e-06,
+      "loss": 4.1031,
+      "step": 8088
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618884323640961e-06,
+      "loss": 3.1428,
+      "step": 8089
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618836915297092e-06,
+      "loss": 3.7677,
+      "step": 8090
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618789506953224e-06,
+      "loss": 3.4141,
+      "step": 8091
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618742098609355e-06,
+      "loss": 3.6979,
+      "step": 8092
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618694690265487e-06,
+      "loss": 3.0378,
+      "step": 8093
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6186472819216185e-06,
+      "loss": 4.0742,
+      "step": 8094
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.61859987357775e-06,
+      "loss": 3.8159,
+      "step": 8095
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618552465233881e-06,
+      "loss": 3.6949,
+      "step": 8096
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618505056890013e-06,
+      "loss": 3.9588,
+      "step": 8097
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618457648546144e-06,
+      "loss": 3.5144,
+      "step": 8098
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6184102402022758e-06,
+      "loss": 3.1753,
+      "step": 8099
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6183628318584075e-06,
+      "loss": 3.9405,
+      "step": 8100
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6183154235145388e-06,
+      "loss": 3.0451,
+      "step": 8101
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.61826801517067e-06,
+      "loss": 3.8088,
+      "step": 8102
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6182206068268013e-06,
+      "loss": 3.0751,
+      "step": 8103
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618173198482933e-06,
+      "loss": 3.9261,
+      "step": 8104
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6181257901390643e-06,
+      "loss": 3.8003,
+      "step": 8105
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.618078381795196e-06,
+      "loss": 3.7302,
+      "step": 8106
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6180309734513277e-06,
+      "loss": 3.7453,
+      "step": 8107
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617983565107459e-06,
+      "loss": 3.4109,
+      "step": 8108
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6179361567635903e-06,
+      "loss": 4.3914,
+      "step": 8109
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617888748419722e-06,
+      "loss": 3.0559,
+      "step": 8110
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6178413400758533e-06,
+      "loss": 3.9872,
+      "step": 8111
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617793931731985e-06,
+      "loss": 2.9144,
+      "step": 8112
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6177465233881167e-06,
+      "loss": 3.5012,
+      "step": 8113
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617699115044248e-06,
+      "loss": 4.1083,
+      "step": 8114
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6176517067003792e-06,
+      "loss": 3.598,
+      "step": 8115
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617604298356511e-06,
+      "loss": 3.456,
+      "step": 8116
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6175568900126422e-06,
+      "loss": 4.1465,
+      "step": 8117
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6175094816687735e-06,
+      "loss": 3.8744,
+      "step": 8118
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617462073324905e-06,
+      "loss": 3.9178,
+      "step": 8119
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617414664981037e-06,
+      "loss": 3.7048,
+      "step": 8120
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617367256637168e-06,
+      "loss": 4.1046,
+      "step": 8121
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6173198482933e-06,
+      "loss": 3.8141,
+      "step": 8122
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617272439949431e-06,
+      "loss": 3.6439,
+      "step": 8123
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6172250316055625e-06,
+      "loss": 3.4875,
+      "step": 8124
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617177623261694e-06,
+      "loss": 3.2162,
+      "step": 8125
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617130214917826e-06,
+      "loss": 3.7744,
+      "step": 8126
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.617082806573957e-06,
+      "loss": 3.45,
+      "step": 8127
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6170353982300884e-06,
+      "loss": 3.3907,
+      "step": 8128
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.61698798988622e-06,
+      "loss": 3.5688,
+      "step": 8129
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6169405815423514e-06,
+      "loss": 4.0305,
+      "step": 8130
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6168931731984827e-06,
+      "loss": 3.6656,
+      "step": 8131
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6168457648546144e-06,
+      "loss": 3.793,
+      "step": 8132
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.616798356510746e-06,
+      "loss": 3.8477,
+      "step": 8133
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6167509481668774e-06,
+      "loss": 3.0491,
+      "step": 8134
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.616703539823009e-06,
+      "loss": 3.8106,
+      "step": 8135
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6166561314791404e-06,
+      "loss": 3.4439,
+      "step": 8136
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2.6166087231352716e-06,
+      "loss": 3.3295,
+      "step": 8137
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6165613147914033e-06,
+      "loss": 3.6652,
+      "step": 8138
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.616513906447535e-06,
+      "loss": 3.9821,
+      "step": 8139
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6164664981036663e-06,
+      "loss": 3.2929,
+      "step": 8140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.616419089759798e-06,
+      "loss": 3.8073,
+      "step": 8141
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6163716814159293e-06,
+      "loss": 3.2549,
+      "step": 8142
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6163242730720606e-06,
+      "loss": 3.7693,
+      "step": 8143
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6162768647281923e-06,
+      "loss": 3.3459,
+      "step": 8144
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6162294563843236e-06,
+      "loss": 4.0224,
+      "step": 8145
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6161820480404553e-06,
+      "loss": 3.2626,
+      "step": 8146
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6161346396965866e-06,
+      "loss": 3.4471,
+      "step": 8147
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6160872313527183e-06,
+      "loss": 4.0666,
+      "step": 8148
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6160398230088495e-06,
+      "loss": 3.4728,
+      "step": 8149
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.615992414664981e-06,
+      "loss": 3.4897,
+      "step": 8150
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6159450063211125e-06,
+      "loss": 3.8453,
+      "step": 8151
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6158975979772442e-06,
+      "loss": 3.3278,
+      "step": 8152
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6158501896333755e-06,
+      "loss": 2.5842,
+      "step": 8153
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6158027812895072e-06,
+      "loss": 3.3543,
+      "step": 8154
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6157553729456385e-06,
+      "loss": 4.0548,
+      "step": 8155
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6157079646017698e-06,
+      "loss": 3.2361,
+      "step": 8156
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6156605562579015e-06,
+      "loss": 3.8119,
+      "step": 8157
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6156131479140328e-06,
+      "loss": 3.9155,
+      "step": 8158
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6155657395701645e-06,
+      "loss": 3.8268,
+      "step": 8159
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.615518331226296e-06,
+      "loss": 3.5684,
+      "step": 8160
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6154709228824275e-06,
+      "loss": 3.9777,
+      "step": 8161
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6154235145385587e-06,
+      "loss": 3.7916,
+      "step": 8162
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6153761061946904e-06,
+      "loss": 3.9006,
+      "step": 8163
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6153286978508217e-06,
+      "loss": 3.6562,
+      "step": 8164
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6152812895069534e-06,
+      "loss": 3.7796,
+      "step": 8165
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.615233881163085e-06,
+      "loss": 3.5772,
+      "step": 8166
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6151864728192164e-06,
+      "loss": 4.2104,
+      "step": 8167
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6151390644753477e-06,
+      "loss": 4.3592,
+      "step": 8168
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.615091656131479e-06,
+      "loss": 3.7087,
+      "step": 8169
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6150442477876107e-06,
+      "loss": 4.0888,
+      "step": 8170
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614996839443742e-06,
+      "loss": 3.8401,
+      "step": 8171
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6149494310998737e-06,
+      "loss": 3.7999,
+      "step": 8172
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6149020227560054e-06,
+      "loss": 3.411,
+      "step": 8173
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6148546144121366e-06,
+      "loss": 3.5528,
+      "step": 8174
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614807206068268e-06,
+      "loss": 3.3935,
+      "step": 8175
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6147597977243996e-06,
+      "loss": 3.235,
+      "step": 8176
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614712389380531e-06,
+      "loss": 3.6844,
+      "step": 8177
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6146649810366626e-06,
+      "loss": 3.0513,
+      "step": 8178
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6146175726927943e-06,
+      "loss": 3.7471,
+      "step": 8179
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6145701643489256e-06,
+      "loss": 3.4897,
+      "step": 8180
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614522756005057e-06,
+      "loss": 3.4838,
+      "step": 8181
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6144753476611886e-06,
+      "loss": 4.1026,
+      "step": 8182
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.61442793931732e-06,
+      "loss": 3.3893,
+      "step": 8183
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614380530973451e-06,
+      "loss": 3.2501,
+      "step": 8184
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614333122629583e-06,
+      "loss": 3.9829,
+      "step": 8185
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6142857142857145e-06,
+      "loss": 3.4661,
+      "step": 8186
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614238305941846e-06,
+      "loss": 3.539,
+      "step": 8187
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6141908975979775e-06,
+      "loss": 4.1266,
+      "step": 8188
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614143489254109e-06,
+      "loss": 3.6596,
+      "step": 8189
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.61409608091024e-06,
+      "loss": 3.541,
+      "step": 8190
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.614048672566372e-06,
+      "loss": 3.7577,
+      "step": 8191
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6140012642225035e-06,
+      "loss": 3.8359,
+      "step": 8192
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6139538558786348e-06,
+      "loss": 3.5689,
+      "step": 8193
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.613906447534766e-06,
+      "loss": 4.1236,
+      "step": 8194
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6138590391908978e-06,
+      "loss": 3.5904,
+      "step": 8195
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.613811630847029e-06,
+      "loss": 3.4632,
+      "step": 8196
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6137642225031603e-06,
+      "loss": 4.1498,
+      "step": 8197
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.613716814159292e-06,
+      "loss": 3.5807,
+      "step": 8198
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6136694058154237e-06,
+      "loss": 3.2874,
+      "step": 8199
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.613621997471555e-06,
+      "loss": 3.3967,
+      "step": 8200
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.6135745891276867e-06,
+      "loss": 2.8461,
+      "step": 8201
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.613527180783818e-06,
+      "loss": 4.036,
+      "step": 8202
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6134797724399493e-06,
+      "loss": 3.7683,
+      "step": 8203
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.613432364096081e-06,
+      "loss": 4.1026,
+      "step": 8204
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6133849557522127e-06,
+      "loss": 3.859,
+      "step": 8205
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.613337547408344e-06,
+      "loss": 3.0895,
+      "step": 8206
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6132901390644757e-06,
+      "loss": 3.8876,
+      "step": 8207
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.613242730720607e-06,
+      "loss": 3.6913,
+      "step": 8208
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6131953223767382e-06,
+      "loss": 3.4054,
+      "step": 8209
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.61314791403287e-06,
+      "loss": 3.9837,
+      "step": 8210
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6131005056890012e-06,
+      "loss": 3.423,
+      "step": 8211
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.613053097345133e-06,
+      "loss": 3.8784,
+      "step": 8212
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.613005689001264e-06,
+      "loss": 3.4672,
+      "step": 8213
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612958280657396e-06,
+      "loss": 4.1597,
+      "step": 8214
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612910872313527e-06,
+      "loss": 3.4844,
+      "step": 8215
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6128634639696585e-06,
+      "loss": 3.5301,
+      "step": 8216
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.61281605562579e-06,
+      "loss": 3.3405,
+      "step": 8217
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612768647281922e-06,
+      "loss": 3.2056,
+      "step": 8218
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612721238938053e-06,
+      "loss": 3.7929,
+      "step": 8219
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612673830594185e-06,
+      "loss": 3.8526,
+      "step": 8220
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612626422250316e-06,
+      "loss": 3.5395,
+      "step": 8221
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6125790139064474e-06,
+      "loss": 3.1127,
+      "step": 8222
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612531605562579e-06,
+      "loss": 3.6853,
+      "step": 8223
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6124841972187104e-06,
+      "loss": 3.8859,
+      "step": 8224
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612436788874842e-06,
+      "loss": 3.7251,
+      "step": 8225
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612389380530974e-06,
+      "loss": 3.0844,
+      "step": 8226
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612341972187105e-06,
+      "loss": 4.2503,
+      "step": 8227
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6122945638432364e-06,
+      "loss": 3.2507,
+      "step": 8228
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612247155499368e-06,
+      "loss": 3.6807,
+      "step": 8229
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6121997471554994e-06,
+      "loss": 3.823,
+      "step": 8230
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612152338811631e-06,
+      "loss": 3.6743,
+      "step": 8231
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6121049304677628e-06,
+      "loss": 3.8006,
+      "step": 8232
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.612057522123894e-06,
+      "loss": 3.5705,
+      "step": 8233
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6120101137800253e-06,
+      "loss": 4.1781,
+      "step": 8234
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6119627054361566e-06,
+      "loss": 3.5628,
+      "step": 8235
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6119152970922883e-06,
+      "loss": 2.8161,
+      "step": 8236
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6118678887484196e-06,
+      "loss": 3.529,
+      "step": 8237
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6118204804045513e-06,
+      "loss": 2.9087,
+      "step": 8238
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.611773072060683e-06,
+      "loss": 3.496,
+      "step": 8239
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6117256637168143e-06,
+      "loss": 3.7127,
+      "step": 8240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6116782553729456e-06,
+      "loss": 3.7972,
+      "step": 8241
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6116308470290773e-06,
+      "loss": 3.676,
+      "step": 8242
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6115834386852085e-06,
+      "loss": 3.5865,
+      "step": 8243
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.61153603034134e-06,
+      "loss": 3.6964,
+      "step": 8244
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.611488621997472e-06,
+      "loss": 3.3269,
+      "step": 8245
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6114412136536032e-06,
+      "loss": 3.3758,
+      "step": 8246
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6113938053097345e-06,
+      "loss": 3.8964,
+      "step": 8247
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6113463969658662e-06,
+      "loss": 3.492,
+      "step": 8248
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6112989886219975e-06,
+      "loss": 3.9594,
+      "step": 8249
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6112515802781288e-06,
+      "loss": 3.1881,
+      "step": 8250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6112041719342605e-06,
+      "loss": 3.853,
+      "step": 8251
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.611156763590392e-06,
+      "loss": 3.5851,
+      "step": 8252
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6111093552465235e-06,
+      "loss": 3.7946,
+      "step": 8253
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.611061946902655e-06,
+      "loss": 3.801,
+      "step": 8254
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6110145385587864e-06,
+      "loss": 4.2479,
+      "step": 8255
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6109671302149177e-06,
+      "loss": 4.3367,
+      "step": 8256
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.610919721871049e-06,
+      "loss": 3.425,
+      "step": 8257
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6108723135271807e-06,
+      "loss": 3.1921,
+      "step": 8258
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6108249051833124e-06,
+      "loss": 3.8855,
+      "step": 8259
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6107774968394437e-06,
+      "loss": 3.9422,
+      "step": 8260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6107300884955754e-06,
+      "loss": 3.2563,
+      "step": 8261
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6106826801517067e-06,
+      "loss": 3.6251,
+      "step": 8262
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.610635271807838e-06,
+      "loss": 3.8223,
+      "step": 8263
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.6105878634639697e-06,
+      "loss": 3.7069,
+      "step": 8264
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6105404551201014e-06,
+      "loss": 4.0339,
+      "step": 8265
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6104930467762326e-06,
+      "loss": 3.5128,
+      "step": 8266
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6104456384323644e-06,
+      "loss": 3.8162,
+      "step": 8267
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6103982300884956e-06,
+      "loss": 3.2063,
+      "step": 8268
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.610350821744627e-06,
+      "loss": 3.437,
+      "step": 8269
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6103034134007586e-06,
+      "loss": 3.9141,
+      "step": 8270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.61025600505689e-06,
+      "loss": 3.4178,
+      "step": 8271
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6102085967130216e-06,
+      "loss": 3.2162,
+      "step": 8272
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6101611883691533e-06,
+      "loss": 3.5145,
+      "step": 8273
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6101137800252846e-06,
+      "loss": 3.219,
+      "step": 8274
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.610066371681416e-06,
+      "loss": 3.6544,
+      "step": 8275
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6100189633375476e-06,
+      "loss": 3.8216,
+      "step": 8276
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609971554993679e-06,
+      "loss": 4.1601,
+      "step": 8277
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6099241466498106e-06,
+      "loss": 3.8238,
+      "step": 8278
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609876738305942e-06,
+      "loss": 3.4195,
+      "step": 8279
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6098293299620735e-06,
+      "loss": 3.7357,
+      "step": 8280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609781921618205e-06,
+      "loss": 3.2652,
+      "step": 8281
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609734513274336e-06,
+      "loss": 2.7668,
+      "step": 8282
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609687104930468e-06,
+      "loss": 3.2065,
+      "step": 8283
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609639696586599e-06,
+      "loss": 3.6262,
+      "step": 8284
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609592288242731e-06,
+      "loss": 3.6216,
+      "step": 8285
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6095448798988625e-06,
+      "loss": 4.1692,
+      "step": 8286
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6094974715549938e-06,
+      "loss": 4.0836,
+      "step": 8287
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609450063211125e-06,
+      "loss": 3.5735,
+      "step": 8288
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6094026548672568e-06,
+      "loss": 2.9808,
+      "step": 8289
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609355246523388e-06,
+      "loss": 3.8101,
+      "step": 8290
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6093078381795197e-06,
+      "loss": 4.1294,
+      "step": 8291
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6092604298356514e-06,
+      "loss": 3.3656,
+      "step": 8292
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6092130214917827e-06,
+      "loss": 3.3226,
+      "step": 8293
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609165613147914e-06,
+      "loss": 3.634,
+      "step": 8294
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6091182048040457e-06,
+      "loss": 3.3875,
+      "step": 8295
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.609070796460177e-06,
+      "loss": 3.9428,
+      "step": 8296
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6090233881163083e-06,
+      "loss": 3.4924,
+      "step": 8297
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.60897597977244e-06,
+      "loss": 4.3128,
+      "step": 8298
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6089285714285717e-06,
+      "loss": 3.9038,
+      "step": 8299
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608881163084703e-06,
+      "loss": 3.209,
+      "step": 8300
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6088337547408342e-06,
+      "loss": 2.8397,
+      "step": 8301
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608786346396966e-06,
+      "loss": 3.9152,
+      "step": 8302
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6087389380530972e-06,
+      "loss": 3.7559,
+      "step": 8303
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608691529709229e-06,
+      "loss": 3.6677,
+      "step": 8304
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6086441213653606e-06,
+      "loss": 3.6308,
+      "step": 8305
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608596713021492e-06,
+      "loss": 3.3906,
+      "step": 8306
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608549304677623e-06,
+      "loss": 3.9401,
+      "step": 8307
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608501896333755e-06,
+      "loss": 3.5908,
+      "step": 8308
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608454487989886e-06,
+      "loss": 3.4255,
+      "step": 8309
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6084070796460175e-06,
+      "loss": 3.4085,
+      "step": 8310
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608359671302149e-06,
+      "loss": 3.6183,
+      "step": 8311
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608312262958281e-06,
+      "loss": 3.683,
+      "step": 8312
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608264854614412e-06,
+      "loss": 3.6975,
+      "step": 8313
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608217446270544e-06,
+      "loss": 4.4216,
+      "step": 8314
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608170037926675e-06,
+      "loss": 4.1103,
+      "step": 8315
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6081226295828064e-06,
+      "loss": 4.0723,
+      "step": 8316
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.608075221238938e-06,
+      "loss": 2.7141,
+      "step": 8317
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.60802781289507e-06,
+      "loss": 3.8819,
+      "step": 8318
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.607980404551201e-06,
+      "loss": 4.2072,
+      "step": 8319
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.607932996207333e-06,
+      "loss": 3.3692,
+      "step": 8320
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.607885587863464e-06,
+      "loss": 3.0513,
+      "step": 8321
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6078381795195954e-06,
+      "loss": 3.6442,
+      "step": 8322
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6077907711757266e-06,
+      "loss": 3.4393,
+      "step": 8323
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6077433628318583e-06,
+      "loss": 3.6363,
+      "step": 8324
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.60769595448799e-06,
+      "loss": 3.5791,
+      "step": 8325
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.6076485461441213e-06,
+      "loss": 3.8843,
+      "step": 8326
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.607601137800253e-06,
+      "loss": 3.8189,
+      "step": 8327
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6075537294563843e-06,
+      "loss": 4.1075,
+      "step": 8328
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6075063211125156e-06,
+      "loss": 3.5163,
+      "step": 8329
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6074589127686473e-06,
+      "loss": 3.8114,
+      "step": 8330
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.607411504424779e-06,
+      "loss": 3.6667,
+      "step": 8331
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6073640960809103e-06,
+      "loss": 3.9193,
+      "step": 8332
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.607316687737042e-06,
+      "loss": 3.8909,
+      "step": 8333
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6072692793931733e-06,
+      "loss": 3.905,
+      "step": 8334
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6072218710493046e-06,
+      "loss": 3.7946,
+      "step": 8335
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6071744627054363e-06,
+      "loss": 3.8616,
+      "step": 8336
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6071270543615675e-06,
+      "loss": 3.3908,
+      "step": 8337
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6070796460176992e-06,
+      "loss": 3.9755,
+      "step": 8338
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.607032237673831e-06,
+      "loss": 3.8134,
+      "step": 8339
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6069848293299622e-06,
+      "loss": 4.1641,
+      "step": 8340
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6069374209860935e-06,
+      "loss": 3.0485,
+      "step": 8341
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.606890012642225e-06,
+      "loss": 3.6895,
+      "step": 8342
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6068426042983565e-06,
+      "loss": 3.9619,
+      "step": 8343
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.606795195954488e-06,
+      "loss": 4.0787,
+      "step": 8344
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6067477876106195e-06,
+      "loss": 3.7569,
+      "step": 8345
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.606700379266751e-06,
+      "loss": 3.1246,
+      "step": 8346
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6066529709228825e-06,
+      "loss": 4.021,
+      "step": 8347
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6066055625790137e-06,
+      "loss": 4.1026,
+      "step": 8348
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6065581542351454e-06,
+      "loss": 3.296,
+      "step": 8349
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6065107458912767e-06,
+      "loss": 3.3124,
+      "step": 8350
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6064633375474084e-06,
+      "loss": 3.7752,
+      "step": 8351
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.60641592920354e-06,
+      "loss": 3.908,
+      "step": 8352
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6063685208596714e-06,
+      "loss": 3.5274,
+      "step": 8353
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6063211125158027e-06,
+      "loss": 3.6578,
+      "step": 8354
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6062737041719344e-06,
+      "loss": 3.1075,
+      "step": 8355
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6062262958280657e-06,
+      "loss": 3.5206,
+      "step": 8356
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6061788874841974e-06,
+      "loss": 3.3029,
+      "step": 8357
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.606131479140329e-06,
+      "loss": 4.0715,
+      "step": 8358
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6060840707964604e-06,
+      "loss": 3.084,
+      "step": 8359
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6060366624525916e-06,
+      "loss": 3.1601,
+      "step": 8360
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6059892541087233e-06,
+      "loss": 4.0239,
+      "step": 8361
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6059418457648546e-06,
+      "loss": 3.7578,
+      "step": 8362
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605894437420986e-06,
+      "loss": 4.0882,
+      "step": 8363
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6058470290771176e-06,
+      "loss": 3.0889,
+      "step": 8364
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6057996207332493e-06,
+      "loss": 3.7143,
+      "step": 8365
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6057522123893806e-06,
+      "loss": 3.0774,
+      "step": 8366
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605704804045512e-06,
+      "loss": 3.6334,
+      "step": 8367
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6056573957016436e-06,
+      "loss": 3.9487,
+      "step": 8368
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605609987357775e-06,
+      "loss": 3.8109,
+      "step": 8369
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6055625790139066e-06,
+      "loss": 2.606,
+      "step": 8370
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6055151706700383e-06,
+      "loss": 4.2858,
+      "step": 8371
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6054677623261695e-06,
+      "loss": 3.2486,
+      "step": 8372
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605420353982301e-06,
+      "loss": 3.6579,
+      "step": 8373
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6053729456384325e-06,
+      "loss": 3.8902,
+      "step": 8374
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605325537294564e-06,
+      "loss": 4.168,
+      "step": 8375
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605278128950695e-06,
+      "loss": 2.8821,
+      "step": 8376
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.605230720606827e-06,
+      "loss": 3.4684,
+      "step": 8377
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6051833122629585e-06,
+      "loss": 3.4911,
+      "step": 8378
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6051359039190898e-06,
+      "loss": 3.7066,
+      "step": 8379
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6050884955752215e-06,
+      "loss": 3.5897,
+      "step": 8380
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6050410872313528e-06,
+      "loss": 4.1923,
+      "step": 8381
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.604993678887484e-06,
+      "loss": 3.5155,
+      "step": 8382
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6049462705436158e-06,
+      "loss": 3.4322,
+      "step": 8383
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6048988621997475e-06,
+      "loss": 3.5266,
+      "step": 8384
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6048514538558787e-06,
+      "loss": 3.6273,
+      "step": 8385
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6048040455120104e-06,
+      "loss": 3.8765,
+      "step": 8386
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6047566371681417e-06,
+      "loss": 3.7787,
+      "step": 8387
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.604709228824273e-06,
+      "loss": 3.0701,
+      "step": 8388
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6046618204804043e-06,
+      "loss": 3.8722,
+      "step": 8389
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.604614412136536e-06,
+      "loss": 4.1431,
+      "step": 8390
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.6045670037926677e-06,
+      "loss": 3.4964,
+      "step": 8391
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604519595448799e-06,
+      "loss": 3.2161,
+      "step": 8392
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6044721871049307e-06,
+      "loss": 2.9461,
+      "step": 8393
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604424778761062e-06,
+      "loss": 3.7568,
+      "step": 8394
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6043773704171932e-06,
+      "loss": 3.253,
+      "step": 8395
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604329962073325e-06,
+      "loss": 3.4156,
+      "step": 8396
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6042825537294566e-06,
+      "loss": 3.1221,
+      "step": 8397
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604235145385588e-06,
+      "loss": 2.8938,
+      "step": 8398
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6041877370417196e-06,
+      "loss": 3.3412,
+      "step": 8399
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604140328697851e-06,
+      "loss": 3.6173,
+      "step": 8400
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604092920353982e-06,
+      "loss": 4.1289,
+      "step": 8401
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.604045512010114e-06,
+      "loss": 3.5735,
+      "step": 8402
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603998103666245e-06,
+      "loss": 3.5315,
+      "step": 8403
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603950695322377e-06,
+      "loss": 3.7689,
+      "step": 8404
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6039032869785086e-06,
+      "loss": 3.5504,
+      "step": 8405
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.60385587863464e-06,
+      "loss": 3.7259,
+      "step": 8406
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603808470290771e-06,
+      "loss": 3.6738,
+      "step": 8407
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6037610619469024e-06,
+      "loss": 3.8433,
+      "step": 8408
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603713653603034e-06,
+      "loss": 3.5315,
+      "step": 8409
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603666245259166e-06,
+      "loss": 3.5026,
+      "step": 8410
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603618836915297e-06,
+      "loss": 3.0323,
+      "step": 8411
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603571428571429e-06,
+      "loss": 3.6014,
+      "step": 8412
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.60352402022756e-06,
+      "loss": 2.8856,
+      "step": 8413
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6034766118836914e-06,
+      "loss": 3.8125,
+      "step": 8414
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603429203539823e-06,
+      "loss": 3.97,
+      "step": 8415
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6033817951959544e-06,
+      "loss": 3.4546,
+      "step": 8416
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603334386852086e-06,
+      "loss": 4.0013,
+      "step": 8417
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6032869785082178e-06,
+      "loss": 4.1274,
+      "step": 8418
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603239570164349e-06,
+      "loss": 4.0766,
+      "step": 8419
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6031921618204803e-06,
+      "loss": 3.7342,
+      "step": 8420
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603144753476612e-06,
+      "loss": 3.5473,
+      "step": 8421
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6030973451327433e-06,
+      "loss": 3.6529,
+      "step": 8422
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.603049936788875e-06,
+      "loss": 3.0859,
+      "step": 8423
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6030025284450067e-06,
+      "loss": 3.6653,
+      "step": 8424
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602955120101138e-06,
+      "loss": 3.9545,
+      "step": 8425
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6029077117572693e-06,
+      "loss": 3.5208,
+      "step": 8426
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602860303413401e-06,
+      "loss": 3.0287,
+      "step": 8427
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6028128950695323e-06,
+      "loss": 3.8083,
+      "step": 8428
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6027654867256635e-06,
+      "loss": 3.3307,
+      "step": 8429
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6027180783817952e-06,
+      "loss": 3.5718,
+      "step": 8430
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602670670037927e-06,
+      "loss": 3.9301,
+      "step": 8431
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6026232616940582e-06,
+      "loss": 3.469,
+      "step": 8432
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6025758533501895e-06,
+      "loss": 3.3562,
+      "step": 8433
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6025284450063212e-06,
+      "loss": 3.5424,
+      "step": 8434
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6024810366624525e-06,
+      "loss": 3.9502,
+      "step": 8435
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602433628318584e-06,
+      "loss": 3.4172,
+      "step": 8436
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602386219974716e-06,
+      "loss": 3.7491,
+      "step": 8437
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602338811630847e-06,
+      "loss": 3.8214,
+      "step": 8438
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6022914032869785e-06,
+      "loss": 3.8932,
+      "step": 8439
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.60224399494311e-06,
+      "loss": 3.5008,
+      "step": 8440
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6021965865992415e-06,
+      "loss": 3.9296,
+      "step": 8441
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6021491782553727e-06,
+      "loss": 3.4672,
+      "step": 8442
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6021017699115044e-06,
+      "loss": 3.3306,
+      "step": 8443
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.602054361567636e-06,
+      "loss": 3.4288,
+      "step": 8444
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6020069532237674e-06,
+      "loss": 3.4829,
+      "step": 8445
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.601959544879899e-06,
+      "loss": 3.3054,
+      "step": 8446
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6019121365360304e-06,
+      "loss": 4.0385,
+      "step": 8447
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6018647281921617e-06,
+      "loss": 3.0643,
+      "step": 8448
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6018173198482934e-06,
+      "loss": 2.8635,
+      "step": 8449
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.601769911504425e-06,
+      "loss": 3.6931,
+      "step": 8450
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6017225031605564e-06,
+      "loss": 3.8363,
+      "step": 8451
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6016750948166877e-06,
+      "loss": 3.3925,
+      "step": 8452
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6016276864728194e-06,
+      "loss": 3.9545,
+      "step": 8453
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.6015802781289506e-06,
+      "loss": 4.0192,
+      "step": 8454
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.601532869785082e-06,
+      "loss": 2.9201,
+      "step": 8455
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6014854614412136e-06,
+      "loss": 3.9405,
+      "step": 8456
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6014380530973453e-06,
+      "loss": 3.4767,
+      "step": 8457
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6013906447534766e-06,
+      "loss": 2.9241,
+      "step": 8458
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6013432364096083e-06,
+      "loss": 3.9048,
+      "step": 8459
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6012958280657396e-06,
+      "loss": 3.7943,
+      "step": 8460
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.601248419721871e-06,
+      "loss": 3.8762,
+      "step": 8461
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6012010113780026e-06,
+      "loss": 4.444,
+      "step": 8462
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6011536030341343e-06,
+      "loss": 3.0035,
+      "step": 8463
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6011061946902656e-06,
+      "loss": 3.7354,
+      "step": 8464
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6010587863463973e-06,
+      "loss": 3.1669,
+      "step": 8465
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6010113780025285e-06,
+      "loss": 3.6144,
+      "step": 8466
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.60096396965866e-06,
+      "loss": 3.5854,
+      "step": 8467
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6009165613147915e-06,
+      "loss": 3.6994,
+      "step": 8468
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.600869152970923e-06,
+      "loss": 3.2186,
+      "step": 8469
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6008217446270545e-06,
+      "loss": 3.2575,
+      "step": 8470
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6007743362831862e-06,
+      "loss": 3.8588,
+      "step": 8471
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6007269279393175e-06,
+      "loss": 4.2607,
+      "step": 8472
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6006795195954488e-06,
+      "loss": 3.958,
+      "step": 8473
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.60063211125158e-06,
+      "loss": 3.6666,
+      "step": 8474
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6005847029077118e-06,
+      "loss": 3.8489,
+      "step": 8475
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6005372945638435e-06,
+      "loss": 3.9618,
+      "step": 8476
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6004898862199747e-06,
+      "loss": 4.0365,
+      "step": 8477
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6004424778761064e-06,
+      "loss": 3.4645,
+      "step": 8478
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6003950695322377e-06,
+      "loss": 3.0222,
+      "step": 8479
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.600347661188369e-06,
+      "loss": 3.6894,
+      "step": 8480
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6003002528445007e-06,
+      "loss": 4.2583,
+      "step": 8481
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.600252844500632e-06,
+      "loss": 3.7974,
+      "step": 8482
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6002054361567637e-06,
+      "loss": 4.2597,
+      "step": 8483
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6001580278128954e-06,
+      "loss": 3.7232,
+      "step": 8484
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6001106194690267e-06,
+      "loss": 3.6406,
+      "step": 8485
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.600063211125158e-06,
+      "loss": 3.5999,
+      "step": 8486
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.6000158027812897e-06,
+      "loss": 3.7245,
+      "step": 8487
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599968394437421e-06,
+      "loss": 3.8176,
+      "step": 8488
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5999209860935527e-06,
+      "loss": 4.1699,
+      "step": 8489
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5998735777496844e-06,
+      "loss": 4.0285,
+      "step": 8490
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5998261694058156e-06,
+      "loss": 3.4686,
+      "step": 8491
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599778761061947e-06,
+      "loss": 3.4828,
+      "step": 8492
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5997313527180786e-06,
+      "loss": 3.7603,
+      "step": 8493
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.59968394437421e-06,
+      "loss": 3.8914,
+      "step": 8494
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599636536030341e-06,
+      "loss": 3.6535,
+      "step": 8495
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599589127686473e-06,
+      "loss": 4.1777,
+      "step": 8496
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5995417193426046e-06,
+      "loss": 3.6478,
+      "step": 8497
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599494310998736e-06,
+      "loss": 3.3209,
+      "step": 8498
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599446902654867e-06,
+      "loss": 3.7012,
+      "step": 8499
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599399494310999e-06,
+      "loss": 4.0946,
+      "step": 8500
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.59935208596713e-06,
+      "loss": 3.0884,
+      "step": 8501
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599304677623262e-06,
+      "loss": 3.8504,
+      "step": 8502
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5992572692793935e-06,
+      "loss": 3.1875,
+      "step": 8503
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599209860935525e-06,
+      "loss": 3.3222,
+      "step": 8504
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599162452591656e-06,
+      "loss": 3.553,
+      "step": 8505
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599115044247788e-06,
+      "loss": 3.7499,
+      "step": 8506
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.599067635903919e-06,
+      "loss": 3.9304,
+      "step": 8507
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5990202275600504e-06,
+      "loss": 3.65,
+      "step": 8508
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.598972819216182e-06,
+      "loss": 3.3583,
+      "step": 8509
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5989254108723138e-06,
+      "loss": 3.6516,
+      "step": 8510
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.598878002528445e-06,
+      "loss": 3.2529,
+      "step": 8511
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5988305941845768e-06,
+      "loss": 3.3917,
+      "step": 8512
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.598783185840708e-06,
+      "loss": 3.1542,
+      "step": 8513
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5987357774968393e-06,
+      "loss": 3.1928,
+      "step": 8514
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.598688369152971e-06,
+      "loss": 3.4517,
+      "step": 8515
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.5986409608091027e-06,
+      "loss": 3.9619,
+      "step": 8516
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.598593552465234e-06,
+      "loss": 3.4363,
+      "step": 8517
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5985461441213653e-06,
+      "loss": 3.1496,
+      "step": 8518
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.598498735777497e-06,
+      "loss": 3.9678,
+      "step": 8519
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5984513274336283e-06,
+      "loss": 3.4883,
+      "step": 8520
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5984039190897596e-06,
+      "loss": 3.408,
+      "step": 8521
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5983565107458913e-06,
+      "loss": 3.6905,
+      "step": 8522
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.598309102402023e-06,
+      "loss": 3.974,
+      "step": 8523
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5982616940581542e-06,
+      "loss": 2.8835,
+      "step": 8524
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.598214285714286e-06,
+      "loss": 3.8438,
+      "step": 8525
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5981668773704172e-06,
+      "loss": 4.0627,
+      "step": 8526
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5981194690265485e-06,
+      "loss": 3.2742,
+      "step": 8527
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.59807206068268e-06,
+      "loss": 3.5035,
+      "step": 8528
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5980246523388115e-06,
+      "loss": 3.7944,
+      "step": 8529
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597977243994943e-06,
+      "loss": 4.0328,
+      "step": 8530
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597929835651075e-06,
+      "loss": 3.733,
+      "step": 8531
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597882427307206e-06,
+      "loss": 3.419,
+      "step": 8532
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5978350189633375e-06,
+      "loss": 3.4496,
+      "step": 8533
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597787610619469e-06,
+      "loss": 3.2947,
+      "step": 8534
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5977402022756004e-06,
+      "loss": 3.3669,
+      "step": 8535
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597692793931732e-06,
+      "loss": 3.6046,
+      "step": 8536
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597645385587864e-06,
+      "loss": 3.6585,
+      "step": 8537
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597597977243995e-06,
+      "loss": 3.7597,
+      "step": 8538
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5975505689001264e-06,
+      "loss": 3.4192,
+      "step": 8539
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5975031605562577e-06,
+      "loss": 3.2838,
+      "step": 8540
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5974557522123894e-06,
+      "loss": 4.1833,
+      "step": 8541
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5974083438685207e-06,
+      "loss": 4.0907,
+      "step": 8542
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5973609355246524e-06,
+      "loss": 3.8785,
+      "step": 8543
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597313527180784e-06,
+      "loss": 4.0276,
+      "step": 8544
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5972661188369154e-06,
+      "loss": 3.7909,
+      "step": 8545
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5972187104930466e-06,
+      "loss": 3.2546,
+      "step": 8546
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5971713021491784e-06,
+      "loss": 4.072,
+      "step": 8547
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5971238938053096e-06,
+      "loss": 4.102,
+      "step": 8548
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5970764854614413e-06,
+      "loss": 3.0687,
+      "step": 8549
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.597029077117573e-06,
+      "loss": 4.1509,
+      "step": 8550
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5969816687737043e-06,
+      "loss": 3.2821,
+      "step": 8551
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5969342604298356e-06,
+      "loss": 3.5433,
+      "step": 8552
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5968868520859673e-06,
+      "loss": 3.5826,
+      "step": 8553
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5968394437420986e-06,
+      "loss": 3.4637,
+      "step": 8554
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.59679203539823e-06,
+      "loss": 3.2989,
+      "step": 8555
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5967446270543616e-06,
+      "loss": 3.3035,
+      "step": 8556
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5966972187104933e-06,
+      "loss": 4.0321,
+      "step": 8557
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5966498103666246e-06,
+      "loss": 3.194,
+      "step": 8558
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5966024020227563e-06,
+      "loss": 3.08,
+      "step": 8559
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5965549936788875e-06,
+      "loss": 3.568,
+      "step": 8560
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.596507585335019e-06,
+      "loss": 3.0766,
+      "step": 8561
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5964601769911505e-06,
+      "loss": 3.4164,
+      "step": 8562
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5964127686472822e-06,
+      "loss": 3.9335,
+      "step": 8563
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5963653603034135e-06,
+      "loss": 3.2822,
+      "step": 8564
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5963179519595448e-06,
+      "loss": 3.6606,
+      "step": 8565
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5962705436156765e-06,
+      "loss": 3.3351,
+      "step": 8566
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5962231352718078e-06,
+      "loss": 2.8165,
+      "step": 8567
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.596175726927939e-06,
+      "loss": 3.937,
+      "step": 8568
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5961283185840708e-06,
+      "loss": 3.7164,
+      "step": 8569
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5960809102402025e-06,
+      "loss": 3.6172,
+      "step": 8570
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5960335018963337e-06,
+      "loss": 3.7144,
+      "step": 8571
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5959860935524654e-06,
+      "loss": 3.0441,
+      "step": 8572
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5959386852085967e-06,
+      "loss": 3.3143,
+      "step": 8573
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.595891276864728e-06,
+      "loss": 3.5068,
+      "step": 8574
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5958438685208597e-06,
+      "loss": 4.0923,
+      "step": 8575
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5957964601769914e-06,
+      "loss": 3.3686,
+      "step": 8576
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5957490518331227e-06,
+      "loss": 3.4809,
+      "step": 8577
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5957016434892544e-06,
+      "loss": 3.9111,
+      "step": 8578
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5956542351453857e-06,
+      "loss": 3.626,
+      "step": 8579
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.595606826801517e-06,
+      "loss": 3.1098,
+      "step": 8580
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.5955594184576487e-06,
+      "loss": 4.0984,
+      "step": 8581
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.59551201011378e-06,
+      "loss": 3.9271,
+      "step": 8582
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5954646017699116e-06,
+      "loss": 4.0739,
+      "step": 8583
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.595417193426043e-06,
+      "loss": 3.7659,
+      "step": 8584
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5953697850821746e-06,
+      "loss": 3.2834,
+      "step": 8585
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.595322376738306e-06,
+      "loss": 4.2544,
+      "step": 8586
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.595274968394437e-06,
+      "loss": 3.6035,
+      "step": 8587
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.595227560050569e-06,
+      "loss": 4.3115,
+      "step": 8588
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5951801517067006e-06,
+      "loss": 3.492,
+      "step": 8589
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.595132743362832e-06,
+      "loss": 3.3941,
+      "step": 8590
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5950853350189636e-06,
+      "loss": 4.1878,
+      "step": 8591
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.595037926675095e-06,
+      "loss": 3.6244,
+      "step": 8592
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594990518331226e-06,
+      "loss": 3.7042,
+      "step": 8593
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594943109987358e-06,
+      "loss": 3.6371,
+      "step": 8594
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594895701643489e-06,
+      "loss": 3.0762,
+      "step": 8595
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594848293299621e-06,
+      "loss": 3.7478,
+      "step": 8596
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5948008849557525e-06,
+      "loss": 3.1582,
+      "step": 8597
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594753476611884e-06,
+      "loss": 3.3849,
+      "step": 8598
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594706068268015e-06,
+      "loss": 3.3305,
+      "step": 8599
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594658659924147e-06,
+      "loss": 3.556,
+      "step": 8600
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594611251580278e-06,
+      "loss": 3.7195,
+      "step": 8601
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5945638432364098e-06,
+      "loss": 3.3002,
+      "step": 8602
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5945164348925415e-06,
+      "loss": 3.7996,
+      "step": 8603
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5944690265486728e-06,
+      "loss": 3.9338,
+      "step": 8604
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594421618204804e-06,
+      "loss": 3.7929,
+      "step": 8605
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5943742098609353e-06,
+      "loss": 4.0078,
+      "step": 8606
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594326801517067e-06,
+      "loss": 3.0706,
+      "step": 8607
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5942793931731983e-06,
+      "loss": 3.3678,
+      "step": 8608
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.59423198482933e-06,
+      "loss": 3.7366,
+      "step": 8609
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5941845764854617e-06,
+      "loss": 3.7797,
+      "step": 8610
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594137168141593e-06,
+      "loss": 3.2024,
+      "step": 8611
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5940897597977243e-06,
+      "loss": 2.7878,
+      "step": 8612
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.594042351453856e-06,
+      "loss": 3.2409,
+      "step": 8613
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5939949431099873e-06,
+      "loss": 4.2072,
+      "step": 8614
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593947534766119e-06,
+      "loss": 3.926,
+      "step": 8615
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5939001264222507e-06,
+      "loss": 3.9745,
+      "step": 8616
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593852718078382e-06,
+      "loss": 3.6417,
+      "step": 8617
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5938053097345132e-06,
+      "loss": 3.8826,
+      "step": 8618
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593757901390645e-06,
+      "loss": 3.7641,
+      "step": 8619
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5937104930467762e-06,
+      "loss": 3.6947,
+      "step": 8620
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5936630847029075e-06,
+      "loss": 3.7151,
+      "step": 8621
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593615676359039e-06,
+      "loss": 3.308,
+      "step": 8622
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593568268015171e-06,
+      "loss": 4.2046,
+      "step": 8623
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593520859671302e-06,
+      "loss": 3.3983,
+      "step": 8624
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593473451327434e-06,
+      "loss": 2.7657,
+      "step": 8625
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593426042983565e-06,
+      "loss": 3.6015,
+      "step": 8626
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5933786346396965e-06,
+      "loss": 4.106,
+      "step": 8627
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593331226295828e-06,
+      "loss": 4.0927,
+      "step": 8628
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.59328381795196e-06,
+      "loss": 3.2838,
+      "step": 8629
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593236409608091e-06,
+      "loss": 3.6193,
+      "step": 8630
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5931890012642224e-06,
+      "loss": 3.7509,
+      "step": 8631
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.593141592920354e-06,
+      "loss": 3.485,
+      "step": 8632
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5930941845764854e-06,
+      "loss": 3.7732,
+      "step": 8633
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5930467762326167e-06,
+      "loss": 3.8491,
+      "step": 8634
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5929993678887484e-06,
+      "loss": 4.0554,
+      "step": 8635
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.59295195954488e-06,
+      "loss": 3.343,
+      "step": 8636
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5929045512010114e-06,
+      "loss": 4.1818,
+      "step": 8637
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.592857142857143e-06,
+      "loss": 4.0226,
+      "step": 8638
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5928097345132744e-06,
+      "loss": 3.8982,
+      "step": 8639
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5927623261694056e-06,
+      "loss": 3.2733,
+      "step": 8640
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5927149178255373e-06,
+      "loss": 3.9675,
+      "step": 8641
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.592667509481669e-06,
+      "loss": 3.5548,
+      "step": 8642
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.5926201011378003e-06,
+      "loss": 3.472,
+      "step": 8643
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.592572692793932e-06,
+      "loss": 3.5254,
+      "step": 8644
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5925252844500633e-06,
+      "loss": 2.9622,
+      "step": 8645
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5924778761061946e-06,
+      "loss": 3.5809,
+      "step": 8646
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5924304677623263e-06,
+      "loss": 3.9443,
+      "step": 8647
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5923830594184576e-06,
+      "loss": 3.6917,
+      "step": 8648
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5923356510745893e-06,
+      "loss": 3.1693,
+      "step": 8649
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5922882427307206e-06,
+      "loss": 3.4121,
+      "step": 8650
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5922408343868523e-06,
+      "loss": 3.9471,
+      "step": 8651
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5921934260429835e-06,
+      "loss": 2.4348,
+      "step": 8652
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.592146017699115e-06,
+      "loss": 3.5289,
+      "step": 8653
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5920986093552465e-06,
+      "loss": 3.099,
+      "step": 8654
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5920512010113782e-06,
+      "loss": 3.462,
+      "step": 8655
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5920037926675095e-06,
+      "loss": 3.9086,
+      "step": 8656
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5919563843236412e-06,
+      "loss": 3.6471,
+      "step": 8657
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5919089759797725e-06,
+      "loss": 4.0297,
+      "step": 8658
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5918615676359038e-06,
+      "loss": 3.6959,
+      "step": 8659
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5918141592920355e-06,
+      "loss": 4.0442,
+      "step": 8660
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5917667509481668e-06,
+      "loss": 2.9792,
+      "step": 8661
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5917193426042985e-06,
+      "loss": 3.9445,
+      "step": 8662
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.59167193426043e-06,
+      "loss": 3.6058,
+      "step": 8663
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5916245259165615e-06,
+      "loss": 3.6091,
+      "step": 8664
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5915771175726927e-06,
+      "loss": 3.4814,
+      "step": 8665
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5915297092288244e-06,
+      "loss": 2.6984,
+      "step": 8666
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5914823008849557e-06,
+      "loss": 3.625,
+      "step": 8667
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5914348925410874e-06,
+      "loss": 3.3765,
+      "step": 8668
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.591387484197219e-06,
+      "loss": 3.5752,
+      "step": 8669
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5913400758533504e-06,
+      "loss": 4.4371,
+      "step": 8670
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5912926675094817e-06,
+      "loss": 3.1664,
+      "step": 8671
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.591245259165613e-06,
+      "loss": 3.3972,
+      "step": 8672
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5911978508217447e-06,
+      "loss": 4.0699,
+      "step": 8673
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.591150442477876e-06,
+      "loss": 2.548,
+      "step": 8674
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5911030341340077e-06,
+      "loss": 3.8078,
+      "step": 8675
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5910556257901394e-06,
+      "loss": 4.2351,
+      "step": 8676
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5910082174462706e-06,
+      "loss": 3.3007,
+      "step": 8677
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590960809102402e-06,
+      "loss": 4.1875,
+      "step": 8678
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5909134007585336e-06,
+      "loss": 3.8149,
+      "step": 8679
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590865992414665e-06,
+      "loss": 3.2974,
+      "step": 8680
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5908185840707966e-06,
+      "loss": 4.3984,
+      "step": 8681
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5907711757269283e-06,
+      "loss": 3.7106,
+      "step": 8682
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5907237673830596e-06,
+      "loss": 4.0689,
+      "step": 8683
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590676359039191e-06,
+      "loss": 3.2917,
+      "step": 8684
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5906289506953226e-06,
+      "loss": 3.4326,
+      "step": 8685
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590581542351454e-06,
+      "loss": 3.7954,
+      "step": 8686
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590534134007585e-06,
+      "loss": 3.7797,
+      "step": 8687
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590486725663717e-06,
+      "loss": 3.8086,
+      "step": 8688
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5904393173198485e-06,
+      "loss": 3.2232,
+      "step": 8689
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.59039190897598e-06,
+      "loss": 3.9794,
+      "step": 8690
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5903445006321115e-06,
+      "loss": 3.3917,
+      "step": 8691
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590297092288243e-06,
+      "loss": 4.1298,
+      "step": 8692
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590249683944374e-06,
+      "loss": 3.0733,
+      "step": 8693
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.590202275600506e-06,
+      "loss": 3.3067,
+      "step": 8694
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5901548672566375e-06,
+      "loss": 3.5896,
+      "step": 8695
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5901074589127688e-06,
+      "loss": 3.23,
+      "step": 8696
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5900600505689e-06,
+      "loss": 3.2119,
+      "step": 8697
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5900126422250318e-06,
+      "loss": 3.988,
+      "step": 8698
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.589965233881163e-06,
+      "loss": 4.1245,
+      "step": 8699
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5899178255372943e-06,
+      "loss": 3.9343,
+      "step": 8700
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.589870417193426e-06,
+      "loss": 2.9463,
+      "step": 8701
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5898230088495577e-06,
+      "loss": 2.6202,
+      "step": 8702
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.589775600505689e-06,
+      "loss": 3.8321,
+      "step": 8703
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5897281921618207e-06,
+      "loss": 3.6382,
+      "step": 8704
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.589680783817952e-06,
+      "loss": 3.8183,
+      "step": 8705
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.5896333754740833e-06,
+      "loss": 2.9068,
+      "step": 8706
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.589585967130215e-06,
+      "loss": 3.6039,
+      "step": 8707
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5895385587863467e-06,
+      "loss": 3.5475,
+      "step": 8708
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.589491150442478e-06,
+      "loss": 3.254,
+      "step": 8709
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5894437420986097e-06,
+      "loss": 3.4933,
+      "step": 8710
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.589396333754741e-06,
+      "loss": 3.9873,
+      "step": 8711
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5893489254108722e-06,
+      "loss": 3.216,
+      "step": 8712
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.589301517067004e-06,
+      "loss": 3.2806,
+      "step": 8713
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5892541087231352e-06,
+      "loss": 3.9472,
+      "step": 8714
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.589206700379267e-06,
+      "loss": 3.079,
+      "step": 8715
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.589159292035398e-06,
+      "loss": 3.6177,
+      "step": 8716
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.58911188369153e-06,
+      "loss": 3.9243,
+      "step": 8717
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.589064475347661e-06,
+      "loss": 2.9245,
+      "step": 8718
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5890170670037925e-06,
+      "loss": 2.7637,
+      "step": 8719
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588969658659924e-06,
+      "loss": 3.464,
+      "step": 8720
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588922250316056e-06,
+      "loss": 3.2557,
+      "step": 8721
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588874841972187e-06,
+      "loss": 3.1594,
+      "step": 8722
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588827433628319e-06,
+      "loss": 4.0711,
+      "step": 8723
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.58878002528445e-06,
+      "loss": 3.0769,
+      "step": 8724
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5887326169405814e-06,
+      "loss": 3.606,
+      "step": 8725
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588685208596713e-06,
+      "loss": 4.2904,
+      "step": 8726
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5886378002528444e-06,
+      "loss": 3.8894,
+      "step": 8727
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588590391908976e-06,
+      "loss": 3.8777,
+      "step": 8728
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588542983565108e-06,
+      "loss": 4.0643,
+      "step": 8729
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588495575221239e-06,
+      "loss": 3.4594,
+      "step": 8730
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5884481668773704e-06,
+      "loss": 3.6332,
+      "step": 8731
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588400758533502e-06,
+      "loss": 3.749,
+      "step": 8732
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5883533501896334e-06,
+      "loss": 3.7899,
+      "step": 8733
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588305941845765e-06,
+      "loss": 2.8412,
+      "step": 8734
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5882585335018968e-06,
+      "loss": 4.3445,
+      "step": 8735
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.588211125158028e-06,
+      "loss": 3.9055,
+      "step": 8736
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5881637168141593e-06,
+      "loss": 3.361,
+      "step": 8737
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5881163084702906e-06,
+      "loss": 3.831,
+      "step": 8738
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5880689001264223e-06,
+      "loss": 4.1818,
+      "step": 8739
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5880214917825536e-06,
+      "loss": 3.6194,
+      "step": 8740
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5879740834386853e-06,
+      "loss": 3.6663,
+      "step": 8741
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.587926675094817e-06,
+      "loss": 4.1247,
+      "step": 8742
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5878792667509483e-06,
+      "loss": 3.4482,
+      "step": 8743
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5878318584070796e-06,
+      "loss": 3.4504,
+      "step": 8744
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5877844500632113e-06,
+      "loss": 3.7672,
+      "step": 8745
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5877370417193425e-06,
+      "loss": 3.1988,
+      "step": 8746
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5876896333754742e-06,
+      "loss": 3.5631,
+      "step": 8747
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.587642225031606e-06,
+      "loss": 3.586,
+      "step": 8748
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5875948166877372e-06,
+      "loss": 3.6428,
+      "step": 8749
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5875474083438685e-06,
+      "loss": 3.6637,
+      "step": 8750
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5875000000000002e-06,
+      "loss": 4.0736,
+      "step": 8751
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5874525916561315e-06,
+      "loss": 2.786,
+      "step": 8752
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5874051833122628e-06,
+      "loss": 3.6489,
+      "step": 8753
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5873577749683945e-06,
+      "loss": 3.7677,
+      "step": 8754
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.587310366624526e-06,
+      "loss": 4.1024,
+      "step": 8755
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5872629582806575e-06,
+      "loss": 3.3674,
+      "step": 8756
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.587215549936789e-06,
+      "loss": 4.0515,
+      "step": 8757
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5871681415929204e-06,
+      "loss": 3.8113,
+      "step": 8758
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5871207332490517e-06,
+      "loss": 3.4418,
+      "step": 8759
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5870733249051834e-06,
+      "loss": 3.4952,
+      "step": 8760
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.587025916561315e-06,
+      "loss": 4.1143,
+      "step": 8761
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5869785082174464e-06,
+      "loss": 3.1896,
+      "step": 8762
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5869310998735777e-06,
+      "loss": 3.4377,
+      "step": 8763
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5868836915297094e-06,
+      "loss": 3.4142,
+      "step": 8764
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5868362831858407e-06,
+      "loss": 3.3846,
+      "step": 8765
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.586788874841972e-06,
+      "loss": 2.9822,
+      "step": 8766
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5867414664981037e-06,
+      "loss": 3.7918,
+      "step": 8767
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5866940581542354e-06,
+      "loss": 3.4007,
+      "step": 8768
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5866466498103666e-06,
+      "loss": 3.4444,
+      "step": 8769
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5865992414664984e-06,
+      "loss": 3.4946,
+      "step": 8770
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.5865518331226296e-06,
+      "loss": 3.5317,
+      "step": 8771
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.586504424778761e-06,
+      "loss": 3.8587,
+      "step": 8772
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5864570164348926e-06,
+      "loss": 3.7726,
+      "step": 8773
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5864096080910243e-06,
+      "loss": 3.7993,
+      "step": 8774
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5863621997471556e-06,
+      "loss": 4.0713,
+      "step": 8775
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5863147914032873e-06,
+      "loss": 3.293,
+      "step": 8776
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5862673830594186e-06,
+      "loss": 3.8749,
+      "step": 8777
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.58621997471555e-06,
+      "loss": 3.4462,
+      "step": 8778
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5861725663716816e-06,
+      "loss": 3.5647,
+      "step": 8779
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.586125158027813e-06,
+      "loss": 3.4395,
+      "step": 8780
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5860777496839446e-06,
+      "loss": 3.2623,
+      "step": 8781
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.586030341340076e-06,
+      "loss": 3.4137,
+      "step": 8782
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5859829329962075e-06,
+      "loss": 3.2509,
+      "step": 8783
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585935524652339e-06,
+      "loss": 4.0827,
+      "step": 8784
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.58588811630847e-06,
+      "loss": 3.2645,
+      "step": 8785
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585840707964602e-06,
+      "loss": 2.903,
+      "step": 8786
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5857932996207335e-06,
+      "loss": 2.9079,
+      "step": 8787
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585745891276865e-06,
+      "loss": 4.247,
+      "step": 8788
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5856984829329965e-06,
+      "loss": 3.2329,
+      "step": 8789
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5856510745891278e-06,
+      "loss": 3.341,
+      "step": 8790
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585603666245259e-06,
+      "loss": 2.8525,
+      "step": 8791
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5855562579013908e-06,
+      "loss": 3.9732,
+      "step": 8792
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585508849557522e-06,
+      "loss": 3.3857,
+      "step": 8793
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5854614412136537e-06,
+      "loss": 3.8305,
+      "step": 8794
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5854140328697854e-06,
+      "loss": 3.9103,
+      "step": 8795
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5853666245259167e-06,
+      "loss": 3.2636,
+      "step": 8796
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585319216182048e-06,
+      "loss": 3.8071,
+      "step": 8797
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5852718078381797e-06,
+      "loss": 3.8648,
+      "step": 8798
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585224399494311e-06,
+      "loss": 3.9878,
+      "step": 8799
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5851769911504423e-06,
+      "loss": 3.4954,
+      "step": 8800
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5851295828065744e-06,
+      "loss": 3.8929,
+      "step": 8801
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5850821744627057e-06,
+      "loss": 4.2613,
+      "step": 8802
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.585034766118837e-06,
+      "loss": 4.1546,
+      "step": 8803
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5849873577749682e-06,
+      "loss": 3.2658,
+      "step": 8804
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5849399494311e-06,
+      "loss": 3.6344,
+      "step": 8805
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5848925410872312e-06,
+      "loss": 3.5895,
+      "step": 8806
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584845132743363e-06,
+      "loss": 4.1229,
+      "step": 8807
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5847977243994946e-06,
+      "loss": 3.8341,
+      "step": 8808
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584750316055626e-06,
+      "loss": 3.3078,
+      "step": 8809
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584702907711757e-06,
+      "loss": 3.7543,
+      "step": 8810
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584655499367889e-06,
+      "loss": 3.7708,
+      "step": 8811
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.58460809102402e-06,
+      "loss": 3.5682,
+      "step": 8812
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5845606826801515e-06,
+      "loss": 3.5896,
+      "step": 8813
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584513274336283e-06,
+      "loss": 3.3932,
+      "step": 8814
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584465865992415e-06,
+      "loss": 2.9454,
+      "step": 8815
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584418457648546e-06,
+      "loss": 3.3974,
+      "step": 8816
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584371049304678e-06,
+      "loss": 4.0412,
+      "step": 8817
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584323640960809e-06,
+      "loss": 3.226,
+      "step": 8818
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5842762326169404e-06,
+      "loss": 3.8643,
+      "step": 8819
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584228824273072e-06,
+      "loss": 3.4151,
+      "step": 8820
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584181415929204e-06,
+      "loss": 3.8767,
+      "step": 8821
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584134007585335e-06,
+      "loss": 3.7573,
+      "step": 8822
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584086599241467e-06,
+      "loss": 3.6262,
+      "step": 8823
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.584039190897598e-06,
+      "loss": 4.0796,
+      "step": 8824
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5839917825537294e-06,
+      "loss": 3.9777,
+      "step": 8825
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5839443742098606e-06,
+      "loss": 4.1977,
+      "step": 8826
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5838969658659923e-06,
+      "loss": 3.9769,
+      "step": 8827
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.583849557522124e-06,
+      "loss": 3.8342,
+      "step": 8828
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5838021491782553e-06,
+      "loss": 3.8417,
+      "step": 8829
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.583754740834387e-06,
+      "loss": 3.9085,
+      "step": 8830
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5837073324905183e-06,
+      "loss": 3.8742,
+      "step": 8831
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5836599241466496e-06,
+      "loss": 3.6046,
+      "step": 8832
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.5836125158027813e-06,
+      "loss": 3.7096,
+      "step": 8833
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.583565107458913e-06,
+      "loss": 3.8337,
+      "step": 8834
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5835176991150443e-06,
+      "loss": 3.4292,
+      "step": 8835
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.583470290771176e-06,
+      "loss": 3.3433,
+      "step": 8836
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5834228824273073e-06,
+      "loss": 3.8599,
+      "step": 8837
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5833754740834386e-06,
+      "loss": 3.283,
+      "step": 8838
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5833280657395703e-06,
+      "loss": 3.2371,
+      "step": 8839
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5832806573957015e-06,
+      "loss": 3.3803,
+      "step": 8840
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5832332490518332e-06,
+      "loss": 3.5903,
+      "step": 8841
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.583185840707965e-06,
+      "loss": 3.6937,
+      "step": 8842
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5831384323640962e-06,
+      "loss": 3.3009,
+      "step": 8843
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5830910240202275e-06,
+      "loss": 2.9561,
+      "step": 8844
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5830436156763588e-06,
+      "loss": 3.0168,
+      "step": 8845
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5829962073324905e-06,
+      "loss": 3.2342,
+      "step": 8846
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.582948798988622e-06,
+      "loss": 4.3349,
+      "step": 8847
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5829013906447535e-06,
+      "loss": 3.6345,
+      "step": 8848
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.582853982300885e-06,
+      "loss": 2.698,
+      "step": 8849
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5828065739570165e-06,
+      "loss": 3.8851,
+      "step": 8850
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5827591656131477e-06,
+      "loss": 3.1264,
+      "step": 8851
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5827117572692794e-06,
+      "loss": 3.4781,
+      "step": 8852
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5826643489254107e-06,
+      "loss": 3.7987,
+      "step": 8853
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5826169405815424e-06,
+      "loss": 3.9093,
+      "step": 8854
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.582569532237674e-06,
+      "loss": 4.0328,
+      "step": 8855
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5825221238938054e-06,
+      "loss": 3.6061,
+      "step": 8856
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5824747155499367e-06,
+      "loss": 3.6799,
+      "step": 8857
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5824273072060684e-06,
+      "loss": 3.3207,
+      "step": 8858
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5823798988621997e-06,
+      "loss": 4.1651,
+      "step": 8859
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5823324905183314e-06,
+      "loss": 3.1313,
+      "step": 8860
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.582285082174463e-06,
+      "loss": 2.5161,
+      "step": 8861
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5822376738305944e-06,
+      "loss": 3.3901,
+      "step": 8862
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5821902654867256e-06,
+      "loss": 3.7619,
+      "step": 8863
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5821428571428573e-06,
+      "loss": 3.0554,
+      "step": 8864
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5820954487989886e-06,
+      "loss": 3.5075,
+      "step": 8865
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.58204804045512e-06,
+      "loss": 4.0052,
+      "step": 8866
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5820006321112516e-06,
+      "loss": 3.8054,
+      "step": 8867
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5819532237673833e-06,
+      "loss": 3.2482,
+      "step": 8868
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5819058154235146e-06,
+      "loss": 3.1176,
+      "step": 8869
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581858407079646e-06,
+      "loss": 3.67,
+      "step": 8870
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5818109987357776e-06,
+      "loss": 3.6689,
+      "step": 8871
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581763590391909e-06,
+      "loss": 3.9621,
+      "step": 8872
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5817161820480406e-06,
+      "loss": 3.494,
+      "step": 8873
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5816687737041723e-06,
+      "loss": 3.9368,
+      "step": 8874
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5816213653603035e-06,
+      "loss": 3.2944,
+      "step": 8875
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581573957016435e-06,
+      "loss": 3.5909,
+      "step": 8876
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5815265486725665e-06,
+      "loss": 2.9482,
+      "step": 8877
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581479140328698e-06,
+      "loss": 3.4945,
+      "step": 8878
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581431731984829e-06,
+      "loss": 3.1605,
+      "step": 8879
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581384323640961e-06,
+      "loss": 4.0208,
+      "step": 8880
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5813369152970925e-06,
+      "loss": 3.8112,
+      "step": 8881
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5812895069532238e-06,
+      "loss": 3.8985,
+      "step": 8882
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5812420986093555e-06,
+      "loss": 3.7922,
+      "step": 8883
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5811946902654868e-06,
+      "loss": 3.7652,
+      "step": 8884
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.581147281921618e-06,
+      "loss": 3.0736,
+      "step": 8885
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5810998735777498e-06,
+      "loss": 3.9431,
+      "step": 8886
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5810524652338815e-06,
+      "loss": 4.2316,
+      "step": 8887
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5810050568900127e-06,
+      "loss": 4.2077,
+      "step": 8888
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.580957648546144e-06,
+      "loss": 4.1897,
+      "step": 8889
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5809102402022757e-06,
+      "loss": 4.5041,
+      "step": 8890
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.580862831858407e-06,
+      "loss": 3.1543,
+      "step": 8891
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5808154235145383e-06,
+      "loss": 3.9075,
+      "step": 8892
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.58076801517067e-06,
+      "loss": 3.322,
+      "step": 8893
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5807206068268017e-06,
+      "loss": 4.1562,
+      "step": 8894
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.580673198482933e-06,
+      "loss": 4.0817,
+      "step": 8895
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.5806257901390647e-06,
+      "loss": 3.4854,
+      "step": 8896
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.580578381795196e-06,
+      "loss": 3.695,
+      "step": 8897
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5805309734513272e-06,
+      "loss": 3.899,
+      "step": 8898
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580483565107459e-06,
+      "loss": 3.6491,
+      "step": 8899
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5804361567635906e-06,
+      "loss": 3.9959,
+      "step": 8900
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580388748419722e-06,
+      "loss": 3.4805,
+      "step": 8901
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5803413400758536e-06,
+      "loss": 3.5171,
+      "step": 8902
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580293931731985e-06,
+      "loss": 3.3946,
+      "step": 8903
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580246523388116e-06,
+      "loss": 3.1544,
+      "step": 8904
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580199115044248e-06,
+      "loss": 3.4621,
+      "step": 8905
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580151706700379e-06,
+      "loss": 3.6655,
+      "step": 8906
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580104298356511e-06,
+      "loss": 3.5495,
+      "step": 8907
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5800568900126426e-06,
+      "loss": 3.6493,
+      "step": 8908
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.580009481668774e-06,
+      "loss": 3.3261,
+      "step": 8909
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579962073324905e-06,
+      "loss": 3.428,
+      "step": 8910
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5799146649810364e-06,
+      "loss": 3.7819,
+      "step": 8911
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579867256637168e-06,
+      "loss": 3.8651,
+      "step": 8912
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5798198482933e-06,
+      "loss": 3.9431,
+      "step": 8913
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579772439949431e-06,
+      "loss": 3.5906,
+      "step": 8914
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579725031605563e-06,
+      "loss": 3.1996,
+      "step": 8915
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579677623261694e-06,
+      "loss": 3.497,
+      "step": 8916
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5796302149178254e-06,
+      "loss": 4.0041,
+      "step": 8917
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579582806573957e-06,
+      "loss": 3.008,
+      "step": 8918
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5795353982300884e-06,
+      "loss": 3.8218,
+      "step": 8919
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.57948798988622e-06,
+      "loss": 4.1166,
+      "step": 8920
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5794405815423518e-06,
+      "loss": 2.4253,
+      "step": 8921
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579393173198483e-06,
+      "loss": 3.8963,
+      "step": 8922
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5793457648546143e-06,
+      "loss": 3.2179,
+      "step": 8923
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579298356510746e-06,
+      "loss": 3.8519,
+      "step": 8924
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5792509481668773e-06,
+      "loss": 3.2482,
+      "step": 8925
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579203539823009e-06,
+      "loss": 3.2451,
+      "step": 8926
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5791561314791407e-06,
+      "loss": 3.792,
+      "step": 8927
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579108723135272e-06,
+      "loss": 3.1646,
+      "step": 8928
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5790613147914033e-06,
+      "loss": 2.5004,
+      "step": 8929
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.579013906447535e-06,
+      "loss": 3.6147,
+      "step": 8930
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5789664981036663e-06,
+      "loss": 3.0502,
+      "step": 8931
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5789190897597975e-06,
+      "loss": 3.5201,
+      "step": 8932
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5788716814159292e-06,
+      "loss": 3.368,
+      "step": 8933
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.578824273072061e-06,
+      "loss": 4.1565,
+      "step": 8934
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5787768647281922e-06,
+      "loss": 2.852,
+      "step": 8935
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5787294563843235e-06,
+      "loss": 2.5981,
+      "step": 8936
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5786820480404552e-06,
+      "loss": 3.8776,
+      "step": 8937
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5786346396965865e-06,
+      "loss": 3.8636,
+      "step": 8938
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.578587231352718e-06,
+      "loss": 3.5076,
+      "step": 8939
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.57853982300885e-06,
+      "loss": 3.2171,
+      "step": 8940
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.578492414664981e-06,
+      "loss": 4.5183,
+      "step": 8941
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5784450063211125e-06,
+      "loss": 3.1752,
+      "step": 8942
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.578397597977244e-06,
+      "loss": 3.7911,
+      "step": 8943
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5783501896333755e-06,
+      "loss": 2.7046,
+      "step": 8944
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5783027812895067e-06,
+      "loss": 3.696,
+      "step": 8945
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5782553729456384e-06,
+      "loss": 3.6061,
+      "step": 8946
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.57820796460177e-06,
+      "loss": 3.41,
+      "step": 8947
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5781605562579014e-06,
+      "loss": 3.8626,
+      "step": 8948
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.578113147914033e-06,
+      "loss": 2.7912,
+      "step": 8949
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5780657395701644e-06,
+      "loss": 3.6716,
+      "step": 8950
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5780183312262957e-06,
+      "loss": 3.2164,
+      "step": 8951
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5779709228824274e-06,
+      "loss": 3.7064,
+      "step": 8952
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.577923514538559e-06,
+      "loss": 3.9892,
+      "step": 8953
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5778761061946904e-06,
+      "loss": 3.4262,
+      "step": 8954
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5778286978508217e-06,
+      "loss": 3.3536,
+      "step": 8955
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5777812895069534e-06,
+      "loss": 3.3742,
+      "step": 8956
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5777338811630846e-06,
+      "loss": 3.6291,
+      "step": 8957
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.577686472819216e-06,
+      "loss": 2.9046,
+      "step": 8958
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5776390644753476e-06,
+      "loss": 2.5646,
+      "step": 8959
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5775916561314793e-06,
+      "loss": 2.9568,
+      "step": 8960
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.5775442477876106e-06,
+      "loss": 3.8578,
+      "step": 8961
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5774968394437423e-06,
+      "loss": 3.5123,
+      "step": 8962
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5774494310998736e-06,
+      "loss": 3.1749,
+      "step": 8963
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.577402022756005e-06,
+      "loss": 2.9636,
+      "step": 8964
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5773546144121366e-06,
+      "loss": 3.2741,
+      "step": 8965
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5773072060682683e-06,
+      "loss": 2.8071,
+      "step": 8966
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5772597977243996e-06,
+      "loss": 2.7947,
+      "step": 8967
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5772123893805313e-06,
+      "loss": 4.0146,
+      "step": 8968
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5771649810366625e-06,
+      "loss": 3.6936,
+      "step": 8969
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.577117572692794e-06,
+      "loss": 3.3881,
+      "step": 8970
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5770701643489255e-06,
+      "loss": 3.1803,
+      "step": 8971
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.577022756005057e-06,
+      "loss": 3.454,
+      "step": 8972
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5769753476611885e-06,
+      "loss": 3.8229,
+      "step": 8973
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5769279393173202e-06,
+      "loss": 3.0074,
+      "step": 8974
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5768805309734515e-06,
+      "loss": 3.487,
+      "step": 8975
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5768331226295828e-06,
+      "loss": 4.0834,
+      "step": 8976
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.576785714285714e-06,
+      "loss": 2.5656,
+      "step": 8977
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5767383059418458e-06,
+      "loss": 3.1434,
+      "step": 8978
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5766908975979775e-06,
+      "loss": 3.2945,
+      "step": 8979
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5766434892541087e-06,
+      "loss": 3.8381,
+      "step": 8980
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5765960809102404e-06,
+      "loss": 3.1854,
+      "step": 8981
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5765486725663717e-06,
+      "loss": 3.2071,
+      "step": 8982
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.576501264222503e-06,
+      "loss": 3.4597,
+      "step": 8983
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5764538558786347e-06,
+      "loss": 4.0392,
+      "step": 8984
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.576406447534766e-06,
+      "loss": 4.4654,
+      "step": 8985
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5763590391908977e-06,
+      "loss": 4.0976,
+      "step": 8986
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5763116308470294e-06,
+      "loss": 3.6347,
+      "step": 8987
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5762642225031607e-06,
+      "loss": 3.2913,
+      "step": 8988
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.576216814159292e-06,
+      "loss": 2.6729,
+      "step": 8989
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5761694058154237e-06,
+      "loss": 4.1619,
+      "step": 8990
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.576121997471555e-06,
+      "loss": 3.0526,
+      "step": 8991
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5760745891276867e-06,
+      "loss": 3.1184,
+      "step": 8992
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5760271807838184e-06,
+      "loss": 3.6506,
+      "step": 8993
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5759797724399496e-06,
+      "loss": 3.1945,
+      "step": 8994
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575932364096081e-06,
+      "loss": 3.4664,
+      "step": 8995
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5758849557522126e-06,
+      "loss": 3.9905,
+      "step": 8996
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575837547408344e-06,
+      "loss": 3.8302,
+      "step": 8997
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575790139064475e-06,
+      "loss": 3.8236,
+      "step": 8998
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575742730720607e-06,
+      "loss": 3.7223,
+      "step": 8999
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5756953223767386e-06,
+      "loss": 3.4912,
+      "step": 9000
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.57564791403287e-06,
+      "loss": 3.943,
+      "step": 9001
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575600505689001e-06,
+      "loss": 3.3687,
+      "step": 9002
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575553097345133e-06,
+      "loss": 3.9158,
+      "step": 9003
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575505689001264e-06,
+      "loss": 3.8405,
+      "step": 9004
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575458280657396e-06,
+      "loss": 3.1556,
+      "step": 9005
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5754108723135275e-06,
+      "loss": 3.6141,
+      "step": 9006
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575363463969659e-06,
+      "loss": 4.1224,
+      "step": 9007
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.57531605562579e-06,
+      "loss": 3.5708,
+      "step": 9008
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575268647281922e-06,
+      "loss": 4.3763,
+      "step": 9009
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575221238938053e-06,
+      "loss": 3.4072,
+      "step": 9010
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5751738305941844e-06,
+      "loss": 2.8474,
+      "step": 9011
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575126422250316e-06,
+      "loss": 3.5466,
+      "step": 9012
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5750790139064478e-06,
+      "loss": 3.9539,
+      "step": 9013
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.575031605562579e-06,
+      "loss": 3.4838,
+      "step": 9014
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5749841972187108e-06,
+      "loss": 3.1463,
+      "step": 9015
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.574936788874842e-06,
+      "loss": 3.5729,
+      "step": 9016
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5748893805309733e-06,
+      "loss": 4.0391,
+      "step": 9017
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.574841972187105e-06,
+      "loss": 3.9131,
+      "step": 9018
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5747945638432367e-06,
+      "loss": 2.9841,
+      "step": 9019
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.574747155499368e-06,
+      "loss": 3.2645,
+      "step": 9020
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5746997471554993e-06,
+      "loss": 3.1706,
+      "step": 9021
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.574652338811631e-06,
+      "loss": 3.7839,
+      "step": 9022
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5746049304677623e-06,
+      "loss": 4.0814,
+      "step": 9023
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.5745575221238936e-06,
+      "loss": 3.2508,
+      "step": 9024
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5745101137800253e-06,
+      "loss": 3.0357,
+      "step": 9025
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.574462705436157e-06,
+      "loss": 2.7883,
+      "step": 9026
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5744152970922882e-06,
+      "loss": 4.032,
+      "step": 9027
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.57436788874842e-06,
+      "loss": 3.1105,
+      "step": 9028
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5743204804045512e-06,
+      "loss": 3.2936,
+      "step": 9029
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5742730720606825e-06,
+      "loss": 3.349,
+      "step": 9030
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.574225663716814e-06,
+      "loss": 3.7723,
+      "step": 9031
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.574178255372946e-06,
+      "loss": 3.4356,
+      "step": 9032
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.574130847029077e-06,
+      "loss": 3.4446,
+      "step": 9033
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.574083438685209e-06,
+      "loss": 4.1966,
+      "step": 9034
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.57403603034134e-06,
+      "loss": 3.2266,
+      "step": 9035
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5739886219974715e-06,
+      "loss": 3.8162,
+      "step": 9036
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573941213653603e-06,
+      "loss": 4.2902,
+      "step": 9037
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5738938053097344e-06,
+      "loss": 3.1253,
+      "step": 9038
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573846396965866e-06,
+      "loss": 3.3499,
+      "step": 9039
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573798988621998e-06,
+      "loss": 2.9986,
+      "step": 9040
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573751580278129e-06,
+      "loss": 3.5361,
+      "step": 9041
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5737041719342604e-06,
+      "loss": 3.3065,
+      "step": 9042
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5736567635903917e-06,
+      "loss": 3.6631,
+      "step": 9043
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5736093552465234e-06,
+      "loss": 3.7208,
+      "step": 9044
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573561946902655e-06,
+      "loss": 4.1903,
+      "step": 9045
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5735145385587864e-06,
+      "loss": 4.038,
+      "step": 9046
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573467130214918e-06,
+      "loss": 3.5771,
+      "step": 9047
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5734197218710494e-06,
+      "loss": 3.7238,
+      "step": 9048
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5733723135271806e-06,
+      "loss": 3.8499,
+      "step": 9049
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5733249051833124e-06,
+      "loss": 3.6628,
+      "step": 9050
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5732774968394436e-06,
+      "loss": 2.6212,
+      "step": 9051
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5732300884955753e-06,
+      "loss": 3.3427,
+      "step": 9052
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.573182680151707e-06,
+      "loss": 3.5637,
+      "step": 9053
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5731352718078383e-06,
+      "loss": 3.1526,
+      "step": 9054
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5730878634639696e-06,
+      "loss": 4.0013,
+      "step": 9055
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5730404551201013e-06,
+      "loss": 2.9672,
+      "step": 9056
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5729930467762326e-06,
+      "loss": 4.0776,
+      "step": 9057
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5729456384323643e-06,
+      "loss": 3.7587,
+      "step": 9058
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.572898230088496e-06,
+      "loss": 3.5853,
+      "step": 9059
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5728508217446273e-06,
+      "loss": 3.7174,
+      "step": 9060
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5728034134007586e-06,
+      "loss": 2.9788,
+      "step": 9061
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5727560050568903e-06,
+      "loss": 3.5739,
+      "step": 9062
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5727085967130215e-06,
+      "loss": 2.7887,
+      "step": 9063
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.572661188369153e-06,
+      "loss": 3.5217,
+      "step": 9064
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5726137800252845e-06,
+      "loss": 4.0256,
+      "step": 9065
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5725663716814162e-06,
+      "loss": 3.4581,
+      "step": 9066
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5725189633375475e-06,
+      "loss": 3.9592,
+      "step": 9067
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5724715549936788e-06,
+      "loss": 4.2955,
+      "step": 9068
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5724241466498105e-06,
+      "loss": 3.8001,
+      "step": 9069
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5723767383059418e-06,
+      "loss": 3.7096,
+      "step": 9070
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.572329329962073e-06,
+      "loss": 3.502,
+      "step": 9071
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.572281921618205e-06,
+      "loss": 3.2842,
+      "step": 9072
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5722345132743365e-06,
+      "loss": 4.1187,
+      "step": 9073
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5721871049304677e-06,
+      "loss": 3.5814,
+      "step": 9074
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5721396965865994e-06,
+      "loss": 3.2714,
+      "step": 9075
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5720922882427307e-06,
+      "loss": 3.3923,
+      "step": 9076
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.572044879898862e-06,
+      "loss": 3.9341,
+      "step": 9077
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5719974715549937e-06,
+      "loss": 3.3859,
+      "step": 9078
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5719500632111254e-06,
+      "loss": 3.6178,
+      "step": 9079
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5719026548672567e-06,
+      "loss": 3.1761,
+      "step": 9080
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5718552465233884e-06,
+      "loss": 2.9102,
+      "step": 9081
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5718078381795197e-06,
+      "loss": 4.0657,
+      "step": 9082
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.571760429835651e-06,
+      "loss": 3.7956,
+      "step": 9083
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5717130214917827e-06,
+      "loss": 4.0327,
+      "step": 9084
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.571665613147914e-06,
+      "loss": 3.6167,
+      "step": 9085
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5716182048040456e-06,
+      "loss": 3.7738,
+      "step": 9086
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.571570796460177e-06,
+      "loss": 3.5243,
+      "step": 9087
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5715233881163086e-06,
+      "loss": 4.1154,
+      "step": 9088
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.57147597977244e-06,
+      "loss": 3.6812,
+      "step": 9089
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571428571428571e-06,
+      "loss": 3.7007,
+      "step": 9090
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571381163084703e-06,
+      "loss": 3.6189,
+      "step": 9091
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5713337547408346e-06,
+      "loss": 3.9319,
+      "step": 9092
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571286346396966e-06,
+      "loss": 3.7114,
+      "step": 9093
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5712389380530976e-06,
+      "loss": 3.1523,
+      "step": 9094
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571191529709229e-06,
+      "loss": 3.4983,
+      "step": 9095
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.57114412136536e-06,
+      "loss": 3.6267,
+      "step": 9096
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571096713021492e-06,
+      "loss": 3.7762,
+      "step": 9097
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571049304677623e-06,
+      "loss": 3.6492,
+      "step": 9098
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.571001896333755e-06,
+      "loss": 2.6027,
+      "step": 9099
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5709544879898865e-06,
+      "loss": 3.3028,
+      "step": 9100
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570907079646018e-06,
+      "loss": 3.4777,
+      "step": 9101
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570859671302149e-06,
+      "loss": 2.8162,
+      "step": 9102
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570812262958281e-06,
+      "loss": 3.7527,
+      "step": 9103
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570764854614412e-06,
+      "loss": 2.8695,
+      "step": 9104
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5707174462705438e-06,
+      "loss": 3.8048,
+      "step": 9105
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5706700379266755e-06,
+      "loss": 3.1364,
+      "step": 9106
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5706226295828068e-06,
+      "loss": 3.1747,
+      "step": 9107
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570575221238938e-06,
+      "loss": 2.8471,
+      "step": 9108
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5705278128950693e-06,
+      "loss": 3.4473,
+      "step": 9109
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570480404551201e-06,
+      "loss": 3.8528,
+      "step": 9110
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5704329962073323e-06,
+      "loss": 3.4473,
+      "step": 9111
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570385587863464e-06,
+      "loss": 3.8132,
+      "step": 9112
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5703381795195957e-06,
+      "loss": 3.3869,
+      "step": 9113
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570290771175727e-06,
+      "loss": 3.091,
+      "step": 9114
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5702433628318583e-06,
+      "loss": 3.475,
+      "step": 9115
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.57019595448799e-06,
+      "loss": 3.3472,
+      "step": 9116
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5701485461441213e-06,
+      "loss": 3.335,
+      "step": 9117
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570101137800253e-06,
+      "loss": 3.9255,
+      "step": 9118
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5700537294563847e-06,
+      "loss": 2.9787,
+      "step": 9119
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.570006321112516e-06,
+      "loss": 3.5415,
+      "step": 9120
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5699589127686472e-06,
+      "loss": 3.7274,
+      "step": 9121
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569911504424779e-06,
+      "loss": 3.5334,
+      "step": 9122
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5698640960809102e-06,
+      "loss": 2.2867,
+      "step": 9123
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5698166877370415e-06,
+      "loss": 3.4861,
+      "step": 9124
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569769279393173e-06,
+      "loss": 3.5036,
+      "step": 9125
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569721871049305e-06,
+      "loss": 4.2547,
+      "step": 9126
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569674462705436e-06,
+      "loss": 3.22,
+      "step": 9127
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569627054361568e-06,
+      "loss": 3.4785,
+      "step": 9128
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569579646017699e-06,
+      "loss": 3.3517,
+      "step": 9129
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5695322376738305e-06,
+      "loss": 3.4757,
+      "step": 9130
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569484829329962e-06,
+      "loss": 3.5891,
+      "step": 9131
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569437420986094e-06,
+      "loss": 3.6702,
+      "step": 9132
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569390012642225e-06,
+      "loss": 4.0544,
+      "step": 9133
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5693426042983564e-06,
+      "loss": 3.8327,
+      "step": 9134
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569295195954488e-06,
+      "loss": 3.6079,
+      "step": 9135
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5692477876106194e-06,
+      "loss": 3.5887,
+      "step": 9136
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5692003792667507e-06,
+      "loss": 2.433,
+      "step": 9137
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5691529709228824e-06,
+      "loss": 2.6597,
+      "step": 9138
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569105562579014e-06,
+      "loss": 2.6936,
+      "step": 9139
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5690581542351454e-06,
+      "loss": 3.6217,
+      "step": 9140
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.569010745891277e-06,
+      "loss": 3.6521,
+      "step": 9141
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5689633375474084e-06,
+      "loss": 3.3739,
+      "step": 9142
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5689159292035396e-06,
+      "loss": 3.2007,
+      "step": 9143
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5688685208596713e-06,
+      "loss": 3.7644,
+      "step": 9144
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.568821112515803e-06,
+      "loss": 3.4016,
+      "step": 9145
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5687737041719343e-06,
+      "loss": 3.3519,
+      "step": 9146
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.568726295828066e-06,
+      "loss": 3.5821,
+      "step": 9147
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5686788874841973e-06,
+      "loss": 4.1084,
+      "step": 9148
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5686314791403286e-06,
+      "loss": 3.2937,
+      "step": 9149
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5685840707964603e-06,
+      "loss": 3.6663,
+      "step": 9150
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.5685366624525916e-06,
+      "loss": 3.0814,
+      "step": 9151
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5684892541087233e-06,
+      "loss": 3.4323,
+      "step": 9152
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5684418457648546e-06,
+      "loss": 3.4136,
+      "step": 9153
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5683944374209863e-06,
+      "loss": 3.91,
+      "step": 9154
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5683470290771175e-06,
+      "loss": 3.4728,
+      "step": 9155
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.568299620733249e-06,
+      "loss": 2.1595,
+      "step": 9156
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5682522123893805e-06,
+      "loss": 3.9802,
+      "step": 9157
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5682048040455122e-06,
+      "loss": 3.6919,
+      "step": 9158
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5681573957016435e-06,
+      "loss": 3.114,
+      "step": 9159
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5681099873577752e-06,
+      "loss": 2.8954,
+      "step": 9160
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5680625790139065e-06,
+      "loss": 3.8667,
+      "step": 9161
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5680151706700378e-06,
+      "loss": 2.8507,
+      "step": 9162
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5679677623261695e-06,
+      "loss": 2.7768,
+      "step": 9163
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5679203539823008e-06,
+      "loss": 3.5838,
+      "step": 9164
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5678729456384325e-06,
+      "loss": 3.5884,
+      "step": 9165
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.567825537294564e-06,
+      "loss": 3.9439,
+      "step": 9166
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5677781289506955e-06,
+      "loss": 3.8163,
+      "step": 9167
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5677307206068267e-06,
+      "loss": 2.5909,
+      "step": 9168
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5676833122629584e-06,
+      "loss": 3.5004,
+      "step": 9169
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5676359039190897e-06,
+      "loss": 4.0045,
+      "step": 9170
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5675884955752214e-06,
+      "loss": 3.8128,
+      "step": 9171
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.567541087231353e-06,
+      "loss": 3.0863,
+      "step": 9172
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5674936788874844e-06,
+      "loss": 3.3591,
+      "step": 9173
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5674462705436157e-06,
+      "loss": 3.2145,
+      "step": 9174
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.567398862199747e-06,
+      "loss": 3.1799,
+      "step": 9175
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5673514538558787e-06,
+      "loss": 2.9771,
+      "step": 9176
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.56730404551201e-06,
+      "loss": 4.2461,
+      "step": 9177
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5672566371681417e-06,
+      "loss": 3.413,
+      "step": 9178
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5672092288242734e-06,
+      "loss": 3.1782,
+      "step": 9179
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5671618204804046e-06,
+      "loss": 3.7537,
+      "step": 9180
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.567114412136536e-06,
+      "loss": 3.6694,
+      "step": 9181
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5670670037926676e-06,
+      "loss": 4.0081,
+      "step": 9182
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.567019595448799e-06,
+      "loss": 4.0442,
+      "step": 9183
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5669721871049306e-06,
+      "loss": 2.5673,
+      "step": 9184
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5669247787610623e-06,
+      "loss": 3.3405,
+      "step": 9185
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5668773704171936e-06,
+      "loss": 3.1578,
+      "step": 9186
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566829962073325e-06,
+      "loss": 3.4914,
+      "step": 9187
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5667825537294566e-06,
+      "loss": 3.3978,
+      "step": 9188
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566735145385588e-06,
+      "loss": 3.7407,
+      "step": 9189
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566687737041719e-06,
+      "loss": 3.5454,
+      "step": 9190
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566640328697851e-06,
+      "loss": 3.7412,
+      "step": 9191
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5665929203539825e-06,
+      "loss": 3.7276,
+      "step": 9192
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566545512010114e-06,
+      "loss": 3.648,
+      "step": 9193
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5664981036662455e-06,
+      "loss": 3.3954,
+      "step": 9194
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566450695322377e-06,
+      "loss": 4.0645,
+      "step": 9195
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566403286978508e-06,
+      "loss": 3.0985,
+      "step": 9196
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.56635587863464e-06,
+      "loss": 3.7536,
+      "step": 9197
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5663084702907715e-06,
+      "loss": 3.4215,
+      "step": 9198
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5662610619469028e-06,
+      "loss": 3.4112,
+      "step": 9199
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566213653603034e-06,
+      "loss": 2.9317,
+      "step": 9200
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5661662452591658e-06,
+      "loss": 3.826,
+      "step": 9201
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.566118836915297e-06,
+      "loss": 3.3235,
+      "step": 9202
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5660714285714283e-06,
+      "loss": 3.7562,
+      "step": 9203
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.56602402022756e-06,
+      "loss": 4.1756,
+      "step": 9204
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5659766118836917e-06,
+      "loss": 2.8413,
+      "step": 9205
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.565929203539823e-06,
+      "loss": 3.8432,
+      "step": 9206
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5658817951959547e-06,
+      "loss": 3.2271,
+      "step": 9207
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.565834386852086e-06,
+      "loss": 4.1188,
+      "step": 9208
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5657869785082173e-06,
+      "loss": 3.5122,
+      "step": 9209
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.565739570164349e-06,
+      "loss": 3.3765,
+      "step": 9210
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5656921618204807e-06,
+      "loss": 3.0625,
+      "step": 9211
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.565644753476612e-06,
+      "loss": 3.2893,
+      "step": 9212
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.5655973451327437e-06,
+      "loss": 3.7714,
+      "step": 9213
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.565549936788875e-06,
+      "loss": 3.6652,
+      "step": 9214
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5655025284450062e-06,
+      "loss": 3.4038,
+      "step": 9215
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565455120101138e-06,
+      "loss": 2.3737,
+      "step": 9216
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5654077117572692e-06,
+      "loss": 3.4811,
+      "step": 9217
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565360303413401e-06,
+      "loss": 2.9146,
+      "step": 9218
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565312895069532e-06,
+      "loss": 3.3153,
+      "step": 9219
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565265486725664e-06,
+      "loss": 3.2528,
+      "step": 9220
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565218078381795e-06,
+      "loss": 3.9036,
+      "step": 9221
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5651706700379265e-06,
+      "loss": 4.1158,
+      "step": 9222
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565123261694058e-06,
+      "loss": 3.1864,
+      "step": 9223
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.56507585335019e-06,
+      "loss": 3.509,
+      "step": 9224
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.565028445006321e-06,
+      "loss": 4.3132,
+      "step": 9225
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564981036662453e-06,
+      "loss": 3.9581,
+      "step": 9226
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564933628318584e-06,
+      "loss": 3.1248,
+      "step": 9227
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5648862199747154e-06,
+      "loss": 2.784,
+      "step": 9228
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564838811630847e-06,
+      "loss": 3.9936,
+      "step": 9229
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5647914032869784e-06,
+      "loss": 4.0949,
+      "step": 9230
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.56474399494311e-06,
+      "loss": 4.1829,
+      "step": 9231
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564696586599242e-06,
+      "loss": 4.156,
+      "step": 9232
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564649178255373e-06,
+      "loss": 3.3649,
+      "step": 9233
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5646017699115044e-06,
+      "loss": 3.6892,
+      "step": 9234
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564554361567636e-06,
+      "loss": 3.6252,
+      "step": 9235
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5645069532237674e-06,
+      "loss": 3.1419,
+      "step": 9236
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564459544879899e-06,
+      "loss": 3.5576,
+      "step": 9237
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5644121365360308e-06,
+      "loss": 3.4302,
+      "step": 9238
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564364728192162e-06,
+      "loss": 3.6042,
+      "step": 9239
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5643173198482933e-06,
+      "loss": 4.185,
+      "step": 9240
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5642699115044246e-06,
+      "loss": 3.9963,
+      "step": 9241
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5642225031605563e-06,
+      "loss": 2.9172,
+      "step": 9242
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5641750948166876e-06,
+      "loss": 4.046,
+      "step": 9243
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5641276864728193e-06,
+      "loss": 3.3649,
+      "step": 9244
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.564080278128951e-06,
+      "loss": 3.0823,
+      "step": 9245
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5640328697850823e-06,
+      "loss": 3.2366,
+      "step": 9246
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5639854614412136e-06,
+      "loss": 3.8232,
+      "step": 9247
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5639380530973453e-06,
+      "loss": 3.8996,
+      "step": 9248
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5638906447534765e-06,
+      "loss": 3.6423,
+      "step": 9249
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5638432364096082e-06,
+      "loss": 4.3185,
+      "step": 9250
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.56379582806574e-06,
+      "loss": 3.661,
+      "step": 9251
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5637484197218712e-06,
+      "loss": 3.6561,
+      "step": 9252
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5637010113780025e-06,
+      "loss": 3.0495,
+      "step": 9253
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5636536030341342e-06,
+      "loss": 3.2878,
+      "step": 9254
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5636061946902655e-06,
+      "loss": 3.4789,
+      "step": 9255
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5635587863463968e-06,
+      "loss": 3.376,
+      "step": 9256
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5635113780025285e-06,
+      "loss": 4.0045,
+      "step": 9257
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.56346396965866e-06,
+      "loss": 3.2721,
+      "step": 9258
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5634165613147915e-06,
+      "loss": 3.9438,
+      "step": 9259
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.563369152970923e-06,
+      "loss": 3.0544,
+      "step": 9260
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5633217446270544e-06,
+      "loss": 3.404,
+      "step": 9261
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5632743362831857e-06,
+      "loss": 3.7377,
+      "step": 9262
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5632269279393174e-06,
+      "loss": 3.1686,
+      "step": 9263
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.563179519595449e-06,
+      "loss": 4.198,
+      "step": 9264
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5631321112515804e-06,
+      "loss": 3.7434,
+      "step": 9265
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5630847029077117e-06,
+      "loss": 4.2462,
+      "step": 9266
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5630372945638434e-06,
+      "loss": 3.7796,
+      "step": 9267
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5629898862199747e-06,
+      "loss": 2.778,
+      "step": 9268
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.562942477876106e-06,
+      "loss": 3.5024,
+      "step": 9269
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5628950695322377e-06,
+      "loss": 2.6719,
+      "step": 9270
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5628476611883694e-06,
+      "loss": 3.0674,
+      "step": 9271
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5628002528445006e-06,
+      "loss": 3.2211,
+      "step": 9272
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5627528445006324e-06,
+      "loss": 3.6007,
+      "step": 9273
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5627054361567636e-06,
+      "loss": 3.3598,
+      "step": 9274
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.562658027812895e-06,
+      "loss": 3.5689,
+      "step": 9275
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5626106194690266e-06,
+      "loss": 3.8431,
+      "step": 9276
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.5625632111251583e-06,
+      "loss": 3.5834,
+      "step": 9277
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5625158027812896e-06,
+      "loss": 3.052,
+      "step": 9278
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5624683944374213e-06,
+      "loss": 4.203,
+      "step": 9279
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5624209860935526e-06,
+      "loss": 3.45,
+      "step": 9280
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.562373577749684e-06,
+      "loss": 3.6011,
+      "step": 9281
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.562326169405815e-06,
+      "loss": 3.6373,
+      "step": 9282
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.562278761061947e-06,
+      "loss": 3.6284,
+      "step": 9283
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5622313527180786e-06,
+      "loss": 2.9302,
+      "step": 9284
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.56218394437421e-06,
+      "loss": 4.0284,
+      "step": 9285
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5621365360303415e-06,
+      "loss": 3.4266,
+      "step": 9286
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.562089127686473e-06,
+      "loss": 3.6363,
+      "step": 9287
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.562041719342604e-06,
+      "loss": 3.4159,
+      "step": 9288
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561994310998736e-06,
+      "loss": 3.0754,
+      "step": 9289
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5619469026548675e-06,
+      "loss": 3.7716,
+      "step": 9290
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561899494310999e-06,
+      "loss": 4.1121,
+      "step": 9291
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5618520859671305e-06,
+      "loss": 3.4253,
+      "step": 9292
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5618046776232618e-06,
+      "loss": 3.7036,
+      "step": 9293
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561757269279393e-06,
+      "loss": 3.9599,
+      "step": 9294
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5617098609355248e-06,
+      "loss": 3.3085,
+      "step": 9295
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561662452591656e-06,
+      "loss": 3.8383,
+      "step": 9296
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5616150442477877e-06,
+      "loss": 3.5264,
+      "step": 9297
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5615676359039194e-06,
+      "loss": 3.5213,
+      "step": 9298
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5615202275600507e-06,
+      "loss": 3.8712,
+      "step": 9299
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561472819216182e-06,
+      "loss": 2.9094,
+      "step": 9300
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5614254108723137e-06,
+      "loss": 3.1063,
+      "step": 9301
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561378002528445e-06,
+      "loss": 3.3841,
+      "step": 9302
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5613305941845767e-06,
+      "loss": 3.0167,
+      "step": 9303
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561283185840708e-06,
+      "loss": 3.6756,
+      "step": 9304
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5612357774968397e-06,
+      "loss": 2.7867,
+      "step": 9305
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561188369152971e-06,
+      "loss": 3.7037,
+      "step": 9306
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5611409608091022e-06,
+      "loss": 3.7136,
+      "step": 9307
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.561093552465234e-06,
+      "loss": 3.0798,
+      "step": 9308
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5610461441213652e-06,
+      "loss": 3.8052,
+      "step": 9309
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560998735777497e-06,
+      "loss": 3.179,
+      "step": 9310
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5609513274336286e-06,
+      "loss": 3.9978,
+      "step": 9311
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.56090391908976e-06,
+      "loss": 3.3983,
+      "step": 9312
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560856510745891e-06,
+      "loss": 3.2769,
+      "step": 9313
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560809102402023e-06,
+      "loss": 2.9079,
+      "step": 9314
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560761694058154e-06,
+      "loss": 3.2314,
+      "step": 9315
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560714285714286e-06,
+      "loss": 3.5887,
+      "step": 9316
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5606668773704176e-06,
+      "loss": 3.4736,
+      "step": 9317
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560619469026549e-06,
+      "loss": 2.8513,
+      "step": 9318
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.56057206068268e-06,
+      "loss": 3.7903,
+      "step": 9319
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560524652338812e-06,
+      "loss": 3.8097,
+      "step": 9320
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560477243994943e-06,
+      "loss": 3.1366,
+      "step": 9321
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5604298356510744e-06,
+      "loss": 3.1987,
+      "step": 9322
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560382427307206e-06,
+      "loss": 3.0934,
+      "step": 9323
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560335018963338e-06,
+      "loss": 3.8287,
+      "step": 9324
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560287610619469e-06,
+      "loss": 2.9096,
+      "step": 9325
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5602402022756004e-06,
+      "loss": 3.448,
+      "step": 9326
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560192793931732e-06,
+      "loss": 3.3589,
+      "step": 9327
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5601453855878634e-06,
+      "loss": 3.1159,
+      "step": 9328
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560097977243995e-06,
+      "loss": 4.3066,
+      "step": 9329
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5600505689001268e-06,
+      "loss": 3.0057,
+      "step": 9330
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.560003160556258e-06,
+      "loss": 3.3676,
+      "step": 9331
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5599557522123893e-06,
+      "loss": 3.1961,
+      "step": 9332
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.559908343868521e-06,
+      "loss": 3.2475,
+      "step": 9333
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5598609355246523e-06,
+      "loss": 2.947,
+      "step": 9334
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5598135271807836e-06,
+      "loss": 3.5104,
+      "step": 9335
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5597661188369153e-06,
+      "loss": 3.7717,
+      "step": 9336
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.559718710493047e-06,
+      "loss": 3.1826,
+      "step": 9337
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5596713021491783e-06,
+      "loss": 3.7436,
+      "step": 9338
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.55962389380531e-06,
+      "loss": 3.4109,
+      "step": 9339
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5595764854614413e-06,
+      "loss": 2.5063,
+      "step": 9340
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.5595290771175726e-06,
+      "loss": 3.406,
+      "step": 9341
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5594816687737043e-06,
+      "loss": 3.5455,
+      "step": 9342
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.559434260429836e-06,
+      "loss": 3.4473,
+      "step": 9343
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5593868520859672e-06,
+      "loss": 3.5091,
+      "step": 9344
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.559339443742099e-06,
+      "loss": 3.5151,
+      "step": 9345
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5592920353982302e-06,
+      "loss": 3.4444,
+      "step": 9346
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5592446270543615e-06,
+      "loss": 3.1035,
+      "step": 9347
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5591972187104928e-06,
+      "loss": 2.9624,
+      "step": 9348
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5591498103666245e-06,
+      "loss": 3.3229,
+      "step": 9349
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.559102402022756e-06,
+      "loss": 4.081,
+      "step": 9350
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5590549936788875e-06,
+      "loss": 3.0419,
+      "step": 9351
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.559007585335019e-06,
+      "loss": 3.5997,
+      "step": 9352
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5589601769911505e-06,
+      "loss": 3.711,
+      "step": 9353
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5589127686472817e-06,
+      "loss": 3.3057,
+      "step": 9354
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5588653603034134e-06,
+      "loss": 3.9799,
+      "step": 9355
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5588179519595447e-06,
+      "loss": 3.3953,
+      "step": 9356
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5587705436156764e-06,
+      "loss": 3.2322,
+      "step": 9357
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.558723135271808e-06,
+      "loss": 3.8298,
+      "step": 9358
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5586757269279394e-06,
+      "loss": 3.4788,
+      "step": 9359
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5586283185840707e-06,
+      "loss": 4.1197,
+      "step": 9360
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5585809102402024e-06,
+      "loss": 3.2439,
+      "step": 9361
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5585335018963337e-06,
+      "loss": 3.6515,
+      "step": 9362
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5584860935524654e-06,
+      "loss": 3.4675,
+      "step": 9363
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.558438685208597e-06,
+      "loss": 3.1309,
+      "step": 9364
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5583912768647284e-06,
+      "loss": 3.4247,
+      "step": 9365
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5583438685208596e-06,
+      "loss": 3.6629,
+      "step": 9366
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5582964601769913e-06,
+      "loss": 2.9645,
+      "step": 9367
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5582490518331226e-06,
+      "loss": 3.9256,
+      "step": 9368
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.558201643489254e-06,
+      "loss": 3.7742,
+      "step": 9369
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5581542351453856e-06,
+      "loss": 3.953,
+      "step": 9370
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5581068268015173e-06,
+      "loss": 3.3706,
+      "step": 9371
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5580594184576486e-06,
+      "loss": 3.3868,
+      "step": 9372
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.55801201011378e-06,
+      "loss": 3.4805,
+      "step": 9373
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5579646017699116e-06,
+      "loss": 3.0773,
+      "step": 9374
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557917193426043e-06,
+      "loss": 2.8058,
+      "step": 9375
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5578697850821746e-06,
+      "loss": 3.3054,
+      "step": 9376
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5578223767383063e-06,
+      "loss": 3.2797,
+      "step": 9377
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5577749683944375e-06,
+      "loss": 3.3707,
+      "step": 9378
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557727560050569e-06,
+      "loss": 3.7405,
+      "step": 9379
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5576801517067005e-06,
+      "loss": 3.579,
+      "step": 9380
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557632743362832e-06,
+      "loss": 2.8674,
+      "step": 9381
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557585335018963e-06,
+      "loss": 4.3855,
+      "step": 9382
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557537926675095e-06,
+      "loss": 3.8502,
+      "step": 9383
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5574905183312265e-06,
+      "loss": 3.8151,
+      "step": 9384
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5574431099873578e-06,
+      "loss": 2.9542,
+      "step": 9385
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5573957016434895e-06,
+      "loss": 3.6842,
+      "step": 9386
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5573482932996208e-06,
+      "loss": 3.4953,
+      "step": 9387
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557300884955752e-06,
+      "loss": 2.9242,
+      "step": 9388
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5572534766118838e-06,
+      "loss": 3.677,
+      "step": 9389
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5572060682680155e-06,
+      "loss": 4.1994,
+      "step": 9390
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5571586599241467e-06,
+      "loss": 3.941,
+      "step": 9391
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557111251580278e-06,
+      "loss": 3.765,
+      "step": 9392
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5570638432364097e-06,
+      "loss": 3.4284,
+      "step": 9393
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.557016434892541e-06,
+      "loss": 2.7737,
+      "step": 9394
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5569690265486723e-06,
+      "loss": 2.8875,
+      "step": 9395
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.556921618204804e-06,
+      "loss": 3.8322,
+      "step": 9396
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5568742098609357e-06,
+      "loss": 3.5594,
+      "step": 9397
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.556826801517067e-06,
+      "loss": 4.2078,
+      "step": 9398
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5567793931731987e-06,
+      "loss": 3.7288,
+      "step": 9399
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.55673198482933e-06,
+      "loss": 3.5104,
+      "step": 9400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5566845764854612e-06,
+      "loss": 3.0294,
+      "step": 9401
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.556637168141593e-06,
+      "loss": 3.7282,
+      "step": 9402
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5565897597977246e-06,
+      "loss": 3.4145,
+      "step": 9403
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.556542351453856e-06,
+      "loss": 3.7257,
+      "step": 9404
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5564949431099876e-06,
+      "loss": 4.0526,
+      "step": 9405
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556447534766119e-06,
+      "loss": 3.6766,
+      "step": 9406
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.55640012642225e-06,
+      "loss": 3.2695,
+      "step": 9407
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556352718078382e-06,
+      "loss": 3.2723,
+      "step": 9408
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556305309734513e-06,
+      "loss": 3.7538,
+      "step": 9409
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556257901390645e-06,
+      "loss": 2.22,
+      "step": 9410
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5562104930467766e-06,
+      "loss": 4.0412,
+      "step": 9411
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556163084702908e-06,
+      "loss": 3.4527,
+      "step": 9412
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556115676359039e-06,
+      "loss": 3.3041,
+      "step": 9413
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5560682680151704e-06,
+      "loss": 3.39,
+      "step": 9414
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.556020859671302e-06,
+      "loss": 3.6726,
+      "step": 9415
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555973451327434e-06,
+      "loss": 3.9995,
+      "step": 9416
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555926042983565e-06,
+      "loss": 2.8981,
+      "step": 9417
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555878634639697e-06,
+      "loss": 4.4767,
+      "step": 9418
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555831226295828e-06,
+      "loss": 3.8599,
+      "step": 9419
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5557838179519594e-06,
+      "loss": 3.2614,
+      "step": 9420
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555736409608091e-06,
+      "loss": 4.1929,
+      "step": 9421
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5556890012642224e-06,
+      "loss": 3.4837,
+      "step": 9422
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555641592920354e-06,
+      "loss": 3.6686,
+      "step": 9423
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5555941845764858e-06,
+      "loss": 3.6794,
+      "step": 9424
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555546776232617e-06,
+      "loss": 2.5605,
+      "step": 9425
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5554993678887483e-06,
+      "loss": 3.5988,
+      "step": 9426
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.55545195954488e-06,
+      "loss": 2.8583,
+      "step": 9427
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5554045512010113e-06,
+      "loss": 3.7942,
+      "step": 9428
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555357142857143e-06,
+      "loss": 3.7166,
+      "step": 9429
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5553097345132747e-06,
+      "loss": 4.1449,
+      "step": 9430
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555262326169406e-06,
+      "loss": 3.3908,
+      "step": 9431
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5552149178255373e-06,
+      "loss": 2.6775,
+      "step": 9432
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.555167509481669e-06,
+      "loss": 3.3536,
+      "step": 9433
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5551201011378003e-06,
+      "loss": 3.947,
+      "step": 9434
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5550726927939315e-06,
+      "loss": 3.8523,
+      "step": 9435
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5550252844500632e-06,
+      "loss": 4.1147,
+      "step": 9436
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554977876106195e-06,
+      "loss": 3.135,
+      "step": 9437
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5549304677623262e-06,
+      "loss": 2.852,
+      "step": 9438
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5548830594184575e-06,
+      "loss": 3.3117,
+      "step": 9439
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5548356510745892e-06,
+      "loss": 3.4001,
+      "step": 9440
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5547882427307205e-06,
+      "loss": 3.3207,
+      "step": 9441
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554740834386852e-06,
+      "loss": 3.3145,
+      "step": 9442
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554693426042984e-06,
+      "loss": 3.039,
+      "step": 9443
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554646017699115e-06,
+      "loss": 3.8019,
+      "step": 9444
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5545986093552465e-06,
+      "loss": 3.3528,
+      "step": 9445
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554551201011378e-06,
+      "loss": 3.4248,
+      "step": 9446
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5545037926675095e-06,
+      "loss": 3.4415,
+      "step": 9447
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5544563843236407e-06,
+      "loss": 2.9351,
+      "step": 9448
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5544089759797724e-06,
+      "loss": 3.8767,
+      "step": 9449
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554361567635904e-06,
+      "loss": 3.7724,
+      "step": 9450
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5543141592920354e-06,
+      "loss": 4.0645,
+      "step": 9451
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554266750948167e-06,
+      "loss": 3.5052,
+      "step": 9452
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5542193426042984e-06,
+      "loss": 4.1855,
+      "step": 9453
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5541719342604297e-06,
+      "loss": 4.0561,
+      "step": 9454
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5541245259165614e-06,
+      "loss": 3.218,
+      "step": 9455
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554077117572693e-06,
+      "loss": 4.077,
+      "step": 9456
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5540297092288244e-06,
+      "loss": 3.7401,
+      "step": 9457
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5539823008849557e-06,
+      "loss": 4.0973,
+      "step": 9458
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5539348925410874e-06,
+      "loss": 3.8907,
+      "step": 9459
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5538874841972186e-06,
+      "loss": 3.848,
+      "step": 9460
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.55384007585335e-06,
+      "loss": 2.9621,
+      "step": 9461
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5537926675094816e-06,
+      "loss": 2.7281,
+      "step": 9462
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5537452591656133e-06,
+      "loss": 3.6556,
+      "step": 9463
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5536978508217446e-06,
+      "loss": 3.0385,
+      "step": 9464
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5536504424778763e-06,
+      "loss": 3.9358,
+      "step": 9465
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5536030341340076e-06,
+      "loss": 3.8421,
+      "step": 9466
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.553555625790139e-06,
+      "loss": 3.0065,
+      "step": 9467
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5535082174462706e-06,
+      "loss": 2.5444,
+      "step": 9468
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5534608091024023e-06,
+      "loss": 2.9885,
+      "step": 9469
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5534134007585336e-06,
+      "loss": 2.9919,
+      "step": 9470
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5533659924146653e-06,
+      "loss": 3.6474,
+      "step": 9471
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5533185840707965e-06,
+      "loss": 3.8562,
+      "step": 9472
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.553271175726928e-06,
+      "loss": 3.6352,
+      "step": 9473
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5532237673830595e-06,
+      "loss": 3.6231,
+      "step": 9474
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.553176359039191e-06,
+      "loss": 3.8315,
+      "step": 9475
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5531289506953225e-06,
+      "loss": 3.953,
+      "step": 9476
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5530815423514542e-06,
+      "loss": 2.8585,
+      "step": 9477
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5530341340075855e-06,
+      "loss": 4.0825,
+      "step": 9478
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5529867256637168e-06,
+      "loss": 4.1392,
+      "step": 9479
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.552939317319848e-06,
+      "loss": 3.6642,
+      "step": 9480
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5528919089759798e-06,
+      "loss": 3.2362,
+      "step": 9481
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5528445006321115e-06,
+      "loss": 3.3804,
+      "step": 9482
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5527970922882427e-06,
+      "loss": 3.8491,
+      "step": 9483
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5527496839443745e-06,
+      "loss": 2.8481,
+      "step": 9484
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5527022756005057e-06,
+      "loss": 4.0861,
+      "step": 9485
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.552654867256637e-06,
+      "loss": 3.4017,
+      "step": 9486
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5526074589127687e-06,
+      "loss": 3.7656,
+      "step": 9487
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5525600505689e-06,
+      "loss": 2.9327,
+      "step": 9488
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5525126422250317e-06,
+      "loss": 3.1665,
+      "step": 9489
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5524652338811634e-06,
+      "loss": 3.2891,
+      "step": 9490
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5524178255372947e-06,
+      "loss": 3.7542,
+      "step": 9491
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.552370417193426e-06,
+      "loss": 3.4486,
+      "step": 9492
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5523230088495577e-06,
+      "loss": 2.8864,
+      "step": 9493
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.552275600505689e-06,
+      "loss": 3.7664,
+      "step": 9494
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5522281921618207e-06,
+      "loss": 4.1092,
+      "step": 9495
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5521807838179524e-06,
+      "loss": 4.2021,
+      "step": 9496
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5521333754740836e-06,
+      "loss": 2.901,
+      "step": 9497
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.552085967130215e-06,
+      "loss": 3.3378,
+      "step": 9498
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5520385587863466e-06,
+      "loss": 2.7789,
+      "step": 9499
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551991150442478e-06,
+      "loss": 3.1443,
+      "step": 9500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551943742098609e-06,
+      "loss": 3.7899,
+      "step": 9501
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551896333754741e-06,
+      "loss": 4.1086,
+      "step": 9502
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5518489254108726e-06,
+      "loss": 4.0788,
+      "step": 9503
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551801517067004e-06,
+      "loss": 3.2452,
+      "step": 9504
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551754108723135e-06,
+      "loss": 2.9633,
+      "step": 9505
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551706700379267e-06,
+      "loss": 3.5956,
+      "step": 9506
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551659292035398e-06,
+      "loss": 3.9823,
+      "step": 9507
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.55161188369153e-06,
+      "loss": 3.7491,
+      "step": 9508
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5515644753476615e-06,
+      "loss": 3.6176,
+      "step": 9509
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551517067003793e-06,
+      "loss": 3.2284,
+      "step": 9510
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551469658659924e-06,
+      "loss": 3.6968,
+      "step": 9511
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551422250316056e-06,
+      "loss": 3.6832,
+      "step": 9512
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551374841972187e-06,
+      "loss": 3.4144,
+      "step": 9513
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5513274336283184e-06,
+      "loss": 2.7802,
+      "step": 9514
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.55128002528445e-06,
+      "loss": 3.2372,
+      "step": 9515
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5512326169405818e-06,
+      "loss": 4.0218,
+      "step": 9516
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551185208596713e-06,
+      "loss": 2.3028,
+      "step": 9517
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5511378002528448e-06,
+      "loss": 2.7916,
+      "step": 9518
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.551090391908976e-06,
+      "loss": 3.4247,
+      "step": 9519
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5510429835651073e-06,
+      "loss": 2.9353,
+      "step": 9520
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.550995575221239e-06,
+      "loss": 3.8369,
+      "step": 9521
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5509481668773707e-06,
+      "loss": 4.2296,
+      "step": 9522
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.550900758533502e-06,
+      "loss": 2.752,
+      "step": 9523
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5508533501896333e-06,
+      "loss": 3.748,
+      "step": 9524
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.550805941845765e-06,
+      "loss": 3.814,
+      "step": 9525
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5507585335018963e-06,
+      "loss": 4.0913,
+      "step": 9526
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5507111251580276e-06,
+      "loss": 3.4464,
+      "step": 9527
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5506637168141593e-06,
+      "loss": 3.8347,
+      "step": 9528
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.550616308470291e-06,
+      "loss": 4.1047,
+      "step": 9529
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5505689001264222e-06,
+      "loss": 3.6397,
+      "step": 9530
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.550521491782554e-06,
+      "loss": 3.3902,
+      "step": 9531
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5504740834386852e-06,
+      "loss": 3.3814,
+      "step": 9532
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5504266750948165e-06,
+      "loss": 3.5175,
+      "step": 9533
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5503792667509482e-06,
+      "loss": 3.6612,
+      "step": 9534
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.55033185840708e-06,
+      "loss": 3.9003,
+      "step": 9535
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.550284450063211e-06,
+      "loss": 2.8737,
+      "step": 9536
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.550237041719343e-06,
+      "loss": 4.6086,
+      "step": 9537
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.550189633375474e-06,
+      "loss": 3.6072,
+      "step": 9538
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5501422250316055e-06,
+      "loss": 3.0584,
+      "step": 9539
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.550094816687737e-06,
+      "loss": 3.6906,
+      "step": 9540
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5500474083438684e-06,
+      "loss": 4.1758,
+      "step": 9541
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.55e-06,
+      "loss": 3.6327,
+      "step": 9542
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.549952591656132e-06,
+      "loss": 3.7214,
+      "step": 9543
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.549905183312263e-06,
+      "loss": 3.2502,
+      "step": 9544
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5498577749683944e-06,
+      "loss": 4.2079,
+      "step": 9545
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5498103666245257e-06,
+      "loss": 3.6975,
+      "step": 9546
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5497629582806574e-06,
+      "loss": 3.6668,
+      "step": 9547
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.549715549936789e-06,
+      "loss": 3.2186,
+      "step": 9548
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5496681415929204e-06,
+      "loss": 3.3012,
+      "step": 9549
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.549620733249052e-06,
+      "loss": 3.24,
+      "step": 9550
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5495733249051834e-06,
+      "loss": 3.824,
+      "step": 9551
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5495259165613146e-06,
+      "loss": 2.9089,
+      "step": 9552
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5494785082174464e-06,
+      "loss": 3.4719,
+      "step": 9553
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5494310998735776e-06,
+      "loss": 2.9855,
+      "step": 9554
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5493836915297093e-06,
+      "loss": 3.5215,
+      "step": 9555
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.549336283185841e-06,
+      "loss": 3.4116,
+      "step": 9556
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5492888748419723e-06,
+      "loss": 4.0957,
+      "step": 9557
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5492414664981036e-06,
+      "loss": 3.9999,
+      "step": 9558
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5491940581542353e-06,
+      "loss": 3.6273,
+      "step": 9559
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5491466498103666e-06,
+      "loss": 3.7309,
+      "step": 9560
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5490992414664983e-06,
+      "loss": 3.8734,
+      "step": 9561
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.54905183312263e-06,
+      "loss": 3.6257,
+      "step": 9562
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5490044247787613e-06,
+      "loss": 3.831,
+      "step": 9563
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5489570164348926e-06,
+      "loss": 3.3066,
+      "step": 9564
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5489096080910243e-06,
+      "loss": 3.9179,
+      "step": 9565
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5488621997471555e-06,
+      "loss": 3.7531,
+      "step": 9566
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.548814791403287e-06,
+      "loss": 4.0443,
+      "step": 9567
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5487673830594185e-06,
+      "loss": 3.3373,
+      "step": 9568
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5487199747155502e-06,
+      "loss": 3.7045,
+      "step": 9569
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5486725663716815e-06,
+      "loss": 4.0038,
+      "step": 9570
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5486251580278128e-06,
+      "loss": 3.2796,
+      "step": 9571
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5485777496839445e-06,
+      "loss": 3.7106,
+      "step": 9572
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5485303413400758e-06,
+      "loss": 3.6538,
+      "step": 9573
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5484829329962075e-06,
+      "loss": 3.643,
+      "step": 9574
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.548435524652339e-06,
+      "loss": 3.5608,
+      "step": 9575
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5483881163084705e-06,
+      "loss": 4.0974,
+      "step": 9576
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5483407079646017e-06,
+      "loss": 3.4445,
+      "step": 9577
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5482932996207334e-06,
+      "loss": 3.4382,
+      "step": 9578
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5482458912768647e-06,
+      "loss": 3.8536,
+      "step": 9579
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.548198482932996e-06,
+      "loss": 4.0224,
+      "step": 9580
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5481510745891277e-06,
+      "loss": 3.4112,
+      "step": 9581
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5481036662452594e-06,
+      "loss": 3.4126,
+      "step": 9582
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5480562579013907e-06,
+      "loss": 3.8631,
+      "step": 9583
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5480088495575224e-06,
+      "loss": 3.0605,
+      "step": 9584
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5479614412136537e-06,
+      "loss": 3.8146,
+      "step": 9585
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.547914032869785e-06,
+      "loss": 3.7721,
+      "step": 9586
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5478666245259167e-06,
+      "loss": 3.4213,
+      "step": 9587
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5478192161820484e-06,
+      "loss": 4.1247,
+      "step": 9588
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5477718078381796e-06,
+      "loss": 3.161,
+      "step": 9589
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.547724399494311e-06,
+      "loss": 4.0217,
+      "step": 9590
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5476769911504426e-06,
+      "loss": 3.5293,
+      "step": 9591
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.547629582806574e-06,
+      "loss": 4.0874,
+      "step": 9592
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.547582174462705e-06,
+      "loss": 3.6895,
+      "step": 9593
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.547534766118837e-06,
+      "loss": 3.7095,
+      "step": 9594
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5474873577749686e-06,
+      "loss": 3.6704,
+      "step": 9595
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5474399494311e-06,
+      "loss": 3.1931,
+      "step": 9596
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5473925410872316e-06,
+      "loss": 3.7858,
+      "step": 9597
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.547345132743363e-06,
+      "loss": 4.1146,
+      "step": 9598
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.547297724399494e-06,
+      "loss": 3.4642,
+      "step": 9599
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.547250316055626e-06,
+      "loss": 2.7975,
+      "step": 9600
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5472029077117576e-06,
+      "loss": 3.2178,
+      "step": 9601
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.547155499367889e-06,
+      "loss": 3.3111,
+      "step": 9602
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5471080910240205e-06,
+      "loss": 3.0417,
+      "step": 9603
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.547060682680152e-06,
+      "loss": 3.565,
+      "step": 9604
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.547013274336283e-06,
+      "loss": 3.7535,
+      "step": 9605
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546965865992415e-06,
+      "loss": 4.1265,
+      "step": 9606
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546918457648546e-06,
+      "loss": 3.2332,
+      "step": 9607
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5468710493046778e-06,
+      "loss": 3.7289,
+      "step": 9608
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5468236409608095e-06,
+      "loss": 3.2549,
+      "step": 9609
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5467762326169408e-06,
+      "loss": 3.0569,
+      "step": 9610
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546728824273072e-06,
+      "loss": 4.2638,
+      "step": 9611
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5466814159292033e-06,
+      "loss": 3.7892,
+      "step": 9612
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546634007585335e-06,
+      "loss": 4.0202,
+      "step": 9613
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5465865992414667e-06,
+      "loss": 3.2081,
+      "step": 9614
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546539190897598e-06,
+      "loss": 3.504,
+      "step": 9615
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5464917825537297e-06,
+      "loss": 3.1818,
+      "step": 9616
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546444374209861e-06,
+      "loss": 3.2747,
+      "step": 9617
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5463969658659923e-06,
+      "loss": 3.6288,
+      "step": 9618
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546349557522124e-06,
+      "loss": 2.6222,
+      "step": 9619
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5463021491782553e-06,
+      "loss": 4.0783,
+      "step": 9620
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546254740834387e-06,
+      "loss": 4.0245,
+      "step": 9621
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5462073324905187e-06,
+      "loss": 3.8268,
+      "step": 9622
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.54615992414665e-06,
+      "loss": 3.946,
+      "step": 9623
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5461125158027812e-06,
+      "loss": 3.8516,
+      "step": 9624
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.546065107458913e-06,
+      "loss": 3.2731,
+      "step": 9625
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5460176991150442e-06,
+      "loss": 3.4868,
+      "step": 9626
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5459702907711755e-06,
+      "loss": 3.2292,
+      "step": 9627
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5459228824273076e-06,
+      "loss": 3.3887,
+      "step": 9628
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545875474083439e-06,
+      "loss": 2.8318,
+      "step": 9629
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.54582806573957e-06,
+      "loss": 2.9041,
+      "step": 9630
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545780657395702e-06,
+      "loss": 3.6144,
+      "step": 9631
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545733249051833e-06,
+      "loss": 3.6485,
+      "step": 9632
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5456858407079645e-06,
+      "loss": 3.7529,
+      "step": 9633
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545638432364096e-06,
+      "loss": 2.7796,
+      "step": 9634
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545591024020228e-06,
+      "loss": 3.9511,
+      "step": 9635
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545543615676359e-06,
+      "loss": 3.1714,
+      "step": 9636
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5454962073324904e-06,
+      "loss": 3.403,
+      "step": 9637
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545448798988622e-06,
+      "loss": 2.6617,
+      "step": 9638
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5454013906447534e-06,
+      "loss": 3.6199,
+      "step": 9639
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5453539823008847e-06,
+      "loss": 3.8789,
+      "step": 9640
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5453065739570164e-06,
+      "loss": 3.3616,
+      "step": 9641
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545259165613148e-06,
+      "loss": 3.2418,
+      "step": 9642
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5452117572692794e-06,
+      "loss": 3.4507,
+      "step": 9643
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.545164348925411e-06,
+      "loss": 3.5241,
+      "step": 9644
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5451169405815424e-06,
+      "loss": 3.0703,
+      "step": 9645
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5450695322376736e-06,
+      "loss": 4.3102,
+      "step": 9646
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5450221238938053e-06,
+      "loss": 3.7309,
+      "step": 9647
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.544974715549937e-06,
+      "loss": 3.079,
+      "step": 9648
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5449273072060683e-06,
+      "loss": 3.5327,
+      "step": 9649
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5448798988622e-06,
+      "loss": 4.1638,
+      "step": 9650
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5448324905183313e-06,
+      "loss": 2.7731,
+      "step": 9651
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5447850821744626e-06,
+      "loss": 4.0273,
+      "step": 9652
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5447376738305943e-06,
+      "loss": 3.1081,
+      "step": 9653
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5446902654867256e-06,
+      "loss": 2.6306,
+      "step": 9654
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5446428571428573e-06,
+      "loss": 4.0533,
+      "step": 9655
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5445954487989886e-06,
+      "loss": 3.4924,
+      "step": 9656
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5445480404551203e-06,
+      "loss": 3.4299,
+      "step": 9657
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5445006321112515e-06,
+      "loss": 3.2193,
+      "step": 9658
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.544453223767383e-06,
+      "loss": 2.8909,
+      "step": 9659
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5444058154235145e-06,
+      "loss": 3.8551,
+      "step": 9660
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5443584070796462e-06,
+      "loss": 3.8722,
+      "step": 9661
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5443109987357775e-06,
+      "loss": 2.92,
+      "step": 9662
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5442635903919092e-06,
+      "loss": 2.7665,
+      "step": 9663
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5442161820480405e-06,
+      "loss": 3.7725,
+      "step": 9664
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5441687737041718e-06,
+      "loss": 3.3712,
+      "step": 9665
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5441213653603035e-06,
+      "loss": 3.5642,
+      "step": 9666
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5440739570164348e-06,
+      "loss": 3.5914,
+      "step": 9667
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5440265486725665e-06,
+      "loss": 3.5534,
+      "step": 9668
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.543979140328698e-06,
+      "loss": 4.4377,
+      "step": 9669
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5439317319848295e-06,
+      "loss": 3.8117,
+      "step": 9670
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5438843236409607e-06,
+      "loss": 3.3237,
+      "step": 9671
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5438369152970924e-06,
+      "loss": 3.3326,
+      "step": 9672
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5437895069532237e-06,
+      "loss": 3.02,
+      "step": 9673
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5437420986093554e-06,
+      "loss": 3.2664,
+      "step": 9674
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.543694690265487e-06,
+      "loss": 4.0608,
+      "step": 9675
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5436472819216184e-06,
+      "loss": 3.4927,
+      "step": 9676
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5435998735777497e-06,
+      "loss": 3.6919,
+      "step": 9677
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.543552465233881e-06,
+      "loss": 3.1547,
+      "step": 9678
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5435050568900127e-06,
+      "loss": 3.7017,
+      "step": 9679
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.543457648546144e-06,
+      "loss": 3.6073,
+      "step": 9680
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5434102402022757e-06,
+      "loss": 3.031,
+      "step": 9681
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5433628318584074e-06,
+      "loss": 4.2947,
+      "step": 9682
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5433154235145386e-06,
+      "loss": 3.5329,
+      "step": 9683
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.54326801517067e-06,
+      "loss": 3.8477,
+      "step": 9684
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5432206068268016e-06,
+      "loss": 3.6927,
+      "step": 9685
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.543173198482933e-06,
+      "loss": 3.107,
+      "step": 9686
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5431257901390646e-06,
+      "loss": 3.6333,
+      "step": 9687
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5430783817951963e-06,
+      "loss": 2.9679,
+      "step": 9688
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5430309734513276e-06,
+      "loss": 3.3682,
+      "step": 9689
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542983565107459e-06,
+      "loss": 3.6948,
+      "step": 9690
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5429361567635906e-06,
+      "loss": 3.8873,
+      "step": 9691
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542888748419722e-06,
+      "loss": 3.9664,
+      "step": 9692
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542841340075853e-06,
+      "loss": 2.9135,
+      "step": 9693
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542793931731985e-06,
+      "loss": 2.4926,
+      "step": 9694
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5427465233881165e-06,
+      "loss": 3.8408,
+      "step": 9695
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542699115044248e-06,
+      "loss": 3.6817,
+      "step": 9696
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5426517067003795e-06,
+      "loss": 3.5709,
+      "step": 9697
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542604298356511e-06,
+      "loss": 3.608,
+      "step": 9698
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542556890012642e-06,
+      "loss": 3.6852,
+      "step": 9699
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542509481668774e-06,
+      "loss": 3.4822,
+      "step": 9700
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5424620733249055e-06,
+      "loss": 2.1056,
+      "step": 9701
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5424146649810368e-06,
+      "loss": 2.9076,
+      "step": 9702
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542367256637168e-06,
+      "loss": 3.5042,
+      "step": 9703
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5423198482932998e-06,
+      "loss": 3.673,
+      "step": 9704
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542272439949431e-06,
+      "loss": 3.6723,
+      "step": 9705
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5422250316055623e-06,
+      "loss": 3.8678,
+      "step": 9706
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542177623261694e-06,
+      "loss": 3.4057,
+      "step": 9707
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5421302149178257e-06,
+      "loss": 3.666,
+      "step": 9708
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.542082806573957e-06,
+      "loss": 4.2173,
+      "step": 9709
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5420353982300887e-06,
+      "loss": 2.7084,
+      "step": 9710
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.54198798988622e-06,
+      "loss": 3.9026,
+      "step": 9711
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5419405815423513e-06,
+      "loss": 3.6173,
+      "step": 9712
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.541893173198483e-06,
+      "loss": 4.0284,
+      "step": 9713
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5418457648546147e-06,
+      "loss": 3.5314,
+      "step": 9714
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.541798356510746e-06,
+      "loss": 3.9934,
+      "step": 9715
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5417509481668777e-06,
+      "loss": 3.0078,
+      "step": 9716
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.541703539823009e-06,
+      "loss": 2.5985,
+      "step": 9717
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5416561314791402e-06,
+      "loss": 2.963,
+      "step": 9718
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5416087231352715e-06,
+      "loss": 3.6634,
+      "step": 9719
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5415613147914032e-06,
+      "loss": 3.9175,
+      "step": 9720
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.541513906447535e-06,
+      "loss": 3.9507,
+      "step": 9721
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541466498103666e-06,
+      "loss": 2.838,
+      "step": 9722
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541419089759798e-06,
+      "loss": 2.8901,
+      "step": 9723
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541371681415929e-06,
+      "loss": 2.9788,
+      "step": 9724
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5413242730720605e-06,
+      "loss": 3.6591,
+      "step": 9725
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541276864728192e-06,
+      "loss": 3.7252,
+      "step": 9726
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541229456384324e-06,
+      "loss": 3.1855,
+      "step": 9727
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541182048040455e-06,
+      "loss": 3.7719,
+      "step": 9728
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541134639696587e-06,
+      "loss": 2.8398,
+      "step": 9729
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.541087231352718e-06,
+      "loss": 4.3319,
+      "step": 9730
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5410398230088494e-06,
+      "loss": 3.5178,
+      "step": 9731
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540992414664981e-06,
+      "loss": 3.7709,
+      "step": 9732
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5409450063211124e-06,
+      "loss": 3.2266,
+      "step": 9733
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540897597977244e-06,
+      "loss": 3.5079,
+      "step": 9734
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540850189633376e-06,
+      "loss": 2.8937,
+      "step": 9735
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540802781289507e-06,
+      "loss": 3.7915,
+      "step": 9736
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5407553729456384e-06,
+      "loss": 3.4945,
+      "step": 9737
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.54070796460177e-06,
+      "loss": 3.285,
+      "step": 9738
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5406605562579014e-06,
+      "loss": 3.7788,
+      "step": 9739
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540613147914033e-06,
+      "loss": 3.2597,
+      "step": 9740
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5405657395701643e-06,
+      "loss": 3.11,
+      "step": 9741
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540518331226296e-06,
+      "loss": 4.1607,
+      "step": 9742
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5404709228824273e-06,
+      "loss": 3.0346,
+      "step": 9743
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5404235145385586e-06,
+      "loss": 3.8228,
+      "step": 9744
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5403761061946903e-06,
+      "loss": 3.5064,
+      "step": 9745
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5403286978508216e-06,
+      "loss": 3.8541,
+      "step": 9746
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5402812895069533e-06,
+      "loss": 3.102,
+      "step": 9747
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.540233881163085e-06,
+      "loss": 3.9431,
+      "step": 9748
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5401864728192163e-06,
+      "loss": 3.3898,
+      "step": 9749
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5401390644753476e-06,
+      "loss": 3.9301,
+      "step": 9750
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5400916561314793e-06,
+      "loss": 3.376,
+      "step": 9751
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5400442477876105e-06,
+      "loss": 3.6273,
+      "step": 9752
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5399968394437422e-06,
+      "loss": 3.6198,
+      "step": 9753
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.539949431099874e-06,
+      "loss": 3.604,
+      "step": 9754
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5399020227560052e-06,
+      "loss": 3.2934,
+      "step": 9755
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5398546144121365e-06,
+      "loss": 3.4622,
+      "step": 9756
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5398072060682682e-06,
+      "loss": 2.6117,
+      "step": 9757
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5397597977243995e-06,
+      "loss": 3.5575,
+      "step": 9758
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5397123893805308e-06,
+      "loss": 2.8275,
+      "step": 9759
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5396649810366625e-06,
+      "loss": 3.1516,
+      "step": 9760
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.539617572692794e-06,
+      "loss": 2.6218,
+      "step": 9761
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5395701643489255e-06,
+      "loss": 2.8667,
+      "step": 9762
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5395227560050567e-06,
+      "loss": 3.4976,
+      "step": 9763
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5394753476611884e-06,
+      "loss": 3.9303,
+      "step": 9764
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5394279393173197e-06,
+      "loss": 3.379,
+      "step": 9765
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5393805309734514e-06,
+      "loss": 3.3994,
+      "step": 9766
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.539333122629583e-06,
+      "loss": 3.236,
+      "step": 9767
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5392857142857144e-06,
+      "loss": 4.0522,
+      "step": 9768
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5392383059418457e-06,
+      "loss": 3.5579,
+      "step": 9769
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5391908975979774e-06,
+      "loss": 3.5192,
+      "step": 9770
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5391434892541087e-06,
+      "loss": 3.1253,
+      "step": 9771
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.53909608091024e-06,
+      "loss": 3.0347,
+      "step": 9772
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5390486725663717e-06,
+      "loss": 3.2805,
+      "step": 9773
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5390012642225034e-06,
+      "loss": 3.5168,
+      "step": 9774
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5389538558786347e-06,
+      "loss": 3.1151,
+      "step": 9775
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5389064475347664e-06,
+      "loss": 4.2556,
+      "step": 9776
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5388590391908976e-06,
+      "loss": 3.6069,
+      "step": 9777
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.538811630847029e-06,
+      "loss": 3.6238,
+      "step": 9778
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5387642225031606e-06,
+      "loss": 3.5701,
+      "step": 9779
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5387168141592923e-06,
+      "loss": 3.3705,
+      "step": 9780
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5386694058154236e-06,
+      "loss": 3.1154,
+      "step": 9781
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5386219974715553e-06,
+      "loss": 3.1457,
+      "step": 9782
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5385745891276866e-06,
+      "loss": 3.0075,
+      "step": 9783
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.538527180783818e-06,
+      "loss": 3.8984,
+      "step": 9784
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.538479772439949e-06,
+      "loss": 4.0843,
+      "step": 9785
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.538432364096081e-06,
+      "loss": 3.3445,
+      "step": 9786
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5383849557522126e-06,
+      "loss": 3.3464,
+      "step": 9787
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.538337547408344e-06,
+      "loss": 4.2262,
+      "step": 9788
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5382901390644755e-06,
+      "loss": 4.0256,
+      "step": 9789
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.538242730720607e-06,
+      "loss": 3.9129,
+      "step": 9790
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.538195322376738e-06,
+      "loss": 3.7201,
+      "step": 9791
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.53814791403287e-06,
+      "loss": 4.1442,
+      "step": 9792
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5381005056890015e-06,
+      "loss": 3.1647,
+      "step": 9793
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.538053097345133e-06,
+      "loss": 4.0878,
+      "step": 9794
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5380056890012645e-06,
+      "loss": 3.4057,
+      "step": 9795
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5379582806573958e-06,
+      "loss": 3.8083,
+      "step": 9796
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537910872313527e-06,
+      "loss": 3.4415,
+      "step": 9797
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5378634639696588e-06,
+      "loss": 3.9272,
+      "step": 9798
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.53781605562579e-06,
+      "loss": 3.5236,
+      "step": 9799
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5377686472819217e-06,
+      "loss": 3.6486,
+      "step": 9800
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5377212389380534e-06,
+      "loss": 3.792,
+      "step": 9801
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5376738305941847e-06,
+      "loss": 3.0774,
+      "step": 9802
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537626422250316e-06,
+      "loss": 4.085,
+      "step": 9803
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5375790139064477e-06,
+      "loss": 2.9692,
+      "step": 9804
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537531605562579e-06,
+      "loss": 3.2595,
+      "step": 9805
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5374841972187107e-06,
+      "loss": 2.8377,
+      "step": 9806
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537436788874842e-06,
+      "loss": 2.6816,
+      "step": 9807
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5373893805309737e-06,
+      "loss": 3.2996,
+      "step": 9808
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537341972187105e-06,
+      "loss": 3.2951,
+      "step": 9809
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5372945638432362e-06,
+      "loss": 3.0972,
+      "step": 9810
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537247155499368e-06,
+      "loss": 3.0759,
+      "step": 9811
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5371997471554992e-06,
+      "loss": 3.8262,
+      "step": 9812
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537152338811631e-06,
+      "loss": 3.0809,
+      "step": 9813
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5371049304677626e-06,
+      "loss": 3.8364,
+      "step": 9814
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537057522123894e-06,
+      "loss": 3.6702,
+      "step": 9815
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.537010113780025e-06,
+      "loss": 3.3138,
+      "step": 9816
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536962705436157e-06,
+      "loss": 2.8779,
+      "step": 9817
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536915297092288e-06,
+      "loss": 3.2882,
+      "step": 9818
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.53686788874842e-06,
+      "loss": 4.1205,
+      "step": 9819
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5368204804045516e-06,
+      "loss": 3.8853,
+      "step": 9820
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536773072060683e-06,
+      "loss": 3.8392,
+      "step": 9821
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536725663716814e-06,
+      "loss": 3.2675,
+      "step": 9822
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536678255372946e-06,
+      "loss": 3.9268,
+      "step": 9823
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536630847029077e-06,
+      "loss": 3.736,
+      "step": 9824
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5365834386852084e-06,
+      "loss": 3.3331,
+      "step": 9825
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.53653603034134e-06,
+      "loss": 3.8701,
+      "step": 9826
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536488621997472e-06,
+      "loss": 4.1402,
+      "step": 9827
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536441213653603e-06,
+      "loss": 3.0669,
+      "step": 9828
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5363938053097344e-06,
+      "loss": 3.5822,
+      "step": 9829
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536346396965866e-06,
+      "loss": 3.0233,
+      "step": 9830
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5362989886219974e-06,
+      "loss": 3.1741,
+      "step": 9831
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536251580278129e-06,
+      "loss": 3.064,
+      "step": 9832
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5362041719342608e-06,
+      "loss": 3.9676,
+      "step": 9833
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536156763590392e-06,
+      "loss": 3.8311,
+      "step": 9834
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5361093552465233e-06,
+      "loss": 3.9705,
+      "step": 9835
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.536061946902655e-06,
+      "loss": 3.8918,
+      "step": 9836
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5360145385587863e-06,
+      "loss": 3.5618,
+      "step": 9837
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5359671302149176e-06,
+      "loss": 3.0961,
+      "step": 9838
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5359197218710493e-06,
+      "loss": 3.8521,
+      "step": 9839
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.535872313527181e-06,
+      "loss": 3.6026,
+      "step": 9840
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5358249051833123e-06,
+      "loss": 3.6158,
+      "step": 9841
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.535777496839444e-06,
+      "loss": 3.1158,
+      "step": 9842
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5357300884955753e-06,
+      "loss": 3.7332,
+      "step": 9843
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5356826801517066e-06,
+      "loss": 3.3171,
+      "step": 9844
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5356352718078383e-06,
+      "loss": 4.0342,
+      "step": 9845
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.53558786346397e-06,
+      "loss": 3.6103,
+      "step": 9846
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5355404551201012e-06,
+      "loss": 3.6968,
+      "step": 9847
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.535493046776233e-06,
+      "loss": 3.0854,
+      "step": 9848
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5354456384323642e-06,
+      "loss": 3.5871,
+      "step": 9849
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5353982300884955e-06,
+      "loss": 4.261,
+      "step": 9850
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5353508217446268e-06,
+      "loss": 3.3737,
+      "step": 9851
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5353034134007585e-06,
+      "loss": 3.3889,
+      "step": 9852
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.53525600505689e-06,
+      "loss": 3.4802,
+      "step": 9853
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5352085967130215e-06,
+      "loss": 3.3769,
+      "step": 9854
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.535161188369153e-06,
+      "loss": 4.2001,
+      "step": 9855
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5351137800252845e-06,
+      "loss": 3.3595,
+      "step": 9856
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5350663716814157e-06,
+      "loss": 3.8508,
+      "step": 9857
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5350189633375474e-06,
+      "loss": 4.0341,
+      "step": 9858
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.534971554993679e-06,
+      "loss": 2.6509,
+      "step": 9859
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5349241466498104e-06,
+      "loss": 2.7347,
+      "step": 9860
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.534876738305942e-06,
+      "loss": 2.8583,
+      "step": 9861
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5348293299620734e-06,
+      "loss": 3.2877,
+      "step": 9862
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5347819216182047e-06,
+      "loss": 2.8984,
+      "step": 9863
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5347345132743364e-06,
+      "loss": 3.1705,
+      "step": 9864
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5346871049304677e-06,
+      "loss": 3.1953,
+      "step": 9865
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5346396965865994e-06,
+      "loss": 4.0541,
+      "step": 9866
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.534592288242731e-06,
+      "loss": 2.6947,
+      "step": 9867
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5345448798988624e-06,
+      "loss": 3.1878,
+      "step": 9868
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5344974715549936e-06,
+      "loss": 3.1745,
+      "step": 9869
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5344500632111253e-06,
+      "loss": 3.527,
+      "step": 9870
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5344026548672566e-06,
+      "loss": 3.1918,
+      "step": 9871
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5343552465233883e-06,
+      "loss": 4.1261,
+      "step": 9872
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5343078381795196e-06,
+      "loss": 3.5303,
+      "step": 9873
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5342604298356513e-06,
+      "loss": 3.4322,
+      "step": 9874
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5342130214917826e-06,
+      "loss": 3.5799,
+      "step": 9875
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.534165613147914e-06,
+      "loss": 4.3406,
+      "step": 9876
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5341182048040456e-06,
+      "loss": 3.4112,
+      "step": 9877
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.534070796460177e-06,
+      "loss": 3.8363,
+      "step": 9878
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5340233881163086e-06,
+      "loss": 3.3424,
+      "step": 9879
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5339759797724403e-06,
+      "loss": 3.9237,
+      "step": 9880
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5339285714285716e-06,
+      "loss": 3.5208,
+      "step": 9881
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.533881163084703e-06,
+      "loss": 3.7492,
+      "step": 9882
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5338337547408345e-06,
+      "loss": 2.3791,
+      "step": 9883
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.533786346396966e-06,
+      "loss": 3.5022,
+      "step": 9884
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5337389380530975e-06,
+      "loss": 3.8145,
+      "step": 9885
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5336915297092292e-06,
+      "loss": 3.0748,
+      "step": 9886
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5336441213653605e-06,
+      "loss": 3.2888,
+      "step": 9887
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5335967130214918e-06,
+      "loss": 4.3645,
+      "step": 9888
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5335493046776235e-06,
+      "loss": 3.2317,
+      "step": 9889
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5335018963337548e-06,
+      "loss": 3.369,
+      "step": 9890
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.533454487989886e-06,
+      "loss": 4.0259,
+      "step": 9891
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5334070796460178e-06,
+      "loss": 3.351,
+      "step": 9892
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5333596713021495e-06,
+      "loss": 3.4309,
+      "step": 9893
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5333122629582807e-06,
+      "loss": 3.5062,
+      "step": 9894
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.533264854614412e-06,
+      "loss": 3.3702,
+      "step": 9895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5332174462705437e-06,
+      "loss": 2.869,
+      "step": 9896
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.533170037926675e-06,
+      "loss": 3.6848,
+      "step": 9897
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5331226295828063e-06,
+      "loss": 4.1218,
+      "step": 9898
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5330752212389384e-06,
+      "loss": 3.7051,
+      "step": 9899
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5330278128950697e-06,
+      "loss": 2.5618,
+      "step": 9900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.532980404551201e-06,
+      "loss": 2.795,
+      "step": 9901
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5329329962073327e-06,
+      "loss": 3.3466,
+      "step": 9902
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.532885587863464e-06,
+      "loss": 3.5409,
+      "step": 9903
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5328381795195952e-06,
+      "loss": 3.9089,
+      "step": 9904
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.532790771175727e-06,
+      "loss": 3.6817,
+      "step": 9905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5327433628318586e-06,
+      "loss": 3.1322,
+      "step": 9906
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.53269595448799e-06,
+      "loss": 4.1688,
+      "step": 9907
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5326485461441216e-06,
+      "loss": 3.411,
+      "step": 9908
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.532601137800253e-06,
+      "loss": 3.6293,
+      "step": 9909
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.532553729456384e-06,
+      "loss": 4.1471,
+      "step": 9910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.532506321112516e-06,
+      "loss": 3.9757,
+      "step": 9911
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532458912768647e-06,
+      "loss": 3.5344,
+      "step": 9912
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532411504424779e-06,
+      "loss": 3.227,
+      "step": 9913
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5323640960809106e-06,
+      "loss": 3.4421,
+      "step": 9914
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532316687737042e-06,
+      "loss": 3.1637,
+      "step": 9915
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532269279393173e-06,
+      "loss": 3.6026,
+      "step": 9916
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5322218710493044e-06,
+      "loss": 3.7683,
+      "step": 9917
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532174462705436e-06,
+      "loss": 2.6322,
+      "step": 9918
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532127054361568e-06,
+      "loss": 3.5559,
+      "step": 9919
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532079646017699e-06,
+      "loss": 3.0521,
+      "step": 9920
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.532032237673831e-06,
+      "loss": 3.1215,
+      "step": 9921
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531984829329962e-06,
+      "loss": 3.7231,
+      "step": 9922
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5319374209860934e-06,
+      "loss": 3.2883,
+      "step": 9923
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531890012642225e-06,
+      "loss": 3.0299,
+      "step": 9924
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5318426042983564e-06,
+      "loss": 3.6603,
+      "step": 9925
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531795195954488e-06,
+      "loss": 3.9585,
+      "step": 9926
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5317477876106198e-06,
+      "loss": 3.2943,
+      "step": 9927
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531700379266751e-06,
+      "loss": 2.6872,
+      "step": 9928
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5316529709228823e-06,
+      "loss": 3.323,
+      "step": 9929
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531605562579014e-06,
+      "loss": 4.1878,
+      "step": 9930
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5315581542351453e-06,
+      "loss": 3.6271,
+      "step": 9931
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531510745891277e-06,
+      "loss": 3.4221,
+      "step": 9932
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5314633375474087e-06,
+      "loss": 3.3499,
+      "step": 9933
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.53141592920354e-06,
+      "loss": 2.7178,
+      "step": 9934
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5313685208596713e-06,
+      "loss": 3.536,
+      "step": 9935
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531321112515803e-06,
+      "loss": 4.1093,
+      "step": 9936
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5312737041719343e-06,
+      "loss": 2.9077,
+      "step": 9937
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5312262958280655e-06,
+      "loss": 3.7495,
+      "step": 9938
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5311788874841972e-06,
+      "loss": 3.6711,
+      "step": 9939
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.531131479140329e-06,
+      "loss": 3.5486,
+      "step": 9940
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5310840707964602e-06,
+      "loss": 4.0563,
+      "step": 9941
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5310366624525915e-06,
+      "loss": 2.2036,
+      "step": 9942
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5309892541087232e-06,
+      "loss": 2.8316,
+      "step": 9943
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5309418457648545e-06,
+      "loss": 2.8405,
+      "step": 9944
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530894437420986e-06,
+      "loss": 2.8415,
+      "step": 9945
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530847029077118e-06,
+      "loss": 3.5015,
+      "step": 9946
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530799620733249e-06,
+      "loss": 3.219,
+      "step": 9947
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5307522123893805e-06,
+      "loss": 2.8918,
+      "step": 9948
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530704804045512e-06,
+      "loss": 3.0954,
+      "step": 9949
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5306573957016435e-06,
+      "loss": 3.6625,
+      "step": 9950
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5306099873577747e-06,
+      "loss": 3.0067,
+      "step": 9951
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5305625790139064e-06,
+      "loss": 3.5015,
+      "step": 9952
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530515170670038e-06,
+      "loss": 3.9566,
+      "step": 9953
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5304677623261694e-06,
+      "loss": 3.7704,
+      "step": 9954
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530420353982301e-06,
+      "loss": 3.3991,
+      "step": 9955
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5303729456384324e-06,
+      "loss": 3.8529,
+      "step": 9956
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5303255372945637e-06,
+      "loss": 3.5864,
+      "step": 9957
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5302781289506954e-06,
+      "loss": 3.5456,
+      "step": 9958
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.530230720606827e-06,
+      "loss": 2.7583,
+      "step": 9959
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5301833122629584e-06,
+      "loss": 3.5561,
+      "step": 9960
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5301359039190897e-06,
+      "loss": 3.9929,
+      "step": 9961
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5300884955752214e-06,
+      "loss": 3.3765,
+      "step": 9962
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5300410872313526e-06,
+      "loss": 3.1847,
+      "step": 9963
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.529993678887484e-06,
+      "loss": 4.0599,
+      "step": 9964
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5299462705436156e-06,
+      "loss": 3.576,
+      "step": 9965
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5298988621997473e-06,
+      "loss": 3.6277,
+      "step": 9966
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5298514538558786e-06,
+      "loss": 3.7881,
+      "step": 9967
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5298040455120103e-06,
+      "loss": 3.7838,
+      "step": 9968
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5297566371681416e-06,
+      "loss": 3.9059,
+      "step": 9969
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.529709228824273e-06,
+      "loss": 3.7574,
+      "step": 9970
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5296618204804046e-06,
+      "loss": 3.5503,
+      "step": 9971
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5296144121365363e-06,
+      "loss": 3.2593,
+      "step": 9972
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5295670037926676e-06,
+      "loss": 2.8534,
+      "step": 9973
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5295195954487993e-06,
+      "loss": 4.0194,
+      "step": 9974
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5294721871049305e-06,
+      "loss": 3.5957,
+      "step": 9975
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.529424778761062e-06,
+      "loss": 3.8201,
+      "step": 9976
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5293773704171935e-06,
+      "loss": 3.7962,
+      "step": 9977
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.529329962073325e-06,
+      "loss": 3.3193,
+      "step": 9978
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5292825537294565e-06,
+      "loss": 3.1008,
+      "step": 9979
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5292351453855882e-06,
+      "loss": 3.2044,
+      "step": 9980
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5291877370417195e-06,
+      "loss": 3.752,
+      "step": 9981
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5291403286978508e-06,
+      "loss": 3.4963,
+      "step": 9982
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.529092920353982e-06,
+      "loss": 3.3817,
+      "step": 9983
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5290455120101138e-06,
+      "loss": 2.9741,
+      "step": 9984
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5289981036662455e-06,
+      "loss": 3.3814,
+      "step": 9985
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5289506953223767e-06,
+      "loss": 3.885,
+      "step": 9986
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5289032869785085e-06,
+      "loss": 3.9671,
+      "step": 9987
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5288558786346397e-06,
+      "loss": 3.4742,
+      "step": 9988
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528808470290771e-06,
+      "loss": 3.3674,
+      "step": 9989
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5287610619469027e-06,
+      "loss": 3.9223,
+      "step": 9990
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528713653603034e-06,
+      "loss": 3.642,
+      "step": 9991
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5286662452591657e-06,
+      "loss": 3.4358,
+      "step": 9992
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5286188369152974e-06,
+      "loss": 3.6871,
+      "step": 9993
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5285714285714287e-06,
+      "loss": 2.6907,
+      "step": 9994
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.52852402022756e-06,
+      "loss": 3.1488,
+      "step": 9995
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5284766118836917e-06,
+      "loss": 3.3505,
+      "step": 9996
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528429203539823e-06,
+      "loss": 3.8109,
+      "step": 9997
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5283817951959547e-06,
+      "loss": 3.5282,
+      "step": 9998
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5283343868520864e-06,
+      "loss": 2.776,
+      "step": 9999
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5282869785082176e-06,
+      "loss": 3.2103,
+      "step": 10000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528239570164349e-06,
+      "loss": 3.6401,
+      "step": 10001
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5281921618204806e-06,
+      "loss": 3.6983,
+      "step": 10002
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528144753476612e-06,
+      "loss": 3.4292,
+      "step": 10003
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528097345132743e-06,
+      "loss": 3.5916,
+      "step": 10004
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.528049936788875e-06,
+      "loss": 3.3014,
+      "step": 10005
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5280025284450066e-06,
+      "loss": 3.2417,
+      "step": 10006
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527955120101138e-06,
+      "loss": 3.5924,
+      "step": 10007
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527907711757269e-06,
+      "loss": 3.0537,
+      "step": 10008
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527860303413401e-06,
+      "loss": 4.0107,
+      "step": 10009
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527812895069532e-06,
+      "loss": 3.1788,
+      "step": 10010
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527765486725664e-06,
+      "loss": 3.5243,
+      "step": 10011
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5277180783817955e-06,
+      "loss": 2.8551,
+      "step": 10012
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527670670037927e-06,
+      "loss": 3.5794,
+      "step": 10013
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527623261694058e-06,
+      "loss": 3.3744,
+      "step": 10014
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.52757585335019e-06,
+      "loss": 3.3543,
+      "step": 10015
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527528445006321e-06,
+      "loss": 3.2974,
+      "step": 10016
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5274810366624524e-06,
+      "loss": 3.9167,
+      "step": 10017
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527433628318584e-06,
+      "loss": 2.9052,
+      "step": 10018
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5273862199747158e-06,
+      "loss": 3.202,
+      "step": 10019
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527338811630847e-06,
+      "loss": 3.6242,
+      "step": 10020
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5272914032869788e-06,
+      "loss": 3.5225,
+      "step": 10021
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.52724399494311e-06,
+      "loss": 3.6137,
+      "step": 10022
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5271965865992413e-06,
+      "loss": 3.2182,
+      "step": 10023
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527149178255373e-06,
+      "loss": 4.26,
+      "step": 10024
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5271017699115047e-06,
+      "loss": 3.281,
+      "step": 10025
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.527054361567636e-06,
+      "loss": 3.1406,
+      "step": 10026
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5270069532237673e-06,
+      "loss": 3.5827,
+      "step": 10027
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.526959544879899e-06,
+      "loss": 3.3579,
+      "step": 10028
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5269121365360303e-06,
+      "loss": 3.9151,
+      "step": 10029
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5268647281921616e-06,
+      "loss": 3.6412,
+      "step": 10030
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5268173198482933e-06,
+      "loss": 3.6187,
+      "step": 10031
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.526769911504425e-06,
+      "loss": 2.7154,
+      "step": 10032
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5267225031605562e-06,
+      "loss": 3.5582,
+      "step": 10033
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.526675094816688e-06,
+      "loss": 3.6397,
+      "step": 10034
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5266276864728192e-06,
+      "loss": 3.4391,
+      "step": 10035
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5265802781289505e-06,
+      "loss": 3.7348,
+      "step": 10036
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5265328697850822e-06,
+      "loss": 3.0345,
+      "step": 10037
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526485461441214e-06,
+      "loss": 2.7062,
+      "step": 10038
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526438053097345e-06,
+      "loss": 3.2964,
+      "step": 10039
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526390644753477e-06,
+      "loss": 3.1887,
+      "step": 10040
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526343236409608e-06,
+      "loss": 3.2493,
+      "step": 10041
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5262958280657395e-06,
+      "loss": 2.6632,
+      "step": 10042
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526248419721871e-06,
+      "loss": 2.8945,
+      "step": 10043
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5262010113780024e-06,
+      "loss": 3.936,
+      "step": 10044
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526153603034134e-06,
+      "loss": 3.3548,
+      "step": 10045
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526106194690266e-06,
+      "loss": 3.4054,
+      "step": 10046
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.526058786346397e-06,
+      "loss": 2.9794,
+      "step": 10047
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5260113780025284e-06,
+      "loss": 3.1264,
+      "step": 10048
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5259639696586597e-06,
+      "loss": 2.9336,
+      "step": 10049
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5259165613147914e-06,
+      "loss": 3.7717,
+      "step": 10050
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.525869152970923e-06,
+      "loss": 3.429,
+      "step": 10051
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5258217446270544e-06,
+      "loss": 3.0683,
+      "step": 10052
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.525774336283186e-06,
+      "loss": 2.9378,
+      "step": 10053
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5257269279393174e-06,
+      "loss": 3.2216,
+      "step": 10054
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5256795195954486e-06,
+      "loss": 2.8017,
+      "step": 10055
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5256321112515804e-06,
+      "loss": 3.9768,
+      "step": 10056
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5255847029077116e-06,
+      "loss": 3.9256,
+      "step": 10057
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5255372945638433e-06,
+      "loss": 2.8335,
+      "step": 10058
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.525489886219975e-06,
+      "loss": 4.2159,
+      "step": 10059
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5254424778761063e-06,
+      "loss": 3.2118,
+      "step": 10060
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5253950695322376e-06,
+      "loss": 3.7655,
+      "step": 10061
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5253476611883693e-06,
+      "loss": 2.9013,
+      "step": 10062
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5253002528445006e-06,
+      "loss": 3.6651,
+      "step": 10063
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5252528445006323e-06,
+      "loss": 3.1518,
+      "step": 10064
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.525205436156764e-06,
+      "loss": 3.7946,
+      "step": 10065
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5251580278128953e-06,
+      "loss": 3.3918,
+      "step": 10066
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5251106194690266e-06,
+      "loss": 4.0665,
+      "step": 10067
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5250632111251583e-06,
+      "loss": 3.0378,
+      "step": 10068
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5250158027812895e-06,
+      "loss": 3.7966,
+      "step": 10069
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.524968394437421e-06,
+      "loss": 3.5169,
+      "step": 10070
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5249209860935525e-06,
+      "loss": 2.9581,
+      "step": 10071
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5248735777496842e-06,
+      "loss": 2.955,
+      "step": 10072
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5248261694058155e-06,
+      "loss": 3.5663,
+      "step": 10073
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5247787610619468e-06,
+      "loss": 3.976,
+      "step": 10074
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5247313527180785e-06,
+      "loss": 4.1026,
+      "step": 10075
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5246839443742098e-06,
+      "loss": 3.6182,
+      "step": 10076
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5246365360303415e-06,
+      "loss": 4.3404,
+      "step": 10077
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.524589127686473e-06,
+      "loss": 3.9812,
+      "step": 10078
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5245417193426045e-06,
+      "loss": 3.0585,
+      "step": 10079
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5244943109987357e-06,
+      "loss": 3.6866,
+      "step": 10080
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5244469026548674e-06,
+      "loss": 3.6128,
+      "step": 10081
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5243994943109987e-06,
+      "loss": 3.7012,
+      "step": 10082
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.52435208596713e-06,
+      "loss": 3.4535,
+      "step": 10083
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5243046776232617e-06,
+      "loss": 3.4475,
+      "step": 10084
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5242572692793934e-06,
+      "loss": 3.0527,
+      "step": 10085
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5242098609355247e-06,
+      "loss": 3.3039,
+      "step": 10086
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5241624525916564e-06,
+      "loss": 3.3978,
+      "step": 10087
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5241150442477877e-06,
+      "loss": 3.309,
+      "step": 10088
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.524067635903919e-06,
+      "loss": 3.5148,
+      "step": 10089
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5240202275600507e-06,
+      "loss": 3.6246,
+      "step": 10090
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5239728192161824e-06,
+      "loss": 2.5498,
+      "step": 10091
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5239254108723136e-06,
+      "loss": 3.9274,
+      "step": 10092
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.523878002528445e-06,
+      "loss": 3.7521,
+      "step": 10093
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5238305941845766e-06,
+      "loss": 3.4807,
+      "step": 10094
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.523783185840708e-06,
+      "loss": 3.1816,
+      "step": 10095
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.523735777496839e-06,
+      "loss": 3.6311,
+      "step": 10096
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.523688369152971e-06,
+      "loss": 4.1218,
+      "step": 10097
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5236409608091026e-06,
+      "loss": 2.6164,
+      "step": 10098
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.523593552465234e-06,
+      "loss": 3.2043,
+      "step": 10099
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.5235461441213656e-06,
+      "loss": 4.3701,
+      "step": 10100
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.523498735777497e-06,
+      "loss": 3.1411,
+      "step": 10101
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.523451327433628e-06,
+      "loss": 2.9811,
+      "step": 10102
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.52340391908976e-06,
+      "loss": 3.5582,
+      "step": 10103
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5233565107458916e-06,
+      "loss": 3.9285,
+      "step": 10104
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.523309102402023e-06,
+      "loss": 4.264,
+      "step": 10105
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5232616940581545e-06,
+      "loss": 3.4267,
+      "step": 10106
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.523214285714286e-06,
+      "loss": 4.0958,
+      "step": 10107
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.523166877370417e-06,
+      "loss": 2.8981,
+      "step": 10108
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.523119469026549e-06,
+      "loss": 3.9796,
+      "step": 10109
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.52307206068268e-06,
+      "loss": 3.5491,
+      "step": 10110
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5230246523388118e-06,
+      "loss": 3.6241,
+      "step": 10111
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5229772439949435e-06,
+      "loss": 4.1928,
+      "step": 10112
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5229298356510748e-06,
+      "loss": 3.3527,
+      "step": 10113
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522882427307206e-06,
+      "loss": 3.6983,
+      "step": 10114
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5228350189633373e-06,
+      "loss": 3.4644,
+      "step": 10115
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522787610619469e-06,
+      "loss": 3.3227,
+      "step": 10116
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5227402022756007e-06,
+      "loss": 3.0325,
+      "step": 10117
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522692793931732e-06,
+      "loss": 3.4208,
+      "step": 10118
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5226453855878637e-06,
+      "loss": 4.2964,
+      "step": 10119
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522597977243995e-06,
+      "loss": 2.9663,
+      "step": 10120
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5225505689001263e-06,
+      "loss": 3.4389,
+      "step": 10121
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522503160556258e-06,
+      "loss": 3.8003,
+      "step": 10122
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5224557522123893e-06,
+      "loss": 3.8446,
+      "step": 10123
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522408343868521e-06,
+      "loss": 3.1986,
+      "step": 10124
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5223609355246527e-06,
+      "loss": 4.1162,
+      "step": 10125
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522313527180784e-06,
+      "loss": 3.765,
+      "step": 10126
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5222661188369152e-06,
+      "loss": 3.2183,
+      "step": 10127
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522218710493047e-06,
+      "loss": 3.9338,
+      "step": 10128
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5221713021491782e-06,
+      "loss": 3.7941,
+      "step": 10129
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.52212389380531e-06,
+      "loss": 3.1918,
+      "step": 10130
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5220764854614416e-06,
+      "loss": 2.4955,
+      "step": 10131
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.522029077117573e-06,
+      "loss": 3.1197,
+      "step": 10132
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521981668773704e-06,
+      "loss": 3.7154,
+      "step": 10133
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521934260429836e-06,
+      "loss": 3.5081,
+      "step": 10134
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521886852085967e-06,
+      "loss": 3.3223,
+      "step": 10135
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5218394437420985e-06,
+      "loss": 2.9025,
+      "step": 10136
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.52179203539823e-06,
+      "loss": 3.2446,
+      "step": 10137
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521744627054362e-06,
+      "loss": 3.3253,
+      "step": 10138
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521697218710493e-06,
+      "loss": 3.4494,
+      "step": 10139
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5216498103666244e-06,
+      "loss": 3.3446,
+      "step": 10140
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521602402022756e-06,
+      "loss": 4.3534,
+      "step": 10141
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5215549936788874e-06,
+      "loss": 3.3773,
+      "step": 10142
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521507585335019e-06,
+      "loss": 4.0117,
+      "step": 10143
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521460176991151e-06,
+      "loss": 3.7052,
+      "step": 10144
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521412768647282e-06,
+      "loss": 3.1859,
+      "step": 10145
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5213653603034134e-06,
+      "loss": 3.2562,
+      "step": 10146
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521317951959545e-06,
+      "loss": 3.0919,
+      "step": 10147
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5212705436156764e-06,
+      "loss": 4.0078,
+      "step": 10148
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5212231352718076e-06,
+      "loss": 2.9635,
+      "step": 10149
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5211757269279393e-06,
+      "loss": 3.1731,
+      "step": 10150
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521128318584071e-06,
+      "loss": 2.9324,
+      "step": 10151
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5210809102402023e-06,
+      "loss": 3.5587,
+      "step": 10152
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.521033501896334e-06,
+      "loss": 3.46,
+      "step": 10153
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5209860935524653e-06,
+      "loss": 3.6808,
+      "step": 10154
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5209386852085966e-06,
+      "loss": 4.085,
+      "step": 10155
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5208912768647283e-06,
+      "loss": 3.8442,
+      "step": 10156
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.52084386852086e-06,
+      "loss": 3.8692,
+      "step": 10157
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5207964601769913e-06,
+      "loss": 3.8426,
+      "step": 10158
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5207490518331226e-06,
+      "loss": 3.4666,
+      "step": 10159
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5207016434892543e-06,
+      "loss": 3.102,
+      "step": 10160
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5206542351453855e-06,
+      "loss": 3.5251,
+      "step": 10161
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.520606826801517e-06,
+      "loss": 4.1944,
+      "step": 10162
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5205594184576485e-06,
+      "loss": 3.7526,
+      "step": 10163
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.5205120101137802e-06,
+      "loss": 3.8004,
+      "step": 10164
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5204646017699115e-06,
+      "loss": 3.2211,
+      "step": 10165
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5204171934260432e-06,
+      "loss": 3.8734,
+      "step": 10166
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5203697850821745e-06,
+      "loss": 4.1729,
+      "step": 10167
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5203223767383058e-06,
+      "loss": 2.9921,
+      "step": 10168
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5202749683944375e-06,
+      "loss": 3.3877,
+      "step": 10169
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.520227560050569e-06,
+      "loss": 3.7858,
+      "step": 10170
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5201801517067005e-06,
+      "loss": 3.973,
+      "step": 10171
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.520132743362832e-06,
+      "loss": 2.2585,
+      "step": 10172
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5200853350189635e-06,
+      "loss": 2.6436,
+      "step": 10173
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5200379266750947e-06,
+      "loss": 3.8709,
+      "step": 10174
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5199905183312264e-06,
+      "loss": 3.7975,
+      "step": 10175
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5199431099873577e-06,
+      "loss": 2.9575,
+      "step": 10176
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5198957016434894e-06,
+      "loss": 3.3679,
+      "step": 10177
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5198482932996207e-06,
+      "loss": 3.4172,
+      "step": 10178
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5198008849557524e-06,
+      "loss": 3.3691,
+      "step": 10179
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5197534766118837e-06,
+      "loss": 3.4015,
+      "step": 10180
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.519706068268015e-06,
+      "loss": 3.2978,
+      "step": 10181
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5196586599241467e-06,
+      "loss": 3.881,
+      "step": 10182
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.519611251580278e-06,
+      "loss": 3.1405,
+      "step": 10183
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5195638432364097e-06,
+      "loss": 2.9874,
+      "step": 10184
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5195164348925414e-06,
+      "loss": 3.5504,
+      "step": 10185
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5194690265486726e-06,
+      "loss": 3.3508,
+      "step": 10186
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.519421618204804e-06,
+      "loss": 2.817,
+      "step": 10187
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5193742098609356e-06,
+      "loss": 2.9176,
+      "step": 10188
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.519326801517067e-06,
+      "loss": 3.2577,
+      "step": 10189
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5192793931731986e-06,
+      "loss": 3.931,
+      "step": 10190
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5192319848293303e-06,
+      "loss": 3.2171,
+      "step": 10191
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5191845764854616e-06,
+      "loss": 3.3816,
+      "step": 10192
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.519137168141593e-06,
+      "loss": 3.9189,
+      "step": 10193
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5190897597977246e-06,
+      "loss": 3.8848,
+      "step": 10194
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.519042351453856e-06,
+      "loss": 3.804,
+      "step": 10195
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518994943109987e-06,
+      "loss": 3.6613,
+      "step": 10196
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518947534766119e-06,
+      "loss": 3.0763,
+      "step": 10197
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5189001264222505e-06,
+      "loss": 3.5465,
+      "step": 10198
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518852718078382e-06,
+      "loss": 3.1388,
+      "step": 10199
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518805309734513e-06,
+      "loss": 3.5224,
+      "step": 10200
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518757901390645e-06,
+      "loss": 3.6974,
+      "step": 10201
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518710493046776e-06,
+      "loss": 3.0658,
+      "step": 10202
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518663084702908e-06,
+      "loss": 3.1439,
+      "step": 10203
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5186156763590395e-06,
+      "loss": 3.6705,
+      "step": 10204
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5185682680151708e-06,
+      "loss": 3.467,
+      "step": 10205
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518520859671302e-06,
+      "loss": 3.2824,
+      "step": 10206
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5184734513274338e-06,
+      "loss": 3.4205,
+      "step": 10207
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518426042983565e-06,
+      "loss": 3.1629,
+      "step": 10208
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5183786346396963e-06,
+      "loss": 3.1244,
+      "step": 10209
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518331226295828e-06,
+      "loss": 3.2178,
+      "step": 10210
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5182838179519597e-06,
+      "loss": 3.3324,
+      "step": 10211
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518236409608091e-06,
+      "loss": 2.6888,
+      "step": 10212
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5181890012642227e-06,
+      "loss": 3.2287,
+      "step": 10213
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518141592920354e-06,
+      "loss": 3.4002,
+      "step": 10214
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5180941845764853e-06,
+      "loss": 3.8604,
+      "step": 10215
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.518046776232617e-06,
+      "loss": 3.9693,
+      "step": 10216
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5179993678887487e-06,
+      "loss": 3.1612,
+      "step": 10217
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.51795195954488e-06,
+      "loss": 3.4333,
+      "step": 10218
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5179045512010117e-06,
+      "loss": 3.3768,
+      "step": 10219
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.517857142857143e-06,
+      "loss": 3.9311,
+      "step": 10220
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5178097345132742e-06,
+      "loss": 3.8999,
+      "step": 10221
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5177623261694055e-06,
+      "loss": 3.0567,
+      "step": 10222
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5177149178255372e-06,
+      "loss": 2.9187,
+      "step": 10223
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.517667509481669e-06,
+      "loss": 3.0833,
+      "step": 10224
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.5176201011378e-06,
+      "loss": 3.2242,
+      "step": 10225
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.517572692793932e-06,
+      "loss": 3.6899,
+      "step": 10226
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.517525284450063e-06,
+      "loss": 3.0698,
+      "step": 10227
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5174778761061945e-06,
+      "loss": 3.4973,
+      "step": 10228
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517430467762326e-06,
+      "loss": 3.9351,
+      "step": 10229
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517383059418458e-06,
+      "loss": 2.8213,
+      "step": 10230
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517335651074589e-06,
+      "loss": 4.4898,
+      "step": 10231
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517288242730721e-06,
+      "loss": 3.0333,
+      "step": 10232
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517240834386852e-06,
+      "loss": 4.1067,
+      "step": 10233
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5171934260429834e-06,
+      "loss": 3.2038,
+      "step": 10234
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517146017699115e-06,
+      "loss": 2.9809,
+      "step": 10235
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5170986093552464e-06,
+      "loss": 3.2747,
+      "step": 10236
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.517051201011378e-06,
+      "loss": 3.8144,
+      "step": 10237
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.51700379266751e-06,
+      "loss": 3.7139,
+      "step": 10238
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.516956384323641e-06,
+      "loss": 3.9272,
+      "step": 10239
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5169089759797724e-06,
+      "loss": 3.114,
+      "step": 10240
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.516861567635904e-06,
+      "loss": 3.6065,
+      "step": 10241
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5168141592920354e-06,
+      "loss": 3.1407,
+      "step": 10242
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.516766750948167e-06,
+      "loss": 3.5704,
+      "step": 10243
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5167193426042983e-06,
+      "loss": 3.5794,
+      "step": 10244
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.51667193426043e-06,
+      "loss": 3.8365,
+      "step": 10245
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5166245259165613e-06,
+      "loss": 3.9902,
+      "step": 10246
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5165771175726926e-06,
+      "loss": 4.111,
+      "step": 10247
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5165297092288243e-06,
+      "loss": 2.6765,
+      "step": 10248
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5164823008849556e-06,
+      "loss": 3.4144,
+      "step": 10249
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5164348925410873e-06,
+      "loss": 3.7338,
+      "step": 10250
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.516387484197219e-06,
+      "loss": 3.6279,
+      "step": 10251
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5163400758533503e-06,
+      "loss": 3.3708,
+      "step": 10252
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5162926675094816e-06,
+      "loss": 2.9693,
+      "step": 10253
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5162452591656133e-06,
+      "loss": 2.951,
+      "step": 10254
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5161978508217445e-06,
+      "loss": 3.8871,
+      "step": 10255
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5161504424778762e-06,
+      "loss": 3.1078,
+      "step": 10256
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.516103034134008e-06,
+      "loss": 3.8965,
+      "step": 10257
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5160556257901392e-06,
+      "loss": 2.9817,
+      "step": 10258
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5160082174462705e-06,
+      "loss": 3.8958,
+      "step": 10259
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5159608091024022e-06,
+      "loss": 2.2966,
+      "step": 10260
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5159134007585335e-06,
+      "loss": 3.238,
+      "step": 10261
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5158659924146648e-06,
+      "loss": 3.2757,
+      "step": 10262
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5158185840707965e-06,
+      "loss": 3.487,
+      "step": 10263
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.515771175726928e-06,
+      "loss": 3.3491,
+      "step": 10264
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5157237673830595e-06,
+      "loss": 2.8906,
+      "step": 10265
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5156763590391907e-06,
+      "loss": 3.7495,
+      "step": 10266
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5156289506953224e-06,
+      "loss": 2.4629,
+      "step": 10267
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5155815423514537e-06,
+      "loss": 3.3335,
+      "step": 10268
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5155341340075854e-06,
+      "loss": 3.9924,
+      "step": 10269
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.515486725663717e-06,
+      "loss": 3.2817,
+      "step": 10270
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5154393173198484e-06,
+      "loss": 4.0834,
+      "step": 10271
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5153919089759797e-06,
+      "loss": 3.7996,
+      "step": 10272
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5153445006321114e-06,
+      "loss": 4.4289,
+      "step": 10273
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5152970922882427e-06,
+      "loss": 3.2121,
+      "step": 10274
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.515249683944374e-06,
+      "loss": 3.7196,
+      "step": 10275
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5152022756005057e-06,
+      "loss": 2.952,
+      "step": 10276
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5151548672566374e-06,
+      "loss": 3.5079,
+      "step": 10277
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5151074589127687e-06,
+      "loss": 3.4556,
+      "step": 10278
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5150600505689004e-06,
+      "loss": 3.6393,
+      "step": 10279
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5150126422250316e-06,
+      "loss": 3.6017,
+      "step": 10280
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.514965233881163e-06,
+      "loss": 3.3342,
+      "step": 10281
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5149178255372946e-06,
+      "loss": 4.0091,
+      "step": 10282
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5148704171934263e-06,
+      "loss": 3.3239,
+      "step": 10283
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5148230088495576e-06,
+      "loss": 4.1911,
+      "step": 10284
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5147756005056893e-06,
+      "loss": 3.5721,
+      "step": 10285
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5147281921618206e-06,
+      "loss": 3.3372,
+      "step": 10286
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.514680783817952e-06,
+      "loss": 3.963,
+      "step": 10287
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.514633375474083e-06,
+      "loss": 3.4678,
+      "step": 10288
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.514585967130215e-06,
+      "loss": 4.4331,
+      "step": 10289
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.5145385587863466e-06,
+      "loss": 3.2657,
+      "step": 10290
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.514491150442478e-06,
+      "loss": 3.3022,
+      "step": 10291
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5144437420986095e-06,
+      "loss": 3.7976,
+      "step": 10292
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.514396333754741e-06,
+      "loss": 3.7433,
+      "step": 10293
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.514348925410872e-06,
+      "loss": 2.8179,
+      "step": 10294
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.514301517067004e-06,
+      "loss": 3.2501,
+      "step": 10295
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5142541087231355e-06,
+      "loss": 4.0869,
+      "step": 10296
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.514206700379267e-06,
+      "loss": 3.0142,
+      "step": 10297
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5141592920353985e-06,
+      "loss": 3.4809,
+      "step": 10298
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5141118836915298e-06,
+      "loss": 3.4512,
+      "step": 10299
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.514064475347661e-06,
+      "loss": 3.1995,
+      "step": 10300
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5140170670037928e-06,
+      "loss": 3.6393,
+      "step": 10301
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513969658659924e-06,
+      "loss": 3.0295,
+      "step": 10302
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5139222503160557e-06,
+      "loss": 2.9889,
+      "step": 10303
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5138748419721874e-06,
+      "loss": 3.9986,
+      "step": 10304
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5138274336283187e-06,
+      "loss": 2.8053,
+      "step": 10305
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.51378002528445e-06,
+      "loss": 3.7725,
+      "step": 10306
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5137326169405817e-06,
+      "loss": 2.966,
+      "step": 10307
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513685208596713e-06,
+      "loss": 3.2954,
+      "step": 10308
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5136378002528447e-06,
+      "loss": 4.1515,
+      "step": 10309
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513590391908976e-06,
+      "loss": 3.5814,
+      "step": 10310
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5135429835651077e-06,
+      "loss": 4.1327,
+      "step": 10311
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513495575221239e-06,
+      "loss": 3.7028,
+      "step": 10312
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5134481668773702e-06,
+      "loss": 3.0697,
+      "step": 10313
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513400758533502e-06,
+      "loss": 3.4181,
+      "step": 10314
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5133533501896332e-06,
+      "loss": 2.685,
+      "step": 10315
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513305941845765e-06,
+      "loss": 4.1402,
+      "step": 10316
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5132585335018966e-06,
+      "loss": 2.9879,
+      "step": 10317
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513211125158028e-06,
+      "loss": 4.1267,
+      "step": 10318
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513163716814159e-06,
+      "loss": 3.0476,
+      "step": 10319
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513116308470291e-06,
+      "loss": 2.0807,
+      "step": 10320
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513068900126422e-06,
+      "loss": 3.918,
+      "step": 10321
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.513021491782554e-06,
+      "loss": 3.1228,
+      "step": 10322
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5129740834386856e-06,
+      "loss": 3.1001,
+      "step": 10323
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512926675094817e-06,
+      "loss": 4.2825,
+      "step": 10324
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512879266750948e-06,
+      "loss": 3.8256,
+      "step": 10325
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.51283185840708e-06,
+      "loss": 2.8668,
+      "step": 10326
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512784450063211e-06,
+      "loss": 3.3256,
+      "step": 10327
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5127370417193424e-06,
+      "loss": 3.7264,
+      "step": 10328
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512689633375474e-06,
+      "loss": 3.0309,
+      "step": 10329
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512642225031606e-06,
+      "loss": 4.1033,
+      "step": 10330
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512594816687737e-06,
+      "loss": 3.505,
+      "step": 10331
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5125474083438684e-06,
+      "loss": 4.3037,
+      "step": 10332
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5125e-06,
+      "loss": 3.5069,
+      "step": 10333
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5124525916561314e-06,
+      "loss": 3.7388,
+      "step": 10334
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512405183312263e-06,
+      "loss": 2.9524,
+      "step": 10335
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5123577749683948e-06,
+      "loss": 3.2315,
+      "step": 10336
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512310366624526e-06,
+      "loss": 2.4542,
+      "step": 10337
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5122629582806573e-06,
+      "loss": 3.747,
+      "step": 10338
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512215549936789e-06,
+      "loss": 3.2377,
+      "step": 10339
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5121681415929203e-06,
+      "loss": 3.5547,
+      "step": 10340
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5121207332490516e-06,
+      "loss": 2.7294,
+      "step": 10341
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5120733249051833e-06,
+      "loss": 3.7958,
+      "step": 10342
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.512025916561315e-06,
+      "loss": 2.8115,
+      "step": 10343
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5119785082174463e-06,
+      "loss": 3.2546,
+      "step": 10344
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.511931099873578e-06,
+      "loss": 3.2355,
+      "step": 10345
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5118836915297093e-06,
+      "loss": 3.2657,
+      "step": 10346
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5118362831858406e-06,
+      "loss": 3.5367,
+      "step": 10347
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5117888748419723e-06,
+      "loss": 3.2116,
+      "step": 10348
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.511741466498104e-06,
+      "loss": 3.5572,
+      "step": 10349
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5116940581542352e-06,
+      "loss": 3.7718,
+      "step": 10350
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.511646649810367e-06,
+      "loss": 2.8013,
+      "step": 10351
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5115992414664982e-06,
+      "loss": 3.5872,
+      "step": 10352
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5115518331226295e-06,
+      "loss": 4.1072,
+      "step": 10353
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.5115044247787608e-06,
+      "loss": 4.0642,
+      "step": 10354
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5114570164348925e-06,
+      "loss": 2.8775,
+      "step": 10355
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.511409608091024e-06,
+      "loss": 3.5191,
+      "step": 10356
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5113621997471555e-06,
+      "loss": 4.131,
+      "step": 10357
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.511314791403287e-06,
+      "loss": 2.9787,
+      "step": 10358
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5112673830594185e-06,
+      "loss": 3.0043,
+      "step": 10359
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5112199747155497e-06,
+      "loss": 3.5921,
+      "step": 10360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5111725663716814e-06,
+      "loss": 3.8365,
+      "step": 10361
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.511125158027813e-06,
+      "loss": 3.5773,
+      "step": 10362
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5110777496839444e-06,
+      "loss": 3.3531,
+      "step": 10363
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.511030341340076e-06,
+      "loss": 4.1148,
+      "step": 10364
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5109829329962074e-06,
+      "loss": 3.182,
+      "step": 10365
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5109355246523387e-06,
+      "loss": 3.2034,
+      "step": 10366
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5108881163084704e-06,
+      "loss": 3.4316,
+      "step": 10367
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5108407079646017e-06,
+      "loss": 3.9416,
+      "step": 10368
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5107932996207334e-06,
+      "loss": 3.679,
+      "step": 10369
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.510745891276865e-06,
+      "loss": 4.0361,
+      "step": 10370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5106984829329964e-06,
+      "loss": 4.1863,
+      "step": 10371
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5106510745891276e-06,
+      "loss": 3.0424,
+      "step": 10372
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5106036662452593e-06,
+      "loss": 2.9493,
+      "step": 10373
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5105562579013906e-06,
+      "loss": 4.134,
+      "step": 10374
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5105088495575223e-06,
+      "loss": 3.3613,
+      "step": 10375
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5104614412136536e-06,
+      "loss": 2.7856,
+      "step": 10376
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5104140328697853e-06,
+      "loss": 2.8025,
+      "step": 10377
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5103666245259166e-06,
+      "loss": 3.7171,
+      "step": 10378
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.510319216182048e-06,
+      "loss": 3.1128,
+      "step": 10379
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5102718078381796e-06,
+      "loss": 3.22,
+      "step": 10380
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.510224399494311e-06,
+      "loss": 3.5224,
+      "step": 10381
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5101769911504426e-06,
+      "loss": 3.3687,
+      "step": 10382
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5101295828065743e-06,
+      "loss": 3.7766,
+      "step": 10383
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5100821744627056e-06,
+      "loss": 3.4275,
+      "step": 10384
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.510034766118837e-06,
+      "loss": 3.7833,
+      "step": 10385
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5099873577749685e-06,
+      "loss": 3.7031,
+      "step": 10386
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5099399494311e-06,
+      "loss": 3.4304,
+      "step": 10387
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5098925410872315e-06,
+      "loss": 2.7394,
+      "step": 10388
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5098451327433632e-06,
+      "loss": 3.7276,
+      "step": 10389
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5097977243994945e-06,
+      "loss": 2.8882,
+      "step": 10390
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5097503160556258e-06,
+      "loss": 3.9267,
+      "step": 10391
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5097029077117575e-06,
+      "loss": 4.0093,
+      "step": 10392
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5096554993678888e-06,
+      "loss": 2.9642,
+      "step": 10393
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.50960809102402e-06,
+      "loss": 4.0892,
+      "step": 10394
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5095606826801518e-06,
+      "loss": 3.3999,
+      "step": 10395
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5095132743362835e-06,
+      "loss": 3.2182,
+      "step": 10396
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5094658659924147e-06,
+      "loss": 3.0443,
+      "step": 10397
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.509418457648546e-06,
+      "loss": 3.3899,
+      "step": 10398
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5093710493046777e-06,
+      "loss": 3.288,
+      "step": 10399
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.509323640960809e-06,
+      "loss": 3.3006,
+      "step": 10400
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5092762326169407e-06,
+      "loss": 3.4993,
+      "step": 10401
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5092288242730724e-06,
+      "loss": 3.1,
+      "step": 10402
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5091814159292037e-06,
+      "loss": 2.4215,
+      "step": 10403
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.509134007585335e-06,
+      "loss": 3.8838,
+      "step": 10404
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5090865992414667e-06,
+      "loss": 3.7755,
+      "step": 10405
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.509039190897598e-06,
+      "loss": 2.7254,
+      "step": 10406
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5089917825537292e-06,
+      "loss": 2.6581,
+      "step": 10407
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.508944374209861e-06,
+      "loss": 3.0315,
+      "step": 10408
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5088969658659926e-06,
+      "loss": 3.4712,
+      "step": 10409
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.508849557522124e-06,
+      "loss": 3.2377,
+      "step": 10410
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5088021491782556e-06,
+      "loss": 3.8862,
+      "step": 10411
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.508754740834387e-06,
+      "loss": 3.7242,
+      "step": 10412
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.508707332490518e-06,
+      "loss": 3.7209,
+      "step": 10413
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.50865992414665e-06,
+      "loss": 4.0729,
+      "step": 10414
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5086125158027816e-06,
+      "loss": 3.3383,
+      "step": 10415
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.508565107458913e-06,
+      "loss": 3.4622,
+      "step": 10416
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.5085176991150446e-06,
+      "loss": 3.2414,
+      "step": 10417
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508470290771176e-06,
+      "loss": 3.1678,
+      "step": 10418
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508422882427307e-06,
+      "loss": 3.3714,
+      "step": 10419
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5083754740834384e-06,
+      "loss": 3.8122,
+      "step": 10420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.50832806573957e-06,
+      "loss": 3.4164,
+      "step": 10421
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508280657395702e-06,
+      "loss": 3.8373,
+      "step": 10422
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508233249051833e-06,
+      "loss": 3.546,
+      "step": 10423
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508185840707965e-06,
+      "loss": 3.8113,
+      "step": 10424
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508138432364096e-06,
+      "loss": 4.113,
+      "step": 10425
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5080910240202274e-06,
+      "loss": 3.5965,
+      "step": 10426
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.508043615676359e-06,
+      "loss": 3.3465,
+      "step": 10427
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5079962073324908e-06,
+      "loss": 4.3054,
+      "step": 10428
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507948798988622e-06,
+      "loss": 3.0783,
+      "step": 10429
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5079013906447538e-06,
+      "loss": 3.8581,
+      "step": 10430
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507853982300885e-06,
+      "loss": 3.9262,
+      "step": 10431
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5078065739570163e-06,
+      "loss": 2.4216,
+      "step": 10432
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507759165613148e-06,
+      "loss": 3.4562,
+      "step": 10433
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5077117572692793e-06,
+      "loss": 3.5786,
+      "step": 10434
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507664348925411e-06,
+      "loss": 2.448,
+      "step": 10435
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5076169405815427e-06,
+      "loss": 3.8131,
+      "step": 10436
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507569532237674e-06,
+      "loss": 3.2525,
+      "step": 10437
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5075221238938053e-06,
+      "loss": 3.0592,
+      "step": 10438
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507474715549937e-06,
+      "loss": 3.1371,
+      "step": 10439
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5074273072060683e-06,
+      "loss": 3.3733,
+      "step": 10440
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5073798988622e-06,
+      "loss": 3.2444,
+      "step": 10441
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5073324905183312e-06,
+      "loss": 3.3681,
+      "step": 10442
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507285082174463e-06,
+      "loss": 3.6889,
+      "step": 10443
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5072376738305942e-06,
+      "loss": 3.8103,
+      "step": 10444
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5071902654867255e-06,
+      "loss": 3.8708,
+      "step": 10445
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5071428571428572e-06,
+      "loss": 3.4118,
+      "step": 10446
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5070954487989885e-06,
+      "loss": 3.2057,
+      "step": 10447
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.50704804045512e-06,
+      "loss": 3.3207,
+      "step": 10448
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.507000632111252e-06,
+      "loss": 3.8786,
+      "step": 10449
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506953223767383e-06,
+      "loss": 3.3153,
+      "step": 10450
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5069058154235145e-06,
+      "loss": 3.5688,
+      "step": 10451
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506858407079646e-06,
+      "loss": 3.5142,
+      "step": 10452
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5068109987357775e-06,
+      "loss": 4.2369,
+      "step": 10453
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5067635903919087e-06,
+      "loss": 3.9013,
+      "step": 10454
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506716182048041e-06,
+      "loss": 3.0857,
+      "step": 10455
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506668773704172e-06,
+      "loss": 4.0282,
+      "step": 10456
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5066213653603034e-06,
+      "loss": 3.4698,
+      "step": 10457
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506573957016435e-06,
+      "loss": 2.9268,
+      "step": 10458
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5065265486725664e-06,
+      "loss": 3.9228,
+      "step": 10459
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5064791403286977e-06,
+      "loss": 4.1564,
+      "step": 10460
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5064317319848294e-06,
+      "loss": 3.943,
+      "step": 10461
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506384323640961e-06,
+      "loss": 1.9981,
+      "step": 10462
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5063369152970924e-06,
+      "loss": 3.574,
+      "step": 10463
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5062895069532237e-06,
+      "loss": 3.6256,
+      "step": 10464
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5062420986093554e-06,
+      "loss": 3.1327,
+      "step": 10465
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5061946902654866e-06,
+      "loss": 3.3611,
+      "step": 10466
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.506147281921618e-06,
+      "loss": 3.041,
+      "step": 10467
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5060998735777496e-06,
+      "loss": 3.3708,
+      "step": 10468
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5060524652338813e-06,
+      "loss": 4.5127,
+      "step": 10469
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5060050568900126e-06,
+      "loss": 3.8759,
+      "step": 10470
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5059576485461443e-06,
+      "loss": 3.0588,
+      "step": 10471
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5059102402022756e-06,
+      "loss": 3.2277,
+      "step": 10472
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.505862831858407e-06,
+      "loss": 3.131,
+      "step": 10473
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5058154235145386e-06,
+      "loss": 3.4774,
+      "step": 10474
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5057680151706703e-06,
+      "loss": 3.61,
+      "step": 10475
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5057206068268016e-06,
+      "loss": 3.5971,
+      "step": 10476
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5056731984829333e-06,
+      "loss": 3.3259,
+      "step": 10477
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5056257901390645e-06,
+      "loss": 3.1376,
+      "step": 10478
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.505578381795196e-06,
+      "loss": 4.0069,
+      "step": 10479
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.5055309734513275e-06,
+      "loss": 3.3616,
+      "step": 10480
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.505483565107459e-06,
+      "loss": 3.0937,
+      "step": 10481
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5054361567635905e-06,
+      "loss": 2.9903,
+      "step": 10482
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5053887484197222e-06,
+      "loss": 3.6142,
+      "step": 10483
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5053413400758535e-06,
+      "loss": 4.071,
+      "step": 10484
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5052939317319848e-06,
+      "loss": 3.2665,
+      "step": 10485
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.505246523388116e-06,
+      "loss": 3.2971,
+      "step": 10486
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5051991150442478e-06,
+      "loss": 3.7806,
+      "step": 10487
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5051517067003795e-06,
+      "loss": 3.4303,
+      "step": 10488
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5051042983565107e-06,
+      "loss": 3.8366,
+      "step": 10489
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5050568900126425e-06,
+      "loss": 3.839,
+      "step": 10490
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5050094816687737e-06,
+      "loss": 3.5283,
+      "step": 10491
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504962073324905e-06,
+      "loss": 3.5011,
+      "step": 10492
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5049146649810367e-06,
+      "loss": 3.6913,
+      "step": 10493
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504867256637168e-06,
+      "loss": 3.0133,
+      "step": 10494
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5048198482932997e-06,
+      "loss": 3.6276,
+      "step": 10495
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5047724399494314e-06,
+      "loss": 4.2372,
+      "step": 10496
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5047250316055627e-06,
+      "loss": 3.1996,
+      "step": 10497
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504677623261694e-06,
+      "loss": 3.5974,
+      "step": 10498
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5046302149178257e-06,
+      "loss": 3.2973,
+      "step": 10499
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504582806573957e-06,
+      "loss": 3.6568,
+      "step": 10500
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5045353982300887e-06,
+      "loss": 3.5881,
+      "step": 10501
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5044879898862204e-06,
+      "loss": 3.0338,
+      "step": 10502
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5044405815423516e-06,
+      "loss": 2.3769,
+      "step": 10503
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504393173198483e-06,
+      "loss": 3.0723,
+      "step": 10504
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5043457648546146e-06,
+      "loss": 2.8575,
+      "step": 10505
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504298356510746e-06,
+      "loss": 2.8994,
+      "step": 10506
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504250948166877e-06,
+      "loss": 4.0558,
+      "step": 10507
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504203539823009e-06,
+      "loss": 3.4457,
+      "step": 10508
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5041561314791406e-06,
+      "loss": 4.1293,
+      "step": 10509
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504108723135272e-06,
+      "loss": 2.5853,
+      "step": 10510
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504061314791403e-06,
+      "loss": 3.2819,
+      "step": 10511
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.504013906447535e-06,
+      "loss": 3.2623,
+      "step": 10512
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503966498103666e-06,
+      "loss": 3.2497,
+      "step": 10513
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503919089759798e-06,
+      "loss": 2.9497,
+      "step": 10514
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5038716814159295e-06,
+      "loss": 2.7693,
+      "step": 10515
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503824273072061e-06,
+      "loss": 3.5302,
+      "step": 10516
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503776864728192e-06,
+      "loss": 3.6692,
+      "step": 10517
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503729456384324e-06,
+      "loss": 3.3533,
+      "step": 10518
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503682048040455e-06,
+      "loss": 3.1906,
+      "step": 10519
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5036346396965864e-06,
+      "loss": 2.9635,
+      "step": 10520
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503587231352718e-06,
+      "loss": 3.988,
+      "step": 10521
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5035398230088498e-06,
+      "loss": 3.5683,
+      "step": 10522
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503492414664981e-06,
+      "loss": 3.7112,
+      "step": 10523
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5034450063211128e-06,
+      "loss": 3.6473,
+      "step": 10524
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503397597977244e-06,
+      "loss": 3.8985,
+      "step": 10525
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5033501896333753e-06,
+      "loss": 3.8374,
+      "step": 10526
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503302781289507e-06,
+      "loss": 3.9139,
+      "step": 10527
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5032553729456387e-06,
+      "loss": 4.1669,
+      "step": 10528
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.50320796460177e-06,
+      "loss": 4.184,
+      "step": 10529
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5031605562579013e-06,
+      "loss": 3.4735,
+      "step": 10530
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.503113147914033e-06,
+      "loss": 3.4802,
+      "step": 10531
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5030657395701643e-06,
+      "loss": 3.8486,
+      "step": 10532
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5030183312262956e-06,
+      "loss": 3.232,
+      "step": 10533
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5029709228824273e-06,
+      "loss": 3.9344,
+      "step": 10534
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.502923514538559e-06,
+      "loss": 2.768,
+      "step": 10535
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5028761061946902e-06,
+      "loss": 3.4126,
+      "step": 10536
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.502828697850822e-06,
+      "loss": 2.9475,
+      "step": 10537
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5027812895069532e-06,
+      "loss": 2.5417,
+      "step": 10538
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5027338811630845e-06,
+      "loss": 3.7365,
+      "step": 10539
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.5026864728192162e-06,
+      "loss": 3.8255,
+      "step": 10540
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.502639064475348e-06,
+      "loss": 3.2357,
+      "step": 10541
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.502591656131479e-06,
+      "loss": 3.3133,
+      "step": 10542
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.502544247787611e-06,
+      "loss": 4.0182,
+      "step": 10543
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.502496839443742e-06,
+      "loss": 3.8861,
+      "step": 10544
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5024494310998735e-06,
+      "loss": 3.9265,
+      "step": 10545
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.502402022756005e-06,
+      "loss": 4.1814,
+      "step": 10546
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5023546144121364e-06,
+      "loss": 3.1054,
+      "step": 10547
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.502307206068268e-06,
+      "loss": 2.8615,
+      "step": 10548
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5022597977244e-06,
+      "loss": 3.4373,
+      "step": 10549
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.502212389380531e-06,
+      "loss": 3.51,
+      "step": 10550
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5021649810366624e-06,
+      "loss": 3.5894,
+      "step": 10551
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5021175726927937e-06,
+      "loss": 2.7139,
+      "step": 10552
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5020701643489254e-06,
+      "loss": 3.8582,
+      "step": 10553
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.502022756005057e-06,
+      "loss": 2.6611,
+      "step": 10554
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5019753476611884e-06,
+      "loss": 3.5931,
+      "step": 10555
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.50192793931732e-06,
+      "loss": 2.4938,
+      "step": 10556
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5018805309734514e-06,
+      "loss": 3.8075,
+      "step": 10557
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5018331226295826e-06,
+      "loss": 4.0233,
+      "step": 10558
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5017857142857144e-06,
+      "loss": 3.4976,
+      "step": 10559
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5017383059418456e-06,
+      "loss": 3.4138,
+      "step": 10560
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5016908975979773e-06,
+      "loss": 3.3238,
+      "step": 10561
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.501643489254109e-06,
+      "loss": 3.5241,
+      "step": 10562
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5015960809102403e-06,
+      "loss": 3.0528,
+      "step": 10563
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5015486725663716e-06,
+      "loss": 3.6874,
+      "step": 10564
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5015012642225033e-06,
+      "loss": 4.2922,
+      "step": 10565
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5014538558786346e-06,
+      "loss": 3.2762,
+      "step": 10566
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5014064475347663e-06,
+      "loss": 2.9076,
+      "step": 10567
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.501359039190898e-06,
+      "loss": 3.4143,
+      "step": 10568
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5013116308470293e-06,
+      "loss": 3.2765,
+      "step": 10569
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5012642225031606e-06,
+      "loss": 3.2625,
+      "step": 10570
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5012168141592923e-06,
+      "loss": 3.9076,
+      "step": 10571
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5011694058154235e-06,
+      "loss": 3.3076,
+      "step": 10572
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.501121997471555e-06,
+      "loss": 3.8275,
+      "step": 10573
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5010745891276865e-06,
+      "loss": 3.467,
+      "step": 10574
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5010271807838182e-06,
+      "loss": 2.9262,
+      "step": 10575
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5009797724399495e-06,
+      "loss": 3.4108,
+      "step": 10576
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5009323640960808e-06,
+      "loss": 3.5855,
+      "step": 10577
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5008849557522125e-06,
+      "loss": 3.2405,
+      "step": 10578
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5008375474083438e-06,
+      "loss": 4.217,
+      "step": 10579
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5007901390644755e-06,
+      "loss": 2.8064,
+      "step": 10580
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.500742730720607e-06,
+      "loss": 3.2417,
+      "step": 10581
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5006953223767385e-06,
+      "loss": 3.6901,
+      "step": 10582
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5006479140328697e-06,
+      "loss": 3.8988,
+      "step": 10583
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5006005056890014e-06,
+      "loss": 3.514,
+      "step": 10584
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5005530973451327e-06,
+      "loss": 3.2221,
+      "step": 10585
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.500505689001264e-06,
+      "loss": 3.7452,
+      "step": 10586
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5004582806573957e-06,
+      "loss": 3.4919,
+      "step": 10587
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5004108723135274e-06,
+      "loss": 3.8176,
+      "step": 10588
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5003634639696587e-06,
+      "loss": 4.2656,
+      "step": 10589
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5003160556257904e-06,
+      "loss": 2.8918,
+      "step": 10590
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5002686472819217e-06,
+      "loss": 3.4067,
+      "step": 10591
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.500221238938053e-06,
+      "loss": 3.3078,
+      "step": 10592
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5001738305941847e-06,
+      "loss": 3.7876,
+      "step": 10593
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5001264222503164e-06,
+      "loss": 4.2025,
+      "step": 10594
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.5000790139064476e-06,
+      "loss": 3.0175,
+      "step": 10595
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.500031605562579e-06,
+      "loss": 2.6753,
+      "step": 10596
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.4999841972187106e-06,
+      "loss": 2.9987,
+      "step": 10597
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499936788874842e-06,
+      "loss": 3.3523,
+      "step": 10598
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499889380530973e-06,
+      "loss": 3.5275,
+      "step": 10599
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499841972187105e-06,
+      "loss": 2.6655,
+      "step": 10600
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.4997945638432366e-06,
+      "loss": 2.5653,
+      "step": 10601
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499747155499368e-06,
+      "loss": 3.696,
+      "step": 10602
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.4996997471554996e-06,
+      "loss": 2.7107,
+      "step": 10603
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499652338811631e-06,
+      "loss": 2.7804,
+      "step": 10604
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499604930467762e-06,
+      "loss": 3.8426,
+      "step": 10605
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.499557522123894e-06,
+      "loss": 3.1828,
+      "step": 10606
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.4995101137800256e-06,
+      "loss": 3.6986,
+      "step": 10607
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.499462705436157e-06,
+      "loss": 3.4868,
+      "step": 10608
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4994152970922885e-06,
+      "loss": 3.8614,
+      "step": 10609
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.49936788874842e-06,
+      "loss": 3.5626,
+      "step": 10610
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.499320480404551e-06,
+      "loss": 3.5741,
+      "step": 10611
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.499273072060683e-06,
+      "loss": 2.9971,
+      "step": 10612
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.499225663716814e-06,
+      "loss": 2.6851,
+      "step": 10613
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4991782553729458e-06,
+      "loss": 2.8462,
+      "step": 10614
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.499130847029077e-06,
+      "loss": 3.2367,
+      "step": 10615
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4990834386852088e-06,
+      "loss": 3.8617,
+      "step": 10616
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.49903603034134e-06,
+      "loss": 3.2158,
+      "step": 10617
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4989886219974713e-06,
+      "loss": 4.0749,
+      "step": 10618
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498941213653603e-06,
+      "loss": 4.0209,
+      "step": 10619
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4988938053097347e-06,
+      "loss": 2.9215,
+      "step": 10620
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498846396965866e-06,
+      "loss": 2.943,
+      "step": 10621
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4987989886219977e-06,
+      "loss": 3.0216,
+      "step": 10622
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498751580278129e-06,
+      "loss": 2.5978,
+      "step": 10623
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4987041719342603e-06,
+      "loss": 2.4449,
+      "step": 10624
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498656763590392e-06,
+      "loss": 3.5398,
+      "step": 10625
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4986093552465233e-06,
+      "loss": 3.839,
+      "step": 10626
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498561946902655e-06,
+      "loss": 3.4843,
+      "step": 10627
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4985145385587867e-06,
+      "loss": 3.5398,
+      "step": 10628
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498467130214918e-06,
+      "loss": 4.0382,
+      "step": 10629
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4984197218710492e-06,
+      "loss": 3.6427,
+      "step": 10630
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498372313527181e-06,
+      "loss": 3.9823,
+      "step": 10631
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4983249051833122e-06,
+      "loss": 3.1292,
+      "step": 10632
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498277496839444e-06,
+      "loss": 3.8888,
+      "step": 10633
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4982300884955756e-06,
+      "loss": 4.3574,
+      "step": 10634
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498182680151707e-06,
+      "loss": 3.7332,
+      "step": 10635
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498135271807838e-06,
+      "loss": 3.5037,
+      "step": 10636
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4980878634639695e-06,
+      "loss": 3.7755,
+      "step": 10637
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.498040455120101e-06,
+      "loss": 3.9344,
+      "step": 10638
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4979930467762325e-06,
+      "loss": 3.3789,
+      "step": 10639
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497945638432364e-06,
+      "loss": 3.7662,
+      "step": 10640
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497898230088496e-06,
+      "loss": 3.1287,
+      "step": 10641
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497850821744627e-06,
+      "loss": 2.5703,
+      "step": 10642
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4978034134007584e-06,
+      "loss": 3.6318,
+      "step": 10643
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.49775600505689e-06,
+      "loss": 2.6421,
+      "step": 10644
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4977085967130214e-06,
+      "loss": 4.1634,
+      "step": 10645
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497661188369153e-06,
+      "loss": 4.1142,
+      "step": 10646
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497613780025285e-06,
+      "loss": 3.0412,
+      "step": 10647
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497566371681416e-06,
+      "loss": 3.0605,
+      "step": 10648
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4975189633375474e-06,
+      "loss": 3.2377,
+      "step": 10649
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497471554993679e-06,
+      "loss": 3.0638,
+      "step": 10650
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4974241466498104e-06,
+      "loss": 2.867,
+      "step": 10651
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4973767383059416e-06,
+      "loss": 3.798,
+      "step": 10652
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4973293299620733e-06,
+      "loss": 3.331,
+      "step": 10653
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497281921618205e-06,
+      "loss": 3.6362,
+      "step": 10654
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4972345132743363e-06,
+      "loss": 3.5669,
+      "step": 10655
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.497187104930468e-06,
+      "loss": 3.5368,
+      "step": 10656
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4971396965865993e-06,
+      "loss": 4.05,
+      "step": 10657
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4970922882427306e-06,
+      "loss": 3.4528,
+      "step": 10658
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4970448798988623e-06,
+      "loss": 3.6741,
+      "step": 10659
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.496997471554994e-06,
+      "loss": 3.4658,
+      "step": 10660
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4969500632111253e-06,
+      "loss": 3.8012,
+      "step": 10661
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4969026548672566e-06,
+      "loss": 3.8732,
+      "step": 10662
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4968552465233883e-06,
+      "loss": 2.4531,
+      "step": 10663
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4968078381795195e-06,
+      "loss": 2.8963,
+      "step": 10664
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.496760429835651e-06,
+      "loss": 3.4606,
+      "step": 10665
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4967130214917825e-06,
+      "loss": 2.1117,
+      "step": 10666
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4966656131479142e-06,
+      "loss": 3.6212,
+      "step": 10667
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4966182048040455e-06,
+      "loss": 3.6179,
+      "step": 10668
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4965707964601772e-06,
+      "loss": 2.8793,
+      "step": 10669
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4965233881163085e-06,
+      "loss": 3.7868,
+      "step": 10670
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.4964759797724398e-06,
+      "loss": 2.6166,
+      "step": 10671
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4964285714285715e-06,
+      "loss": 3.395,
+      "step": 10672
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.496381163084703e-06,
+      "loss": 3.8499,
+      "step": 10673
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4963337547408345e-06,
+      "loss": 2.9683,
+      "step": 10674
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.496286346396966e-06,
+      "loss": 2.7094,
+      "step": 10675
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4962389380530975e-06,
+      "loss": 3.416,
+      "step": 10676
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4961915297092287e-06,
+      "loss": 3.9253,
+      "step": 10677
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4961441213653604e-06,
+      "loss": 2.8722,
+      "step": 10678
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4960967130214917e-06,
+      "loss": 3.9351,
+      "step": 10679
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4960493046776234e-06,
+      "loss": 2.9183,
+      "step": 10680
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4960018963337547e-06,
+      "loss": 3.4768,
+      "step": 10681
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4959544879898864e-06,
+      "loss": 3.8475,
+      "step": 10682
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4959070796460177e-06,
+      "loss": 3.8727,
+      "step": 10683
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.495859671302149e-06,
+      "loss": 2.6172,
+      "step": 10684
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4958122629582807e-06,
+      "loss": 3.706,
+      "step": 10685
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4957648546144124e-06,
+      "loss": 2.9965,
+      "step": 10686
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4957174462705437e-06,
+      "loss": 3.2152,
+      "step": 10687
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4956700379266754e-06,
+      "loss": 2.9138,
+      "step": 10688
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4956226295828066e-06,
+      "loss": 2.8422,
+      "step": 10689
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.495575221238938e-06,
+      "loss": 3.945,
+      "step": 10690
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4955278128950696e-06,
+      "loss": 3.4401,
+      "step": 10691
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.495480404551201e-06,
+      "loss": 3.688,
+      "step": 10692
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4954329962073326e-06,
+      "loss": 3.3526,
+      "step": 10693
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4953855878634643e-06,
+      "loss": 4.0429,
+      "step": 10694
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4953381795195956e-06,
+      "loss": 2.873,
+      "step": 10695
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.495290771175727e-06,
+      "loss": 3.8262,
+      "step": 10696
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4952433628318586e-06,
+      "loss": 2.4792,
+      "step": 10697
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.49519595448799e-06,
+      "loss": 3.346,
+      "step": 10698
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4951485461441216e-06,
+      "loss": 3.8441,
+      "step": 10699
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4951011378002533e-06,
+      "loss": 3.4782,
+      "step": 10700
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4950537294563845e-06,
+      "loss": 3.921,
+      "step": 10701
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.495006321112516e-06,
+      "loss": 3.9856,
+      "step": 10702
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494958912768647e-06,
+      "loss": 3.9646,
+      "step": 10703
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494911504424779e-06,
+      "loss": 2.2409,
+      "step": 10704
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.49486409608091e-06,
+      "loss": 2.9916,
+      "step": 10705
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494816687737042e-06,
+      "loss": 3.0825,
+      "step": 10706
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4947692793931735e-06,
+      "loss": 3.3221,
+      "step": 10707
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4947218710493048e-06,
+      "loss": 3.3471,
+      "step": 10708
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494674462705436e-06,
+      "loss": 4.3327,
+      "step": 10709
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4946270543615678e-06,
+      "loss": 3.6718,
+      "step": 10710
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494579646017699e-06,
+      "loss": 3.7177,
+      "step": 10711
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4945322376738307e-06,
+      "loss": 3.3197,
+      "step": 10712
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4944848293299625e-06,
+      "loss": 3.7487,
+      "step": 10713
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4944374209860937e-06,
+      "loss": 3.7022,
+      "step": 10714
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494390012642225e-06,
+      "loss": 3.8446,
+      "step": 10715
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4943426042983567e-06,
+      "loss": 3.3499,
+      "step": 10716
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494295195954488e-06,
+      "loss": 3.3828,
+      "step": 10717
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4942477876106193e-06,
+      "loss": 4.1356,
+      "step": 10718
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494200379266751e-06,
+      "loss": 3.408,
+      "step": 10719
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4941529709228827e-06,
+      "loss": 2.8261,
+      "step": 10720
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494105562579014e-06,
+      "loss": 3.4305,
+      "step": 10721
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4940581542351457e-06,
+      "loss": 4.1084,
+      "step": 10722
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.494010745891277e-06,
+      "loss": 3.2333,
+      "step": 10723
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4939633375474082e-06,
+      "loss": 3.3201,
+      "step": 10724
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4939159292035395e-06,
+      "loss": 3.4205,
+      "step": 10725
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4938685208596716e-06,
+      "loss": 2.9937,
+      "step": 10726
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.493821112515803e-06,
+      "loss": 3.4898,
+      "step": 10727
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.493773704171934e-06,
+      "loss": 3.7889,
+      "step": 10728
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.493726295828066e-06,
+      "loss": 3.5361,
+      "step": 10729
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.493678887484197e-06,
+      "loss": 2.8673,
+      "step": 10730
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.4936314791403285e-06,
+      "loss": 3.5408,
+      "step": 10731
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.49358407079646e-06,
+      "loss": 2.8566,
+      "step": 10732
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.493536662452592e-06,
+      "loss": 3.4047,
+      "step": 10733
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.493489254108723e-06,
+      "loss": 3.6406,
+      "step": 10734
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493441845764855e-06,
+      "loss": 2.5648,
+      "step": 10735
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493394437420986e-06,
+      "loss": 3.531,
+      "step": 10736
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4933470290771174e-06,
+      "loss": 3.9548,
+      "step": 10737
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493299620733249e-06,
+      "loss": 2.8444,
+      "step": 10738
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4932522123893804e-06,
+      "loss": 4.3121,
+      "step": 10739
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493204804045512e-06,
+      "loss": 2.9018,
+      "step": 10740
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493157395701644e-06,
+      "loss": 3.4306,
+      "step": 10741
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493109987357775e-06,
+      "loss": 2.6408,
+      "step": 10742
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4930625790139064e-06,
+      "loss": 3.0455,
+      "step": 10743
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.493015170670038e-06,
+      "loss": 2.879,
+      "step": 10744
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4929677623261694e-06,
+      "loss": 4.1965,
+      "step": 10745
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.492920353982301e-06,
+      "loss": 3.2355,
+      "step": 10746
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4928729456384323e-06,
+      "loss": 3.8136,
+      "step": 10747
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.492825537294564e-06,
+      "loss": 2.7417,
+      "step": 10748
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4927781289506953e-06,
+      "loss": 3.4069,
+      "step": 10749
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4927307206068266e-06,
+      "loss": 3.5673,
+      "step": 10750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4926833122629583e-06,
+      "loss": 3.5645,
+      "step": 10751
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4926359039190896e-06,
+      "loss": 3.9367,
+      "step": 10752
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4925884955752213e-06,
+      "loss": 4.2877,
+      "step": 10753
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.492541087231353e-06,
+      "loss": 3.1433,
+      "step": 10754
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4924936788874843e-06,
+      "loss": 2.8356,
+      "step": 10755
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4924462705436156e-06,
+      "loss": 3.779,
+      "step": 10756
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4923988621997473e-06,
+      "loss": 4.1733,
+      "step": 10757
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4923514538558785e-06,
+      "loss": 2.9018,
+      "step": 10758
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4923040455120102e-06,
+      "loss": 3.6615,
+      "step": 10759
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.492256637168142e-06,
+      "loss": 3.5929,
+      "step": 10760
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4922092288242732e-06,
+      "loss": 3.5877,
+      "step": 10761
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4921618204804045e-06,
+      "loss": 3.4068,
+      "step": 10762
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4921144121365362e-06,
+      "loss": 3.3043,
+      "step": 10763
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4920670037926675e-06,
+      "loss": 3.5357,
+      "step": 10764
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4920195954487988e-06,
+      "loss": 3.5403,
+      "step": 10765
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4919721871049305e-06,
+      "loss": 2.9886,
+      "step": 10766
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.491924778761062e-06,
+      "loss": 3.4102,
+      "step": 10767
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4918773704171935e-06,
+      "loss": 3.7886,
+      "step": 10768
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4918299620733247e-06,
+      "loss": 3.5011,
+      "step": 10769
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4917825537294564e-06,
+      "loss": 3.2381,
+      "step": 10770
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4917351453855877e-06,
+      "loss": 3.1291,
+      "step": 10771
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4916877370417194e-06,
+      "loss": 3.3416,
+      "step": 10772
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.491640328697851e-06,
+      "loss": 3.0797,
+      "step": 10773
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4915929203539824e-06,
+      "loss": 2.7927,
+      "step": 10774
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4915455120101137e-06,
+      "loss": 2.8976,
+      "step": 10775
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4914981036662454e-06,
+      "loss": 3.3755,
+      "step": 10776
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4914506953223767e-06,
+      "loss": 3.5976,
+      "step": 10777
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.491403286978508e-06,
+      "loss": 3.2997,
+      "step": 10778
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4913558786346397e-06,
+      "loss": 2.849,
+      "step": 10779
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4913084702907714e-06,
+      "loss": 3.769,
+      "step": 10780
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4912610619469027e-06,
+      "loss": 4.2372,
+      "step": 10781
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4912136536030344e-06,
+      "loss": 3.7977,
+      "step": 10782
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4911662452591656e-06,
+      "loss": 3.8699,
+      "step": 10783
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.491118836915297e-06,
+      "loss": 3.5368,
+      "step": 10784
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4910714285714286e-06,
+      "loss": 3.2181,
+      "step": 10785
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4910240202275603e-06,
+      "loss": 3.9855,
+      "step": 10786
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4909766118836916e-06,
+      "loss": 3.3022,
+      "step": 10787
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4909292035398233e-06,
+      "loss": 2.3756,
+      "step": 10788
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4908817951959546e-06,
+      "loss": 2.6876,
+      "step": 10789
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.490834386852086e-06,
+      "loss": 3.1823,
+      "step": 10790
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.490786978508217e-06,
+      "loss": 3.4416,
+      "step": 10791
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.490739570164349e-06,
+      "loss": 3.593,
+      "step": 10792
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4906921618204806e-06,
+      "loss": 3.1134,
+      "step": 10793
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.490644753476612e-06,
+      "loss": 3.3531,
+      "step": 10794
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.4905973451327435e-06,
+      "loss": 3.7574,
+      "step": 10795
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.490549936788875e-06,
+      "loss": 3.6431,
+      "step": 10796
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.490502528445006e-06,
+      "loss": 3.0963,
+      "step": 10797
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.490455120101138e-06,
+      "loss": 3.3538,
+      "step": 10798
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4904077117572695e-06,
+      "loss": 3.0926,
+      "step": 10799
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.490360303413401e-06,
+      "loss": 3.615,
+      "step": 10800
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4903128950695325e-06,
+      "loss": 2.9224,
+      "step": 10801
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4902654867256638e-06,
+      "loss": 3.7821,
+      "step": 10802
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.490218078381795e-06,
+      "loss": 2.7439,
+      "step": 10803
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4901706700379268e-06,
+      "loss": 3.1944,
+      "step": 10804
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.490123261694058e-06,
+      "loss": 3.1262,
+      "step": 10805
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4900758533501897e-06,
+      "loss": 3.3373,
+      "step": 10806
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4900284450063214e-06,
+      "loss": 3.0174,
+      "step": 10807
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4899810366624527e-06,
+      "loss": 3.7343,
+      "step": 10808
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489933628318584e-06,
+      "loss": 3.7096,
+      "step": 10809
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4898862199747157e-06,
+      "loss": 3.2513,
+      "step": 10810
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489838811630847e-06,
+      "loss": 3.9634,
+      "step": 10811
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4897914032869787e-06,
+      "loss": 3.3936,
+      "step": 10812
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.48974399494311e-06,
+      "loss": 3.1868,
+      "step": 10813
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4896965865992417e-06,
+      "loss": 3.4018,
+      "step": 10814
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489649178255373e-06,
+      "loss": 3.5999,
+      "step": 10815
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4896017699115042e-06,
+      "loss": 3.1336,
+      "step": 10816
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489554361567636e-06,
+      "loss": 3.9643,
+      "step": 10817
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4895069532237672e-06,
+      "loss": 3.9545,
+      "step": 10818
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489459544879899e-06,
+      "loss": 3.5362,
+      "step": 10819
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4894121365360306e-06,
+      "loss": 3.8855,
+      "step": 10820
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489364728192162e-06,
+      "loss": 3.3038,
+      "step": 10821
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489317319848293e-06,
+      "loss": 3.5064,
+      "step": 10822
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489269911504425e-06,
+      "loss": 3.9297,
+      "step": 10823
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489222503160556e-06,
+      "loss": 3.5782,
+      "step": 10824
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489175094816688e-06,
+      "loss": 3.1664,
+      "step": 10825
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4891276864728196e-06,
+      "loss": 3.0395,
+      "step": 10826
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489080278128951e-06,
+      "loss": 3.6143,
+      "step": 10827
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.489032869785082e-06,
+      "loss": 3.5851,
+      "step": 10828
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488985461441214e-06,
+      "loss": 2.9065,
+      "step": 10829
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488938053097345e-06,
+      "loss": 4.218,
+      "step": 10830
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4888906447534764e-06,
+      "loss": 3.7321,
+      "step": 10831
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488843236409608e-06,
+      "loss": 4.1208,
+      "step": 10832
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.48879582806574e-06,
+      "loss": 2.8677,
+      "step": 10833
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488748419721871e-06,
+      "loss": 3.3612,
+      "step": 10834
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4887010113780024e-06,
+      "loss": 2.8244,
+      "step": 10835
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488653603034134e-06,
+      "loss": 3.8255,
+      "step": 10836
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4886061946902654e-06,
+      "loss": 4.4584,
+      "step": 10837
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488558786346397e-06,
+      "loss": 2.8571,
+      "step": 10838
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4885113780025288e-06,
+      "loss": 3.2048,
+      "step": 10839
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.48846396965866e-06,
+      "loss": 2.9867,
+      "step": 10840
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4884165613147913e-06,
+      "loss": 3.9686,
+      "step": 10841
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488369152970923e-06,
+      "loss": 3.7186,
+      "step": 10842
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4883217446270543e-06,
+      "loss": 2.8098,
+      "step": 10843
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4882743362831856e-06,
+      "loss": 2.5465,
+      "step": 10844
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4882269279393173e-06,
+      "loss": 3.321,
+      "step": 10845
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488179519595449e-06,
+      "loss": 3.1587,
+      "step": 10846
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4881321112515803e-06,
+      "loss": 2.5722,
+      "step": 10847
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.488084702907712e-06,
+      "loss": 3.001,
+      "step": 10848
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4880372945638433e-06,
+      "loss": 3.2559,
+      "step": 10849
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4879898862199746e-06,
+      "loss": 2.9457,
+      "step": 10850
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4879424778761063e-06,
+      "loss": 2.7835,
+      "step": 10851
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.487895069532238e-06,
+      "loss": 3.3268,
+      "step": 10852
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4878476611883692e-06,
+      "loss": 3.1394,
+      "step": 10853
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.487800252844501e-06,
+      "loss": 3.1758,
+      "step": 10854
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4877528445006322e-06,
+      "loss": 4.1521,
+      "step": 10855
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4877054361567635e-06,
+      "loss": 3.0297,
+      "step": 10856
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4876580278128948e-06,
+      "loss": 3.2727,
+      "step": 10857
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4876106194690265e-06,
+      "loss": 3.517,
+      "step": 10858
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.487563211125158e-06,
+      "loss": 3.6122,
+      "step": 10859
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.4875158027812895e-06,
+      "loss": 3.2992,
+      "step": 10860
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.487468394437421e-06,
+      "loss": 2.9989,
+      "step": 10861
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4874209860935525e-06,
+      "loss": 2.9388,
+      "step": 10862
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4873735777496837e-06,
+      "loss": 3.032,
+      "step": 10863
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4873261694058154e-06,
+      "loss": 3.8981,
+      "step": 10864
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.487278761061947e-06,
+      "loss": 2.4732,
+      "step": 10865
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4872313527180784e-06,
+      "loss": 3.3402,
+      "step": 10866
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.48718394437421e-06,
+      "loss": 3.3173,
+      "step": 10867
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4871365360303414e-06,
+      "loss": 2.9083,
+      "step": 10868
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4870891276864727e-06,
+      "loss": 3.4731,
+      "step": 10869
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4870417193426044e-06,
+      "loss": 3.7844,
+      "step": 10870
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4869943109987357e-06,
+      "loss": 3.4237,
+      "step": 10871
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4869469026548674e-06,
+      "loss": 3.4558,
+      "step": 10872
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.486899494310999e-06,
+      "loss": 3.3969,
+      "step": 10873
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4868520859671304e-06,
+      "loss": 3.9226,
+      "step": 10874
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4868046776232616e-06,
+      "loss": 2.4758,
+      "step": 10875
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4867572692793933e-06,
+      "loss": 2.8062,
+      "step": 10876
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4867098609355246e-06,
+      "loss": 3.2425,
+      "step": 10877
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4866624525916563e-06,
+      "loss": 3.8641,
+      "step": 10878
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4866150442477876e-06,
+      "loss": 3.5201,
+      "step": 10879
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4865676359039193e-06,
+      "loss": 2.7242,
+      "step": 10880
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4865202275600506e-06,
+      "loss": 3.231,
+      "step": 10881
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.486472819216182e-06,
+      "loss": 3.8238,
+      "step": 10882
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4864254108723136e-06,
+      "loss": 2.9277,
+      "step": 10883
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.486378002528445e-06,
+      "loss": 3.0132,
+      "step": 10884
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4863305941845766e-06,
+      "loss": 2.7888,
+      "step": 10885
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4862831858407083e-06,
+      "loss": 3.5124,
+      "step": 10886
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4862357774968396e-06,
+      "loss": 3.8699,
+      "step": 10887
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.486188369152971e-06,
+      "loss": 3.7699,
+      "step": 10888
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4861409608091025e-06,
+      "loss": 3.6993,
+      "step": 10889
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.486093552465234e-06,
+      "loss": 3.32,
+      "step": 10890
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4860461441213655e-06,
+      "loss": 3.3687,
+      "step": 10891
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4859987357774972e-06,
+      "loss": 3.5727,
+      "step": 10892
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4859513274336285e-06,
+      "loss": 3.378,
+      "step": 10893
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4859039190897598e-06,
+      "loss": 4.1891,
+      "step": 10894
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4858565107458915e-06,
+      "loss": 4.0287,
+      "step": 10895
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4858091024020228e-06,
+      "loss": 3.2114,
+      "step": 10896
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.485761694058154e-06,
+      "loss": 3.1119,
+      "step": 10897
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4857142857142858e-06,
+      "loss": 4.0415,
+      "step": 10898
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4856668773704175e-06,
+      "loss": 3.3411,
+      "step": 10899
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4856194690265487e-06,
+      "loss": 2.289,
+      "step": 10900
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.48557206068268e-06,
+      "loss": 3.4959,
+      "step": 10901
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4855246523388117e-06,
+      "loss": 3.5833,
+      "step": 10902
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.485477243994943e-06,
+      "loss": 3.9388,
+      "step": 10903
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4854298356510747e-06,
+      "loss": 2.7683,
+      "step": 10904
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4853824273072064e-06,
+      "loss": 3.4805,
+      "step": 10905
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4853350189633377e-06,
+      "loss": 3.0483,
+      "step": 10906
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.485287610619469e-06,
+      "loss": 4.0112,
+      "step": 10907
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4852402022756007e-06,
+      "loss": 2.7452,
+      "step": 10908
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.485192793931732e-06,
+      "loss": 3.4841,
+      "step": 10909
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4851453855878632e-06,
+      "loss": 4.4206,
+      "step": 10910
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.485097977243995e-06,
+      "loss": 3.4109,
+      "step": 10911
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4850505689001266e-06,
+      "loss": 2.3705,
+      "step": 10912
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.485003160556258e-06,
+      "loss": 3.3444,
+      "step": 10913
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4849557522123896e-06,
+      "loss": 3.2959,
+      "step": 10914
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.484908343868521e-06,
+      "loss": 3.9669,
+      "step": 10915
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.484860935524652e-06,
+      "loss": 3.3706,
+      "step": 10916
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.484813527180784e-06,
+      "loss": 3.5795,
+      "step": 10917
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4847661188369156e-06,
+      "loss": 3.3187,
+      "step": 10918
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.484718710493047e-06,
+      "loss": 3.2879,
+      "step": 10919
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4846713021491786e-06,
+      "loss": 3.4381,
+      "step": 10920
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.48462389380531e-06,
+      "loss": 2.5319,
+      "step": 10921
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.484576485461441e-06,
+      "loss": 3.4655,
+      "step": 10922
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.4845290771175724e-06,
+      "loss": 3.0599,
+      "step": 10923
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.484481668773704e-06,
+      "loss": 2.7343,
+      "step": 10924
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.484434260429836e-06,
+      "loss": 3.5451,
+      "step": 10925
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.484386852085967e-06,
+      "loss": 3.5024,
+      "step": 10926
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.484339443742099e-06,
+      "loss": 3.932,
+      "step": 10927
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.48429203539823e-06,
+      "loss": 3.4942,
+      "step": 10928
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4842446270543614e-06,
+      "loss": 3.0447,
+      "step": 10929
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.484197218710493e-06,
+      "loss": 3.4858,
+      "step": 10930
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4841498103666248e-06,
+      "loss": 3.4306,
+      "step": 10931
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.484102402022756e-06,
+      "loss": 3.3847,
+      "step": 10932
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4840549936788878e-06,
+      "loss": 3.2527,
+      "step": 10933
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.484007585335019e-06,
+      "loss": 3.0977,
+      "step": 10934
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4839601769911503e-06,
+      "loss": 3.4217,
+      "step": 10935
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483912768647282e-06,
+      "loss": 3.3853,
+      "step": 10936
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4838653603034133e-06,
+      "loss": 3.5614,
+      "step": 10937
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483817951959545e-06,
+      "loss": 2.8917,
+      "step": 10938
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4837705436156767e-06,
+      "loss": 2.3986,
+      "step": 10939
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483723135271808e-06,
+      "loss": 3.3262,
+      "step": 10940
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4836757269279393e-06,
+      "loss": 3.024,
+      "step": 10941
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483628318584071e-06,
+      "loss": 3.2463,
+      "step": 10942
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4835809102402023e-06,
+      "loss": 3.9124,
+      "step": 10943
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483533501896334e-06,
+      "loss": 3.4916,
+      "step": 10944
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4834860935524653e-06,
+      "loss": 3.2244,
+      "step": 10945
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483438685208597e-06,
+      "loss": 3.811,
+      "step": 10946
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4833912768647282e-06,
+      "loss": 2.9788,
+      "step": 10947
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4833438685208595e-06,
+      "loss": 3.3738,
+      "step": 10948
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4832964601769912e-06,
+      "loss": 3.3312,
+      "step": 10949
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4832490518331225e-06,
+      "loss": 3.5061,
+      "step": 10950
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483201643489254e-06,
+      "loss": 3.6295,
+      "step": 10951
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483154235145386e-06,
+      "loss": 4.1811,
+      "step": 10952
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.483106826801517e-06,
+      "loss": 3.7142,
+      "step": 10953
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4830594184576485e-06,
+      "loss": 3.9034,
+      "step": 10954
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.48301201011378e-06,
+      "loss": 3.6456,
+      "step": 10955
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4829646017699115e-06,
+      "loss": 3.2022,
+      "step": 10956
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482917193426043e-06,
+      "loss": 3.6941,
+      "step": 10957
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482869785082175e-06,
+      "loss": 4.0686,
+      "step": 10958
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482822376738306e-06,
+      "loss": 3.1634,
+      "step": 10959
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4827749683944374e-06,
+      "loss": 3.5248,
+      "step": 10960
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482727560050569e-06,
+      "loss": 3.7852,
+      "step": 10961
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4826801517067004e-06,
+      "loss": 3.4855,
+      "step": 10962
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4826327433628317e-06,
+      "loss": 3.5113,
+      "step": 10963
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4825853350189634e-06,
+      "loss": 3.1119,
+      "step": 10964
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482537926675095e-06,
+      "loss": 3.7374,
+      "step": 10965
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4824905183312264e-06,
+      "loss": 3.6517,
+      "step": 10966
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4824431099873577e-06,
+      "loss": 3.9454,
+      "step": 10967
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4823957016434894e-06,
+      "loss": 3.1444,
+      "step": 10968
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4823482932996206e-06,
+      "loss": 3.4624,
+      "step": 10969
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4823008849557523e-06,
+      "loss": 3.6186,
+      "step": 10970
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482253476611884e-06,
+      "loss": 3.6871,
+      "step": 10971
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4822060682680153e-06,
+      "loss": 3.3625,
+      "step": 10972
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4821586599241466e-06,
+      "loss": 3.6898,
+      "step": 10973
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4821112515802783e-06,
+      "loss": 3.317,
+      "step": 10974
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4820638432364096e-06,
+      "loss": 2.9323,
+      "step": 10975
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.482016434892541e-06,
+      "loss": 3.1013,
+      "step": 10976
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4819690265486726e-06,
+      "loss": 3.5023,
+      "step": 10977
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4819216182048043e-06,
+      "loss": 3.4943,
+      "step": 10978
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4818742098609356e-06,
+      "loss": 2.9587,
+      "step": 10979
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4818268015170673e-06,
+      "loss": 3.622,
+      "step": 10980
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4817793931731985e-06,
+      "loss": 3.7018,
+      "step": 10981
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.48173198482933e-06,
+      "loss": 2.641,
+      "step": 10982
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4816845764854615e-06,
+      "loss": 3.8231,
+      "step": 10983
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4816371681415932e-06,
+      "loss": 2.8279,
+      "step": 10984
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4815897597977245e-06,
+      "loss": 3.5875,
+      "step": 10985
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4815423514538562e-06,
+      "loss": 3.6314,
+      "step": 10986
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.4814949431099875e-06,
+      "loss": 3.5215,
+      "step": 10987
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4814475347661188e-06,
+      "loss": 3.6611,
+      "step": 10988
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.48140012642225e-06,
+      "loss": 2.9618,
+      "step": 10989
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4813527180783818e-06,
+      "loss": 2.7592,
+      "step": 10990
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4813053097345135e-06,
+      "loss": 3.3502,
+      "step": 10991
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4812579013906447e-06,
+      "loss": 2.9871,
+      "step": 10992
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4812104930467765e-06,
+      "loss": 3.0476,
+      "step": 10993
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4811630847029077e-06,
+      "loss": 3.3923,
+      "step": 10994
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.481115676359039e-06,
+      "loss": 3.6046,
+      "step": 10995
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4810682680151707e-06,
+      "loss": 3.1257,
+      "step": 10996
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4810208596713024e-06,
+      "loss": 3.9031,
+      "step": 10997
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4809734513274337e-06,
+      "loss": 3.7777,
+      "step": 10998
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4809260429835654e-06,
+      "loss": 2.7289,
+      "step": 10999
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4808786346396967e-06,
+      "loss": 3.1662,
+      "step": 11000
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480831226295828e-06,
+      "loss": 2.3447,
+      "step": 11001
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4807838179519597e-06,
+      "loss": 3.4323,
+      "step": 11002
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480736409608091e-06,
+      "loss": 3.7594,
+      "step": 11003
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4806890012642227e-06,
+      "loss": 3.5139,
+      "step": 11004
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4806415929203544e-06,
+      "loss": 3.4037,
+      "step": 11005
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4805941845764856e-06,
+      "loss": 2.8354,
+      "step": 11006
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480546776232617e-06,
+      "loss": 3.8497,
+      "step": 11007
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4804993678887486e-06,
+      "loss": 3.7852,
+      "step": 11008
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.48045195954488e-06,
+      "loss": 3.4073,
+      "step": 11009
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480404551201011e-06,
+      "loss": 4.0744,
+      "step": 11010
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480357142857143e-06,
+      "loss": 3.3535,
+      "step": 11011
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4803097345132746e-06,
+      "loss": 2.6255,
+      "step": 11012
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480262326169406e-06,
+      "loss": 2.2891,
+      "step": 11013
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480214917825537e-06,
+      "loss": 3.0848,
+      "step": 11014
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480167509481669e-06,
+      "loss": 2.9667,
+      "step": 11015
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4801201011378e-06,
+      "loss": 4.0462,
+      "step": 11016
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.480072692793932e-06,
+      "loss": 3.7481,
+      "step": 11017
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4800252844500635e-06,
+      "loss": 3.6772,
+      "step": 11018
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479977876106195e-06,
+      "loss": 3.5783,
+      "step": 11019
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479930467762326e-06,
+      "loss": 3.5293,
+      "step": 11020
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479883059418458e-06,
+      "loss": 2.6852,
+      "step": 11021
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479835651074589e-06,
+      "loss": 3.6694,
+      "step": 11022
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4797882427307204e-06,
+      "loss": 3.333,
+      "step": 11023
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479740834386852e-06,
+      "loss": 3.6553,
+      "step": 11024
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4796934260429838e-06,
+      "loss": 3.1796,
+      "step": 11025
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479646017699115e-06,
+      "loss": 3.1991,
+      "step": 11026
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4795986093552468e-06,
+      "loss": 2.7124,
+      "step": 11027
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479551201011378e-06,
+      "loss": 3.6716,
+      "step": 11028
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4795037926675093e-06,
+      "loss": 3.5123,
+      "step": 11029
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479456384323641e-06,
+      "loss": 2.2769,
+      "step": 11030
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4794089759797727e-06,
+      "loss": 3.8984,
+      "step": 11031
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479361567635904e-06,
+      "loss": 4.1429,
+      "step": 11032
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4793141592920353e-06,
+      "loss": 3.2756,
+      "step": 11033
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479266750948167e-06,
+      "loss": 3.3626,
+      "step": 11034
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4792193426042983e-06,
+      "loss": 3.4942,
+      "step": 11035
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4791719342604296e-06,
+      "loss": 3.5348,
+      "step": 11036
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4791245259165613e-06,
+      "loss": 3.2727,
+      "step": 11037
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.479077117572693e-06,
+      "loss": 3.4208,
+      "step": 11038
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4790297092288242e-06,
+      "loss": 3.596,
+      "step": 11039
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478982300884956e-06,
+      "loss": 3.043,
+      "step": 11040
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4789348925410872e-06,
+      "loss": 3.3827,
+      "step": 11041
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4788874841972185e-06,
+      "loss": 3.0114,
+      "step": 11042
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4788400758533502e-06,
+      "loss": 3.3546,
+      "step": 11043
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478792667509482e-06,
+      "loss": 3.3609,
+      "step": 11044
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478745259165613e-06,
+      "loss": 2.9674,
+      "step": 11045
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478697850821745e-06,
+      "loss": 3.6808,
+      "step": 11046
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478650442477876e-06,
+      "loss": 3.0373,
+      "step": 11047
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4786030341340075e-06,
+      "loss": 4.1835,
+      "step": 11048
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478555625790139e-06,
+      "loss": 3.6055,
+      "step": 11049
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.4785082174462704e-06,
+      "loss": 3.3444,
+      "step": 11050
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.478460809102402e-06,
+      "loss": 3.952,
+      "step": 11051
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4784134007585334e-06,
+      "loss": 3.2425,
+      "step": 11052
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.478365992414665e-06,
+      "loss": 3.8912,
+      "step": 11053
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4783185840707964e-06,
+      "loss": 3.0415,
+      "step": 11054
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4782711757269277e-06,
+      "loss": 3.0052,
+      "step": 11055
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4782237673830594e-06,
+      "loss": 3.4203,
+      "step": 11056
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.478176359039191e-06,
+      "loss": 2.1164,
+      "step": 11057
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4781289506953224e-06,
+      "loss": 3.9939,
+      "step": 11058
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.478081542351454e-06,
+      "loss": 3.0852,
+      "step": 11059
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4780341340075854e-06,
+      "loss": 3.1414,
+      "step": 11060
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4779867256637166e-06,
+      "loss": 3.0851,
+      "step": 11061
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4779393173198484e-06,
+      "loss": 3.1157,
+      "step": 11062
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4778919089759796e-06,
+      "loss": 3.0615,
+      "step": 11063
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4778445006321113e-06,
+      "loss": 3.2815,
+      "step": 11064
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.477797092288243e-06,
+      "loss": 3.3607,
+      "step": 11065
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4777496839443743e-06,
+      "loss": 3.4637,
+      "step": 11066
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4777022756005056e-06,
+      "loss": 3.3502,
+      "step": 11067
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4776548672566373e-06,
+      "loss": 3.9194,
+      "step": 11068
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4776074589127686e-06,
+      "loss": 2.82,
+      "step": 11069
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4775600505689003e-06,
+      "loss": 3.5474,
+      "step": 11070
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.477512642225032e-06,
+      "loss": 3.3207,
+      "step": 11071
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4774652338811633e-06,
+      "loss": 2.8804,
+      "step": 11072
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4774178255372946e-06,
+      "loss": 3.884,
+      "step": 11073
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.477370417193426e-06,
+      "loss": 3.9207,
+      "step": 11074
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4773230088495575e-06,
+      "loss": 3.6124,
+      "step": 11075
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.477275600505689e-06,
+      "loss": 3.5193,
+      "step": 11076
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4772281921618205e-06,
+      "loss": 4.1183,
+      "step": 11077
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4771807838179522e-06,
+      "loss": 3.4952,
+      "step": 11078
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4771333754740835e-06,
+      "loss": 3.5303,
+      "step": 11079
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4770859671302148e-06,
+      "loss": 3.196,
+      "step": 11080
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4770385587863465e-06,
+      "loss": 3.5233,
+      "step": 11081
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4769911504424778e-06,
+      "loss": 3.442,
+      "step": 11082
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4769437420986095e-06,
+      "loss": 3.6355,
+      "step": 11083
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.476896333754741e-06,
+      "loss": 3.5333,
+      "step": 11084
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4768489254108725e-06,
+      "loss": 3.5075,
+      "step": 11085
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4768015170670037e-06,
+      "loss": 2.8318,
+      "step": 11086
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4767541087231354e-06,
+      "loss": 4.3352,
+      "step": 11087
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4767067003792667e-06,
+      "loss": 3.4406,
+      "step": 11088
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.476659292035398e-06,
+      "loss": 3.7022,
+      "step": 11089
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4766118836915297e-06,
+      "loss": 4.0223,
+      "step": 11090
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4765644753476614e-06,
+      "loss": 2.7944,
+      "step": 11091
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4765170670037927e-06,
+      "loss": 3.656,
+      "step": 11092
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4764696586599244e-06,
+      "loss": 3.3046,
+      "step": 11093
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4764222503160557e-06,
+      "loss": 3.8807,
+      "step": 11094
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.476374841972187e-06,
+      "loss": 2.6477,
+      "step": 11095
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4763274336283187e-06,
+      "loss": 3.378,
+      "step": 11096
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4762800252844504e-06,
+      "loss": 3.3942,
+      "step": 11097
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4762326169405816e-06,
+      "loss": 4.1062,
+      "step": 11098
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.476185208596713e-06,
+      "loss": 3.2019,
+      "step": 11099
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4761378002528446e-06,
+      "loss": 3.6351,
+      "step": 11100
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.476090391908976e-06,
+      "loss": 3.7263,
+      "step": 11101
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.476042983565107e-06,
+      "loss": 3.0569,
+      "step": 11102
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475995575221239e-06,
+      "loss": 3.2666,
+      "step": 11103
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4759481668773706e-06,
+      "loss": 3.6129,
+      "step": 11104
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475900758533502e-06,
+      "loss": 3.4328,
+      "step": 11105
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4758533501896336e-06,
+      "loss": 3.4579,
+      "step": 11106
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475805941845765e-06,
+      "loss": 3.1493,
+      "step": 11107
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475758533501896e-06,
+      "loss": 3.4983,
+      "step": 11108
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475711125158028e-06,
+      "loss": 3.4318,
+      "step": 11109
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4756637168141596e-06,
+      "loss": 3.7608,
+      "step": 11110
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475616308470291e-06,
+      "loss": 3.7045,
+      "step": 11111
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.4755689001264225e-06,
+      "loss": 3.6305,
+      "step": 11112
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475521491782554e-06,
+      "loss": 3.4837,
+      "step": 11113
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.475474083438685e-06,
+      "loss": 3.3973,
+      "step": 11114
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475426675094817e-06,
+      "loss": 3.8163,
+      "step": 11115
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475379266750948e-06,
+      "loss": 3.0429,
+      "step": 11116
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4753318584070798e-06,
+      "loss": 3.5404,
+      "step": 11117
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475284450063211e-06,
+      "loss": 3.954,
+      "step": 11118
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4752370417193428e-06,
+      "loss": 3.4746,
+      "step": 11119
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475189633375474e-06,
+      "loss": 2.6718,
+      "step": 11120
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4751422250316053e-06,
+      "loss": 3.9801,
+      "step": 11121
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475094816687737e-06,
+      "loss": 3.3289,
+      "step": 11122
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4750474083438687e-06,
+      "loss": 3.275,
+      "step": 11123
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475e-06,
+      "loss": 3.2341,
+      "step": 11124
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4749525916561317e-06,
+      "loss": 3.5895,
+      "step": 11125
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474905183312263e-06,
+      "loss": 4.4355,
+      "step": 11126
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4748577749683943e-06,
+      "loss": 3.6794,
+      "step": 11127
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474810366624526e-06,
+      "loss": 2.9952,
+      "step": 11128
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4747629582806573e-06,
+      "loss": 2.8099,
+      "step": 11129
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474715549936789e-06,
+      "loss": 2.6132,
+      "step": 11130
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4746681415929207e-06,
+      "loss": 3.5205,
+      "step": 11131
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474620733249052e-06,
+      "loss": 3.1138,
+      "step": 11132
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4745733249051832e-06,
+      "loss": 2.8941,
+      "step": 11133
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474525916561315e-06,
+      "loss": 2.9759,
+      "step": 11134
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4744785082174462e-06,
+      "loss": 2.5776,
+      "step": 11135
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474431099873578e-06,
+      "loss": 3.1582,
+      "step": 11136
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4743836915297096e-06,
+      "loss": 2.8385,
+      "step": 11137
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474336283185841e-06,
+      "loss": 2.9868,
+      "step": 11138
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474288874841972e-06,
+      "loss": 3.8162,
+      "step": 11139
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4742414664981035e-06,
+      "loss": 4.1303,
+      "step": 11140
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474194058154235e-06,
+      "loss": 3.7063,
+      "step": 11141
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4741466498103665e-06,
+      "loss": 3.5042,
+      "step": 11142
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474099241466498e-06,
+      "loss": 2.8849,
+      "step": 11143
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.47405183312263e-06,
+      "loss": 3.417,
+      "step": 11144
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.474004424778761e-06,
+      "loss": 3.2346,
+      "step": 11145
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4739570164348924e-06,
+      "loss": 3.8942,
+      "step": 11146
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473909608091024e-06,
+      "loss": 3.246,
+      "step": 11147
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4738621997471554e-06,
+      "loss": 3.3425,
+      "step": 11148
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473814791403287e-06,
+      "loss": 3.521,
+      "step": 11149
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473767383059419e-06,
+      "loss": 2.7221,
+      "step": 11150
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.47371997471555e-06,
+      "loss": 3.4105,
+      "step": 11151
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4736725663716814e-06,
+      "loss": 3.729,
+      "step": 11152
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473625158027813e-06,
+      "loss": 3.4389,
+      "step": 11153
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4735777496839444e-06,
+      "loss": 3.3525,
+      "step": 11154
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4735303413400756e-06,
+      "loss": 3.3817,
+      "step": 11155
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4734829329962073e-06,
+      "loss": 3.4653,
+      "step": 11156
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473435524652339e-06,
+      "loss": 2.8259,
+      "step": 11157
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4733881163084703e-06,
+      "loss": 2.785,
+      "step": 11158
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473340707964602e-06,
+      "loss": 3.8255,
+      "step": 11159
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4732932996207333e-06,
+      "loss": 3.0543,
+      "step": 11160
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4732458912768646e-06,
+      "loss": 3.1701,
+      "step": 11161
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4731984829329963e-06,
+      "loss": 3.0008,
+      "step": 11162
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473151074589128e-06,
+      "loss": 3.4927,
+      "step": 11163
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4731036662452593e-06,
+      "loss": 3.1933,
+      "step": 11164
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4730562579013906e-06,
+      "loss": 4.0854,
+      "step": 11165
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4730088495575223e-06,
+      "loss": 3.7914,
+      "step": 11166
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4729614412136535e-06,
+      "loss": 3.3978,
+      "step": 11167
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.472914032869785e-06,
+      "loss": 3.3907,
+      "step": 11168
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4728666245259165e-06,
+      "loss": 3.2334,
+      "step": 11169
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4728192161820482e-06,
+      "loss": 2.7127,
+      "step": 11170
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4727718078381795e-06,
+      "loss": 3.2448,
+      "step": 11171
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4727243994943112e-06,
+      "loss": 3.1855,
+      "step": 11172
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4726769911504425e-06,
+      "loss": 3.9681,
+      "step": 11173
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4726295828065738e-06,
+      "loss": 3.7082,
+      "step": 11174
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4725821744627055e-06,
+      "loss": 3.1338,
+      "step": 11175
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.472534766118837e-06,
+      "loss": 3.6569,
+      "step": 11176
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4724873577749685e-06,
+      "loss": 3.3491,
+      "step": 11177
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4724399494311e-06,
+      "loss": 3.4152,
+      "step": 11178
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4723925410872315e-06,
+      "loss": 3.6367,
+      "step": 11179
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4723451327433627e-06,
+      "loss": 2.5617,
+      "step": 11180
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4722977243994944e-06,
+      "loss": 2.9746,
+      "step": 11181
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4722503160556257e-06,
+      "loss": 3.8886,
+      "step": 11182
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4722029077117574e-06,
+      "loss": 1.9018,
+      "step": 11183
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4721554993678887e-06,
+      "loss": 3.1386,
+      "step": 11184
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4721080910240204e-06,
+      "loss": 3.6549,
+      "step": 11185
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4720606826801517e-06,
+      "loss": 2.9484,
+      "step": 11186
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.472013274336283e-06,
+      "loss": 3.3367,
+      "step": 11187
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4719658659924147e-06,
+      "loss": 3.3932,
+      "step": 11188
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4719184576485464e-06,
+      "loss": 2.692,
+      "step": 11189
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4718710493046777e-06,
+      "loss": 3.2424,
+      "step": 11190
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4718236409608094e-06,
+      "loss": 2.4606,
+      "step": 11191
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4717762326169406e-06,
+      "loss": 3.3984,
+      "step": 11192
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471728824273072e-06,
+      "loss": 3.6814,
+      "step": 11193
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4716814159292036e-06,
+      "loss": 3.2773,
+      "step": 11194
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471634007585335e-06,
+      "loss": 2.8456,
+      "step": 11195
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4715865992414666e-06,
+      "loss": 3.0789,
+      "step": 11196
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4715391908975983e-06,
+      "loss": 3.2818,
+      "step": 11197
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4714917825537296e-06,
+      "loss": 3.7232,
+      "step": 11198
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471444374209861e-06,
+      "loss": 3.3403,
+      "step": 11199
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4713969658659926e-06,
+      "loss": 3.4133,
+      "step": 11200
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471349557522124e-06,
+      "loss": 4.3395,
+      "step": 11201
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4713021491782556e-06,
+      "loss": 2.7252,
+      "step": 11202
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4712547408343873e-06,
+      "loss": 2.811,
+      "step": 11203
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4712073324905185e-06,
+      "loss": 3.0246,
+      "step": 11204
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.47115992414665e-06,
+      "loss": 3.7133,
+      "step": 11205
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471112515802781e-06,
+      "loss": 2.8508,
+      "step": 11206
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471065107458913e-06,
+      "loss": 3.2645,
+      "step": 11207
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.471017699115044e-06,
+      "loss": 3.5603,
+      "step": 11208
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470970290771176e-06,
+      "loss": 3.2822,
+      "step": 11209
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4709228824273075e-06,
+      "loss": 3.56,
+      "step": 11210
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4708754740834388e-06,
+      "loss": 3.3703,
+      "step": 11211
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.47082806573957e-06,
+      "loss": 2.5327,
+      "step": 11212
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4707806573957018e-06,
+      "loss": 3.0466,
+      "step": 11213
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470733249051833e-06,
+      "loss": 2.6263,
+      "step": 11214
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4706858407079647e-06,
+      "loss": 2.9688,
+      "step": 11215
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4706384323640965e-06,
+      "loss": 2.6642,
+      "step": 11216
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4705910240202277e-06,
+      "loss": 3.3794,
+      "step": 11217
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470543615676359e-06,
+      "loss": 4.2523,
+      "step": 11218
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4704962073324907e-06,
+      "loss": 3.9515,
+      "step": 11219
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470448798988622e-06,
+      "loss": 3.9412,
+      "step": 11220
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4704013906447533e-06,
+      "loss": 3.8785,
+      "step": 11221
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470353982300885e-06,
+      "loss": 3.4162,
+      "step": 11222
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4703065739570167e-06,
+      "loss": 3.5783,
+      "step": 11223
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470259165613148e-06,
+      "loss": 3.0637,
+      "step": 11224
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4702117572692797e-06,
+      "loss": 3.6483,
+      "step": 11225
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470164348925411e-06,
+      "loss": 3.6125,
+      "step": 11226
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4701169405815422e-06,
+      "loss": 3.2772,
+      "step": 11227
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.470069532237674e-06,
+      "loss": 3.2415,
+      "step": 11228
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4700221238938056e-06,
+      "loss": 3.8554,
+      "step": 11229
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469974715549937e-06,
+      "loss": 2.9009,
+      "step": 11230
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469927307206068e-06,
+      "loss": 4.0172,
+      "step": 11231
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4698798988622e-06,
+      "loss": 3.2257,
+      "step": 11232
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469832490518331e-06,
+      "loss": 4.0169,
+      "step": 11233
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4697850821744625e-06,
+      "loss": 3.4827,
+      "step": 11234
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469737673830594e-06,
+      "loss": 3.0603,
+      "step": 11235
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469690265486726e-06,
+      "loss": 3.1105,
+      "step": 11236
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469642857142857e-06,
+      "loss": 3.7974,
+      "step": 11237
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469595448798989e-06,
+      "loss": 3.7889,
+      "step": 11238
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.46954804045512e-06,
+      "loss": 3.9458,
+      "step": 11239
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4695006321112514e-06,
+      "loss": 3.2192,
+      "step": 11240
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.469453223767383e-06,
+      "loss": 3.3903,
+      "step": 11241
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.469405815423515e-06,
+      "loss": 3.607,
+      "step": 11242
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.469358407079646e-06,
+      "loss": 2.3005,
+      "step": 11243
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.469310998735778e-06,
+      "loss": 3.1018,
+      "step": 11244
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.469263590391909e-06,
+      "loss": 3.6464,
+      "step": 11245
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4692161820480404e-06,
+      "loss": 3.3772,
+      "step": 11246
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.469168773704172e-06,
+      "loss": 2.9141,
+      "step": 11247
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4691213653603034e-06,
+      "loss": 3.1963,
+      "step": 11248
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.469073957016435e-06,
+      "loss": 3.4434,
+      "step": 11249
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4690265486725663e-06,
+      "loss": 3.8818,
+      "step": 11250
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468979140328698e-06,
+      "loss": 3.4961,
+      "step": 11251
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4689317319848293e-06,
+      "loss": 3.8251,
+      "step": 11252
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4688843236409606e-06,
+      "loss": 3.4521,
+      "step": 11253
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4688369152970923e-06,
+      "loss": 3.3937,
+      "step": 11254
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468789506953224e-06,
+      "loss": 3.2977,
+      "step": 11255
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4687420986093553e-06,
+      "loss": 3.2716,
+      "step": 11256
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468694690265487e-06,
+      "loss": 3.7695,
+      "step": 11257
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4686472819216183e-06,
+      "loss": 3.5151,
+      "step": 11258
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4685998735777496e-06,
+      "loss": 3.8264,
+      "step": 11259
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4685524652338813e-06,
+      "loss": 4.0622,
+      "step": 11260
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4685050568900125e-06,
+      "loss": 3.2174,
+      "step": 11261
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4684576485461442e-06,
+      "loss": 3.3712,
+      "step": 11262
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468410240202276e-06,
+      "loss": 3.6948,
+      "step": 11263
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4683628318584072e-06,
+      "loss": 4.067,
+      "step": 11264
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4683154235145385e-06,
+      "loss": 3.6356,
+      "step": 11265
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4682680151706702e-06,
+      "loss": 2.8228,
+      "step": 11266
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4682206068268015e-06,
+      "loss": 3.4478,
+      "step": 11267
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468173198482933e-06,
+      "loss": 3.3908,
+      "step": 11268
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468125790139065e-06,
+      "loss": 4.0377,
+      "step": 11269
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.468078381795196e-06,
+      "loss": 3.4934,
+      "step": 11270
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4680309734513275e-06,
+      "loss": 3.0195,
+      "step": 11271
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4679835651074587e-06,
+      "loss": 3.5634,
+      "step": 11272
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4679361567635904e-06,
+      "loss": 3.1778,
+      "step": 11273
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4678887484197217e-06,
+      "loss": 2.0655,
+      "step": 11274
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4678413400758534e-06,
+      "loss": 3.293,
+      "step": 11275
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.467793931731985e-06,
+      "loss": 3.9065,
+      "step": 11276
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4677465233881164e-06,
+      "loss": 3.9387,
+      "step": 11277
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4676991150442477e-06,
+      "loss": 2.7347,
+      "step": 11278
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4676517067003794e-06,
+      "loss": 2.3108,
+      "step": 11279
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4676042983565107e-06,
+      "loss": 2.8002,
+      "step": 11280
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4675568900126424e-06,
+      "loss": 3.2219,
+      "step": 11281
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.467509481668774e-06,
+      "loss": 2.5535,
+      "step": 11282
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4674620733249054e-06,
+      "loss": 2.1019,
+      "step": 11283
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4674146649810367e-06,
+      "loss": 3.3206,
+      "step": 11284
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4673672566371684e-06,
+      "loss": 2.8602,
+      "step": 11285
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4673198482932996e-06,
+      "loss": 4.3183,
+      "step": 11286
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.467272439949431e-06,
+      "loss": 4.1914,
+      "step": 11287
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4672250316055626e-06,
+      "loss": 3.1804,
+      "step": 11288
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4671776232616943e-06,
+      "loss": 3.0632,
+      "step": 11289
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4671302149178256e-06,
+      "loss": 4.033,
+      "step": 11290
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4670828065739573e-06,
+      "loss": 3.1843,
+      "step": 11291
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4670353982300886e-06,
+      "loss": 2.0578,
+      "step": 11292
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.46698798988622e-06,
+      "loss": 3.4668,
+      "step": 11293
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.466940581542351e-06,
+      "loss": 3.0412,
+      "step": 11294
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4668931731984833e-06,
+      "loss": 4.0637,
+      "step": 11295
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4668457648546146e-06,
+      "loss": 3.6337,
+      "step": 11296
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.466798356510746e-06,
+      "loss": 3.6261,
+      "step": 11297
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4667509481668775e-06,
+      "loss": 4.0154,
+      "step": 11298
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.466703539823009e-06,
+      "loss": 2.6911,
+      "step": 11299
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.46665613147914e-06,
+      "loss": 3.0751,
+      "step": 11300
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.466608723135272e-06,
+      "loss": 4.115,
+      "step": 11301
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4665613147914035e-06,
+      "loss": 3.246,
+      "step": 11302
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.466513906447535e-06,
+      "loss": 4.0001,
+      "step": 11303
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.4664664981036665e-06,
+      "loss": 3.0712,
+      "step": 11304
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4664190897597978e-06,
+      "loss": 4.3228,
+      "step": 11305
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.466371681415929e-06,
+      "loss": 3.298,
+      "step": 11306
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4663242730720608e-06,
+      "loss": 3.6587,
+      "step": 11307
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.466276864728192e-06,
+      "loss": 4.2154,
+      "step": 11308
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4662294563843237e-06,
+      "loss": 3.5191,
+      "step": 11309
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4661820480404554e-06,
+      "loss": 4.0919,
+      "step": 11310
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4661346396965867e-06,
+      "loss": 2.3006,
+      "step": 11311
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.466087231352718e-06,
+      "loss": 2.9412,
+      "step": 11312
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4660398230088497e-06,
+      "loss": 2.8741,
+      "step": 11313
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465992414664981e-06,
+      "loss": 3.3729,
+      "step": 11314
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4659450063211127e-06,
+      "loss": 3.6898,
+      "step": 11315
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465897597977244e-06,
+      "loss": 3.3477,
+      "step": 11316
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4658501896333757e-06,
+      "loss": 2.8326,
+      "step": 11317
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465802781289507e-06,
+      "loss": 3.083,
+      "step": 11318
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4657553729456382e-06,
+      "loss": 3.2676,
+      "step": 11319
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.46570796460177e-06,
+      "loss": 3.3728,
+      "step": 11320
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4656605562579012e-06,
+      "loss": 2.3997,
+      "step": 11321
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465613147914033e-06,
+      "loss": 3.7274,
+      "step": 11322
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4655657395701646e-06,
+      "loss": 3.237,
+      "step": 11323
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465518331226296e-06,
+      "loss": 3.5422,
+      "step": 11324
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465470922882427e-06,
+      "loss": 3.6205,
+      "step": 11325
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465423514538559e-06,
+      "loss": 3.5913,
+      "step": 11326
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.46537610619469e-06,
+      "loss": 3.553,
+      "step": 11327
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465328697850822e-06,
+      "loss": 2.4421,
+      "step": 11328
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4652812895069536e-06,
+      "loss": 3.9128,
+      "step": 11329
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465233881163085e-06,
+      "loss": 3.4256,
+      "step": 11330
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465186472819216e-06,
+      "loss": 3.2243,
+      "step": 11331
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465139064475348e-06,
+      "loss": 3.3255,
+      "step": 11332
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.465091656131479e-06,
+      "loss": 3.7616,
+      "step": 11333
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4650442477876104e-06,
+      "loss": 3.9607,
+      "step": 11334
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464996839443742e-06,
+      "loss": 3.6337,
+      "step": 11335
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464949431099874e-06,
+      "loss": 3.4077,
+      "step": 11336
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464902022756005e-06,
+      "loss": 3.2257,
+      "step": 11337
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4648546144121364e-06,
+      "loss": 3.7357,
+      "step": 11338
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464807206068268e-06,
+      "loss": 3.8151,
+      "step": 11339
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4647597977243994e-06,
+      "loss": 3.611,
+      "step": 11340
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464712389380531e-06,
+      "loss": 2.4944,
+      "step": 11341
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4646649810366628e-06,
+      "loss": 3.9643,
+      "step": 11342
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464617572692794e-06,
+      "loss": 3.1794,
+      "step": 11343
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4645701643489253e-06,
+      "loss": 3.0554,
+      "step": 11344
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464522756005057e-06,
+      "loss": 2.6456,
+      "step": 11345
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4644753476611883e-06,
+      "loss": 3.4101,
+      "step": 11346
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4644279393173196e-06,
+      "loss": 3.0142,
+      "step": 11347
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4643805309734513e-06,
+      "loss": 3.1907,
+      "step": 11348
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464333122629583e-06,
+      "loss": 3.7921,
+      "step": 11349
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4642857142857143e-06,
+      "loss": 3.3787,
+      "step": 11350
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464238305941846e-06,
+      "loss": 3.8624,
+      "step": 11351
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4641908975979773e-06,
+      "loss": 3.6735,
+      "step": 11352
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4641434892541086e-06,
+      "loss": 3.1215,
+      "step": 11353
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4640960809102403e-06,
+      "loss": 4.0672,
+      "step": 11354
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.464048672566372e-06,
+      "loss": 4.0242,
+      "step": 11355
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4640012642225032e-06,
+      "loss": 3.6716,
+      "step": 11356
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.463953855878635e-06,
+      "loss": 3.5028,
+      "step": 11357
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4639064475347662e-06,
+      "loss": 2.7306,
+      "step": 11358
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4638590391908975e-06,
+      "loss": 4.0186,
+      "step": 11359
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4638116308470288e-06,
+      "loss": 3.36,
+      "step": 11360
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4637642225031605e-06,
+      "loss": 3.491,
+      "step": 11361
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.463716814159292e-06,
+      "loss": 3.9843,
+      "step": 11362
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4636694058154235e-06,
+      "loss": 3.6788,
+      "step": 11363
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.463621997471555e-06,
+      "loss": 2.9941,
+      "step": 11364
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4635745891276865e-06,
+      "loss": 2.8653,
+      "step": 11365
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4635271807838177e-06,
+      "loss": 2.9585,
+      "step": 11366
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.4634797724399494e-06,
+      "loss": 2.6412,
+      "step": 11367
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.463432364096081e-06,
+      "loss": 3.9943,
+      "step": 11368
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4633849557522124e-06,
+      "loss": 3.4682,
+      "step": 11369
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.463337547408344e-06,
+      "loss": 3.5391,
+      "step": 11370
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4632901390644754e-06,
+      "loss": 3.1629,
+      "step": 11371
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4632427307206067e-06,
+      "loss": 3.033,
+      "step": 11372
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4631953223767384e-06,
+      "loss": 3.0331,
+      "step": 11373
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4631479140328697e-06,
+      "loss": 2.6205,
+      "step": 11374
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4631005056890014e-06,
+      "loss": 2.8052,
+      "step": 11375
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.463053097345133e-06,
+      "loss": 3.9712,
+      "step": 11376
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4630056890012644e-06,
+      "loss": 2.7183,
+      "step": 11377
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4629582806573956e-06,
+      "loss": 3.7103,
+      "step": 11378
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4629108723135273e-06,
+      "loss": 3.504,
+      "step": 11379
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4628634639696586e-06,
+      "loss": 3.385,
+      "step": 11380
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4628160556257903e-06,
+      "loss": 3.5528,
+      "step": 11381
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4627686472819216e-06,
+      "loss": 2.6719,
+      "step": 11382
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4627212389380533e-06,
+      "loss": 3.569,
+      "step": 11383
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4626738305941846e-06,
+      "loss": 3.0655,
+      "step": 11384
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.462626422250316e-06,
+      "loss": 2.916,
+      "step": 11385
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4625790139064476e-06,
+      "loss": 3.7058,
+      "step": 11386
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.462531605562579e-06,
+      "loss": 3.6748,
+      "step": 11387
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4624841972187106e-06,
+      "loss": 2.934,
+      "step": 11388
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4624367888748423e-06,
+      "loss": 3.453,
+      "step": 11389
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4623893805309736e-06,
+      "loss": 3.535,
+      "step": 11390
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.462341972187105e-06,
+      "loss": 3.7522,
+      "step": 11391
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4622945638432365e-06,
+      "loss": 3.488,
+      "step": 11392
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.462247155499368e-06,
+      "loss": 2.7272,
+      "step": 11393
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4621997471554995e-06,
+      "loss": 2.9442,
+      "step": 11394
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4621523388116312e-06,
+      "loss": 2.7452,
+      "step": 11395
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4621049304677625e-06,
+      "loss": 3.1594,
+      "step": 11396
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4620575221238938e-06,
+      "loss": 3.9752,
+      "step": 11397
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4620101137800255e-06,
+      "loss": 3.1659,
+      "step": 11398
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4619627054361568e-06,
+      "loss": 3.7546,
+      "step": 11399
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461915297092288e-06,
+      "loss": 3.2474,
+      "step": 11400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4618678887484198e-06,
+      "loss": 3.4478,
+      "step": 11401
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4618204804045515e-06,
+      "loss": 3.4725,
+      "step": 11402
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4617730720606827e-06,
+      "loss": 3.5356,
+      "step": 11403
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461725663716814e-06,
+      "loss": 3.4148,
+      "step": 11404
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4616782553729457e-06,
+      "loss": 2.4262,
+      "step": 11405
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461630847029077e-06,
+      "loss": 3.6835,
+      "step": 11406
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4615834386852087e-06,
+      "loss": 3.6827,
+      "step": 11407
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4615360303413404e-06,
+      "loss": 2.9199,
+      "step": 11408
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4614886219974717e-06,
+      "loss": 3.0708,
+      "step": 11409
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461441213653603e-06,
+      "loss": 3.0393,
+      "step": 11410
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4613938053097347e-06,
+      "loss": 2.8469,
+      "step": 11411
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461346396965866e-06,
+      "loss": 3.529,
+      "step": 11412
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4612989886219972e-06,
+      "loss": 3.7071,
+      "step": 11413
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461251580278129e-06,
+      "loss": 2.6084,
+      "step": 11414
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4612041719342606e-06,
+      "loss": 3.9977,
+      "step": 11415
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461156763590392e-06,
+      "loss": 3.2266,
+      "step": 11416
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4611093552465236e-06,
+      "loss": 4.059,
+      "step": 11417
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461061946902655e-06,
+      "loss": 3.4036,
+      "step": 11418
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.461014538558786e-06,
+      "loss": 3.5909,
+      "step": 11419
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460967130214918e-06,
+      "loss": 3.2062,
+      "step": 11420
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4609197218710496e-06,
+      "loss": 3.8898,
+      "step": 11421
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460872313527181e-06,
+      "loss": 3.1686,
+      "step": 11422
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4608249051833126e-06,
+      "loss": 3.3606,
+      "step": 11423
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460777496839444e-06,
+      "loss": 4.048,
+      "step": 11424
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460730088495575e-06,
+      "loss": 3.0353,
+      "step": 11425
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.4606826801517064e-06,
+      "loss": 2.3502,
+      "step": 11426
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460635271807838e-06,
+      "loss": 2.2608,
+      "step": 11427
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.46058786346397e-06,
+      "loss": 2.8385,
+      "step": 11428
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460540455120101e-06,
+      "loss": 2.6377,
+      "step": 11429
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460493046776233e-06,
+      "loss": 2.8971,
+      "step": 11430
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.460445638432364e-06,
+      "loss": 3.6421,
+      "step": 11431
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4603982300884954e-06,
+      "loss": 3.6529,
+      "step": 11432
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.460350821744627e-06,
+      "loss": 3.6738,
+      "step": 11433
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4603034134007588e-06,
+      "loss": 2.9576,
+      "step": 11434
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.46025600505689e-06,
+      "loss": 2.8154,
+      "step": 11435
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4602085967130218e-06,
+      "loss": 3.6602,
+      "step": 11436
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.460161188369153e-06,
+      "loss": 2.8905,
+      "step": 11437
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4601137800252843e-06,
+      "loss": 3.5701,
+      "step": 11438
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.460066371681416e-06,
+      "loss": 2.6713,
+      "step": 11439
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4600189633375473e-06,
+      "loss": 2.8118,
+      "step": 11440
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459971554993679e-06,
+      "loss": 2.7176,
+      "step": 11441
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4599241466498107e-06,
+      "loss": 3.0151,
+      "step": 11442
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459876738305942e-06,
+      "loss": 3.5212,
+      "step": 11443
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4598293299620733e-06,
+      "loss": 3.107,
+      "step": 11444
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459781921618205e-06,
+      "loss": 2.116,
+      "step": 11445
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4597345132743363e-06,
+      "loss": 2.5869,
+      "step": 11446
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459687104930468e-06,
+      "loss": 3.4611,
+      "step": 11447
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4596396965865993e-06,
+      "loss": 3.6577,
+      "step": 11448
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459592288242731e-06,
+      "loss": 3.1588,
+      "step": 11449
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4595448798988622e-06,
+      "loss": 2.9504,
+      "step": 11450
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4594974715549935e-06,
+      "loss": 2.6012,
+      "step": 11451
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4594500632111252e-06,
+      "loss": 4.0687,
+      "step": 11452
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4594026548672565e-06,
+      "loss": 3.8218,
+      "step": 11453
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459355246523388e-06,
+      "loss": 3.8343,
+      "step": 11454
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.45930783817952e-06,
+      "loss": 3.4523,
+      "step": 11455
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459260429835651e-06,
+      "loss": 3.676,
+      "step": 11456
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4592130214917825e-06,
+      "loss": 3.3957,
+      "step": 11457
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459165613147914e-06,
+      "loss": 2.7978,
+      "step": 11458
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4591182048040455e-06,
+      "loss": 2.4665,
+      "step": 11459
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459070796460177e-06,
+      "loss": 3.3702,
+      "step": 11460
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.459023388116309e-06,
+      "loss": 3.0971,
+      "step": 11461
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.45897597977244e-06,
+      "loss": 3.1759,
+      "step": 11462
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4589285714285714e-06,
+      "loss": 3.2414,
+      "step": 11463
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.458881163084703e-06,
+      "loss": 3.1271,
+      "step": 11464
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4588337547408344e-06,
+      "loss": 3.5075,
+      "step": 11465
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4587863463969657e-06,
+      "loss": 3.1687,
+      "step": 11466
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4587389380530974e-06,
+      "loss": 3.3638,
+      "step": 11467
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.458691529709229e-06,
+      "loss": 3.8644,
+      "step": 11468
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4586441213653604e-06,
+      "loss": 2.094,
+      "step": 11469
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4585967130214917e-06,
+      "loss": 3.205,
+      "step": 11470
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4585493046776234e-06,
+      "loss": 3.6382,
+      "step": 11471
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4585018963337546e-06,
+      "loss": 2.7325,
+      "step": 11472
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4584544879898863e-06,
+      "loss": 2.6216,
+      "step": 11473
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.458407079646018e-06,
+      "loss": 3.8941,
+      "step": 11474
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4583596713021493e-06,
+      "loss": 3.659,
+      "step": 11475
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4583122629582806e-06,
+      "loss": 3.5079,
+      "step": 11476
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4582648546144123e-06,
+      "loss": 3.1927,
+      "step": 11477
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4582174462705436e-06,
+      "loss": 3.2414,
+      "step": 11478
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.458170037926675e-06,
+      "loss": 2.9456,
+      "step": 11479
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4581226295828066e-06,
+      "loss": 2.5444,
+      "step": 11480
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4580752212389383e-06,
+      "loss": 3.2029,
+      "step": 11481
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4580278128950696e-06,
+      "loss": 3.0764,
+      "step": 11482
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4579804045512013e-06,
+      "loss": 3.9103,
+      "step": 11483
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4579329962073325e-06,
+      "loss": 3.7353,
+      "step": 11484
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.457885587863464e-06,
+      "loss": 2.9052,
+      "step": 11485
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4578381795195955e-06,
+      "loss": 3.8723,
+      "step": 11486
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4577907711757272e-06,
+      "loss": 3.0265,
+      "step": 11487
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4577433628318585e-06,
+      "loss": 4.5111,
+      "step": 11488
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.45769595448799e-06,
+      "loss": 3.3625,
+      "step": 11489
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4576485461441215e-06,
+      "loss": 3.0652,
+      "step": 11490
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4576011378002528e-06,
+      "loss": 2.6185,
+      "step": 11491
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.457553729456384e-06,
+      "loss": 3.578,
+      "step": 11492
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4575063211125158e-06,
+      "loss": 3.8085,
+      "step": 11493
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.4574589127686475e-06,
+      "loss": 3.4691,
+      "step": 11494
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4574115044247787e-06,
+      "loss": 2.7726,
+      "step": 11495
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4573640960809105e-06,
+      "loss": 2.9854,
+      "step": 11496
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4573166877370417e-06,
+      "loss": 3.7368,
+      "step": 11497
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.457269279393173e-06,
+      "loss": 3.3982,
+      "step": 11498
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4572218710493047e-06,
+      "loss": 3.6801,
+      "step": 11499
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4571744627054364e-06,
+      "loss": 2.9557,
+      "step": 11500
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4571270543615677e-06,
+      "loss": 2.5053,
+      "step": 11501
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4570796460176994e-06,
+      "loss": 2.5554,
+      "step": 11502
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4570322376738307e-06,
+      "loss": 2.7766,
+      "step": 11503
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456984829329962e-06,
+      "loss": 3.5365,
+      "step": 11504
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4569374209860937e-06,
+      "loss": 3.5869,
+      "step": 11505
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456890012642225e-06,
+      "loss": 4.1458,
+      "step": 11506
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4568426042983567e-06,
+      "loss": 3.854,
+      "step": 11507
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4567951959544884e-06,
+      "loss": 3.7286,
+      "step": 11508
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4567477876106196e-06,
+      "loss": 3.7364,
+      "step": 11509
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456700379266751e-06,
+      "loss": 3.1922,
+      "step": 11510
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456652970922882e-06,
+      "loss": 3.3899,
+      "step": 11511
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456605562579014e-06,
+      "loss": 3.6155,
+      "step": 11512
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4565581542351456e-06,
+      "loss": 3.1366,
+      "step": 11513
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456510745891277e-06,
+      "loss": 2.9464,
+      "step": 11514
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4564633375474086e-06,
+      "loss": 3.2109,
+      "step": 11515
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.45641592920354e-06,
+      "loss": 3.8008,
+      "step": 11516
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456368520859671e-06,
+      "loss": 2.9008,
+      "step": 11517
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456321112515803e-06,
+      "loss": 2.6937,
+      "step": 11518
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456273704171934e-06,
+      "loss": 3.2628,
+      "step": 11519
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456226295828066e-06,
+      "loss": 3.7257,
+      "step": 11520
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4561788874841975e-06,
+      "loss": 4.0524,
+      "step": 11521
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456131479140329e-06,
+      "loss": 2.5059,
+      "step": 11522
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.45608407079646e-06,
+      "loss": 3.0298,
+      "step": 11523
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.456036662452592e-06,
+      "loss": 2.7502,
+      "step": 11524
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455989254108723e-06,
+      "loss": 3.0437,
+      "step": 11525
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455941845764855e-06,
+      "loss": 3.5087,
+      "step": 11526
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4558944374209865e-06,
+      "loss": 3.5769,
+      "step": 11527
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4558470290771178e-06,
+      "loss": 3.0641,
+      "step": 11528
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455799620733249e-06,
+      "loss": 3.2067,
+      "step": 11529
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4557522123893808e-06,
+      "loss": 2.6715,
+      "step": 11530
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455704804045512e-06,
+      "loss": 3.3219,
+      "step": 11531
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4556573957016433e-06,
+      "loss": 3.3686,
+      "step": 11532
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455609987357775e-06,
+      "loss": 3.5833,
+      "step": 11533
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4555625790139067e-06,
+      "loss": 3.4886,
+      "step": 11534
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455515170670038e-06,
+      "loss": 3.9335,
+      "step": 11535
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4554677623261693e-06,
+      "loss": 4.2273,
+      "step": 11536
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455420353982301e-06,
+      "loss": 3.2955,
+      "step": 11537
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4553729456384323e-06,
+      "loss": 2.9771,
+      "step": 11538
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455325537294564e-06,
+      "loss": 2.2496,
+      "step": 11539
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4552781289506957e-06,
+      "loss": 2.1957,
+      "step": 11540
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.455230720606827e-06,
+      "loss": 3.8531,
+      "step": 11541
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4551833122629582e-06,
+      "loss": 2.8559,
+      "step": 11542
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.45513590391909e-06,
+      "loss": 2.8505,
+      "step": 11543
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4550884955752212e-06,
+      "loss": 3.8171,
+      "step": 11544
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4550410872313525e-06,
+      "loss": 2.2021,
+      "step": 11545
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4549936788874842e-06,
+      "loss": 3.2318,
+      "step": 11546
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454946270543616e-06,
+      "loss": 4.2492,
+      "step": 11547
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454898862199747e-06,
+      "loss": 3.5648,
+      "step": 11548
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454851453855879e-06,
+      "loss": 2.7247,
+      "step": 11549
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.45480404551201e-06,
+      "loss": 2.9921,
+      "step": 11550
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4547566371681415e-06,
+      "loss": 3.9305,
+      "step": 11551
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454709228824273e-06,
+      "loss": 3.7424,
+      "step": 11552
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454661820480405e-06,
+      "loss": 2.5802,
+      "step": 11553
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454614412136536e-06,
+      "loss": 3.7211,
+      "step": 11554
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4545670037926674e-06,
+      "loss": 4.0322,
+      "step": 11555
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.454519595448799e-06,
+      "loss": 3.0545,
+      "step": 11556
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.4544721871049304e-06,
+      "loss": 4.4984,
+      "step": 11557
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4544247787610617e-06,
+      "loss": 2.4713,
+      "step": 11558
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4543773704171934e-06,
+      "loss": 3.3305,
+      "step": 11559
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.454329962073325e-06,
+      "loss": 3.0923,
+      "step": 11560
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4542825537294564e-06,
+      "loss": 3.0349,
+      "step": 11561
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.454235145385588e-06,
+      "loss": 3.2012,
+      "step": 11562
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4541877370417194e-06,
+      "loss": 3.5117,
+      "step": 11563
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4541403286978506e-06,
+      "loss": 3.5727,
+      "step": 11564
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4540929203539824e-06,
+      "loss": 3.9434,
+      "step": 11565
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.454045512010114e-06,
+      "loss": 3.7013,
+      "step": 11566
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4539981036662453e-06,
+      "loss": 3.7239,
+      "step": 11567
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.453950695322377e-06,
+      "loss": 2.9517,
+      "step": 11568
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4539032869785083e-06,
+      "loss": 3.0915,
+      "step": 11569
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4538558786346396e-06,
+      "loss": 3.5215,
+      "step": 11570
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4538084702907713e-06,
+      "loss": 3.1402,
+      "step": 11571
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4537610619469026e-06,
+      "loss": 3.7208,
+      "step": 11572
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4537136536030343e-06,
+      "loss": 3.3975,
+      "step": 11573
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.453666245259166e-06,
+      "loss": 3.5174,
+      "step": 11574
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4536188369152973e-06,
+      "loss": 2.9721,
+      "step": 11575
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4535714285714286e-06,
+      "loss": 3.2031,
+      "step": 11576
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.45352402022756e-06,
+      "loss": 3.1956,
+      "step": 11577
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4534766118836915e-06,
+      "loss": 3.6078,
+      "step": 11578
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.453429203539823e-06,
+      "loss": 3.8299,
+      "step": 11579
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4533817951959545e-06,
+      "loss": 3.1794,
+      "step": 11580
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4533343868520862e-06,
+      "loss": 3.4099,
+      "step": 11581
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4532869785082175e-06,
+      "loss": 4.2594,
+      "step": 11582
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.453239570164349e-06,
+      "loss": 3.3294,
+      "step": 11583
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4531921618204805e-06,
+      "loss": 3.7178,
+      "step": 11584
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4531447534766118e-06,
+      "loss": 3.0566,
+      "step": 11585
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4530973451327435e-06,
+      "loss": 3.5608,
+      "step": 11586
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.453049936788875e-06,
+      "loss": 3.9603,
+      "step": 11587
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4530025284450065e-06,
+      "loss": 2.9701,
+      "step": 11588
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4529551201011377e-06,
+      "loss": 2.728,
+      "step": 11589
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4529077117572694e-06,
+      "loss": 2.8355,
+      "step": 11590
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4528603034134007e-06,
+      "loss": 3.4695,
+      "step": 11591
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.452812895069532e-06,
+      "loss": 4.376,
+      "step": 11592
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4527654867256637e-06,
+      "loss": 2.5805,
+      "step": 11593
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4527180783817954e-06,
+      "loss": 3.6141,
+      "step": 11594
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4526706700379267e-06,
+      "loss": 3.3721,
+      "step": 11595
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4526232616940584e-06,
+      "loss": 3.6887,
+      "step": 11596
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4525758533501897e-06,
+      "loss": 3.3641,
+      "step": 11597
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.452528445006321e-06,
+      "loss": 3.5835,
+      "step": 11598
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4524810366624527e-06,
+      "loss": 3.4961,
+      "step": 11599
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4524336283185844e-06,
+      "loss": 3.1961,
+      "step": 11600
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4523862199747156e-06,
+      "loss": 3.1609,
+      "step": 11601
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.452338811630847e-06,
+      "loss": 3.0085,
+      "step": 11602
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4522914032869786e-06,
+      "loss": 3.8882,
+      "step": 11603
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.45224399494311e-06,
+      "loss": 3.2917,
+      "step": 11604
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.452196586599241e-06,
+      "loss": 2.2973,
+      "step": 11605
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.452149178255373e-06,
+      "loss": 3.2125,
+      "step": 11606
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4521017699115046e-06,
+      "loss": 3.1116,
+      "step": 11607
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.452054361567636e-06,
+      "loss": 3.68,
+      "step": 11608
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4520069532237676e-06,
+      "loss": 3.8817,
+      "step": 11609
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451959544879899e-06,
+      "loss": 4.0184,
+      "step": 11610
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.45191213653603e-06,
+      "loss": 3.6989,
+      "step": 11611
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451864728192162e-06,
+      "loss": 3.3715,
+      "step": 11612
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4518173198482936e-06,
+      "loss": 4.2582,
+      "step": 11613
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451769911504425e-06,
+      "loss": 3.9235,
+      "step": 11614
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4517225031605565e-06,
+      "loss": 3.276,
+      "step": 11615
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451675094816688e-06,
+      "loss": 3.0402,
+      "step": 11616
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451627686472819e-06,
+      "loss": 4.085,
+      "step": 11617
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451580278128951e-06,
+      "loss": 3.0643,
+      "step": 11618
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451532869785082e-06,
+      "loss": 3.1241,
+      "step": 11619
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.4514854614412138e-06,
+      "loss": 3.5631,
+      "step": 11620
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.451438053097345e-06,
+      "loss": 3.575,
+      "step": 11621
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4513906447534768e-06,
+      "loss": 3.3616,
+      "step": 11622
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.451343236409608e-06,
+      "loss": 2.9207,
+      "step": 11623
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4512958280657393e-06,
+      "loss": 3.2704,
+      "step": 11624
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.451248419721871e-06,
+      "loss": 3.3959,
+      "step": 11625
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4512010113780027e-06,
+      "loss": 3.4325,
+      "step": 11626
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.451153603034134e-06,
+      "loss": 3.7925,
+      "step": 11627
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4511061946902657e-06,
+      "loss": 2.5628,
+      "step": 11628
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.451058786346397e-06,
+      "loss": 3.2638,
+      "step": 11629
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4510113780025283e-06,
+      "loss": 2.6874,
+      "step": 11630
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.45096396965866e-06,
+      "loss": 3.707,
+      "step": 11631
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4509165613147913e-06,
+      "loss": 3.5675,
+      "step": 11632
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450869152970923e-06,
+      "loss": 3.4267,
+      "step": 11633
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4508217446270547e-06,
+      "loss": 3.3925,
+      "step": 11634
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450774336283186e-06,
+      "loss": 3.5148,
+      "step": 11635
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4507269279393172e-06,
+      "loss": 2.6117,
+      "step": 11636
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450679519595449e-06,
+      "loss": 3.2295,
+      "step": 11637
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4506321112515802e-06,
+      "loss": 2.6436,
+      "step": 11638
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450584702907712e-06,
+      "loss": 3.2539,
+      "step": 11639
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4505372945638436e-06,
+      "loss": 3.849,
+      "step": 11640
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450489886219975e-06,
+      "loss": 3.2995,
+      "step": 11641
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450442477876106e-06,
+      "loss": 2.1077,
+      "step": 11642
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4503950695322375e-06,
+      "loss": 4.0477,
+      "step": 11643
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450347661188369e-06,
+      "loss": 3.3003,
+      "step": 11644
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4503002528445005e-06,
+      "loss": 3.8594,
+      "step": 11645
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450252844500632e-06,
+      "loss": 3.299,
+      "step": 11646
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450205436156764e-06,
+      "loss": 3.4124,
+      "step": 11647
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450158027812895e-06,
+      "loss": 3.7412,
+      "step": 11648
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4501106194690264e-06,
+      "loss": 3.4576,
+      "step": 11649
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.450063211125158e-06,
+      "loss": 3.124,
+      "step": 11650
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4500158027812894e-06,
+      "loss": 3.0249,
+      "step": 11651
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449968394437421e-06,
+      "loss": 3.1113,
+      "step": 11652
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449920986093553e-06,
+      "loss": 3.1087,
+      "step": 11653
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449873577749684e-06,
+      "loss": 3.2126,
+      "step": 11654
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4498261694058154e-06,
+      "loss": 2.4333,
+      "step": 11655
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449778761061947e-06,
+      "loss": 3.9856,
+      "step": 11656
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4497313527180784e-06,
+      "loss": 2.9485,
+      "step": 11657
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4496839443742096e-06,
+      "loss": 3.4319,
+      "step": 11658
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4496365360303413e-06,
+      "loss": 3.7502,
+      "step": 11659
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449589127686473e-06,
+      "loss": 3.9998,
+      "step": 11660
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4495417193426043e-06,
+      "loss": 3.1035,
+      "step": 11661
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449494310998736e-06,
+      "loss": 3.2968,
+      "step": 11662
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4494469026548673e-06,
+      "loss": 3.4221,
+      "step": 11663
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4493994943109986e-06,
+      "loss": 2.6487,
+      "step": 11664
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4493520859671303e-06,
+      "loss": 2.6639,
+      "step": 11665
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449304677623262e-06,
+      "loss": 2.5078,
+      "step": 11666
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4492572692793933e-06,
+      "loss": 3.2104,
+      "step": 11667
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4492098609355246e-06,
+      "loss": 4.0278,
+      "step": 11668
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4491624525916563e-06,
+      "loss": 3.2523,
+      "step": 11669
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4491150442477875e-06,
+      "loss": 2.9822,
+      "step": 11670
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.449067635903919e-06,
+      "loss": 2.6199,
+      "step": 11671
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4490202275600505e-06,
+      "loss": 3.603,
+      "step": 11672
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4489728192161822e-06,
+      "loss": 3.7825,
+      "step": 11673
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4489254108723135e-06,
+      "loss": 4.0108,
+      "step": 11674
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4488780025284452e-06,
+      "loss": 3.8122,
+      "step": 11675
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4488305941845765e-06,
+      "loss": 3.3868,
+      "step": 11676
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4487831858407078e-06,
+      "loss": 3.2294,
+      "step": 11677
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4487357774968395e-06,
+      "loss": 3.6501,
+      "step": 11678
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.448688369152971e-06,
+      "loss": 2.7866,
+      "step": 11679
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4486409608091025e-06,
+      "loss": 3.66,
+      "step": 11680
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.448593552465234e-06,
+      "loss": 3.7757,
+      "step": 11681
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4485461441213655e-06,
+      "loss": 3.3252,
+      "step": 11682
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4484987357774967e-06,
+      "loss": 3.6617,
+      "step": 11683
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.4484513274336284e-06,
+      "loss": 3.1926,
+      "step": 11684
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4484039190897597e-06,
+      "loss": 3.9309,
+      "step": 11685
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4483565107458914e-06,
+      "loss": 3.7402,
+      "step": 11686
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4483091024020227e-06,
+      "loss": 2.6724,
+      "step": 11687
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4482616940581544e-06,
+      "loss": 3.4989,
+      "step": 11688
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4482142857142857e-06,
+      "loss": 3.6331,
+      "step": 11689
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.448166877370417e-06,
+      "loss": 3.2754,
+      "step": 11690
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4481194690265487e-06,
+      "loss": 3.0659,
+      "step": 11691
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4480720606826804e-06,
+      "loss": 3.7051,
+      "step": 11692
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4480246523388117e-06,
+      "loss": 3.4559,
+      "step": 11693
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4479772439949434e-06,
+      "loss": 3.4152,
+      "step": 11694
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4479298356510746e-06,
+      "loss": 4.3006,
+      "step": 11695
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447882427307206e-06,
+      "loss": 3.2804,
+      "step": 11696
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4478350189633376e-06,
+      "loss": 2.666,
+      "step": 11697
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447787610619469e-06,
+      "loss": 2.9525,
+      "step": 11698
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4477402022756006e-06,
+      "loss": 3.643,
+      "step": 11699
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4476927939317323e-06,
+      "loss": 2.7206,
+      "step": 11700
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4476453855878636e-06,
+      "loss": 4.0395,
+      "step": 11701
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447597977243995e-06,
+      "loss": 3.4174,
+      "step": 11702
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4475505689001266e-06,
+      "loss": 3.5873,
+      "step": 11703
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447503160556258e-06,
+      "loss": 4.2904,
+      "step": 11704
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4474557522123896e-06,
+      "loss": 3.1513,
+      "step": 11705
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4474083438685213e-06,
+      "loss": 2.2478,
+      "step": 11706
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4473609355246525e-06,
+      "loss": 2.6921,
+      "step": 11707
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447313527180784e-06,
+      "loss": 3.9235,
+      "step": 11708
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447266118836915e-06,
+      "loss": 1.7763,
+      "step": 11709
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447218710493047e-06,
+      "loss": 4.1755,
+      "step": 11710
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.447171302149178e-06,
+      "loss": 3.7979,
+      "step": 11711
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.44712389380531e-06,
+      "loss": 3.2714,
+      "step": 11712
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4470764854614415e-06,
+      "loss": 3.3916,
+      "step": 11713
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4470290771175728e-06,
+      "loss": 2.8354,
+      "step": 11714
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446981668773704e-06,
+      "loss": 3.6963,
+      "step": 11715
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4469342604298358e-06,
+      "loss": 3.0419,
+      "step": 11716
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446886852085967e-06,
+      "loss": 3.0157,
+      "step": 11717
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4468394437420987e-06,
+      "loss": 2.5379,
+      "step": 11718
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4467920353982305e-06,
+      "loss": 3.6472,
+      "step": 11719
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4467446270543617e-06,
+      "loss": 3.0554,
+      "step": 11720
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446697218710493e-06,
+      "loss": 3.6442,
+      "step": 11721
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4466498103666247e-06,
+      "loss": 2.3241,
+      "step": 11722
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446602402022756e-06,
+      "loss": 3.7335,
+      "step": 11723
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4465549936788873e-06,
+      "loss": 3.1792,
+      "step": 11724
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446507585335019e-06,
+      "loss": 3.5375,
+      "step": 11725
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4464601769911507e-06,
+      "loss": 3.1511,
+      "step": 11726
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446412768647282e-06,
+      "loss": 3.8354,
+      "step": 11727
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4463653603034137e-06,
+      "loss": 3.9139,
+      "step": 11728
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446317951959545e-06,
+      "loss": 3.0767,
+      "step": 11729
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4462705436156762e-06,
+      "loss": 3.9177,
+      "step": 11730
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446223135271808e-06,
+      "loss": 3.7577,
+      "step": 11731
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4461757269279396e-06,
+      "loss": 3.3802,
+      "step": 11732
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446128318584071e-06,
+      "loss": 3.4104,
+      "step": 11733
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446080910240202e-06,
+      "loss": 3.4686,
+      "step": 11734
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.446033501896334e-06,
+      "loss": 3.191,
+      "step": 11735
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445986093552465e-06,
+      "loss": 3.6365,
+      "step": 11736
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4459386852085965e-06,
+      "loss": 3.2847,
+      "step": 11737
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445891276864728e-06,
+      "loss": 4.4113,
+      "step": 11738
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.44584386852086e-06,
+      "loss": 2.7104,
+      "step": 11739
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445796460176991e-06,
+      "loss": 2.8939,
+      "step": 11740
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445749051833123e-06,
+      "loss": 3.7303,
+      "step": 11741
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445701643489254e-06,
+      "loss": 2.59,
+      "step": 11742
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.4456542351453854e-06,
+      "loss": 2.7646,
+      "step": 11743
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445606826801517e-06,
+      "loss": 3.0681,
+      "step": 11744
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445559418457649e-06,
+      "loss": 3.2409,
+      "step": 11745
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.44551201011378e-06,
+      "loss": 3.3039,
+      "step": 11746
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.445464601769912e-06,
+      "loss": 3.648,
+      "step": 11747
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.445417193426043e-06,
+      "loss": 3.5776,
+      "step": 11748
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4453697850821744e-06,
+      "loss": 3.5819,
+      "step": 11749
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.445322376738306e-06,
+      "loss": 3.497,
+      "step": 11750
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4452749683944374e-06,
+      "loss": 3.7631,
+      "step": 11751
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.445227560050569e-06,
+      "loss": 2.924,
+      "step": 11752
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4451801517067003e-06,
+      "loss": 3.8635,
+      "step": 11753
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.445132743362832e-06,
+      "loss": 3.381,
+      "step": 11754
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4450853350189633e-06,
+      "loss": 2.4927,
+      "step": 11755
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4450379266750946e-06,
+      "loss": 3.6432,
+      "step": 11756
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4449905183312263e-06,
+      "loss": 2.6302,
+      "step": 11757
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.444943109987358e-06,
+      "loss": 3.3336,
+      "step": 11758
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4448957016434893e-06,
+      "loss": 3.8809,
+      "step": 11759
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.444848293299621e-06,
+      "loss": 3.8064,
+      "step": 11760
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4448008849557523e-06,
+      "loss": 3.548,
+      "step": 11761
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4447534766118836e-06,
+      "loss": 3.1294,
+      "step": 11762
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4447060682680153e-06,
+      "loss": 2.9923,
+      "step": 11763
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4446586599241465e-06,
+      "loss": 3.6078,
+      "step": 11764
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4446112515802782e-06,
+      "loss": 3.7606,
+      "step": 11765
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.44456384323641e-06,
+      "loss": 3.5296,
+      "step": 11766
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4445164348925412e-06,
+      "loss": 3.2579,
+      "step": 11767
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4444690265486725e-06,
+      "loss": 3.1906,
+      "step": 11768
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4444216182048042e-06,
+      "loss": 3.5199,
+      "step": 11769
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4443742098609355e-06,
+      "loss": 3.1422,
+      "step": 11770
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.444326801517067e-06,
+      "loss": 3.5993,
+      "step": 11771
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.444279393173199e-06,
+      "loss": 4.0102,
+      "step": 11772
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.44423198482933e-06,
+      "loss": 3.2286,
+      "step": 11773
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4441845764854615e-06,
+      "loss": 3.3233,
+      "step": 11774
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4441371681415927e-06,
+      "loss": 3.6517,
+      "step": 11775
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4440897597977244e-06,
+      "loss": 3.1567,
+      "step": 11776
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4440423514538557e-06,
+      "loss": 3.7702,
+      "step": 11777
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4439949431099874e-06,
+      "loss": 3.304,
+      "step": 11778
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.443947534766119e-06,
+      "loss": 3.8896,
+      "step": 11779
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4439001264222504e-06,
+      "loss": 3.9172,
+      "step": 11780
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4438527180783817e-06,
+      "loss": 3.1631,
+      "step": 11781
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4438053097345134e-06,
+      "loss": 3.3509,
+      "step": 11782
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4437579013906447e-06,
+      "loss": 2.9181,
+      "step": 11783
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4437104930467764e-06,
+      "loss": 4.0904,
+      "step": 11784
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.443663084702908e-06,
+      "loss": 3.378,
+      "step": 11785
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4436156763590394e-06,
+      "loss": 4.0482,
+      "step": 11786
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4435682680151707e-06,
+      "loss": 2.9366,
+      "step": 11787
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4435208596713024e-06,
+      "loss": 3.2385,
+      "step": 11788
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4434734513274336e-06,
+      "loss": 3.5924,
+      "step": 11789
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.443426042983565e-06,
+      "loss": 3.4972,
+      "step": 11790
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4433786346396966e-06,
+      "loss": 3.1244,
+      "step": 11791
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4433312262958283e-06,
+      "loss": 3.5969,
+      "step": 11792
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4432838179519596e-06,
+      "loss": 3.9062,
+      "step": 11793
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4432364096080913e-06,
+      "loss": 3.3643,
+      "step": 11794
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4431890012642226e-06,
+      "loss": 4.1968,
+      "step": 11795
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.443141592920354e-06,
+      "loss": 3.593,
+      "step": 11796
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4430941845764856e-06,
+      "loss": 2.7011,
+      "step": 11797
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4430467762326173e-06,
+      "loss": 2.6768,
+      "step": 11798
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4429993678887486e-06,
+      "loss": 3.4169,
+      "step": 11799
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.44295195954488e-06,
+      "loss": 3.8249,
+      "step": 11800
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4429045512010115e-06,
+      "loss": 3.0054,
+      "step": 11801
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.442857142857143e-06,
+      "loss": 3.5651,
+      "step": 11802
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.442809734513274e-06,
+      "loss": 3.2345,
+      "step": 11803
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.442762326169406e-06,
+      "loss": 3.0931,
+      "step": 11804
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4427149178255375e-06,
+      "loss": 3.5204,
+      "step": 11805
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.442667509481669e-06,
+      "loss": 2.6644,
+      "step": 11806
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4426201011378005e-06,
+      "loss": 2.3904,
+      "step": 11807
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4425726927939318e-06,
+      "loss": 3.2188,
+      "step": 11808
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.442525284450063e-06,
+      "loss": 2.7931,
+      "step": 11809
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4424778761061948e-06,
+      "loss": 3.5477,
+      "step": 11810
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.4424304677623265e-06,
+      "loss": 2.9515,
+      "step": 11811
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4423830594184577e-06,
+      "loss": 2.0978,
+      "step": 11812
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4423356510745894e-06,
+      "loss": 3.4692,
+      "step": 11813
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4422882427307207e-06,
+      "loss": 3.5979,
+      "step": 11814
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.442240834386852e-06,
+      "loss": 3.1468,
+      "step": 11815
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4421934260429837e-06,
+      "loss": 3.9575,
+      "step": 11816
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.442146017699115e-06,
+      "loss": 3.7333,
+      "step": 11817
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4420986093552467e-06,
+      "loss": 2.6838,
+      "step": 11818
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.442051201011378e-06,
+      "loss": 2.4661,
+      "step": 11819
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4420037926675097e-06,
+      "loss": 2.8562,
+      "step": 11820
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441956384323641e-06,
+      "loss": 2.3083,
+      "step": 11821
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4419089759797722e-06,
+      "loss": 3.6489,
+      "step": 11822
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441861567635904e-06,
+      "loss": 3.6704,
+      "step": 11823
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4418141592920357e-06,
+      "loss": 2.7467,
+      "step": 11824
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441766750948167e-06,
+      "loss": 2.5795,
+      "step": 11825
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4417193426042986e-06,
+      "loss": 2.8893,
+      "step": 11826
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.44167193426043e-06,
+      "loss": 3.8077,
+      "step": 11827
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441624525916561e-06,
+      "loss": 2.5445,
+      "step": 11828
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441577117572693e-06,
+      "loss": 3.4708,
+      "step": 11829
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441529709228824e-06,
+      "loss": 3.5235,
+      "step": 11830
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441482300884956e-06,
+      "loss": 3.6225,
+      "step": 11831
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4414348925410876e-06,
+      "loss": 3.0858,
+      "step": 11832
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441387484197219e-06,
+      "loss": 4.064,
+      "step": 11833
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.44134007585335e-06,
+      "loss": 4.0119,
+      "step": 11834
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441292667509482e-06,
+      "loss": 3.4811,
+      "step": 11835
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441245259165613e-06,
+      "loss": 3.1777,
+      "step": 11836
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441197850821745e-06,
+      "loss": 3.3919,
+      "step": 11837
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4411504424778765e-06,
+      "loss": 3.8198,
+      "step": 11838
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441103034134008e-06,
+      "loss": 2.5542,
+      "step": 11839
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.441055625790139e-06,
+      "loss": 3.7544,
+      "step": 11840
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4410082174462704e-06,
+      "loss": 2.9753,
+      "step": 11841
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440960809102402e-06,
+      "loss": 2.6297,
+      "step": 11842
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4409134007585334e-06,
+      "loss": 3.767,
+      "step": 11843
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440865992414665e-06,
+      "loss": 4.237,
+      "step": 11844
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4408185840707968e-06,
+      "loss": 3.5499,
+      "step": 11845
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440771175726928e-06,
+      "loss": 2.772,
+      "step": 11846
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4407237673830593e-06,
+      "loss": 3.5819,
+      "step": 11847
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440676359039191e-06,
+      "loss": 2.492,
+      "step": 11848
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4406289506953223e-06,
+      "loss": 3.2895,
+      "step": 11849
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4405815423514536e-06,
+      "loss": 4.4841,
+      "step": 11850
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4405341340075857e-06,
+      "loss": 3.4467,
+      "step": 11851
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440486725663717e-06,
+      "loss": 3.3266,
+      "step": 11852
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4404393173198483e-06,
+      "loss": 3.4857,
+      "step": 11853
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.44039190897598e-06,
+      "loss": 3.0062,
+      "step": 11854
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4403445006321113e-06,
+      "loss": 3.6339,
+      "step": 11855
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4402970922882426e-06,
+      "loss": 3.3633,
+      "step": 11856
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4402496839443743e-06,
+      "loss": 2.6616,
+      "step": 11857
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440202275600506e-06,
+      "loss": 3.1434,
+      "step": 11858
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4401548672566372e-06,
+      "loss": 3.2067,
+      "step": 11859
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.440107458912769e-06,
+      "loss": 4.0582,
+      "step": 11860
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4400600505689002e-06,
+      "loss": 3.3187,
+      "step": 11861
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4400126422250315e-06,
+      "loss": 3.5541,
+      "step": 11862
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4399652338811628e-06,
+      "loss": 2.9073,
+      "step": 11863
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4399178255372945e-06,
+      "loss": 3.1959,
+      "step": 11864
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.439870417193426e-06,
+      "loss": 3.3075,
+      "step": 11865
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4398230088495575e-06,
+      "loss": 3.5838,
+      "step": 11866
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.439775600505689e-06,
+      "loss": 3.2317,
+      "step": 11867
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4397281921618205e-06,
+      "loss": 3.3615,
+      "step": 11868
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4396807838179517e-06,
+      "loss": 3.2601,
+      "step": 11869
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4396333754740834e-06,
+      "loss": 2.2159,
+      "step": 11870
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.439585967130215e-06,
+      "loss": 3.6761,
+      "step": 11871
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4395385587863464e-06,
+      "loss": 2.915,
+      "step": 11872
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.439491150442478e-06,
+      "loss": 3.3904,
+      "step": 11873
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.4394437420986094e-06,
+      "loss": 3.1203,
+      "step": 11874
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4393963337547407e-06,
+      "loss": 3.605,
+      "step": 11875
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4393489254108724e-06,
+      "loss": 3.3596,
+      "step": 11876
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4393015170670037e-06,
+      "loss": 2.8437,
+      "step": 11877
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4392541087231354e-06,
+      "loss": 2.8894,
+      "step": 11878
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.439206700379267e-06,
+      "loss": 3.218,
+      "step": 11879
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4391592920353984e-06,
+      "loss": 3.5635,
+      "step": 11880
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4391118836915296e-06,
+      "loss": 3.8786,
+      "step": 11881
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4390644753476613e-06,
+      "loss": 3.1888,
+      "step": 11882
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4390170670037926e-06,
+      "loss": 3.4961,
+      "step": 11883
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4389696586599243e-06,
+      "loss": 3.4304,
+      "step": 11884
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4389222503160556e-06,
+      "loss": 3.5063,
+      "step": 11885
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4388748419721873e-06,
+      "loss": 3.0182,
+      "step": 11886
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4388274336283186e-06,
+      "loss": 2.6291,
+      "step": 11887
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.43878002528445e-06,
+      "loss": 2.6968,
+      "step": 11888
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4387326169405816e-06,
+      "loss": 2.7829,
+      "step": 11889
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.438685208596713e-06,
+      "loss": 3.3727,
+      "step": 11890
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4386378002528446e-06,
+      "loss": 3.1336,
+      "step": 11891
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4385903919089763e-06,
+      "loss": 3.1277,
+      "step": 11892
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4385429835651076e-06,
+      "loss": 3.0095,
+      "step": 11893
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.438495575221239e-06,
+      "loss": 3.739,
+      "step": 11894
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4384481668773705e-06,
+      "loss": 2.8679,
+      "step": 11895
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.438400758533502e-06,
+      "loss": 3.1218,
+      "step": 11896
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4383533501896335e-06,
+      "loss": 3.1295,
+      "step": 11897
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4383059418457652e-06,
+      "loss": 2.921,
+      "step": 11898
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4382585335018965e-06,
+      "loss": 2.918,
+      "step": 11899
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4382111251580278e-06,
+      "loss": 3.5204,
+      "step": 11900
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4381637168141595e-06,
+      "loss": 3.2276,
+      "step": 11901
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4381163084702908e-06,
+      "loss": 2.1559,
+      "step": 11902
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.438068900126422e-06,
+      "loss": 3.8094,
+      "step": 11903
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4380214917825538e-06,
+      "loss": 2.8267,
+      "step": 11904
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4379740834386855e-06,
+      "loss": 2.6767,
+      "step": 11905
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4379266750948167e-06,
+      "loss": 2.5969,
+      "step": 11906
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437879266750948e-06,
+      "loss": 3.2599,
+      "step": 11907
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4378318584070797e-06,
+      "loss": 3.4958,
+      "step": 11908
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437784450063211e-06,
+      "loss": 3.7398,
+      "step": 11909
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4377370417193427e-06,
+      "loss": 3.4471,
+      "step": 11910
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4376896333754744e-06,
+      "loss": 3.8594,
+      "step": 11911
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4376422250316057e-06,
+      "loss": 3.8627,
+      "step": 11912
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437594816687737e-06,
+      "loss": 2.9911,
+      "step": 11913
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4375474083438687e-06,
+      "loss": 3.6549,
+      "step": 11914
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4375e-06,
+      "loss": 2.9591,
+      "step": 11915
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4374525916561312e-06,
+      "loss": 3.2476,
+      "step": 11916
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437405183312263e-06,
+      "loss": 3.6382,
+      "step": 11917
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4373577749683946e-06,
+      "loss": 2.9789,
+      "step": 11918
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437310366624526e-06,
+      "loss": 3.0601,
+      "step": 11919
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4372629582806576e-06,
+      "loss": 2.5367,
+      "step": 11920
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437215549936789e-06,
+      "loss": 3.0202,
+      "step": 11921
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.43716814159292e-06,
+      "loss": 3.0055,
+      "step": 11922
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437120733249052e-06,
+      "loss": 3.2508,
+      "step": 11923
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4370733249051836e-06,
+      "loss": 3.86,
+      "step": 11924
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.437025916561315e-06,
+      "loss": 3.277,
+      "step": 11925
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436978508217446e-06,
+      "loss": 3.6731,
+      "step": 11926
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436931099873578e-06,
+      "loss": 3.8701,
+      "step": 11927
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436883691529709e-06,
+      "loss": 3.1555,
+      "step": 11928
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4368362831858404e-06,
+      "loss": 3.0902,
+      "step": 11929
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436788874841972e-06,
+      "loss": 3.3941,
+      "step": 11930
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436741466498104e-06,
+      "loss": 3.9487,
+      "step": 11931
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436694058154235e-06,
+      "loss": 3.3519,
+      "step": 11932
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436646649810367e-06,
+      "loss": 3.2969,
+      "step": 11933
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436599241466498e-06,
+      "loss": 2.8321,
+      "step": 11934
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4365518331226294e-06,
+      "loss": 3.4444,
+      "step": 11935
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.436504424778761e-06,
+      "loss": 3.5682,
+      "step": 11936
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.4364570164348928e-06,
+      "loss": 2.9705,
+      "step": 11937
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.436409608091024e-06,
+      "loss": 3.7723,
+      "step": 11938
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4363621997471558e-06,
+      "loss": 2.0956,
+      "step": 11939
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.436314791403287e-06,
+      "loss": 2.8285,
+      "step": 11940
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4362673830594183e-06,
+      "loss": 3.6536,
+      "step": 11941
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.43621997471555e-06,
+      "loss": 3.8364,
+      "step": 11942
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4361725663716813e-06,
+      "loss": 3.3032,
+      "step": 11943
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.436125158027813e-06,
+      "loss": 2.5471,
+      "step": 11944
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4360777496839447e-06,
+      "loss": 3.4804,
+      "step": 11945
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.436030341340076e-06,
+      "loss": 4.143,
+      "step": 11946
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4359829329962073e-06,
+      "loss": 3.8239,
+      "step": 11947
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4359355246523386e-06,
+      "loss": 2.9925,
+      "step": 11948
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4358881163084703e-06,
+      "loss": 3.3924,
+      "step": 11949
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435840707964602e-06,
+      "loss": 3.2689,
+      "step": 11950
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4357932996207333e-06,
+      "loss": 3.2389,
+      "step": 11951
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435745891276865e-06,
+      "loss": 2.783,
+      "step": 11952
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4356984829329962e-06,
+      "loss": 4.0453,
+      "step": 11953
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4356510745891275e-06,
+      "loss": 2.8191,
+      "step": 11954
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4356036662452592e-06,
+      "loss": 3.2934,
+      "step": 11955
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4355562579013905e-06,
+      "loss": 3.6298,
+      "step": 11956
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435508849557522e-06,
+      "loss": 3.3488,
+      "step": 11957
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435461441213654e-06,
+      "loss": 3.8382,
+      "step": 11958
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435414032869785e-06,
+      "loss": 3.1394,
+      "step": 11959
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4353666245259165e-06,
+      "loss": 3.3294,
+      "step": 11960
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435319216182048e-06,
+      "loss": 3.4402,
+      "step": 11961
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4352718078381795e-06,
+      "loss": 2.6107,
+      "step": 11962
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435224399494311e-06,
+      "loss": 2.9445,
+      "step": 11963
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435176991150443e-06,
+      "loss": 2.8868,
+      "step": 11964
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435129582806574e-06,
+      "loss": 3.2823,
+      "step": 11965
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4350821744627054e-06,
+      "loss": 2.6924,
+      "step": 11966
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.435034766118837e-06,
+      "loss": 3.755,
+      "step": 11967
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4349873577749684e-06,
+      "loss": 2.537,
+      "step": 11968
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4349399494310997e-06,
+      "loss": 2.5136,
+      "step": 11969
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4348925410872314e-06,
+      "loss": 3.1765,
+      "step": 11970
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.434845132743363e-06,
+      "loss": 3.7579,
+      "step": 11971
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4347977243994944e-06,
+      "loss": 3.5322,
+      "step": 11972
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4347503160556257e-06,
+      "loss": 2.5626,
+      "step": 11973
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4347029077117574e-06,
+      "loss": 3.7498,
+      "step": 11974
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4346554993678886e-06,
+      "loss": 3.1082,
+      "step": 11975
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4346080910240203e-06,
+      "loss": 2.8421,
+      "step": 11976
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.434560682680152e-06,
+      "loss": 3.4014,
+      "step": 11977
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4345132743362833e-06,
+      "loss": 2.9267,
+      "step": 11978
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4344658659924146e-06,
+      "loss": 2.579,
+      "step": 11979
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4344184576485463e-06,
+      "loss": 2.2227,
+      "step": 11980
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4343710493046776e-06,
+      "loss": 3.1881,
+      "step": 11981
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.434323640960809e-06,
+      "loss": 3.0057,
+      "step": 11982
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4342762326169406e-06,
+      "loss": 2.5724,
+      "step": 11983
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4342288242730723e-06,
+      "loss": 3.9005,
+      "step": 11984
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4341814159292036e-06,
+      "loss": 3.1072,
+      "step": 11985
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4341340075853353e-06,
+      "loss": 4.2105,
+      "step": 11986
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4340865992414665e-06,
+      "loss": 3.2737,
+      "step": 11987
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.434039190897598e-06,
+      "loss": 3.7537,
+      "step": 11988
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4339917825537295e-06,
+      "loss": 3.3788,
+      "step": 11989
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4339443742098612e-06,
+      "loss": 2.8568,
+      "step": 11990
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4338969658659925e-06,
+      "loss": 3.0577,
+      "step": 11991
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.433849557522124e-06,
+      "loss": 3.7293,
+      "step": 11992
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4338021491782555e-06,
+      "loss": 3.2958,
+      "step": 11993
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4337547408343868e-06,
+      "loss": 3.5055,
+      "step": 11994
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.433707332490518e-06,
+      "loss": 2.7957,
+      "step": 11995
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4336599241466498e-06,
+      "loss": 3.0821,
+      "step": 11996
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4336125158027815e-06,
+      "loss": 4.0833,
+      "step": 11997
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4335651074589127e-06,
+      "loss": 3.6215,
+      "step": 11998
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4335176991150445e-06,
+      "loss": 3.1969,
+      "step": 11999
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.4334702907711757e-06,
+      "loss": 3.9868,
+      "step": 12000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.433422882427307e-06,
+      "loss": 3.6151,
+      "step": 12001
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4333754740834387e-06,
+      "loss": 2.7592,
+      "step": 12002
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4333280657395704e-06,
+      "loss": 2.8543,
+      "step": 12003
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4332806573957017e-06,
+      "loss": 3.5867,
+      "step": 12004
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4332332490518334e-06,
+      "loss": 2.6666,
+      "step": 12005
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4331858407079647e-06,
+      "loss": 3.3501,
+      "step": 12006
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.433138432364096e-06,
+      "loss": 3.1551,
+      "step": 12007
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4330910240202277e-06,
+      "loss": 3.5485,
+      "step": 12008
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.433043615676359e-06,
+      "loss": 3.3798,
+      "step": 12009
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4329962073324907e-06,
+      "loss": 3.2362,
+      "step": 12010
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4329487989886224e-06,
+      "loss": 3.311,
+      "step": 12011
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4329013906447536e-06,
+      "loss": 2.5504,
+      "step": 12012
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432853982300885e-06,
+      "loss": 3.8627,
+      "step": 12013
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432806573957016e-06,
+      "loss": 3.1954,
+      "step": 12014
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432759165613148e-06,
+      "loss": 3.3473,
+      "step": 12015
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4327117572692796e-06,
+      "loss": 3.0855,
+      "step": 12016
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432664348925411e-06,
+      "loss": 3.4181,
+      "step": 12017
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4326169405815426e-06,
+      "loss": 3.1993,
+      "step": 12018
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432569532237674e-06,
+      "loss": 2.8454,
+      "step": 12019
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432522123893805e-06,
+      "loss": 3.3742,
+      "step": 12020
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432474715549937e-06,
+      "loss": 3.2377,
+      "step": 12021
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432427307206068e-06,
+      "loss": 3.7781,
+      "step": 12022
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4323798988622e-06,
+      "loss": 3.755,
+      "step": 12023
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4323324905183315e-06,
+      "loss": 3.6718,
+      "step": 12024
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432285082174463e-06,
+      "loss": 3.1193,
+      "step": 12025
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432237673830594e-06,
+      "loss": 3.7216,
+      "step": 12026
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432190265486726e-06,
+      "loss": 3.7893,
+      "step": 12027
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432142857142857e-06,
+      "loss": 2.6192,
+      "step": 12028
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.432095448798989e-06,
+      "loss": 2.7785,
+      "step": 12029
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4320480404551205e-06,
+      "loss": 2.4883,
+      "step": 12030
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4320006321112518e-06,
+      "loss": 4.0336,
+      "step": 12031
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431953223767383e-06,
+      "loss": 3.5814,
+      "step": 12032
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4319058154235148e-06,
+      "loss": 3.5155,
+      "step": 12033
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431858407079646e-06,
+      "loss": 3.0259,
+      "step": 12034
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4318109987357773e-06,
+      "loss": 2.8153,
+      "step": 12035
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431763590391909e-06,
+      "loss": 4.0988,
+      "step": 12036
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4317161820480407e-06,
+      "loss": 2.9125,
+      "step": 12037
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431668773704172e-06,
+      "loss": 4.2878,
+      "step": 12038
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4316213653603033e-06,
+      "loss": 2.7723,
+      "step": 12039
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431573957016435e-06,
+      "loss": 3.7453,
+      "step": 12040
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4315265486725663e-06,
+      "loss": 3.2503,
+      "step": 12041
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431479140328698e-06,
+      "loss": 3.4633,
+      "step": 12042
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4314317319848297e-06,
+      "loss": 3.1996,
+      "step": 12043
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431384323640961e-06,
+      "loss": 2.8778,
+      "step": 12044
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4313369152970922e-06,
+      "loss": 3.7091,
+      "step": 12045
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431289506953224e-06,
+      "loss": 3.1728,
+      "step": 12046
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4312420986093552e-06,
+      "loss": 2.9827,
+      "step": 12047
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4311946902654865e-06,
+      "loss": 3.3921,
+      "step": 12048
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4311472819216182e-06,
+      "loss": 2.7573,
+      "step": 12049
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.43109987357775e-06,
+      "loss": 3.1464,
+      "step": 12050
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431052465233881e-06,
+      "loss": 2.8265,
+      "step": 12051
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.431005056890013e-06,
+      "loss": 2.574,
+      "step": 12052
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.430957648546144e-06,
+      "loss": 3.9477,
+      "step": 12053
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4309102402022755e-06,
+      "loss": 3.329,
+      "step": 12054
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.430862831858407e-06,
+      "loss": 3.8842,
+      "step": 12055
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.430815423514539e-06,
+      "loss": 3.198,
+      "step": 12056
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.43076801517067e-06,
+      "loss": 2.8058,
+      "step": 12057
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4307206068268014e-06,
+      "loss": 2.3513,
+      "step": 12058
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.430673198482933e-06,
+      "loss": 4.0911,
+      "step": 12059
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4306257901390644e-06,
+      "loss": 3.1438,
+      "step": 12060
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4305783817951957e-06,
+      "loss": 3.5802,
+      "step": 12061
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4305309734513274e-06,
+      "loss": 3.476,
+      "step": 12062
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.430483565107459e-06,
+      "loss": 3.1568,
+      "step": 12063
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.4304361567635904e-06,
+      "loss": 2.7948,
+      "step": 12064
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.430388748419722e-06,
+      "loss": 3.4723,
+      "step": 12065
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4303413400758534e-06,
+      "loss": 2.4804,
+      "step": 12066
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4302939317319846e-06,
+      "loss": 3.022,
+      "step": 12067
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4302465233881164e-06,
+      "loss": 2.7062,
+      "step": 12068
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.430199115044248e-06,
+      "loss": 3.8834,
+      "step": 12069
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4301517067003793e-06,
+      "loss": 4.2561,
+      "step": 12070
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.430104298356511e-06,
+      "loss": 2.5545,
+      "step": 12071
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4300568900126423e-06,
+      "loss": 3.0535,
+      "step": 12072
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4300094816687736e-06,
+      "loss": 3.6273,
+      "step": 12073
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4299620733249053e-06,
+      "loss": 3.3144,
+      "step": 12074
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4299146649810366e-06,
+      "loss": 3.4025,
+      "step": 12075
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4298672566371683e-06,
+      "loss": 3.283,
+      "step": 12076
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4298198482933e-06,
+      "loss": 3.1984,
+      "step": 12077
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4297724399494313e-06,
+      "loss": 3.056,
+      "step": 12078
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4297250316055626e-06,
+      "loss": 2.9207,
+      "step": 12079
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.429677623261694e-06,
+      "loss": 3.1668,
+      "step": 12080
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4296302149178255e-06,
+      "loss": 3.4223,
+      "step": 12081
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4295828065739572e-06,
+      "loss": 3.2968,
+      "step": 12082
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4295353982300885e-06,
+      "loss": 2.3869,
+      "step": 12083
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4294879898862202e-06,
+      "loss": 3.345,
+      "step": 12084
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4294405815423515e-06,
+      "loss": 2.716,
+      "step": 12085
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.429393173198483e-06,
+      "loss": 2.8114,
+      "step": 12086
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4293457648546145e-06,
+      "loss": 2.5998,
+      "step": 12087
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4292983565107458e-06,
+      "loss": 2.8822,
+      "step": 12088
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4292509481668775e-06,
+      "loss": 3.768,
+      "step": 12089
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.429203539823009e-06,
+      "loss": 2.5564,
+      "step": 12090
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4291561314791405e-06,
+      "loss": 3.942,
+      "step": 12091
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4291087231352717e-06,
+      "loss": 3.0628,
+      "step": 12092
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4290613147914034e-06,
+      "loss": 3.3747,
+      "step": 12093
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4290139064475347e-06,
+      "loss": 3.0665,
+      "step": 12094
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4289664981036664e-06,
+      "loss": 3.1168,
+      "step": 12095
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428919089759798e-06,
+      "loss": 2.7272,
+      "step": 12096
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4288716814159294e-06,
+      "loss": 3.4157,
+      "step": 12097
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4288242730720607e-06,
+      "loss": 3.6157,
+      "step": 12098
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4287768647281924e-06,
+      "loss": 4.0753,
+      "step": 12099
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4287294563843237e-06,
+      "loss": 3.0569,
+      "step": 12100
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428682048040455e-06,
+      "loss": 3.0995,
+      "step": 12101
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4286346396965867e-06,
+      "loss": 2.5323,
+      "step": 12102
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4285872313527184e-06,
+      "loss": 3.0703,
+      "step": 12103
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4285398230088496e-06,
+      "loss": 3.1765,
+      "step": 12104
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428492414664981e-06,
+      "loss": 3.8854,
+      "step": 12105
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4284450063211126e-06,
+      "loss": 3.4209,
+      "step": 12106
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428397597977244e-06,
+      "loss": 3.3551,
+      "step": 12107
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4283501896333756e-06,
+      "loss": 3.6002,
+      "step": 12108
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4283027812895073e-06,
+      "loss": 3.5192,
+      "step": 12109
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4282553729456386e-06,
+      "loss": 2.9917,
+      "step": 12110
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.42820796460177e-06,
+      "loss": 3.2607,
+      "step": 12111
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4281605562579016e-06,
+      "loss": 3.1799,
+      "step": 12112
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428113147914033e-06,
+      "loss": 2.9842,
+      "step": 12113
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428065739570164e-06,
+      "loss": 2.8595,
+      "step": 12114
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.428018331226296e-06,
+      "loss": 3.624,
+      "step": 12115
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4279709228824276e-06,
+      "loss": 2.186,
+      "step": 12116
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.427923514538559e-06,
+      "loss": 2.7965,
+      "step": 12117
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4278761061946905e-06,
+      "loss": 3.3846,
+      "step": 12118
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.427828697850822e-06,
+      "loss": 3.1932,
+      "step": 12119
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.427781289506953e-06,
+      "loss": 3.0961,
+      "step": 12120
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.427733881163085e-06,
+      "loss": 2.6241,
+      "step": 12121
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4276864728192165e-06,
+      "loss": 3.2657,
+      "step": 12122
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4276390644753478e-06,
+      "loss": 3.0364,
+      "step": 12123
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.427591656131479e-06,
+      "loss": 3.3051,
+      "step": 12124
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4275442477876108e-06,
+      "loss": 3.7646,
+      "step": 12125
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.427496839443742e-06,
+      "loss": 3.3537,
+      "step": 12126
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.4274494310998733e-06,
+      "loss": 4.2529,
+      "step": 12127
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.427402022756005e-06,
+      "loss": 3.67,
+      "step": 12128
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4273546144121367e-06,
+      "loss": 2.3701,
+      "step": 12129
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.427307206068268e-06,
+      "loss": 3.6689,
+      "step": 12130
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4272597977243997e-06,
+      "loss": 3.5,
+      "step": 12131
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.427212389380531e-06,
+      "loss": 3.113,
+      "step": 12132
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4271649810366623e-06,
+      "loss": 2.9346,
+      "step": 12133
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.427117572692794e-06,
+      "loss": 2.5601,
+      "step": 12134
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4270701643489253e-06,
+      "loss": 2.8163,
+      "step": 12135
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.427022756005057e-06,
+      "loss": 3.9651,
+      "step": 12136
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4269753476611887e-06,
+      "loss": 3.5046,
+      "step": 12137
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.42692793931732e-06,
+      "loss": 3.6971,
+      "step": 12138
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4268805309734512e-06,
+      "loss": 3.8147,
+      "step": 12139
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426833122629583e-06,
+      "loss": 3.5535,
+      "step": 12140
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4267857142857142e-06,
+      "loss": 3.2348,
+      "step": 12141
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426738305941846e-06,
+      "loss": 2.4091,
+      "step": 12142
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4266908975979776e-06,
+      "loss": 3.5822,
+      "step": 12143
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426643489254109e-06,
+      "loss": 3.6622,
+      "step": 12144
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.42659608091024e-06,
+      "loss": 3.1424,
+      "step": 12145
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4265486725663715e-06,
+      "loss": 3.8154,
+      "step": 12146
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426501264222503e-06,
+      "loss": 3.6053,
+      "step": 12147
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4264538558786345e-06,
+      "loss": 1.6627,
+      "step": 12148
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426406447534766e-06,
+      "loss": 2.9208,
+      "step": 12149
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426359039190898e-06,
+      "loss": 2.4979,
+      "step": 12150
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426311630847029e-06,
+      "loss": 3.4541,
+      "step": 12151
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4262642225031604e-06,
+      "loss": 3.7522,
+      "step": 12152
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426216814159292e-06,
+      "loss": 3.5139,
+      "step": 12153
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4261694058154234e-06,
+      "loss": 3.7942,
+      "step": 12154
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426121997471555e-06,
+      "loss": 3.2532,
+      "step": 12155
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426074589127687e-06,
+      "loss": 4.0171,
+      "step": 12156
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.426027180783818e-06,
+      "loss": 2.3265,
+      "step": 12157
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4259797724399494e-06,
+      "loss": 4.3902,
+      "step": 12158
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.425932364096081e-06,
+      "loss": 4.1181,
+      "step": 12159
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4258849557522124e-06,
+      "loss": 3.5722,
+      "step": 12160
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4258375474083436e-06,
+      "loss": 3.7984,
+      "step": 12161
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4257901390644753e-06,
+      "loss": 2.8198,
+      "step": 12162
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.425742730720607e-06,
+      "loss": 2.2285,
+      "step": 12163
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4256953223767383e-06,
+      "loss": 2.3816,
+      "step": 12164
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.42564791403287e-06,
+      "loss": 2.7589,
+      "step": 12165
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4256005056890013e-06,
+      "loss": 3.4389,
+      "step": 12166
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4255530973451326e-06,
+      "loss": 3.6231,
+      "step": 12167
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4255056890012643e-06,
+      "loss": 2.991,
+      "step": 12168
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.425458280657396e-06,
+      "loss": 3.4563,
+      "step": 12169
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4254108723135273e-06,
+      "loss": 3.1774,
+      "step": 12170
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4253634639696586e-06,
+      "loss": 2.6962,
+      "step": 12171
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4253160556257903e-06,
+      "loss": 2.9135,
+      "step": 12172
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4252686472819215e-06,
+      "loss": 3.4228,
+      "step": 12173
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.425221238938053e-06,
+      "loss": 2.9847,
+      "step": 12174
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4251738305941845e-06,
+      "loss": 2.8079,
+      "step": 12175
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4251264222503162e-06,
+      "loss": 2.1586,
+      "step": 12176
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4250790139064475e-06,
+      "loss": 3.3664,
+      "step": 12177
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4250316055625792e-06,
+      "loss": 3.3303,
+      "step": 12178
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4249841972187105e-06,
+      "loss": 3.6832,
+      "step": 12179
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4249367888748418e-06,
+      "loss": 4.7981,
+      "step": 12180
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4248893805309735e-06,
+      "loss": 2.8423,
+      "step": 12181
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.424841972187105e-06,
+      "loss": 3.2013,
+      "step": 12182
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4247945638432365e-06,
+      "loss": 3.4472,
+      "step": 12183
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.424747155499368e-06,
+      "loss": 3.4181,
+      "step": 12184
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4246997471554995e-06,
+      "loss": 3.9481,
+      "step": 12185
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4246523388116307e-06,
+      "loss": 3.0971,
+      "step": 12186
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4246049304677624e-06,
+      "loss": 3.3078,
+      "step": 12187
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4245575221238937e-06,
+      "loss": 4.0781,
+      "step": 12188
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4245101137800254e-06,
+      "loss": 2.4233,
+      "step": 12189
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4244627054361567e-06,
+      "loss": 3.0515,
+      "step": 12190
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.4244152970922884e-06,
+      "loss": 3.859,
+      "step": 12191
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4243678887484197e-06,
+      "loss": 3.731,
+      "step": 12192
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.424320480404551e-06,
+      "loss": 2.857,
+      "step": 12193
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4242730720606827e-06,
+      "loss": 3.5267,
+      "step": 12194
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4242256637168144e-06,
+      "loss": 3.9697,
+      "step": 12195
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4241782553729457e-06,
+      "loss": 3.3428,
+      "step": 12196
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4241308470290774e-06,
+      "loss": 3.0268,
+      "step": 12197
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4240834386852086e-06,
+      "loss": 3.2158,
+      "step": 12198
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.42403603034134e-06,
+      "loss": 3.479,
+      "step": 12199
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4239886219974716e-06,
+      "loss": 3.4638,
+      "step": 12200
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423941213653603e-06,
+      "loss": 2.9981,
+      "step": 12201
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4238938053097346e-06,
+      "loss": 3.3874,
+      "step": 12202
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4238463969658663e-06,
+      "loss": 3.5927,
+      "step": 12203
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4237989886219976e-06,
+      "loss": 2.2544,
+      "step": 12204
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423751580278129e-06,
+      "loss": 3.0152,
+      "step": 12205
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4237041719342606e-06,
+      "loss": 3.2902,
+      "step": 12206
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423656763590392e-06,
+      "loss": 3.6981,
+      "step": 12207
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4236093552465236e-06,
+      "loss": 2.7527,
+      "step": 12208
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4235619469026553e-06,
+      "loss": 3.447,
+      "step": 12209
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4235145385587865e-06,
+      "loss": 2.6406,
+      "step": 12210
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423467130214918e-06,
+      "loss": 3.4915,
+      "step": 12211
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423419721871049e-06,
+      "loss": 3.6055,
+      "step": 12212
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423372313527181e-06,
+      "loss": 2.7915,
+      "step": 12213
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423324905183312e-06,
+      "loss": 3.5324,
+      "step": 12214
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423277496839444e-06,
+      "loss": 3.0386,
+      "step": 12215
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4232300884955755e-06,
+      "loss": 2.5292,
+      "step": 12216
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4231826801517068e-06,
+      "loss": 2.0389,
+      "step": 12217
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423135271807838e-06,
+      "loss": 4.0806,
+      "step": 12218
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4230878634639698e-06,
+      "loss": 2.6232,
+      "step": 12219
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.423040455120101e-06,
+      "loss": 2.807,
+      "step": 12220
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4229930467762328e-06,
+      "loss": 2.7643,
+      "step": 12221
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4229456384323645e-06,
+      "loss": 2.9369,
+      "step": 12222
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4228982300884957e-06,
+      "loss": 2.393,
+      "step": 12223
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422850821744627e-06,
+      "loss": 3.1947,
+      "step": 12224
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4228034134007587e-06,
+      "loss": 4.0728,
+      "step": 12225
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.42275600505689e-06,
+      "loss": 3.0271,
+      "step": 12226
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4227085967130213e-06,
+      "loss": 3.302,
+      "step": 12227
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422661188369153e-06,
+      "loss": 3.0553,
+      "step": 12228
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4226137800252847e-06,
+      "loss": 3.393,
+      "step": 12229
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422566371681416e-06,
+      "loss": 3.2921,
+      "step": 12230
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4225189633375477e-06,
+      "loss": 3.8432,
+      "step": 12231
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422471554993679e-06,
+      "loss": 3.0606,
+      "step": 12232
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4224241466498102e-06,
+      "loss": 3.2406,
+      "step": 12233
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422376738305942e-06,
+      "loss": 2.5239,
+      "step": 12234
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4223293299620736e-06,
+      "loss": 3.2125,
+      "step": 12235
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422281921618205e-06,
+      "loss": 3.3793,
+      "step": 12236
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422234513274336e-06,
+      "loss": 3.2972,
+      "step": 12237
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422187104930468e-06,
+      "loss": 3.8428,
+      "step": 12238
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422139696586599e-06,
+      "loss": 4.4183,
+      "step": 12239
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4220922882427305e-06,
+      "loss": 3.7541,
+      "step": 12240
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.422044879898862e-06,
+      "loss": 4.0042,
+      "step": 12241
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421997471554994e-06,
+      "loss": 3.509,
+      "step": 12242
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421950063211125e-06,
+      "loss": 3.6087,
+      "step": 12243
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421902654867257e-06,
+      "loss": 4.1486,
+      "step": 12244
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421855246523388e-06,
+      "loss": 2.8277,
+      "step": 12245
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4218078381795194e-06,
+      "loss": 3.4293,
+      "step": 12246
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421760429835651e-06,
+      "loss": 3.3847,
+      "step": 12247
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421713021491783e-06,
+      "loss": 3.7601,
+      "step": 12248
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421665613147914e-06,
+      "loss": 3.4007,
+      "step": 12249
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421618204804046e-06,
+      "loss": 3.1724,
+      "step": 12250
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.421570796460177e-06,
+      "loss": 4.35,
+      "step": 12251
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4215233881163084e-06,
+      "loss": 2.6639,
+      "step": 12252
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.42147597977244e-06,
+      "loss": 2.9626,
+      "step": 12253
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4214285714285714e-06,
+      "loss": 3.1348,
+      "step": 12254
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.421381163084703e-06,
+      "loss": 3.1453,
+      "step": 12255
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4213337547408343e-06,
+      "loss": 2.3161,
+      "step": 12256
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.421286346396966e-06,
+      "loss": 2.3291,
+      "step": 12257
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4212389380530973e-06,
+      "loss": 3.3363,
+      "step": 12258
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4211915297092286e-06,
+      "loss": 3.083,
+      "step": 12259
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4211441213653603e-06,
+      "loss": 3.3824,
+      "step": 12260
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.421096713021492e-06,
+      "loss": 3.897,
+      "step": 12261
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4210493046776233e-06,
+      "loss": 3.5667,
+      "step": 12262
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.421001896333755e-06,
+      "loss": 3.5895,
+      "step": 12263
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4209544879898863e-06,
+      "loss": 2.5513,
+      "step": 12264
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4209070796460176e-06,
+      "loss": 3.2952,
+      "step": 12265
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4208596713021493e-06,
+      "loss": 3.086,
+      "step": 12266
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4208122629582805e-06,
+      "loss": 3.4531,
+      "step": 12267
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4207648546144122e-06,
+      "loss": 2.8693,
+      "step": 12268
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.420717446270544e-06,
+      "loss": 3.5335,
+      "step": 12269
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4206700379266752e-06,
+      "loss": 3.6661,
+      "step": 12270
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4206226295828065e-06,
+      "loss": 2.6698,
+      "step": 12271
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4205752212389382e-06,
+      "loss": 4.123,
+      "step": 12272
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4205278128950695e-06,
+      "loss": 3.7043,
+      "step": 12273
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.420480404551201e-06,
+      "loss": 2.5045,
+      "step": 12274
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.420432996207333e-06,
+      "loss": 3.3226,
+      "step": 12275
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.420385587863464e-06,
+      "loss": 2.8235,
+      "step": 12276
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4203381795195955e-06,
+      "loss": 2.7001,
+      "step": 12277
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4202907711757267e-06,
+      "loss": 3.1997,
+      "step": 12278
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4202433628318584e-06,
+      "loss": 4.3335,
+      "step": 12279
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4201959544879897e-06,
+      "loss": 3.2439,
+      "step": 12280
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4201485461441214e-06,
+      "loss": 2.391,
+      "step": 12281
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.420101137800253e-06,
+      "loss": 3.9723,
+      "step": 12282
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4200537294563844e-06,
+      "loss": 3.8741,
+      "step": 12283
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4200063211125157e-06,
+      "loss": 2.7133,
+      "step": 12284
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4199589127686474e-06,
+      "loss": 3.1744,
+      "step": 12285
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4199115044247787e-06,
+      "loss": 2.7781,
+      "step": 12286
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4198640960809104e-06,
+      "loss": 3.1054,
+      "step": 12287
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.419816687737042e-06,
+      "loss": 2.8899,
+      "step": 12288
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4197692793931734e-06,
+      "loss": 3.7072,
+      "step": 12289
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4197218710493047e-06,
+      "loss": 3.5896,
+      "step": 12290
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4196744627054364e-06,
+      "loss": 3.0638,
+      "step": 12291
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4196270543615676e-06,
+      "loss": 4.0218,
+      "step": 12292
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.419579646017699e-06,
+      "loss": 3.0997,
+      "step": 12293
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4195322376738306e-06,
+      "loss": 3.5719,
+      "step": 12294
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4194848293299623e-06,
+      "loss": 4.0064,
+      "step": 12295
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4194374209860936e-06,
+      "loss": 3.2439,
+      "step": 12296
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4193900126422253e-06,
+      "loss": 3.781,
+      "step": 12297
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4193426042983566e-06,
+      "loss": 2.7341,
+      "step": 12298
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.419295195954488e-06,
+      "loss": 3.2635,
+      "step": 12299
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4192477876106196e-06,
+      "loss": 2.8736,
+      "step": 12300
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4192003792667513e-06,
+      "loss": 3.8784,
+      "step": 12301
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4191529709228826e-06,
+      "loss": 3.3937,
+      "step": 12302
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.419105562579014e-06,
+      "loss": 2.9852,
+      "step": 12303
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4190581542351455e-06,
+      "loss": 3.7189,
+      "step": 12304
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.419010745891277e-06,
+      "loss": 3.0445,
+      "step": 12305
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.418963337547408e-06,
+      "loss": 3.7273,
+      "step": 12306
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.41891592920354e-06,
+      "loss": 2.5613,
+      "step": 12307
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4188685208596715e-06,
+      "loss": 3.6379,
+      "step": 12308
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.418821112515803e-06,
+      "loss": 3.469,
+      "step": 12309
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4187737041719345e-06,
+      "loss": 3.5824,
+      "step": 12310
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4187262958280658e-06,
+      "loss": 3.6794,
+      "step": 12311
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.418678887484197e-06,
+      "loss": 3.2105,
+      "step": 12312
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4186314791403288e-06,
+      "loss": 3.0904,
+      "step": 12313
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4185840707964605e-06,
+      "loss": 2.574,
+      "step": 12314
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4185366624525917e-06,
+      "loss": 3.4733,
+      "step": 12315
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4184892541087234e-06,
+      "loss": 3.5947,
+      "step": 12316
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.4184418457648547e-06,
+      "loss": 2.8012,
+      "step": 12317
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.418394437420986e-06,
+      "loss": 3.3976,
+      "step": 12318
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4183470290771177e-06,
+      "loss": 3.1379,
+      "step": 12319
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.418299620733249e-06,
+      "loss": 3.7416,
+      "step": 12320
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4182522123893807e-06,
+      "loss": 2.8154,
+      "step": 12321
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.418204804045512e-06,
+      "loss": 2.7352,
+      "step": 12322
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4181573957016437e-06,
+      "loss": 3.4755,
+      "step": 12323
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.418109987357775e-06,
+      "loss": 2.7402,
+      "step": 12324
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4180625790139062e-06,
+      "loss": 3.5886,
+      "step": 12325
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.418015170670038e-06,
+      "loss": 2.9181,
+      "step": 12326
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4179677623261697e-06,
+      "loss": 3.1744,
+      "step": 12327
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417920353982301e-06,
+      "loss": 3.086,
+      "step": 12328
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4178729456384326e-06,
+      "loss": 2.9724,
+      "step": 12329
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417825537294564e-06,
+      "loss": 2.6366,
+      "step": 12330
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417778128950695e-06,
+      "loss": 3.2479,
+      "step": 12331
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417730720606827e-06,
+      "loss": 3.9179,
+      "step": 12332
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417683312262958e-06,
+      "loss": 2.5791,
+      "step": 12333
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.41763590391909e-06,
+      "loss": 3.5023,
+      "step": 12334
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4175884955752216e-06,
+      "loss": 3.3323,
+      "step": 12335
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417541087231353e-06,
+      "loss": 2.9282,
+      "step": 12336
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417493678887484e-06,
+      "loss": 3.4803,
+      "step": 12337
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417446270543616e-06,
+      "loss": 2.2496,
+      "step": 12338
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417398862199747e-06,
+      "loss": 3.1114,
+      "step": 12339
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417351453855879e-06,
+      "loss": 4.0392,
+      "step": 12340
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4173040455120105e-06,
+      "loss": 3.9149,
+      "step": 12341
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417256637168142e-06,
+      "loss": 2.405,
+      "step": 12342
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417209228824273e-06,
+      "loss": 3.5124,
+      "step": 12343
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4171618204804044e-06,
+      "loss": 2.3571,
+      "step": 12344
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417114412136536e-06,
+      "loss": 3.7107,
+      "step": 12345
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4170670037926674e-06,
+      "loss": 2.5691,
+      "step": 12346
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.417019595448799e-06,
+      "loss": 3.2463,
+      "step": 12347
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4169721871049308e-06,
+      "loss": 3.5623,
+      "step": 12348
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416924778761062e-06,
+      "loss": 3.9104,
+      "step": 12349
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4168773704171933e-06,
+      "loss": 3.1971,
+      "step": 12350
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416829962073325e-06,
+      "loss": 2.251,
+      "step": 12351
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4167825537294563e-06,
+      "loss": 3.2553,
+      "step": 12352
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416735145385588e-06,
+      "loss": 2.9607,
+      "step": 12353
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4166877370417197e-06,
+      "loss": 3.0531,
+      "step": 12354
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416640328697851e-06,
+      "loss": 3.9656,
+      "step": 12355
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4165929203539823e-06,
+      "loss": 2.6649,
+      "step": 12356
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416545512010114e-06,
+      "loss": 4.1395,
+      "step": 12357
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4164981036662453e-06,
+      "loss": 2.9985,
+      "step": 12358
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4164506953223766e-06,
+      "loss": 2.5959,
+      "step": 12359
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4164032869785083e-06,
+      "loss": 2.9555,
+      "step": 12360
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.41635587863464e-06,
+      "loss": 3.4799,
+      "step": 12361
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4163084702907712e-06,
+      "loss": 3.2701,
+      "step": 12362
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4162610619469025e-06,
+      "loss": 3.0748,
+      "step": 12363
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4162136536030342e-06,
+      "loss": 3.6009,
+      "step": 12364
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4161662452591655e-06,
+      "loss": 3.0342,
+      "step": 12365
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416118836915297e-06,
+      "loss": 2.2772,
+      "step": 12366
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.416071428571429e-06,
+      "loss": 2.4128,
+      "step": 12367
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.41602402022756e-06,
+      "loss": 3.2136,
+      "step": 12368
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4159766118836915e-06,
+      "loss": 2.7804,
+      "step": 12369
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.415929203539823e-06,
+      "loss": 2.8356,
+      "step": 12370
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4158817951959545e-06,
+      "loss": 3.378,
+      "step": 12371
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4158343868520857e-06,
+      "loss": 3.2881,
+      "step": 12372
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4157869785082174e-06,
+      "loss": 4.2315,
+      "step": 12373
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.415739570164349e-06,
+      "loss": 3.2143,
+      "step": 12374
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4156921618204804e-06,
+      "loss": 3.7016,
+      "step": 12375
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.415644753476612e-06,
+      "loss": 3.2718,
+      "step": 12376
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4155973451327434e-06,
+      "loss": 3.7643,
+      "step": 12377
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4155499367888747e-06,
+      "loss": 3.2037,
+      "step": 12378
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4155025284450064e-06,
+      "loss": 3.6176,
+      "step": 12379
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.415455120101138e-06,
+      "loss": 3.8184,
+      "step": 12380
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4154077117572694e-06,
+      "loss": 3.1571,
+      "step": 12381
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.415360303413401e-06,
+      "loss": 3.5079,
+      "step": 12382
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4153128950695324e-06,
+      "loss": 2.8959,
+      "step": 12383
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4152654867256636e-06,
+      "loss": 3.2004,
+      "step": 12384
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.415218078381795e-06,
+      "loss": 3.6816,
+      "step": 12385
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4151706700379266e-06,
+      "loss": 4.2548,
+      "step": 12386
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4151232616940583e-06,
+      "loss": 3.8574,
+      "step": 12387
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4150758533501896e-06,
+      "loss": 3.4176,
+      "step": 12388
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4150284450063213e-06,
+      "loss": 3.2011,
+      "step": 12389
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4149810366624526e-06,
+      "loss": 3.3015,
+      "step": 12390
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.414933628318584e-06,
+      "loss": 3.5977,
+      "step": 12391
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4148862199747156e-06,
+      "loss": 4.0509,
+      "step": 12392
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4148388116308473e-06,
+      "loss": 3.166,
+      "step": 12393
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4147914032869786e-06,
+      "loss": 2.9721,
+      "step": 12394
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4147439949431103e-06,
+      "loss": 3.0225,
+      "step": 12395
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4146965865992416e-06,
+      "loss": 3.1257,
+      "step": 12396
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.414649178255373e-06,
+      "loss": 3.4658,
+      "step": 12397
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4146017699115045e-06,
+      "loss": 3.9161,
+      "step": 12398
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.414554361567636e-06,
+      "loss": 3.4512,
+      "step": 12399
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4145069532237675e-06,
+      "loss": 3.3076,
+      "step": 12400
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4144595448798992e-06,
+      "loss": 2.9146,
+      "step": 12401
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4144121365360305e-06,
+      "loss": 3.8489,
+      "step": 12402
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4143647281921618e-06,
+      "loss": 2.7587,
+      "step": 12403
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4143173198482935e-06,
+      "loss": 4.0239,
+      "step": 12404
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4142699115044248e-06,
+      "loss": 3.9766,
+      "step": 12405
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.414222503160556e-06,
+      "loss": 3.1033,
+      "step": 12406
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4141750948166878e-06,
+      "loss": 3.3171,
+      "step": 12407
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4141276864728195e-06,
+      "loss": 3.8488,
+      "step": 12408
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4140802781289507e-06,
+      "loss": 3.7318,
+      "step": 12409
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.414032869785082e-06,
+      "loss": 3.0254,
+      "step": 12410
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4139854614412137e-06,
+      "loss": 3.9543,
+      "step": 12411
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413938053097345e-06,
+      "loss": 3.2761,
+      "step": 12412
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4138906447534767e-06,
+      "loss": 2.9393,
+      "step": 12413
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4138432364096084e-06,
+      "loss": 3.1164,
+      "step": 12414
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4137958280657397e-06,
+      "loss": 2.8826,
+      "step": 12415
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413748419721871e-06,
+      "loss": 3.4898,
+      "step": 12416
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4137010113780027e-06,
+      "loss": 2.7814,
+      "step": 12417
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413653603034134e-06,
+      "loss": 3.1082,
+      "step": 12418
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4136061946902652e-06,
+      "loss": 3.4719,
+      "step": 12419
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413558786346397e-06,
+      "loss": 3.3937,
+      "step": 12420
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4135113780025286e-06,
+      "loss": 3.5547,
+      "step": 12421
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.41346396965866e-06,
+      "loss": 3.4943,
+      "step": 12422
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4134165613147916e-06,
+      "loss": 3.787,
+      "step": 12423
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413369152970923e-06,
+      "loss": 3.353,
+      "step": 12424
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413321744627054e-06,
+      "loss": 4.1186,
+      "step": 12425
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413274336283186e-06,
+      "loss": 4.1824,
+      "step": 12426
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4132269279393176e-06,
+      "loss": 3.6922,
+      "step": 12427
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413179519595449e-06,
+      "loss": 3.2885,
+      "step": 12428
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.41313211125158e-06,
+      "loss": 2.962,
+      "step": 12429
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413084702907712e-06,
+      "loss": 3.0147,
+      "step": 12430
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.413037294563843e-06,
+      "loss": 4.0394,
+      "step": 12431
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4129898862199744e-06,
+      "loss": 2.4436,
+      "step": 12432
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412942477876106e-06,
+      "loss": 3.1205,
+      "step": 12433
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412895069532238e-06,
+      "loss": 3.7261,
+      "step": 12434
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412847661188369e-06,
+      "loss": 3.0734,
+      "step": 12435
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412800252844501e-06,
+      "loss": 3.3888,
+      "step": 12436
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412752844500632e-06,
+      "loss": 2.6708,
+      "step": 12437
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4127054361567634e-06,
+      "loss": 2.5499,
+      "step": 12438
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412658027812895e-06,
+      "loss": 2.9706,
+      "step": 12439
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4126106194690268e-06,
+      "loss": 3.824,
+      "step": 12440
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412563211125158e-06,
+      "loss": 3.4349,
+      "step": 12441
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4125158027812898e-06,
+      "loss": 3.4562,
+      "step": 12442
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.412468394437421e-06,
+      "loss": 3.691,
+      "step": 12443
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.4124209860935523e-06,
+      "loss": 2.3183,
+      "step": 12444
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.412373577749684e-06,
+      "loss": 3.2365,
+      "step": 12445
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4123261694058153e-06,
+      "loss": 3.4508,
+      "step": 12446
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.412278761061947e-06,
+      "loss": 3.3749,
+      "step": 12447
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4122313527180787e-06,
+      "loss": 3.1587,
+      "step": 12448
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.41218394437421e-06,
+      "loss": 3.794,
+      "step": 12449
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4121365360303413e-06,
+      "loss": 3.2977,
+      "step": 12450
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4120891276864726e-06,
+      "loss": 3.1497,
+      "step": 12451
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4120417193426043e-06,
+      "loss": 3.8304,
+      "step": 12452
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411994310998736e-06,
+      "loss": 3.5169,
+      "step": 12453
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4119469026548673e-06,
+      "loss": 4.1102,
+      "step": 12454
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411899494310999e-06,
+      "loss": 3.7012,
+      "step": 12455
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4118520859671302e-06,
+      "loss": 2.9182,
+      "step": 12456
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4118046776232615e-06,
+      "loss": 3.001,
+      "step": 12457
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4117572692793932e-06,
+      "loss": 3.3005,
+      "step": 12458
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4117098609355245e-06,
+      "loss": 2.2986,
+      "step": 12459
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411662452591656e-06,
+      "loss": 3.5293,
+      "step": 12460
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411615044247788e-06,
+      "loss": 3.5997,
+      "step": 12461
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411567635903919e-06,
+      "loss": 2.8692,
+      "step": 12462
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4115202275600505e-06,
+      "loss": 3.403,
+      "step": 12463
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411472819216182e-06,
+      "loss": 3.3464,
+      "step": 12464
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4114254108723135e-06,
+      "loss": 2.6113,
+      "step": 12465
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411378002528445e-06,
+      "loss": 2.8124,
+      "step": 12466
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411330594184577e-06,
+      "loss": 3.4187,
+      "step": 12467
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411283185840708e-06,
+      "loss": 2.7068,
+      "step": 12468
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4112357774968394e-06,
+      "loss": 3.5248,
+      "step": 12469
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.411188369152971e-06,
+      "loss": 3.1537,
+      "step": 12470
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4111409608091024e-06,
+      "loss": 4.1319,
+      "step": 12471
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4110935524652337e-06,
+      "loss": 3.3292,
+      "step": 12472
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4110461441213654e-06,
+      "loss": 3.3377,
+      "step": 12473
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.410998735777497e-06,
+      "loss": 3.4699,
+      "step": 12474
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4109513274336284e-06,
+      "loss": 3.6887,
+      "step": 12475
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4109039190897597e-06,
+      "loss": 2.9176,
+      "step": 12476
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4108565107458914e-06,
+      "loss": 3.0052,
+      "step": 12477
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4108091024020226e-06,
+      "loss": 3.3117,
+      "step": 12478
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4107616940581543e-06,
+      "loss": 3.0246,
+      "step": 12479
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.410714285714286e-06,
+      "loss": 3.0895,
+      "step": 12480
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4106668773704173e-06,
+      "loss": 3.434,
+      "step": 12481
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4106194690265486e-06,
+      "loss": 3.6186,
+      "step": 12482
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4105720606826803e-06,
+      "loss": 4.3058,
+      "step": 12483
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4105246523388116e-06,
+      "loss": 3.9231,
+      "step": 12484
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.410477243994943e-06,
+      "loss": 4.2157,
+      "step": 12485
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4104298356510746e-06,
+      "loss": 2.7129,
+      "step": 12486
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4103824273072063e-06,
+      "loss": 4.0531,
+      "step": 12487
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4103350189633376e-06,
+      "loss": 3.8556,
+      "step": 12488
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4102876106194693e-06,
+      "loss": 2.7975,
+      "step": 12489
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4102402022756005e-06,
+      "loss": 3.3996,
+      "step": 12490
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.410192793931732e-06,
+      "loss": 3.5944,
+      "step": 12491
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4101453855878635e-06,
+      "loss": 3.0678,
+      "step": 12492
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4100979772439952e-06,
+      "loss": 3.0032,
+      "step": 12493
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4100505689001265e-06,
+      "loss": 3.2872,
+      "step": 12494
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.410003160556258e-06,
+      "loss": 2.9647,
+      "step": 12495
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4099557522123895e-06,
+      "loss": 2.7743,
+      "step": 12496
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4099083438685208e-06,
+      "loss": 3.8916,
+      "step": 12497
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.409860935524652e-06,
+      "loss": 3.0793,
+      "step": 12498
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4098135271807838e-06,
+      "loss": 3.0471,
+      "step": 12499
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4097661188369155e-06,
+      "loss": 2.7519,
+      "step": 12500
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4097187104930467e-06,
+      "loss": 2.5366,
+      "step": 12501
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4096713021491785e-06,
+      "loss": 3.5917,
+      "step": 12502
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4096238938053097e-06,
+      "loss": 3.3606,
+      "step": 12503
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.409576485461441e-06,
+      "loss": 3.128,
+      "step": 12504
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4095290771175727e-06,
+      "loss": 3.03,
+      "step": 12505
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4094816687737044e-06,
+      "loss": 3.4274,
+      "step": 12506
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.4094342604298357e-06,
+      "loss": 3.5591,
+      "step": 12507
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4093868520859674e-06,
+      "loss": 2.8169,
+      "step": 12508
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4093394437420987e-06,
+      "loss": 3.4518,
+      "step": 12509
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.40929203539823e-06,
+      "loss": 3.3403,
+      "step": 12510
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4092446270543617e-06,
+      "loss": 3.3528,
+      "step": 12511
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.409197218710493e-06,
+      "loss": 3.9285,
+      "step": 12512
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4091498103666247e-06,
+      "loss": 3.1603,
+      "step": 12513
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4091024020227564e-06,
+      "loss": 3.1056,
+      "step": 12514
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4090549936788876e-06,
+      "loss": 2.4941,
+      "step": 12515
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.409007585335019e-06,
+      "loss": 3.2474,
+      "step": 12516
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.40896017699115e-06,
+      "loss": 3.2896,
+      "step": 12517
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408912768647282e-06,
+      "loss": 2.5376,
+      "step": 12518
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4088653603034136e-06,
+      "loss": 3.2425,
+      "step": 12519
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408817951959545e-06,
+      "loss": 3.7849,
+      "step": 12520
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4087705436156766e-06,
+      "loss": 3.0501,
+      "step": 12521
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408723135271808e-06,
+      "loss": 3.2259,
+      "step": 12522
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408675726927939e-06,
+      "loss": 3.4841,
+      "step": 12523
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408628318584071e-06,
+      "loss": 3.6296,
+      "step": 12524
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408580910240202e-06,
+      "loss": 3.9335,
+      "step": 12525
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408533501896334e-06,
+      "loss": 4.1028,
+      "step": 12526
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4084860935524655e-06,
+      "loss": 4.3385,
+      "step": 12527
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408438685208597e-06,
+      "loss": 3.2497,
+      "step": 12528
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408391276864728e-06,
+      "loss": 3.4862,
+      "step": 12529
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.40834386852086e-06,
+      "loss": 4.4519,
+      "step": 12530
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408296460176991e-06,
+      "loss": 3.183,
+      "step": 12531
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408249051833123e-06,
+      "loss": 4.1429,
+      "step": 12532
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4082016434892545e-06,
+      "loss": 3.1712,
+      "step": 12533
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4081542351453858e-06,
+      "loss": 3.0904,
+      "step": 12534
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.408106826801517e-06,
+      "loss": 2.8221,
+      "step": 12535
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4080594184576488e-06,
+      "loss": 3.0384,
+      "step": 12536
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.40801201011378e-06,
+      "loss": 1.8501,
+      "step": 12537
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4079646017699113e-06,
+      "loss": 3.2555,
+      "step": 12538
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407917193426043e-06,
+      "loss": 3.306,
+      "step": 12539
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4078697850821747e-06,
+      "loss": 2.5944,
+      "step": 12540
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407822376738306e-06,
+      "loss": 2.8438,
+      "step": 12541
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4077749683944373e-06,
+      "loss": 2.5329,
+      "step": 12542
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407727560050569e-06,
+      "loss": 4.1615,
+      "step": 12543
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4076801517067003e-06,
+      "loss": 3.3733,
+      "step": 12544
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407632743362832e-06,
+      "loss": 2.5401,
+      "step": 12545
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4075853350189637e-06,
+      "loss": 3.2134,
+      "step": 12546
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407537926675095e-06,
+      "loss": 3.0927,
+      "step": 12547
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4074905183312262e-06,
+      "loss": 3.7733,
+      "step": 12548
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407443109987358e-06,
+      "loss": 4.0265,
+      "step": 12549
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4073957016434892e-06,
+      "loss": 3.0745,
+      "step": 12550
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4073482932996205e-06,
+      "loss": 2.8227,
+      "step": 12551
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4073008849557522e-06,
+      "loss": 3.2207,
+      "step": 12552
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407253476611884e-06,
+      "loss": 3.0667,
+      "step": 12553
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407206068268015e-06,
+      "loss": 3.1046,
+      "step": 12554
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407158659924147e-06,
+      "loss": 2.7509,
+      "step": 12555
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407111251580278e-06,
+      "loss": 3.8738,
+      "step": 12556
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4070638432364095e-06,
+      "loss": 3.0356,
+      "step": 12557
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.407016434892541e-06,
+      "loss": 3.3567,
+      "step": 12558
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.406969026548673e-06,
+      "loss": 3.8433,
+      "step": 12559
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.406921618204804e-06,
+      "loss": 3.0131,
+      "step": 12560
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4068742098609354e-06,
+      "loss": 2.0296,
+      "step": 12561
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.406826801517067e-06,
+      "loss": 2.7795,
+      "step": 12562
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4067793931731984e-06,
+      "loss": 2.734,
+      "step": 12563
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4067319848293297e-06,
+      "loss": 2.8745,
+      "step": 12564
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4066845764854614e-06,
+      "loss": 3.2575,
+      "step": 12565
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.406637168141593e-06,
+      "loss": 2.9513,
+      "step": 12566
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4065897597977244e-06,
+      "loss": 3.3615,
+      "step": 12567
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.406542351453856e-06,
+      "loss": 3.3041,
+      "step": 12568
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4064949431099874e-06,
+      "loss": 3.5407,
+      "step": 12569
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4064475347661186e-06,
+      "loss": 3.834,
+      "step": 12570
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.4064001264222504e-06,
+      "loss": 2.8192,
+      "step": 12571
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.406352718078382e-06,
+      "loss": 2.8534,
+      "step": 12572
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4063053097345133e-06,
+      "loss": 2.7957,
+      "step": 12573
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.406257901390645e-06,
+      "loss": 3.6776,
+      "step": 12574
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4062104930467763e-06,
+      "loss": 3.3779,
+      "step": 12575
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4061630847029076e-06,
+      "loss": 3.1418,
+      "step": 12576
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4061156763590393e-06,
+      "loss": 3.7026,
+      "step": 12577
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4060682680151706e-06,
+      "loss": 2.6046,
+      "step": 12578
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4060208596713023e-06,
+      "loss": 3.521,
+      "step": 12579
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.405973451327434e-06,
+      "loss": 4.0126,
+      "step": 12580
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4059260429835653e-06,
+      "loss": 2.4048,
+      "step": 12581
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4058786346396966e-06,
+      "loss": 3.6619,
+      "step": 12582
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.405831226295828e-06,
+      "loss": 2.6646,
+      "step": 12583
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4057838179519595e-06,
+      "loss": 3.7423,
+      "step": 12584
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4057364096080912e-06,
+      "loss": 2.441,
+      "step": 12585
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4056890012642225e-06,
+      "loss": 4.0852,
+      "step": 12586
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4056415929203542e-06,
+      "loss": 3.0737,
+      "step": 12587
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4055941845764855e-06,
+      "loss": 3.0545,
+      "step": 12588
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.405546776232617e-06,
+      "loss": 2.4842,
+      "step": 12589
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4054993678887485e-06,
+      "loss": 3.4699,
+      "step": 12590
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4054519595448798e-06,
+      "loss": 3.6499,
+      "step": 12591
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4054045512010115e-06,
+      "loss": 3.772,
+      "step": 12592
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.405357142857143e-06,
+      "loss": 4.3945,
+      "step": 12593
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4053097345132745e-06,
+      "loss": 2.9633,
+      "step": 12594
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4052623261694057e-06,
+      "loss": 3.4885,
+      "step": 12595
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4052149178255374e-06,
+      "loss": 3.2251,
+      "step": 12596
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4051675094816687e-06,
+      "loss": 3.1191,
+      "step": 12597
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4051201011378004e-06,
+      "loss": 3.2983,
+      "step": 12598
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.405072692793932e-06,
+      "loss": 3.991,
+      "step": 12599
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4050252844500634e-06,
+      "loss": 2.8902,
+      "step": 12600
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4049778761061947e-06,
+      "loss": 2.4196,
+      "step": 12601
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4049304677623264e-06,
+      "loss": 3.1321,
+      "step": 12602
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4048830594184577e-06,
+      "loss": 3.7298,
+      "step": 12603
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404835651074589e-06,
+      "loss": 3.209,
+      "step": 12604
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4047882427307207e-06,
+      "loss": 2.8254,
+      "step": 12605
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4047408343868524e-06,
+      "loss": 3.2055,
+      "step": 12606
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4046934260429836e-06,
+      "loss": 2.8248,
+      "step": 12607
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404646017699115e-06,
+      "loss": 3.5094,
+      "step": 12608
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4045986093552466e-06,
+      "loss": 3.167,
+      "step": 12609
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404551201011378e-06,
+      "loss": 3.0152,
+      "step": 12610
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4045037926675096e-06,
+      "loss": 2.1935,
+      "step": 12611
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4044563843236413e-06,
+      "loss": 3.5585,
+      "step": 12612
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4044089759797726e-06,
+      "loss": 2.9701,
+      "step": 12613
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404361567635904e-06,
+      "loss": 3.2195,
+      "step": 12614
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4043141592920356e-06,
+      "loss": 2.8382,
+      "step": 12615
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404266750948167e-06,
+      "loss": 3.1754,
+      "step": 12616
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404219342604298e-06,
+      "loss": 3.2398,
+      "step": 12617
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.40417193426043e-06,
+      "loss": 3.9153,
+      "step": 12618
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4041245259165616e-06,
+      "loss": 3.7777,
+      "step": 12619
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.404077117572693e-06,
+      "loss": 2.3268,
+      "step": 12620
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4040297092288245e-06,
+      "loss": 3.9965,
+      "step": 12621
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403982300884956e-06,
+      "loss": 3.5294,
+      "step": 12622
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403934892541087e-06,
+      "loss": 3.8918,
+      "step": 12623
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403887484197219e-06,
+      "loss": 2.9529,
+      "step": 12624
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4038400758533505e-06,
+      "loss": 2.8172,
+      "step": 12625
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4037926675094818e-06,
+      "loss": 3.5814,
+      "step": 12626
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403745259165613e-06,
+      "loss": 3.8024,
+      "step": 12627
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4036978508217448e-06,
+      "loss": 3.1798,
+      "step": 12628
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403650442477876e-06,
+      "loss": 2.8733,
+      "step": 12629
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4036030341340073e-06,
+      "loss": 3.2067,
+      "step": 12630
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403555625790139e-06,
+      "loss": 2.3248,
+      "step": 12631
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4035082174462707e-06,
+      "loss": 3.5057,
+      "step": 12632
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.403460809102402e-06,
+      "loss": 3.8879,
+      "step": 12633
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4034134007585337e-06,
+      "loss": 3.6706,
+      "step": 12634
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.403365992414665e-06,
+      "loss": 3.1773,
+      "step": 12635
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4033185840707963e-06,
+      "loss": 2.7056,
+      "step": 12636
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.403271175726928e-06,
+      "loss": 3.4595,
+      "step": 12637
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4032237673830597e-06,
+      "loss": 3.5297,
+      "step": 12638
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.403176359039191e-06,
+      "loss": 3.4079,
+      "step": 12639
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4031289506953227e-06,
+      "loss": 3.0616,
+      "step": 12640
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.403081542351454e-06,
+      "loss": 2.9899,
+      "step": 12641
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4030341340075852e-06,
+      "loss": 3.1579,
+      "step": 12642
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402986725663717e-06,
+      "loss": 3.6735,
+      "step": 12643
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4029393173198482e-06,
+      "loss": 3.4534,
+      "step": 12644
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.40289190897598e-06,
+      "loss": 2.7849,
+      "step": 12645
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4028445006321116e-06,
+      "loss": 3.3251,
+      "step": 12646
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402797092288243e-06,
+      "loss": 3.5143,
+      "step": 12647
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402749683944374e-06,
+      "loss": 2.0884,
+      "step": 12648
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4027022756005055e-06,
+      "loss": 3.8711,
+      "step": 12649
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402654867256637e-06,
+      "loss": 3.0422,
+      "step": 12650
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402607458912769e-06,
+      "loss": 2.981,
+      "step": 12651
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4025600505689e-06,
+      "loss": 3.2771,
+      "step": 12652
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402512642225032e-06,
+      "loss": 3.0608,
+      "step": 12653
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402465233881163e-06,
+      "loss": 3.7739,
+      "step": 12654
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4024178255372944e-06,
+      "loss": 2.219,
+      "step": 12655
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402370417193426e-06,
+      "loss": 3.7183,
+      "step": 12656
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4023230088495574e-06,
+      "loss": 2.5776,
+      "step": 12657
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402275600505689e-06,
+      "loss": 3.5427,
+      "step": 12658
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402228192161821e-06,
+      "loss": 3.9893,
+      "step": 12659
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402180783817952e-06,
+      "loss": 1.8975,
+      "step": 12660
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4021333754740834e-06,
+      "loss": 3.1879,
+      "step": 12661
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.402085967130215e-06,
+      "loss": 3.0483,
+      "step": 12662
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4020385587863464e-06,
+      "loss": 3.7972,
+      "step": 12663
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.401991150442478e-06,
+      "loss": 3.5616,
+      "step": 12664
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4019437420986098e-06,
+      "loss": 3.0648,
+      "step": 12665
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.401896333754741e-06,
+      "loss": 1.5806,
+      "step": 12666
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.2547069271758437,
+      "eval_loss": 3.143479824066162,
+      "eval_runtime": 154.94,
+      "eval_samples_per_second": 18.168,
+      "eval_steps_per_second": 4.544,
+      "step": 12666
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4018489254108723e-06,
+      "loss": 2.7812,
+      "step": 12667
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.401801517067004e-06,
+      "loss": 3.3634,
+      "step": 12668
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4017541087231353e-06,
+      "loss": 3.4373,
+      "step": 12669
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4017067003792666e-06,
+      "loss": 3.4352,
+      "step": 12670
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4016592920353983e-06,
+      "loss": 2.8256,
+      "step": 12671
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.40161188369153e-06,
+      "loss": 3.0124,
+      "step": 12672
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4015644753476613e-06,
+      "loss": 3.0161,
+      "step": 12673
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4015170670037926e-06,
+      "loss": 3.2804,
+      "step": 12674
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4014696586599243e-06,
+      "loss": 3.23,
+      "step": 12675
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4014222503160555e-06,
+      "loss": 3.0268,
+      "step": 12676
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.401374841972187e-06,
+      "loss": 2.6072,
+      "step": 12677
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.401327433628319e-06,
+      "loss": 3.1807,
+      "step": 12678
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4012800252844502e-06,
+      "loss": 3.9128,
+      "step": 12679
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4012326169405815e-06,
+      "loss": 3.086,
+      "step": 12680
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4011852085967132e-06,
+      "loss": 2.2119,
+      "step": 12681
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4011378002528445e-06,
+      "loss": 3.4971,
+      "step": 12682
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4010903919089758e-06,
+      "loss": 2.9233,
+      "step": 12683
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4010429835651075e-06,
+      "loss": 3.3025,
+      "step": 12684
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.400995575221239e-06,
+      "loss": 3.5534,
+      "step": 12685
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4009481668773705e-06,
+      "loss": 4.0928,
+      "step": 12686
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.400900758533502e-06,
+      "loss": 3.4008,
+      "step": 12687
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4008533501896335e-06,
+      "loss": 2.6162,
+      "step": 12688
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4008059418457647e-06,
+      "loss": 3.5949,
+      "step": 12689
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4007585335018964e-06,
+      "loss": 2.9343,
+      "step": 12690
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4007111251580277e-06,
+      "loss": 2.7219,
+      "step": 12691
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4006637168141594e-06,
+      "loss": 2.9101,
+      "step": 12692
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4006163084702907e-06,
+      "loss": 2.9639,
+      "step": 12693
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4005689001264224e-06,
+      "loss": 3.9464,
+      "step": 12694
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4005214917825537e-06,
+      "loss": 4.063,
+      "step": 12695
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.400474083438685e-06,
+      "loss": 3.0631,
+      "step": 12696
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4004266750948167e-06,
+      "loss": 2.6359,
+      "step": 12697
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4003792667509484e-06,
+      "loss": 2.6397,
+      "step": 12698
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4003318584070797e-06,
+      "loss": 3.8292,
+      "step": 12699
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4002844500632114e-06,
+      "loss": 3.829,
+      "step": 12700
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4002370417193426e-06,
+      "loss": 3.0291,
+      "step": 12701
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.400189633375474e-06,
+      "loss": 3.3548,
+      "step": 12702
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4001422250316056e-06,
+      "loss": 3.2618,
+      "step": 12703
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.400094816687737e-06,
+      "loss": 3.764,
+      "step": 12704
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4000474083438686e-06,
+      "loss": 3.1883,
+      "step": 12705
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 2.6594,
+      "step": 12706
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3999525916561316e-06,
+      "loss": 3.0813,
+      "step": 12707
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399905183312263e-06,
+      "loss": 3.9481,
+      "step": 12708
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3998577749683946e-06,
+      "loss": 3.5548,
+      "step": 12709
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399810366624526e-06,
+      "loss": 2.9564,
+      "step": 12710
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3997629582806576e-06,
+      "loss": 3.4799,
+      "step": 12711
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3997155499367893e-06,
+      "loss": 2.864,
+      "step": 12712
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3996681415929205e-06,
+      "loss": 2.6666,
+      "step": 12713
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399620733249052e-06,
+      "loss": 3.2655,
+      "step": 12714
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399573324905183e-06,
+      "loss": 3.3813,
+      "step": 12715
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399525916561315e-06,
+      "loss": 2.9353,
+      "step": 12716
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399478508217446e-06,
+      "loss": 4.052,
+      "step": 12717
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399431099873578e-06,
+      "loss": 3.6221,
+      "step": 12718
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3993836915297095e-06,
+      "loss": 3.4509,
+      "step": 12719
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3993362831858408e-06,
+      "loss": 3.8788,
+      "step": 12720
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399288874841972e-06,
+      "loss": 2.8574,
+      "step": 12721
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3992414664981038e-06,
+      "loss": 3.38,
+      "step": 12722
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399194058154235e-06,
+      "loss": 3.453,
+      "step": 12723
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3991466498103668e-06,
+      "loss": 3.2675,
+      "step": 12724
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3990992414664985e-06,
+      "loss": 2.7754,
+      "step": 12725
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3990518331226297e-06,
+      "loss": 3.9181,
+      "step": 12726
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.399004424778761e-06,
+      "loss": 3.6856,
+      "step": 12727
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3989570164348927e-06,
+      "loss": 3.2062,
+      "step": 12728
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398909608091024e-06,
+      "loss": 3.1797,
+      "step": 12729
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3988621997471553e-06,
+      "loss": 3.138,
+      "step": 12730
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398814791403287e-06,
+      "loss": 2.9135,
+      "step": 12731
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3987673830594187e-06,
+      "loss": 2.9395,
+      "step": 12732
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.39871997471555e-06,
+      "loss": 2.6178,
+      "step": 12733
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3986725663716817e-06,
+      "loss": 3.6391,
+      "step": 12734
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398625158027813e-06,
+      "loss": 3.6705,
+      "step": 12735
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3985777496839442e-06,
+      "loss": 4.0934,
+      "step": 12736
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398530341340076e-06,
+      "loss": 3.3819,
+      "step": 12737
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3984829329962076e-06,
+      "loss": 2.8559,
+      "step": 12738
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398435524652339e-06,
+      "loss": 2.6456,
+      "step": 12739
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.39838811630847e-06,
+      "loss": 3.1441,
+      "step": 12740
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398340707964602e-06,
+      "loss": 3.0065,
+      "step": 12741
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398293299620733e-06,
+      "loss": 2.7546,
+      "step": 12742
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3982458912768645e-06,
+      "loss": 3.1838,
+      "step": 12743
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398198482932996e-06,
+      "loss": 2.6644,
+      "step": 12744
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398151074589128e-06,
+      "loss": 2.994,
+      "step": 12745
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398103666245259e-06,
+      "loss": 2.9148,
+      "step": 12746
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398056257901391e-06,
+      "loss": 3.1994,
+      "step": 12747
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.398008849557522e-06,
+      "loss": 3.5606,
+      "step": 12748
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3979614412136534e-06,
+      "loss": 3.4252,
+      "step": 12749
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.397914032869785e-06,
+      "loss": 3.9369,
+      "step": 12750
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.397866624525917e-06,
+      "loss": 2.9773,
+      "step": 12751
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.397819216182048e-06,
+      "loss": 3.8093,
+      "step": 12752
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.39777180783818e-06,
+      "loss": 3.9136,
+      "step": 12753
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.397724399494311e-06,
+      "loss": 2.6378,
+      "step": 12754
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3976769911504424e-06,
+      "loss": 3.9027,
+      "step": 12755
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.397629582806574e-06,
+      "loss": 3.8737,
+      "step": 12756
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3975821744627054e-06,
+      "loss": 3.2834,
+      "step": 12757
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.397534766118837e-06,
+      "loss": 3.5977,
+      "step": 12758
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3974873577749683e-06,
+      "loss": 2.97,
+      "step": 12759
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.3974399494311e-06,
+      "loss": 2.9111,
+      "step": 12760
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3973925410872313e-06,
+      "loss": 3.3875,
+      "step": 12761
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3973451327433626e-06,
+      "loss": 3.3842,
+      "step": 12762
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3972977243994943e-06,
+      "loss": 3.3691,
+      "step": 12763
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.397250316055626e-06,
+      "loss": 3.7884,
+      "step": 12764
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3972029077117573e-06,
+      "loss": 1.9487,
+      "step": 12765
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.397155499367889e-06,
+      "loss": 3.1127,
+      "step": 12766
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3971080910240203e-06,
+      "loss": 3.2601,
+      "step": 12767
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3970606826801516e-06,
+      "loss": 4.1402,
+      "step": 12768
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3970132743362833e-06,
+      "loss": 2.8793,
+      "step": 12769
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3969658659924145e-06,
+      "loss": 3.7951,
+      "step": 12770
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3969184576485462e-06,
+      "loss": 3.2644,
+      "step": 12771
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.396871049304678e-06,
+      "loss": 3.4396,
+      "step": 12772
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3968236409608092e-06,
+      "loss": 2.811,
+      "step": 12773
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3967762326169405e-06,
+      "loss": 2.6846,
+      "step": 12774
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3967288242730722e-06,
+      "loss": 2.7639,
+      "step": 12775
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3966814159292035e-06,
+      "loss": 3.5662,
+      "step": 12776
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.396634007585335e-06,
+      "loss": 3.6123,
+      "step": 12777
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.396586599241467e-06,
+      "loss": 2.1825,
+      "step": 12778
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.396539190897598e-06,
+      "loss": 3.3915,
+      "step": 12779
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3964917825537295e-06,
+      "loss": 3.2255,
+      "step": 12780
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3964443742098607e-06,
+      "loss": 3.9186,
+      "step": 12781
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3963969658659924e-06,
+      "loss": 3.3391,
+      "step": 12782
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3963495575221237e-06,
+      "loss": 3.2365,
+      "step": 12783
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3963021491782554e-06,
+      "loss": 3.6428,
+      "step": 12784
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.396254740834387e-06,
+      "loss": 3.6897,
+      "step": 12785
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3962073324905184e-06,
+      "loss": 3.42,
+      "step": 12786
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3961599241466497e-06,
+      "loss": 2.7793,
+      "step": 12787
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3961125158027814e-06,
+      "loss": 2.3872,
+      "step": 12788
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3960651074589127e-06,
+      "loss": 3.072,
+      "step": 12789
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3960176991150444e-06,
+      "loss": 2.7038,
+      "step": 12790
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395970290771176e-06,
+      "loss": 2.5907,
+      "step": 12791
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3959228824273074e-06,
+      "loss": 3.7812,
+      "step": 12792
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3958754740834387e-06,
+      "loss": 3.076,
+      "step": 12793
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3958280657395704e-06,
+      "loss": 3.0076,
+      "step": 12794
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3957806573957016e-06,
+      "loss": 3.4107,
+      "step": 12795
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395733249051833e-06,
+      "loss": 3.4187,
+      "step": 12796
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3956858407079646e-06,
+      "loss": 4.0266,
+      "step": 12797
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3956384323640963e-06,
+      "loss": 3.1095,
+      "step": 12798
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3955910240202276e-06,
+      "loss": 3.6544,
+      "step": 12799
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395543615676359e-06,
+      "loss": 3.6823,
+      "step": 12800
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3954962073324906e-06,
+      "loss": 2.9222,
+      "step": 12801
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395448798988622e-06,
+      "loss": 3.9442,
+      "step": 12802
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3954013906447536e-06,
+      "loss": 3.0691,
+      "step": 12803
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3953539823008853e-06,
+      "loss": 1.9007,
+      "step": 12804
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3953065739570166e-06,
+      "loss": 3.8799,
+      "step": 12805
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395259165613148e-06,
+      "loss": 3.9454,
+      "step": 12806
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3952117572692795e-06,
+      "loss": 3.1556,
+      "step": 12807
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395164348925411e-06,
+      "loss": 2.7301,
+      "step": 12808
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395116940581542e-06,
+      "loss": 3.5173,
+      "step": 12809
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.395069532237674e-06,
+      "loss": 2.8582,
+      "step": 12810
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3950221238938055e-06,
+      "loss": 3.0721,
+      "step": 12811
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.394974715549937e-06,
+      "loss": 4.1725,
+      "step": 12812
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3949273072060685e-06,
+      "loss": 3.4731,
+      "step": 12813
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3948798988621998e-06,
+      "loss": 2.2007,
+      "step": 12814
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.394832490518331e-06,
+      "loss": 3.288,
+      "step": 12815
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3947850821744628e-06,
+      "loss": 3.7779,
+      "step": 12816
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3947376738305945e-06,
+      "loss": 3.4722,
+      "step": 12817
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3946902654867257e-06,
+      "loss": 4.1726,
+      "step": 12818
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3946428571428574e-06,
+      "loss": 2.9834,
+      "step": 12819
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3945954487989887e-06,
+      "loss": 3.1552,
+      "step": 12820
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.39454804045512e-06,
+      "loss": 3.2233,
+      "step": 12821
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3945006321112513e-06,
+      "loss": 2.6896,
+      "step": 12822
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.394453223767383e-06,
+      "loss": 3.9071,
+      "step": 12823
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.3944058154235147e-06,
+      "loss": 2.5475,
+      "step": 12824
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.394358407079646e-06,
+      "loss": 3.528,
+      "step": 12825
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3943109987357777e-06,
+      "loss": 2.3979,
+      "step": 12826
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.394263590391909e-06,
+      "loss": 3.0236,
+      "step": 12827
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3942161820480402e-06,
+      "loss": 3.7035,
+      "step": 12828
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.394168773704172e-06,
+      "loss": 2.7519,
+      "step": 12829
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3941213653603037e-06,
+      "loss": 3.3943,
+      "step": 12830
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.394073957016435e-06,
+      "loss": 3.2356,
+      "step": 12831
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3940265486725666e-06,
+      "loss": 2.8878,
+      "step": 12832
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393979140328698e-06,
+      "loss": 3.4306,
+      "step": 12833
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393931731984829e-06,
+      "loss": 3.2597,
+      "step": 12834
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393884323640961e-06,
+      "loss": 2.8945,
+      "step": 12835
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393836915297092e-06,
+      "loss": 2.853,
+      "step": 12836
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393789506953224e-06,
+      "loss": 2.1693,
+      "step": 12837
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3937420986093556e-06,
+      "loss": 3.8662,
+      "step": 12838
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393694690265487e-06,
+      "loss": 3.5809,
+      "step": 12839
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393647281921618e-06,
+      "loss": 3.538,
+      "step": 12840
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.39359987357775e-06,
+      "loss": 3.3085,
+      "step": 12841
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393552465233881e-06,
+      "loss": 2.9565,
+      "step": 12842
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393505056890013e-06,
+      "loss": 2.4344,
+      "step": 12843
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393457648546144e-06,
+      "loss": 4.0091,
+      "step": 12844
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393410240202276e-06,
+      "loss": 3.0247,
+      "step": 12845
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393362831858407e-06,
+      "loss": 2.3391,
+      "step": 12846
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3933154235145384e-06,
+      "loss": 3.6576,
+      "step": 12847
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.39326801517067e-06,
+      "loss": 2.9495,
+      "step": 12848
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3932206068268014e-06,
+      "loss": 3.2123,
+      "step": 12849
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393173198482933e-06,
+      "loss": 3.4065,
+      "step": 12850
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3931257901390648e-06,
+      "loss": 4.1645,
+      "step": 12851
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.393078381795196e-06,
+      "loss": 3.8418,
+      "step": 12852
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3930309734513273e-06,
+      "loss": 3.9017,
+      "step": 12853
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392983565107459e-06,
+      "loss": 4.488,
+      "step": 12854
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3929361567635903e-06,
+      "loss": 2.693,
+      "step": 12855
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392888748419722e-06,
+      "loss": 2.986,
+      "step": 12856
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3928413400758537e-06,
+      "loss": 3.1725,
+      "step": 12857
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392793931731985e-06,
+      "loss": 2.5132,
+      "step": 12858
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3927465233881163e-06,
+      "loss": 3.6579,
+      "step": 12859
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392699115044248e-06,
+      "loss": 3.2323,
+      "step": 12860
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3926517067003793e-06,
+      "loss": 3.1782,
+      "step": 12861
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3926042983565106e-06,
+      "loss": 2.9537,
+      "step": 12862
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3925568900126423e-06,
+      "loss": 3.8738,
+      "step": 12863
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392509481668774e-06,
+      "loss": 2.7185,
+      "step": 12864
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3924620733249052e-06,
+      "loss": 3.0826,
+      "step": 12865
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3924146649810365e-06,
+      "loss": 3.9049,
+      "step": 12866
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3923672566371682e-06,
+      "loss": 2.6354,
+      "step": 12867
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3923198482932995e-06,
+      "loss": 3.3767,
+      "step": 12868
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392272439949431e-06,
+      "loss": 4.0588,
+      "step": 12869
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392225031605563e-06,
+      "loss": 3.3369,
+      "step": 12870
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392177623261694e-06,
+      "loss": 4.2139,
+      "step": 12871
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3921302149178255e-06,
+      "loss": 3.2558,
+      "step": 12872
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.392082806573957e-06,
+      "loss": 3.5954,
+      "step": 12873
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3920353982300885e-06,
+      "loss": 3.3526,
+      "step": 12874
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3919879898862197e-06,
+      "loss": 3.775,
+      "step": 12875
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3919405815423514e-06,
+      "loss": 3.3045,
+      "step": 12876
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.391893173198483e-06,
+      "loss": 2.5915,
+      "step": 12877
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3918457648546144e-06,
+      "loss": 3.5666,
+      "step": 12878
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.391798356510746e-06,
+      "loss": 2.8207,
+      "step": 12879
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3917509481668774e-06,
+      "loss": 3.4479,
+      "step": 12880
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3917035398230087e-06,
+      "loss": 2.7874,
+      "step": 12881
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3916561314791404e-06,
+      "loss": 3.2252,
+      "step": 12882
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.391608723135272e-06,
+      "loss": 3.3297,
+      "step": 12883
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3915613147914034e-06,
+      "loss": 2.8143,
+      "step": 12884
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.391513906447535e-06,
+      "loss": 1.9785,
+      "step": 12885
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3914664981036664e-06,
+      "loss": 3.3796,
+      "step": 12886
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.3914190897597976e-06,
+      "loss": 3.1507,
+      "step": 12887
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.391371681415929e-06,
+      "loss": 3.7822,
+      "step": 12888
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3913242730720606e-06,
+      "loss": 2.8977,
+      "step": 12889
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3912768647281923e-06,
+      "loss": 2.9793,
+      "step": 12890
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3912294563843236e-06,
+      "loss": 2.7302,
+      "step": 12891
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3911820480404553e-06,
+      "loss": 3.1337,
+      "step": 12892
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3911346396965866e-06,
+      "loss": 3.6146,
+      "step": 12893
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.391087231352718e-06,
+      "loss": 3.6504,
+      "step": 12894
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3910398230088496e-06,
+      "loss": 2.8651,
+      "step": 12895
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3909924146649813e-06,
+      "loss": 3.0612,
+      "step": 12896
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3909450063211126e-06,
+      "loss": 2.8062,
+      "step": 12897
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3908975979772443e-06,
+      "loss": 4.0199,
+      "step": 12898
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3908501896333756e-06,
+      "loss": 2.8902,
+      "step": 12899
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.390802781289507e-06,
+      "loss": 3.8561,
+      "step": 12900
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3907553729456385e-06,
+      "loss": 3.1958,
+      "step": 12901
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.39070796460177e-06,
+      "loss": 3.9249,
+      "step": 12902
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3906605562579015e-06,
+      "loss": 2.9013,
+      "step": 12903
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3906131479140332e-06,
+      "loss": 3.0832,
+      "step": 12904
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3905657395701645e-06,
+      "loss": 4.0722,
+      "step": 12905
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3905183312262958e-06,
+      "loss": 2.7476,
+      "step": 12906
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3904709228824275e-06,
+      "loss": 2.0282,
+      "step": 12907
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3904235145385588e-06,
+      "loss": 2.2903,
+      "step": 12908
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3903761061946905e-06,
+      "loss": 3.1609,
+      "step": 12909
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3903286978508218e-06,
+      "loss": 2.4554,
+      "step": 12910
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3902812895069535e-06,
+      "loss": 2.7655,
+      "step": 12911
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3902338811630847e-06,
+      "loss": 2.7924,
+      "step": 12912
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.390186472819216e-06,
+      "loss": 2.8044,
+      "step": 12913
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3901390644753477e-06,
+      "loss": 3.3003,
+      "step": 12914
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.390091656131479e-06,
+      "loss": 3.4594,
+      "step": 12915
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3900442477876107e-06,
+      "loss": 2.5762,
+      "step": 12916
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3899968394437424e-06,
+      "loss": 3.5084,
+      "step": 12917
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3899494310998737e-06,
+      "loss": 3.7714,
+      "step": 12918
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389902022756005e-06,
+      "loss": 3.2822,
+      "step": 12919
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3898546144121367e-06,
+      "loss": 3.7063,
+      "step": 12920
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389807206068268e-06,
+      "loss": 2.6748,
+      "step": 12921
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3897597977243997e-06,
+      "loss": 2.8327,
+      "step": 12922
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3897123893805314e-06,
+      "loss": 3.7691,
+      "step": 12923
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3896649810366626e-06,
+      "loss": 3.7988,
+      "step": 12924
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389617572692794e-06,
+      "loss": 2.9895,
+      "step": 12925
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3895701643489256e-06,
+      "loss": 3.9279,
+      "step": 12926
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389522756005057e-06,
+      "loss": 3.847,
+      "step": 12927
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389475347661188e-06,
+      "loss": 2.9337,
+      "step": 12928
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.38942793931732e-06,
+      "loss": 2.6821,
+      "step": 12929
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3893805309734516e-06,
+      "loss": 2.8262,
+      "step": 12930
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389333122629583e-06,
+      "loss": 3.8716,
+      "step": 12931
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389285714285714e-06,
+      "loss": 2.8231,
+      "step": 12932
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389238305941846e-06,
+      "loss": 2.454,
+      "step": 12933
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389190897597977e-06,
+      "loss": 3.9398,
+      "step": 12934
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389143489254109e-06,
+      "loss": 2.6322,
+      "step": 12935
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3890960809102406e-06,
+      "loss": 2.4408,
+      "step": 12936
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389048672566372e-06,
+      "loss": 3.1659,
+      "step": 12937
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.389001264222503e-06,
+      "loss": 2.6289,
+      "step": 12938
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388953855878635e-06,
+      "loss": 3.0554,
+      "step": 12939
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388906447534766e-06,
+      "loss": 3.9465,
+      "step": 12940
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3888590391908974e-06,
+      "loss": 3.1883,
+      "step": 12941
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388811630847029e-06,
+      "loss": 2.5347,
+      "step": 12942
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3887642225031608e-06,
+      "loss": 3.1978,
+      "step": 12943
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388716814159292e-06,
+      "loss": 3.1426,
+      "step": 12944
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3886694058154238e-06,
+      "loss": 3.065,
+      "step": 12945
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388621997471555e-06,
+      "loss": 3.1123,
+      "step": 12946
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3885745891276863e-06,
+      "loss": 3.7793,
+      "step": 12947
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388527180783818e-06,
+      "loss": 2.304,
+      "step": 12948
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.3884797724399497e-06,
+      "loss": 3.1551,
+      "step": 12949
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.388432364096081e-06,
+      "loss": 3.9088,
+      "step": 12950
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3883849557522127e-06,
+      "loss": 2.7499,
+      "step": 12951
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.388337547408344e-06,
+      "loss": 3.424,
+      "step": 12952
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3882901390644753e-06,
+      "loss": 3.2267,
+      "step": 12953
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3882427307206066e-06,
+      "loss": 2.7802,
+      "step": 12954
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3881953223767383e-06,
+      "loss": 2.7336,
+      "step": 12955
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.38814791403287e-06,
+      "loss": 2.5374,
+      "step": 12956
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3881005056890013e-06,
+      "loss": 2.9638,
+      "step": 12957
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.388053097345133e-06,
+      "loss": 3.4809,
+      "step": 12958
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3880056890012642e-06,
+      "loss": 3.6939,
+      "step": 12959
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3879582806573955e-06,
+      "loss": 3.7304,
+      "step": 12960
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3879108723135272e-06,
+      "loss": 3.1724,
+      "step": 12961
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3878634639696585e-06,
+      "loss": 4.2636,
+      "step": 12962
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.38781605562579e-06,
+      "loss": 4.134,
+      "step": 12963
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387768647281922e-06,
+      "loss": 2.6974,
+      "step": 12964
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387721238938053e-06,
+      "loss": 2.7685,
+      "step": 12965
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3876738305941845e-06,
+      "loss": 2.891,
+      "step": 12966
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387626422250316e-06,
+      "loss": 2.3014,
+      "step": 12967
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3875790139064475e-06,
+      "loss": 3.036,
+      "step": 12968
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387531605562579e-06,
+      "loss": 3.0715,
+      "step": 12969
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387484197218711e-06,
+      "loss": 3.5835,
+      "step": 12970
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387436788874842e-06,
+      "loss": 3.2216,
+      "step": 12971
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3873893805309734e-06,
+      "loss": 3.1901,
+      "step": 12972
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387341972187105e-06,
+      "loss": 3.2535,
+      "step": 12973
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3872945638432364e-06,
+      "loss": 3.7609,
+      "step": 12974
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3872471554993677e-06,
+      "loss": 3.1607,
+      "step": 12975
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3871997471554994e-06,
+      "loss": 3.0277,
+      "step": 12976
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.387152338811631e-06,
+      "loss": 3.0933,
+      "step": 12977
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3871049304677624e-06,
+      "loss": 2.6235,
+      "step": 12978
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3870575221238937e-06,
+      "loss": 3.05,
+      "step": 12979
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3870101137800254e-06,
+      "loss": 2.7716,
+      "step": 12980
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3869627054361566e-06,
+      "loss": 3.1742,
+      "step": 12981
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3869152970922883e-06,
+      "loss": 2.4618,
+      "step": 12982
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.38686788874842e-06,
+      "loss": 2.2049,
+      "step": 12983
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3868204804045513e-06,
+      "loss": 3.6026,
+      "step": 12984
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3867730720606826e-06,
+      "loss": 3.76,
+      "step": 12985
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3867256637168143e-06,
+      "loss": 3.4374,
+      "step": 12986
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3866782553729456e-06,
+      "loss": 3.3812,
+      "step": 12987
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.386630847029077e-06,
+      "loss": 2.7442,
+      "step": 12988
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3865834386852086e-06,
+      "loss": 2.493,
+      "step": 12989
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3865360303413403e-06,
+      "loss": 3.2354,
+      "step": 12990
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3864886219974716e-06,
+      "loss": 3.0606,
+      "step": 12991
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3864412136536033e-06,
+      "loss": 3.4842,
+      "step": 12992
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3863938053097345e-06,
+      "loss": 2.6013,
+      "step": 12993
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.386346396965866e-06,
+      "loss": 3.0542,
+      "step": 12994
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3862989886219975e-06,
+      "loss": 2.7563,
+      "step": 12995
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3862515802781292e-06,
+      "loss": 3.0724,
+      "step": 12996
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3862041719342605e-06,
+      "loss": 4.3901,
+      "step": 12997
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.386156763590392e-06,
+      "loss": 2.5123,
+      "step": 12998
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3861093552465235e-06,
+      "loss": 3.2817,
+      "step": 12999
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3860619469026548e-06,
+      "loss": 3.5285,
+      "step": 13000
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.386014538558786e-06,
+      "loss": 3.2475,
+      "step": 13001
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3859671302149178e-06,
+      "loss": 2.8778,
+      "step": 13002
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3859197218710495e-06,
+      "loss": 3.0352,
+      "step": 13003
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3858723135271807e-06,
+      "loss": 3.2065,
+      "step": 13004
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3858249051833125e-06,
+      "loss": 3.0443,
+      "step": 13005
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3857774968394437e-06,
+      "loss": 2.9358,
+      "step": 13006
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.385730088495575e-06,
+      "loss": 3.9983,
+      "step": 13007
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3856826801517067e-06,
+      "loss": 3.1259,
+      "step": 13008
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3856352718078384e-06,
+      "loss": 2.6813,
+      "step": 13009
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3855878634639697e-06,
+      "loss": 4.3878,
+      "step": 13010
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3855404551201014e-06,
+      "loss": 3.1504,
+      "step": 13011
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3854930467762327e-06,
+      "loss": 3.2254,
+      "step": 13012
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.385445638432364e-06,
+      "loss": 2.9506,
+      "step": 13013
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.3853982300884957e-06,
+      "loss": 3.2604,
+      "step": 13014
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.385350821744627e-06,
+      "loss": 3.827,
+      "step": 13015
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3853034134007587e-06,
+      "loss": 3.2472,
+      "step": 13016
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3852560050568904e-06,
+      "loss": 4.3836,
+      "step": 13017
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3852085967130216e-06,
+      "loss": 3.2584,
+      "step": 13018
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.385161188369153e-06,
+      "loss": 2.273,
+      "step": 13019
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.385113780025284e-06,
+      "loss": 2.6437,
+      "step": 13020
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.385066371681416e-06,
+      "loss": 2.5302,
+      "step": 13021
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3850189633375476e-06,
+      "loss": 3.4648,
+      "step": 13022
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384971554993679e-06,
+      "loss": 3.2291,
+      "step": 13023
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3849241466498106e-06,
+      "loss": 4.0133,
+      "step": 13024
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384876738305942e-06,
+      "loss": 3.1105,
+      "step": 13025
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384829329962073e-06,
+      "loss": 3.2599,
+      "step": 13026
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384781921618205e-06,
+      "loss": 3.3045,
+      "step": 13027
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384734513274336e-06,
+      "loss": 2.3393,
+      "step": 13028
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384687104930468e-06,
+      "loss": 3.3221,
+      "step": 13029
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3846396965865995e-06,
+      "loss": 4.077,
+      "step": 13030
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384592288242731e-06,
+      "loss": 3.8852,
+      "step": 13031
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384544879898862e-06,
+      "loss": 2.8089,
+      "step": 13032
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384497471554994e-06,
+      "loss": 3.4129,
+      "step": 13033
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384450063211125e-06,
+      "loss": 1.8908,
+      "step": 13034
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384402654867257e-06,
+      "loss": 3.0366,
+      "step": 13035
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3843552465233885e-06,
+      "loss": 2.3917,
+      "step": 13036
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3843078381795198e-06,
+      "loss": 3.2801,
+      "step": 13037
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384260429835651e-06,
+      "loss": 3.6053,
+      "step": 13038
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3842130214917828e-06,
+      "loss": 3.493,
+      "step": 13039
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384165613147914e-06,
+      "loss": 3.5525,
+      "step": 13040
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3841182048040453e-06,
+      "loss": 3.8765,
+      "step": 13041
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.384070796460177e-06,
+      "loss": 2.9426,
+      "step": 13042
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3840233881163087e-06,
+      "loss": 2.5646,
+      "step": 13043
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.38397597977244e-06,
+      "loss": 2.0008,
+      "step": 13044
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3839285714285713e-06,
+      "loss": 3.5237,
+      "step": 13045
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383881163084703e-06,
+      "loss": 3.944,
+      "step": 13046
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3838337547408343e-06,
+      "loss": 3.2565,
+      "step": 13047
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383786346396966e-06,
+      "loss": 3.238,
+      "step": 13048
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3837389380530977e-06,
+      "loss": 2.9102,
+      "step": 13049
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383691529709229e-06,
+      "loss": 2.8343,
+      "step": 13050
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3836441213653602e-06,
+      "loss": 4.4731,
+      "step": 13051
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383596713021492e-06,
+      "loss": 1.9868,
+      "step": 13052
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3835493046776232e-06,
+      "loss": 1.8562,
+      "step": 13053
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3835018963337545e-06,
+      "loss": 3.6973,
+      "step": 13054
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3834544879898862e-06,
+      "loss": 3.6273,
+      "step": 13055
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383407079646018e-06,
+      "loss": 2.7966,
+      "step": 13056
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383359671302149e-06,
+      "loss": 4.432,
+      "step": 13057
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383312262958281e-06,
+      "loss": 1.9132,
+      "step": 13058
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383264854614412e-06,
+      "loss": 3.2222,
+      "step": 13059
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3832174462705435e-06,
+      "loss": 3.4912,
+      "step": 13060
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383170037926675e-06,
+      "loss": 2.7677,
+      "step": 13061
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383122629582807e-06,
+      "loss": 3.0224,
+      "step": 13062
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.383075221238938e-06,
+      "loss": 3.7376,
+      "step": 13063
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3830278128950694e-06,
+      "loss": 3.4896,
+      "step": 13064
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.382980404551201e-06,
+      "loss": 2.9053,
+      "step": 13065
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3829329962073324e-06,
+      "loss": 3.1075,
+      "step": 13066
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3828855878634637e-06,
+      "loss": 3.8871,
+      "step": 13067
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3828381795195954e-06,
+      "loss": 2.8666,
+      "step": 13068
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.382790771175727e-06,
+      "loss": 3.0124,
+      "step": 13069
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3827433628318584e-06,
+      "loss": 3.2263,
+      "step": 13070
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.38269595448799e-06,
+      "loss": 2.0335,
+      "step": 13071
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3826485461441214e-06,
+      "loss": 2.2875,
+      "step": 13072
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3826011378002526e-06,
+      "loss": 4.0019,
+      "step": 13073
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3825537294563844e-06,
+      "loss": 3.4535,
+      "step": 13074
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.382506321112516e-06,
+      "loss": 4.5866,
+      "step": 13075
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.3824589127686473e-06,
+      "loss": 2.6471,
+      "step": 13076
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.382411504424779e-06,
+      "loss": 3.6269,
+      "step": 13077
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3823640960809103e-06,
+      "loss": 2.5123,
+      "step": 13078
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3823166877370416e-06,
+      "loss": 4.3577,
+      "step": 13079
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3822692793931733e-06,
+      "loss": 3.2489,
+      "step": 13080
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3822218710493046e-06,
+      "loss": 3.3803,
+      "step": 13081
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3821744627054363e-06,
+      "loss": 3.4099,
+      "step": 13082
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.382127054361568e-06,
+      "loss": 3.7557,
+      "step": 13083
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3820796460176993e-06,
+      "loss": 3.746,
+      "step": 13084
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3820322376738306e-06,
+      "loss": 3.075,
+      "step": 13085
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.381984829329962e-06,
+      "loss": 3.4725,
+      "step": 13086
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3819374209860935e-06,
+      "loss": 3.1139,
+      "step": 13087
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3818900126422252e-06,
+      "loss": 3.3856,
+      "step": 13088
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3818426042983565e-06,
+      "loss": 3.6916,
+      "step": 13089
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3817951959544882e-06,
+      "loss": 2.9369,
+      "step": 13090
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3817477876106195e-06,
+      "loss": 4.165,
+      "step": 13091
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.381700379266751e-06,
+      "loss": 2.6118,
+      "step": 13092
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3816529709228825e-06,
+      "loss": 2.3862,
+      "step": 13093
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3816055625790138e-06,
+      "loss": 3.307,
+      "step": 13094
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3815581542351455e-06,
+      "loss": 3.5899,
+      "step": 13095
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.381510745891277e-06,
+      "loss": 3.3862,
+      "step": 13096
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3814633375474085e-06,
+      "loss": 3.4364,
+      "step": 13097
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3814159292035397e-06,
+      "loss": 3.0192,
+      "step": 13098
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3813685208596714e-06,
+      "loss": 4.1861,
+      "step": 13099
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3813211125158027e-06,
+      "loss": 3.151,
+      "step": 13100
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3812737041719344e-06,
+      "loss": 3.2213,
+      "step": 13101
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.381226295828066e-06,
+      "loss": 3.3785,
+      "step": 13102
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3811788874841974e-06,
+      "loss": 3.3321,
+      "step": 13103
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3811314791403287e-06,
+      "loss": 3.3891,
+      "step": 13104
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3810840707964604e-06,
+      "loss": 2.6621,
+      "step": 13105
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3810366624525917e-06,
+      "loss": 3.7361,
+      "step": 13106
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380989254108723e-06,
+      "loss": 4.3326,
+      "step": 13107
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3809418457648547e-06,
+      "loss": 2.6064,
+      "step": 13108
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3808944374209864e-06,
+      "loss": 3.3842,
+      "step": 13109
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3808470290771176e-06,
+      "loss": 3.381,
+      "step": 13110
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380799620733249e-06,
+      "loss": 1.4779,
+      "step": 13111
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3807522123893806e-06,
+      "loss": 2.7448,
+      "step": 13112
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380704804045512e-06,
+      "loss": 4.1684,
+      "step": 13113
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3806573957016436e-06,
+      "loss": 2.8606,
+      "step": 13114
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3806099873577753e-06,
+      "loss": 3.1912,
+      "step": 13115
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3805625790139066e-06,
+      "loss": 2.7935,
+      "step": 13116
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380515170670038e-06,
+      "loss": 2.8882,
+      "step": 13117
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3804677623261696e-06,
+      "loss": 2.8075,
+      "step": 13118
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380420353982301e-06,
+      "loss": 3.9724,
+      "step": 13119
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380372945638432e-06,
+      "loss": 3.6091,
+      "step": 13120
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380325537294564e-06,
+      "loss": 3.6156,
+      "step": 13121
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3802781289506956e-06,
+      "loss": 2.5005,
+      "step": 13122
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380230720606827e-06,
+      "loss": 3.9999,
+      "step": 13123
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3801833122629585e-06,
+      "loss": 2.3061,
+      "step": 13124
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.38013590391909e-06,
+      "loss": 4.4263,
+      "step": 13125
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380088495575221e-06,
+      "loss": 2.8264,
+      "step": 13126
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.380041087231353e-06,
+      "loss": 3.9313,
+      "step": 13127
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3799936788874845e-06,
+      "loss": 3.7237,
+      "step": 13128
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3799462705436158e-06,
+      "loss": 4.3016,
+      "step": 13129
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.379898862199747e-06,
+      "loss": 3.2055,
+      "step": 13130
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3798514538558788e-06,
+      "loss": 1.6498,
+      "step": 13131
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.37980404551201e-06,
+      "loss": 3.1544,
+      "step": 13132
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3797566371681413e-06,
+      "loss": 2.8816,
+      "step": 13133
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.379709228824273e-06,
+      "loss": 3.6527,
+      "step": 13134
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3796618204804047e-06,
+      "loss": 3.4355,
+      "step": 13135
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.379614412136536e-06,
+      "loss": 3.5661,
+      "step": 13136
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3795670037926677e-06,
+      "loss": 3.6042,
+      "step": 13137
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.379519595448799e-06,
+      "loss": 3.8868,
+      "step": 13138
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.3794721871049303e-06,
+      "loss": 2.6629,
+      "step": 13139
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.379424778761062e-06,
+      "loss": 2.4072,
+      "step": 13140
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3793773704171937e-06,
+      "loss": 3.45,
+      "step": 13141
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.379329962073325e-06,
+      "loss": 2.7582,
+      "step": 13142
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3792825537294567e-06,
+      "loss": 3.3638,
+      "step": 13143
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.379235145385588e-06,
+      "loss": 2.7142,
+      "step": 13144
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3791877370417192e-06,
+      "loss": 3.8302,
+      "step": 13145
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.379140328697851e-06,
+      "loss": 2.0341,
+      "step": 13146
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3790929203539822e-06,
+      "loss": 3.0207,
+      "step": 13147
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.379045512010114e-06,
+      "loss": 2.2541,
+      "step": 13148
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3789981036662456e-06,
+      "loss": 2.8492,
+      "step": 13149
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378950695322377e-06,
+      "loss": 3.5046,
+      "step": 13150
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378903286978508e-06,
+      "loss": 3.0959,
+      "step": 13151
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3788558786346395e-06,
+      "loss": 4.1286,
+      "step": 13152
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378808470290771e-06,
+      "loss": 3.7938,
+      "step": 13153
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378761061946903e-06,
+      "loss": 2.9466,
+      "step": 13154
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378713653603034e-06,
+      "loss": 2.6074,
+      "step": 13155
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378666245259166e-06,
+      "loss": 3.761,
+      "step": 13156
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378618836915297e-06,
+      "loss": 3.5671,
+      "step": 13157
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3785714285714284e-06,
+      "loss": 3.4545,
+      "step": 13158
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.37852402022756e-06,
+      "loss": 4.3122,
+      "step": 13159
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3784766118836914e-06,
+      "loss": 2.9511,
+      "step": 13160
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378429203539823e-06,
+      "loss": 3.1373,
+      "step": 13161
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378381795195955e-06,
+      "loss": 3.4091,
+      "step": 13162
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378334386852086e-06,
+      "loss": 2.899,
+      "step": 13163
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3782869785082174e-06,
+      "loss": 3.0218,
+      "step": 13164
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378239570164349e-06,
+      "loss": 2.8347,
+      "step": 13165
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3781921618204804e-06,
+      "loss": 3.8028,
+      "step": 13166
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378144753476612e-06,
+      "loss": 2.6695,
+      "step": 13167
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3780973451327438e-06,
+      "loss": 2.4022,
+      "step": 13168
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.378049936788875e-06,
+      "loss": 3.7244,
+      "step": 13169
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3780025284450063e-06,
+      "loss": 2.6249,
+      "step": 13170
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.377955120101138e-06,
+      "loss": 3.8194,
+      "step": 13171
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3779077117572693e-06,
+      "loss": 3.2865,
+      "step": 13172
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3778603034134006e-06,
+      "loss": 2.7735,
+      "step": 13173
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3778128950695323e-06,
+      "loss": 3.1747,
+      "step": 13174
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.377765486725664e-06,
+      "loss": 3.1762,
+      "step": 13175
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3777180783817953e-06,
+      "loss": 3.8622,
+      "step": 13176
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3776706700379266e-06,
+      "loss": 2.8327,
+      "step": 13177
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3776232616940583e-06,
+      "loss": 2.8742,
+      "step": 13178
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3775758533501895e-06,
+      "loss": 2.8335,
+      "step": 13179
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3775284450063213e-06,
+      "loss": 2.558,
+      "step": 13180
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.377481036662453e-06,
+      "loss": 2.5561,
+      "step": 13181
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3774336283185842e-06,
+      "loss": 2.0779,
+      "step": 13182
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3773862199747155e-06,
+      "loss": 2.8518,
+      "step": 13183
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3773388116308472e-06,
+      "loss": 3.8609,
+      "step": 13184
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3772914032869785e-06,
+      "loss": 2.209,
+      "step": 13185
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3772439949431098e-06,
+      "loss": 2.8015,
+      "step": 13186
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3771965865992415e-06,
+      "loss": 3.0136,
+      "step": 13187
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.377149178255373e-06,
+      "loss": 3.1745,
+      "step": 13188
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3771017699115045e-06,
+      "loss": 2.5129,
+      "step": 13189
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.377054361567636e-06,
+      "loss": 2.6332,
+      "step": 13190
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3770069532237675e-06,
+      "loss": 3.585,
+      "step": 13191
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3769595448798987e-06,
+      "loss": 3.8665,
+      "step": 13192
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3769121365360304e-06,
+      "loss": 3.3826,
+      "step": 13193
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.376864728192162e-06,
+      "loss": 3.7094,
+      "step": 13194
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3768173198482934e-06,
+      "loss": 3.417,
+      "step": 13195
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3767699115044247e-06,
+      "loss": 4.077,
+      "step": 13196
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3767225031605564e-06,
+      "loss": 3.2547,
+      "step": 13197
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3766750948166877e-06,
+      "loss": 3.1517,
+      "step": 13198
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.376627686472819e-06,
+      "loss": 3.941,
+      "step": 13199
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3765802781289507e-06,
+      "loss": 3.0888,
+      "step": 13200
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3765328697850824e-06,
+      "loss": 3.1898,
+      "step": 13201
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3764854614412137e-06,
+      "loss": 3.6954,
+      "step": 13202
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3764380530973454e-06,
+      "loss": 3.7604,
+      "step": 13203
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.3763906447534766e-06,
+      "loss": 2.6457,
+      "step": 13204
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.376343236409608e-06,
+      "loss": 2.5369,
+      "step": 13205
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3762958280657396e-06,
+      "loss": 2.6411,
+      "step": 13206
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3762484197218713e-06,
+      "loss": 3.9055,
+      "step": 13207
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3762010113780026e-06,
+      "loss": 2.6476,
+      "step": 13208
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3761536030341343e-06,
+      "loss": 3.6041,
+      "step": 13209
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3761061946902656e-06,
+      "loss": 3.0873,
+      "step": 13210
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.376058786346397e-06,
+      "loss": 4.0092,
+      "step": 13211
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3760113780025286e-06,
+      "loss": 3.2782,
+      "step": 13212
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.37596396965866e-06,
+      "loss": 3.8817,
+      "step": 13213
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3759165613147916e-06,
+      "loss": 2.9536,
+      "step": 13214
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3758691529709233e-06,
+      "loss": 3.7392,
+      "step": 13215
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3758217446270545e-06,
+      "loss": 3.5929,
+      "step": 13216
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375774336283186e-06,
+      "loss": 3.8824,
+      "step": 13217
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375726927939317e-06,
+      "loss": 3.3341,
+      "step": 13218
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375679519595449e-06,
+      "loss": 3.8858,
+      "step": 13219
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3756321112515805e-06,
+      "loss": 2.1659,
+      "step": 13220
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375584702907712e-06,
+      "loss": 3.3881,
+      "step": 13221
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3755372945638435e-06,
+      "loss": 3.2412,
+      "step": 13222
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3754898862199748e-06,
+      "loss": 3.0365,
+      "step": 13223
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375442477876106e-06,
+      "loss": 3.723,
+      "step": 13224
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3753950695322378e-06,
+      "loss": 1.9939,
+      "step": 13225
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375347661188369e-06,
+      "loss": 4.2302,
+      "step": 13226
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3753002528445008e-06,
+      "loss": 2.7947,
+      "step": 13227
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3752528445006325e-06,
+      "loss": 2.8004,
+      "step": 13228
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3752054361567637e-06,
+      "loss": 2.6276,
+      "step": 13229
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375158027812895e-06,
+      "loss": 3.2285,
+      "step": 13230
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3751106194690267e-06,
+      "loss": 3.0009,
+      "step": 13231
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.375063211125158e-06,
+      "loss": 2.345,
+      "step": 13232
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3750158027812893e-06,
+      "loss": 2.5772,
+      "step": 13233
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3749683944374214e-06,
+      "loss": 3.3887,
+      "step": 13234
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3749209860935527e-06,
+      "loss": 2.7754,
+      "step": 13235
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374873577749684e-06,
+      "loss": 3.0917,
+      "step": 13236
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3748261694058152e-06,
+      "loss": 3.6924,
+      "step": 13237
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374778761061947e-06,
+      "loss": 2.722,
+      "step": 13238
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3747313527180782e-06,
+      "loss": 2.4522,
+      "step": 13239
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.37468394437421e-06,
+      "loss": 3.2614,
+      "step": 13240
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3746365360303416e-06,
+      "loss": 3.5149,
+      "step": 13241
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374589127686473e-06,
+      "loss": 2.6897,
+      "step": 13242
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374541719342604e-06,
+      "loss": 3.242,
+      "step": 13243
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374494310998736e-06,
+      "loss": 4.0309,
+      "step": 13244
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374446902654867e-06,
+      "loss": 3.7865,
+      "step": 13245
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3743994943109985e-06,
+      "loss": 3.5346,
+      "step": 13246
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.37435208596713e-06,
+      "loss": 3.5418,
+      "step": 13247
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374304677623262e-06,
+      "loss": 3.2256,
+      "step": 13248
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374257269279393e-06,
+      "loss": 2.7625,
+      "step": 13249
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374209860935525e-06,
+      "loss": 3.2955,
+      "step": 13250
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374162452591656e-06,
+      "loss": 3.6972,
+      "step": 13251
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3741150442477874e-06,
+      "loss": 2.7351,
+      "step": 13252
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374067635903919e-06,
+      "loss": 3.7133,
+      "step": 13253
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.374020227560051e-06,
+      "loss": 4.1579,
+      "step": 13254
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.373972819216182e-06,
+      "loss": 3.068,
+      "step": 13255
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.373925410872314e-06,
+      "loss": 2.9655,
+      "step": 13256
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.373878002528445e-06,
+      "loss": 3.4848,
+      "step": 13257
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3738305941845764e-06,
+      "loss": 3.2585,
+      "step": 13258
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3737831858407077e-06,
+      "loss": 3.0322,
+      "step": 13259
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3737357774968394e-06,
+      "loss": 2.2232,
+      "step": 13260
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.373688369152971e-06,
+      "loss": 3.5073,
+      "step": 13261
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3736409608091023e-06,
+      "loss": 2.5604,
+      "step": 13262
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.373593552465234e-06,
+      "loss": 3.3391,
+      "step": 13263
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3735461441213653e-06,
+      "loss": 3.6855,
+      "step": 13264
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3734987357774966e-06,
+      "loss": 3.2788,
+      "step": 13265
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.3734513274336283e-06,
+      "loss": 2.5204,
+      "step": 13266
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.37340391908976e-06,
+      "loss": 1.8737,
+      "step": 13267
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3733565107458913e-06,
+      "loss": 3.3719,
+      "step": 13268
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.373309102402023e-06,
+      "loss": 2.8642,
+      "step": 13269
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3732616940581543e-06,
+      "loss": 2.807,
+      "step": 13270
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3732142857142856e-06,
+      "loss": 3.7066,
+      "step": 13271
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3731668773704173e-06,
+      "loss": 3.5391,
+      "step": 13272
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3731194690265485e-06,
+      "loss": 2.1903,
+      "step": 13273
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3730720606826802e-06,
+      "loss": 3.3808,
+      "step": 13274
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.373024652338812e-06,
+      "loss": 3.0589,
+      "step": 13275
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3729772439949432e-06,
+      "loss": 3.5802,
+      "step": 13276
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3729298356510745e-06,
+      "loss": 3.5284,
+      "step": 13277
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3728824273072062e-06,
+      "loss": 4.4379,
+      "step": 13278
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3728350189633375e-06,
+      "loss": 3.5091,
+      "step": 13279
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.372787610619469e-06,
+      "loss": 3.0659,
+      "step": 13280
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3727402022756005e-06,
+      "loss": 3.133,
+      "step": 13281
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.372692793931732e-06,
+      "loss": 2.9355,
+      "step": 13282
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3726453855878635e-06,
+      "loss": 3.2002,
+      "step": 13283
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3725979772439947e-06,
+      "loss": 3.3137,
+      "step": 13284
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3725505689001265e-06,
+      "loss": 2.3253,
+      "step": 13285
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3725031605562577e-06,
+      "loss": 3.2997,
+      "step": 13286
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3724557522123894e-06,
+      "loss": 2.3726,
+      "step": 13287
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.372408343868521e-06,
+      "loss": 3.3393,
+      "step": 13288
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3723609355246524e-06,
+      "loss": 2.7526,
+      "step": 13289
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3723135271807837e-06,
+      "loss": 2.4106,
+      "step": 13290
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3722661188369154e-06,
+      "loss": 4.3539,
+      "step": 13291
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3722187104930467e-06,
+      "loss": 1.9596,
+      "step": 13292
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3721713021491784e-06,
+      "loss": 3.0546,
+      "step": 13293
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.37212389380531e-06,
+      "loss": 3.5515,
+      "step": 13294
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3720764854614414e-06,
+      "loss": 3.2287,
+      "step": 13295
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3720290771175727e-06,
+      "loss": 3.1904,
+      "step": 13296
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3719816687737044e-06,
+      "loss": 3.9842,
+      "step": 13297
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3719342604298356e-06,
+      "loss": 3.3768,
+      "step": 13298
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371886852085967e-06,
+      "loss": 3.7528,
+      "step": 13299
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3718394437420986e-06,
+      "loss": 4.4704,
+      "step": 13300
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3717920353982303e-06,
+      "loss": 2.8048,
+      "step": 13301
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3717446270543616e-06,
+      "loss": 2.3336,
+      "step": 13302
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371697218710493e-06,
+      "loss": 3.8486,
+      "step": 13303
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3716498103666246e-06,
+      "loss": 3.3144,
+      "step": 13304
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371602402022756e-06,
+      "loss": 3.8301,
+      "step": 13305
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3715549936788876e-06,
+      "loss": 3.9856,
+      "step": 13306
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3715075853350193e-06,
+      "loss": 3.1693,
+      "step": 13307
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3714601769911506e-06,
+      "loss": 2.9705,
+      "step": 13308
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371412768647282e-06,
+      "loss": 3.3196,
+      "step": 13309
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3713653603034135e-06,
+      "loss": 3.3125,
+      "step": 13310
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371317951959545e-06,
+      "loss": 2.3787,
+      "step": 13311
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371270543615676e-06,
+      "loss": 3.2837,
+      "step": 13312
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371223135271808e-06,
+      "loss": 3.1709,
+      "step": 13313
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3711757269279395e-06,
+      "loss": 3.1258,
+      "step": 13314
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.371128318584071e-06,
+      "loss": 3.4043,
+      "step": 13315
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3710809102402025e-06,
+      "loss": 3.3317,
+      "step": 13316
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3710335018963338e-06,
+      "loss": 2.6778,
+      "step": 13317
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.370986093552465e-06,
+      "loss": 4.2262,
+      "step": 13318
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3709386852085968e-06,
+      "loss": 3.0358,
+      "step": 13319
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3708912768647285e-06,
+      "loss": 3.3946,
+      "step": 13320
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3708438685208597e-06,
+      "loss": 3.1835,
+      "step": 13321
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3707964601769914e-06,
+      "loss": 3.7223,
+      "step": 13322
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3707490518331227e-06,
+      "loss": 2.245,
+      "step": 13323
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.370701643489254e-06,
+      "loss": 2.2594,
+      "step": 13324
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3706542351453853e-06,
+      "loss": 3.3459,
+      "step": 13325
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.370606826801517e-06,
+      "loss": 2.7848,
+      "step": 13326
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3705594184576487e-06,
+      "loss": 3.78,
+      "step": 13327
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.37051201011378e-06,
+      "loss": 2.6256,
+      "step": 13328
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.3704646017699117e-06,
+      "loss": 2.9657,
+      "step": 13329
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.370417193426043e-06,
+      "loss": 3.6182,
+      "step": 13330
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3703697850821742e-06,
+      "loss": 3.4147,
+      "step": 13331
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.370322376738306e-06,
+      "loss": 2.6315,
+      "step": 13332
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3702749683944377e-06,
+      "loss": 2.9204,
+      "step": 13333
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.370227560050569e-06,
+      "loss": 2.8069,
+      "step": 13334
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3701801517067006e-06,
+      "loss": 3.2403,
+      "step": 13335
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.370132743362832e-06,
+      "loss": 3.0556,
+      "step": 13336
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.370085335018963e-06,
+      "loss": 3.4077,
+      "step": 13337
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.370037926675095e-06,
+      "loss": 2.5901,
+      "step": 13338
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369990518331226e-06,
+      "loss": 3.1917,
+      "step": 13339
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369943109987358e-06,
+      "loss": 2.4178,
+      "step": 13340
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3698957016434896e-06,
+      "loss": 2.6647,
+      "step": 13341
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369848293299621e-06,
+      "loss": 3.709,
+      "step": 13342
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369800884955752e-06,
+      "loss": 2.1758,
+      "step": 13343
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369753476611884e-06,
+      "loss": 3.095,
+      "step": 13344
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369706068268015e-06,
+      "loss": 3.636,
+      "step": 13345
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369658659924147e-06,
+      "loss": 2.8355,
+      "step": 13346
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369611251580278e-06,
+      "loss": 2.5865,
+      "step": 13347
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.36956384323641e-06,
+      "loss": 3.2665,
+      "step": 13348
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369516434892541e-06,
+      "loss": 3.4037,
+      "step": 13349
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3694690265486724e-06,
+      "loss": 2.7553,
+      "step": 13350
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369421618204804e-06,
+      "loss": 3.5459,
+      "step": 13351
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3693742098609354e-06,
+      "loss": 3.5094,
+      "step": 13352
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369326801517067e-06,
+      "loss": 3.1286,
+      "step": 13353
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3692793931731988e-06,
+      "loss": 3.4153,
+      "step": 13354
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.36923198482933e-06,
+      "loss": 4.2475,
+      "step": 13355
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3691845764854613e-06,
+      "loss": 2.3769,
+      "step": 13356
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369137168141593e-06,
+      "loss": 3.4591,
+      "step": 13357
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3690897597977243e-06,
+      "loss": 3.8258,
+      "step": 13358
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.369042351453856e-06,
+      "loss": 3.0879,
+      "step": 13359
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3689949431099877e-06,
+      "loss": 3.0712,
+      "step": 13360
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368947534766119e-06,
+      "loss": 3.1791,
+      "step": 13361
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3689001264222503e-06,
+      "loss": 3.4169,
+      "step": 13362
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368852718078382e-06,
+      "loss": 3.3534,
+      "step": 13363
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3688053097345133e-06,
+      "loss": 3.0387,
+      "step": 13364
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3687579013906446e-06,
+      "loss": 2.1966,
+      "step": 13365
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3687104930467763e-06,
+      "loss": 3.1094,
+      "step": 13366
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368663084702908e-06,
+      "loss": 3.1447,
+      "step": 13367
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3686156763590392e-06,
+      "loss": 2.9442,
+      "step": 13368
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3685682680151705e-06,
+      "loss": 3.3055,
+      "step": 13369
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3685208596713022e-06,
+      "loss": 3.1077,
+      "step": 13370
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3684734513274335e-06,
+      "loss": 3.9832,
+      "step": 13371
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368426042983565e-06,
+      "loss": 3.0089,
+      "step": 13372
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368378634639697e-06,
+      "loss": 3.7995,
+      "step": 13373
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368331226295828e-06,
+      "loss": 2.5325,
+      "step": 13374
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3682838179519595e-06,
+      "loss": 3.2588,
+      "step": 13375
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368236409608091e-06,
+      "loss": 3.3738,
+      "step": 13376
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3681890012642225e-06,
+      "loss": 3.002,
+      "step": 13377
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3681415929203537e-06,
+      "loss": 3.1221,
+      "step": 13378
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3680941845764854e-06,
+      "loss": 2.1469,
+      "step": 13379
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.368046776232617e-06,
+      "loss": 3.3032,
+      "step": 13380
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3679993678887484e-06,
+      "loss": 2.8497,
+      "step": 13381
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.36795195954488e-06,
+      "loss": 3.7569,
+      "step": 13382
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3679045512010114e-06,
+      "loss": 2.119,
+      "step": 13383
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3678571428571427e-06,
+      "loss": 3.849,
+      "step": 13384
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3678097345132744e-06,
+      "loss": 2.6687,
+      "step": 13385
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.367762326169406e-06,
+      "loss": 2.5485,
+      "step": 13386
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3677149178255374e-06,
+      "loss": 2.7005,
+      "step": 13387
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.367667509481669e-06,
+      "loss": 2.5116,
+      "step": 13388
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3676201011378004e-06,
+      "loss": 2.9548,
+      "step": 13389
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3675726927939316e-06,
+      "loss": 3.2286,
+      "step": 13390
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.367525284450063e-06,
+      "loss": 3.7287,
+      "step": 13391
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3674778761061946e-06,
+      "loss": 3.1948,
+      "step": 13392
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3674304677623263e-06,
+      "loss": 2.6572,
+      "step": 13393
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.3673830594184576e-06,
+      "loss": 3.5129,
+      "step": 13394
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3673356510745893e-06,
+      "loss": 3.2147,
+      "step": 13395
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3672882427307206e-06,
+      "loss": 3.0001,
+      "step": 13396
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.367240834386852e-06,
+      "loss": 2.251,
+      "step": 13397
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3671934260429836e-06,
+      "loss": 3.1839,
+      "step": 13398
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3671460176991153e-06,
+      "loss": 3.475,
+      "step": 13399
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3670986093552466e-06,
+      "loss": 3.1944,
+      "step": 13400
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3670512010113783e-06,
+      "loss": 3.2721,
+      "step": 13401
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3670037926675096e-06,
+      "loss": 3.1494,
+      "step": 13402
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.366956384323641e-06,
+      "loss": 2.3599,
+      "step": 13403
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3669089759797725e-06,
+      "loss": 2.4638,
+      "step": 13404
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.366861567635904e-06,
+      "loss": 2.42,
+      "step": 13405
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3668141592920355e-06,
+      "loss": 3.8829,
+      "step": 13406
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3667667509481672e-06,
+      "loss": 4.367,
+      "step": 13407
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3667193426042985e-06,
+      "loss": 2.2415,
+      "step": 13408
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3666719342604298e-06,
+      "loss": 2.8134,
+      "step": 13409
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3666245259165615e-06,
+      "loss": 3.508,
+      "step": 13410
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3665771175726928e-06,
+      "loss": 3.7198,
+      "step": 13411
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3665297092288245e-06,
+      "loss": 3.3907,
+      "step": 13412
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3664823008849558e-06,
+      "loss": 2.9697,
+      "step": 13413
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3664348925410875e-06,
+      "loss": 3.7702,
+      "step": 13414
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3663874841972187e-06,
+      "loss": 3.9387,
+      "step": 13415
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.36634007585335e-06,
+      "loss": 2.9186,
+      "step": 13416
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3662926675094817e-06,
+      "loss": 3.1114,
+      "step": 13417
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.366245259165613e-06,
+      "loss": 2.7182,
+      "step": 13418
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3661978508217447e-06,
+      "loss": 3.1386,
+      "step": 13419
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3661504424778764e-06,
+      "loss": 2.3835,
+      "step": 13420
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3661030341340077e-06,
+      "loss": 2.2433,
+      "step": 13421
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.366055625790139e-06,
+      "loss": 2.8716,
+      "step": 13422
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3660082174462707e-06,
+      "loss": 2.8759,
+      "step": 13423
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365960809102402e-06,
+      "loss": 2.8405,
+      "step": 13424
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3659134007585337e-06,
+      "loss": 2.9456,
+      "step": 13425
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3658659924146654e-06,
+      "loss": 3.7885,
+      "step": 13426
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3658185840707966e-06,
+      "loss": 3.4732,
+      "step": 13427
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365771175726928e-06,
+      "loss": 3.594,
+      "step": 13428
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3657237673830596e-06,
+      "loss": 3.0211,
+      "step": 13429
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365676359039191e-06,
+      "loss": 2.0773,
+      "step": 13430
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365628950695322e-06,
+      "loss": 3.5974,
+      "step": 13431
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365581542351454e-06,
+      "loss": 2.7777,
+      "step": 13432
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3655341340075856e-06,
+      "loss": 3.1188,
+      "step": 13433
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365486725663717e-06,
+      "loss": 2.3888,
+      "step": 13434
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365439317319848e-06,
+      "loss": 3.3989,
+      "step": 13435
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.36539190897598e-06,
+      "loss": 3.7925,
+      "step": 13436
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365344500632111e-06,
+      "loss": 2.4806,
+      "step": 13437
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365297092288243e-06,
+      "loss": 4.0533,
+      "step": 13438
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3652496839443746e-06,
+      "loss": 3.1075,
+      "step": 13439
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365202275600506e-06,
+      "loss": 3.9052,
+      "step": 13440
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365154867256637e-06,
+      "loss": 2.9616,
+      "step": 13441
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.365107458912769e-06,
+      "loss": 3.5879,
+      "step": 13442
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3650600505689e-06,
+      "loss": 3.6555,
+      "step": 13443
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3650126422250314e-06,
+      "loss": 3.4388,
+      "step": 13444
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.364965233881163e-06,
+      "loss": 2.0593,
+      "step": 13445
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3649178255372948e-06,
+      "loss": 3.1176,
+      "step": 13446
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.364870417193426e-06,
+      "loss": 2.7962,
+      "step": 13447
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3648230088495578e-06,
+      "loss": 2.2395,
+      "step": 13448
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.364775600505689e-06,
+      "loss": 2.7165,
+      "step": 13449
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3647281921618203e-06,
+      "loss": 3.8419,
+      "step": 13450
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.364680783817952e-06,
+      "loss": 3.3222,
+      "step": 13451
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3646333754740837e-06,
+      "loss": 2.7127,
+      "step": 13452
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.364585967130215e-06,
+      "loss": 3.7732,
+      "step": 13453
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3645385587863467e-06,
+      "loss": 3.4488,
+      "step": 13454
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.364491150442478e-06,
+      "loss": 2.8963,
+      "step": 13455
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3644437420986093e-06,
+      "loss": 2.7963,
+      "step": 13456
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.3643963337547406e-06,
+      "loss": 3.1776,
+      "step": 13457
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3643489254108723e-06,
+      "loss": 2.7055,
+      "step": 13458
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.364301517067004e-06,
+      "loss": 1.9481,
+      "step": 13459
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3642541087231353e-06,
+      "loss": 3.2652,
+      "step": 13460
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.364206700379267e-06,
+      "loss": 2.6583,
+      "step": 13461
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3641592920353982e-06,
+      "loss": 3.8415,
+      "step": 13462
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3641118836915295e-06,
+      "loss": 3.3045,
+      "step": 13463
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3640644753476612e-06,
+      "loss": 3.5469,
+      "step": 13464
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.364017067003793e-06,
+      "loss": 2.9394,
+      "step": 13465
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363969658659924e-06,
+      "loss": 2.5159,
+      "step": 13466
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363922250316056e-06,
+      "loss": 2.9181,
+      "step": 13467
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363874841972187e-06,
+      "loss": 2.6813,
+      "step": 13468
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3638274336283185e-06,
+      "loss": 3.1478,
+      "step": 13469
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.36378002528445e-06,
+      "loss": 3.404,
+      "step": 13470
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3637326169405815e-06,
+      "loss": 2.5081,
+      "step": 13471
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363685208596713e-06,
+      "loss": 3.0972,
+      "step": 13472
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363637800252845e-06,
+      "loss": 3.9667,
+      "step": 13473
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363590391908976e-06,
+      "loss": 3.1399,
+      "step": 13474
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3635429835651074e-06,
+      "loss": 3.1233,
+      "step": 13475
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363495575221239e-06,
+      "loss": 2.565,
+      "step": 13476
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3634481668773704e-06,
+      "loss": 2.3052,
+      "step": 13477
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363400758533502e-06,
+      "loss": 2.7668,
+      "step": 13478
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3633533501896334e-06,
+      "loss": 3.6151,
+      "step": 13479
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363305941845765e-06,
+      "loss": 3.2383,
+      "step": 13480
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3632585335018964e-06,
+      "loss": 3.7448,
+      "step": 13481
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3632111251580277e-06,
+      "loss": 2.7685,
+      "step": 13482
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3631637168141594e-06,
+      "loss": 2.7348,
+      "step": 13483
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3631163084702906e-06,
+      "loss": 2.8795,
+      "step": 13484
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3630689001264223e-06,
+      "loss": 3.2696,
+      "step": 13485
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.363021491782554e-06,
+      "loss": 3.2303,
+      "step": 13486
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3629740834386853e-06,
+      "loss": 2.5899,
+      "step": 13487
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3629266750948166e-06,
+      "loss": 3.1005,
+      "step": 13488
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3628792667509483e-06,
+      "loss": 3.1427,
+      "step": 13489
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3628318584070796e-06,
+      "loss": 3.9121,
+      "step": 13490
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3627844500632113e-06,
+      "loss": 2.5092,
+      "step": 13491
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.362737041719343e-06,
+      "loss": 2.974,
+      "step": 13492
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3626896333754743e-06,
+      "loss": 3.4947,
+      "step": 13493
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3626422250316056e-06,
+      "loss": 3.4946,
+      "step": 13494
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3625948166877373e-06,
+      "loss": 3.4207,
+      "step": 13495
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3625474083438685e-06,
+      "loss": 3.004,
+      "step": 13496
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3625e-06,
+      "loss": 3.695,
+      "step": 13497
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3624525916561315e-06,
+      "loss": 3.437,
+      "step": 13498
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3624051833122632e-06,
+      "loss": 4.3883,
+      "step": 13499
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3623577749683945e-06,
+      "loss": 2.2995,
+      "step": 13500
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.362310366624526e-06,
+      "loss": 2.7526,
+      "step": 13501
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3622629582806575e-06,
+      "loss": 3.5059,
+      "step": 13502
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3622155499367888e-06,
+      "loss": 3.1227,
+      "step": 13503
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.36216814159292e-06,
+      "loss": 3.6868,
+      "step": 13504
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.362120733249052e-06,
+      "loss": 3.4738,
+      "step": 13505
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3620733249051835e-06,
+      "loss": 3.1592,
+      "step": 13506
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3620259165613147e-06,
+      "loss": 3.7154,
+      "step": 13507
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3619785082174465e-06,
+      "loss": 3.5955,
+      "step": 13508
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3619310998735777e-06,
+      "loss": 3.2894,
+      "step": 13509
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.361883691529709e-06,
+      "loss": 3.2013,
+      "step": 13510
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3618362831858407e-06,
+      "loss": 3.7223,
+      "step": 13511
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3617888748419724e-06,
+      "loss": 3.0017,
+      "step": 13512
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3617414664981037e-06,
+      "loss": 3.1127,
+      "step": 13513
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3616940581542354e-06,
+      "loss": 2.9805,
+      "step": 13514
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3616466498103667e-06,
+      "loss": 3.0772,
+      "step": 13515
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.361599241466498e-06,
+      "loss": 3.754,
+      "step": 13516
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3615518331226297e-06,
+      "loss": 3.4633,
+      "step": 13517
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.361504424778761e-06,
+      "loss": 3.6926,
+      "step": 13518
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3614570164348927e-06,
+      "loss": 2.5787,
+      "step": 13519
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.3614096080910244e-06,
+      "loss": 3.2397,
+      "step": 13520
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3613621997471556e-06,
+      "loss": 2.6611,
+      "step": 13521
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.361314791403287e-06,
+      "loss": 3.2964,
+      "step": 13522
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.361267383059418e-06,
+      "loss": 1.7515,
+      "step": 13523
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.36121997471555e-06,
+      "loss": 3.3542,
+      "step": 13524
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3611725663716816e-06,
+      "loss": 2.347,
+      "step": 13525
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.361125158027813e-06,
+      "loss": 3.0753,
+      "step": 13526
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3610777496839446e-06,
+      "loss": 2.9145,
+      "step": 13527
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.361030341340076e-06,
+      "loss": 3.5566,
+      "step": 13528
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360982932996207e-06,
+      "loss": 3.9841,
+      "step": 13529
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360935524652339e-06,
+      "loss": 3.04,
+      "step": 13530
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.36088811630847e-06,
+      "loss": 2.6332,
+      "step": 13531
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360840707964602e-06,
+      "loss": 3.3467,
+      "step": 13532
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3607932996207335e-06,
+      "loss": 3.0235,
+      "step": 13533
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360745891276865e-06,
+      "loss": 3.8934,
+      "step": 13534
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360698482932996e-06,
+      "loss": 3.1733,
+      "step": 13535
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360651074589128e-06,
+      "loss": 3.4316,
+      "step": 13536
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360603666245259e-06,
+      "loss": 2.3648,
+      "step": 13537
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360556257901391e-06,
+      "loss": 3.2263,
+      "step": 13538
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3605088495575225e-06,
+      "loss": 3.2025,
+      "step": 13539
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3604614412136538e-06,
+      "loss": 4.0709,
+      "step": 13540
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360414032869785e-06,
+      "loss": 3.5577,
+      "step": 13541
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3603666245259168e-06,
+      "loss": 3.8634,
+      "step": 13542
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360319216182048e-06,
+      "loss": 2.7677,
+      "step": 13543
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3602718078381793e-06,
+      "loss": 2.2378,
+      "step": 13544
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360224399494311e-06,
+      "loss": 2.3964,
+      "step": 13545
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3601769911504427e-06,
+      "loss": 2.9078,
+      "step": 13546
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360129582806574e-06,
+      "loss": 2.4348,
+      "step": 13547
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3600821744627053e-06,
+      "loss": 2.9946,
+      "step": 13548
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.360034766118837e-06,
+      "loss": 3.2266,
+      "step": 13549
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3599873577749683e-06,
+      "loss": 3.3525,
+      "step": 13550
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3599399494311e-06,
+      "loss": 3.3218,
+      "step": 13551
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3598925410872317e-06,
+      "loss": 4.1757,
+      "step": 13552
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359845132743363e-06,
+      "loss": 4.3212,
+      "step": 13553
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3597977243994942e-06,
+      "loss": 3.0916,
+      "step": 13554
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359750316055626e-06,
+      "loss": 2.6488,
+      "step": 13555
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3597029077117572e-06,
+      "loss": 3.1668,
+      "step": 13556
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3596554993678885e-06,
+      "loss": 3.0085,
+      "step": 13557
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3596080910240202e-06,
+      "loss": 2.4803,
+      "step": 13558
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359560682680152e-06,
+      "loss": 2.4932,
+      "step": 13559
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359513274336283e-06,
+      "loss": 3.5055,
+      "step": 13560
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359465865992415e-06,
+      "loss": 2.4585,
+      "step": 13561
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359418457648546e-06,
+      "loss": 2.9503,
+      "step": 13562
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3593710493046775e-06,
+      "loss": 2.5458,
+      "step": 13563
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359323640960809e-06,
+      "loss": 3.3324,
+      "step": 13564
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359276232616941e-06,
+      "loss": 3.5169,
+      "step": 13565
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359228824273072e-06,
+      "loss": 2.2607,
+      "step": 13566
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3591814159292034e-06,
+      "loss": 3.1426,
+      "step": 13567
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.359134007585335e-06,
+      "loss": 2.2749,
+      "step": 13568
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3590865992414664e-06,
+      "loss": 3.5956,
+      "step": 13569
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3590391908975977e-06,
+      "loss": 4.0112,
+      "step": 13570
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3589917825537294e-06,
+      "loss": 2.9863,
+      "step": 13571
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.358944374209861e-06,
+      "loss": 3.0516,
+      "step": 13572
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3588969658659924e-06,
+      "loss": 4.1559,
+      "step": 13573
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.358849557522124e-06,
+      "loss": 2.7822,
+      "step": 13574
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3588021491782554e-06,
+      "loss": 3.7562,
+      "step": 13575
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3587547408343867e-06,
+      "loss": 3.7387,
+      "step": 13576
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3587073324905184e-06,
+      "loss": 3.5583,
+      "step": 13577
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.35865992414665e-06,
+      "loss": 3.8857,
+      "step": 13578
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3586125158027813e-06,
+      "loss": 2.6964,
+      "step": 13579
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.358565107458913e-06,
+      "loss": 3.2385,
+      "step": 13580
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3585176991150443e-06,
+      "loss": 2.7176,
+      "step": 13581
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3584702907711756e-06,
+      "loss": 3.5259,
+      "step": 13582
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3584228824273073e-06,
+      "loss": 2.3142,
+      "step": 13583
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.3583754740834386e-06,
+      "loss": 3.3526,
+      "step": 13584
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3583280657395703e-06,
+      "loss": 3.4028,
+      "step": 13585
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.358280657395702e-06,
+      "loss": 3.9211,
+      "step": 13586
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3582332490518333e-06,
+      "loss": 3.4286,
+      "step": 13587
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3581858407079646e-06,
+      "loss": 3.2612,
+      "step": 13588
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.358138432364096e-06,
+      "loss": 3.051,
+      "step": 13589
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3580910240202275e-06,
+      "loss": 3.9136,
+      "step": 13590
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3580436156763592e-06,
+      "loss": 3.5607,
+      "step": 13591
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3579962073324905e-06,
+      "loss": 3.2494,
+      "step": 13592
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3579487989886222e-06,
+      "loss": 3.5324,
+      "step": 13593
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3579013906447535e-06,
+      "loss": 2.7678,
+      "step": 13594
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.357853982300885e-06,
+      "loss": 1.8699,
+      "step": 13595
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3578065739570165e-06,
+      "loss": 3.4044,
+      "step": 13596
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3577591656131478e-06,
+      "loss": 2.8324,
+      "step": 13597
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3577117572692795e-06,
+      "loss": 2.8955,
+      "step": 13598
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.357664348925411e-06,
+      "loss": 2.7669,
+      "step": 13599
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3576169405815425e-06,
+      "loss": 2.4212,
+      "step": 13600
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3575695322376737e-06,
+      "loss": 3.3456,
+      "step": 13601
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3575221238938054e-06,
+      "loss": 2.4853,
+      "step": 13602
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3574747155499367e-06,
+      "loss": 2.9049,
+      "step": 13603
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3574273072060684e-06,
+      "loss": 3.3227,
+      "step": 13604
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3573798988622e-06,
+      "loss": 2.8535,
+      "step": 13605
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3573324905183314e-06,
+      "loss": 3.1976,
+      "step": 13606
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3572850821744627e-06,
+      "loss": 3.0116,
+      "step": 13607
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3572376738305944e-06,
+      "loss": 2.4396,
+      "step": 13608
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3571902654867257e-06,
+      "loss": 3.5061,
+      "step": 13609
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.357142857142857e-06,
+      "loss": 3.5664,
+      "step": 13610
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3570954487989887e-06,
+      "loss": 3.6863,
+      "step": 13611
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3570480404551204e-06,
+      "loss": 3.9713,
+      "step": 13612
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3570006321112516e-06,
+      "loss": 3.3724,
+      "step": 13613
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356953223767383e-06,
+      "loss": 3.1068,
+      "step": 13614
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3569058154235146e-06,
+      "loss": 3.4937,
+      "step": 13615
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356858407079646e-06,
+      "loss": 3.2969,
+      "step": 13616
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3568109987357776e-06,
+      "loss": 3.2048,
+      "step": 13617
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3567635903919093e-06,
+      "loss": 3.5017,
+      "step": 13618
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3567161820480406e-06,
+      "loss": 1.9845,
+      "step": 13619
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356668773704172e-06,
+      "loss": 3.7954,
+      "step": 13620
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3566213653603036e-06,
+      "loss": 3.1851,
+      "step": 13621
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356573957016435e-06,
+      "loss": 2.8844,
+      "step": 13622
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356526548672566e-06,
+      "loss": 3.093,
+      "step": 13623
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356479140328698e-06,
+      "loss": 2.8532,
+      "step": 13624
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3564317319848296e-06,
+      "loss": 2.6167,
+      "step": 13625
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356384323640961e-06,
+      "loss": 3.2584,
+      "step": 13626
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3563369152970925e-06,
+      "loss": 3.6704,
+      "step": 13627
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356289506953224e-06,
+      "loss": 3.8223,
+      "step": 13628
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356242098609355e-06,
+      "loss": 3.136,
+      "step": 13629
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356194690265487e-06,
+      "loss": 3.205,
+      "step": 13630
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3561472819216185e-06,
+      "loss": 2.5832,
+      "step": 13631
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.35609987357775e-06,
+      "loss": 3.2909,
+      "step": 13632
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.356052465233881e-06,
+      "loss": 3.2892,
+      "step": 13633
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3560050568900128e-06,
+      "loss": 2.8528,
+      "step": 13634
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.355957648546144e-06,
+      "loss": 3.7053,
+      "step": 13635
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3559102402022753e-06,
+      "loss": 2.987,
+      "step": 13636
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.355862831858407e-06,
+      "loss": 3.4782,
+      "step": 13637
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3558154235145387e-06,
+      "loss": 3.6201,
+      "step": 13638
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.35576801517067e-06,
+      "loss": 3.5305,
+      "step": 13639
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3557206068268017e-06,
+      "loss": 3.6319,
+      "step": 13640
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.355673198482933e-06,
+      "loss": 2.5502,
+      "step": 13641
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3556257901390643e-06,
+      "loss": 2.6506,
+      "step": 13642
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.355578381795196e-06,
+      "loss": 3.0914,
+      "step": 13643
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3555309734513277e-06,
+      "loss": 3.3495,
+      "step": 13644
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.355483565107459e-06,
+      "loss": 2.9771,
+      "step": 13645
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.3554361567635907e-06,
+      "loss": 3.1194,
+      "step": 13646
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.355388748419722e-06,
+      "loss": 2.2827,
+      "step": 13647
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3553413400758532e-06,
+      "loss": 3.4583,
+      "step": 13648
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.355293931731985e-06,
+      "loss": 3.266,
+      "step": 13649
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3552465233881162e-06,
+      "loss": 2.5508,
+      "step": 13650
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.355199115044248e-06,
+      "loss": 3.4606,
+      "step": 13651
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3551517067003796e-06,
+      "loss": 2.9845,
+      "step": 13652
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.355104298356511e-06,
+      "loss": 3.3976,
+      "step": 13653
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.355056890012642e-06,
+      "loss": 3.4737,
+      "step": 13654
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3550094816687735e-06,
+      "loss": 3.4482,
+      "step": 13655
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354962073324905e-06,
+      "loss": 3.2754,
+      "step": 13656
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354914664981037e-06,
+      "loss": 2.6297,
+      "step": 13657
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354867256637168e-06,
+      "loss": 3.4336,
+      "step": 13658
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3548198482933e-06,
+      "loss": 3.5779,
+      "step": 13659
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354772439949431e-06,
+      "loss": 3.1509,
+      "step": 13660
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3547250316055624e-06,
+      "loss": 4.0424,
+      "step": 13661
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354677623261694e-06,
+      "loss": 3.6574,
+      "step": 13662
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3546302149178254e-06,
+      "loss": 3.4989,
+      "step": 13663
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354582806573957e-06,
+      "loss": 3.6691,
+      "step": 13664
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354535398230089e-06,
+      "loss": 3.3215,
+      "step": 13665
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.35448798988622e-06,
+      "loss": 2.557,
+      "step": 13666
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3544405815423514e-06,
+      "loss": 3.4098,
+      "step": 13667
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354393173198483e-06,
+      "loss": 3.96,
+      "step": 13668
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3543457648546144e-06,
+      "loss": 2.2731,
+      "step": 13669
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354298356510746e-06,
+      "loss": 3.3747,
+      "step": 13670
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3542509481668778e-06,
+      "loss": 2.1203,
+      "step": 13671
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.354203539823009e-06,
+      "loss": 2.2589,
+      "step": 13672
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3541561314791403e-06,
+      "loss": 3.1971,
+      "step": 13673
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3541087231352716e-06,
+      "loss": 2.4295,
+      "step": 13674
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3540613147914033e-06,
+      "loss": 2.77,
+      "step": 13675
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3540139064475346e-06,
+      "loss": 3.4775,
+      "step": 13676
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3539664981036663e-06,
+      "loss": 2.5156,
+      "step": 13677
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.353919089759798e-06,
+      "loss": 3.8285,
+      "step": 13678
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3538716814159293e-06,
+      "loss": 2.8115,
+      "step": 13679
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3538242730720606e-06,
+      "loss": 3.1761,
+      "step": 13680
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3537768647281923e-06,
+      "loss": 2.6501,
+      "step": 13681
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3537294563843236e-06,
+      "loss": 4.0931,
+      "step": 13682
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3536820480404553e-06,
+      "loss": 3.2212,
+      "step": 13683
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.353634639696587e-06,
+      "loss": 3.1289,
+      "step": 13684
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3535872313527182e-06,
+      "loss": 3.9915,
+      "step": 13685
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3535398230088495e-06,
+      "loss": 3.896,
+      "step": 13686
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3534924146649812e-06,
+      "loss": 3.033,
+      "step": 13687
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3534450063211125e-06,
+      "loss": 3.3992,
+      "step": 13688
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3533975979772438e-06,
+      "loss": 3.5586,
+      "step": 13689
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3533501896333755e-06,
+      "loss": 3.7544,
+      "step": 13690
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.353302781289507e-06,
+      "loss": 2.5714,
+      "step": 13691
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3532553729456385e-06,
+      "loss": 3.0187,
+      "step": 13692
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.35320796460177e-06,
+      "loss": 2.7839,
+      "step": 13693
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3531605562579015e-06,
+      "loss": 2.7205,
+      "step": 13694
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3531131479140327e-06,
+      "loss": 2.5308,
+      "step": 13695
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3530657395701644e-06,
+      "loss": 3.4854,
+      "step": 13696
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.353018331226296e-06,
+      "loss": 3.6176,
+      "step": 13697
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3529709228824274e-06,
+      "loss": 3.054,
+      "step": 13698
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3529235145385587e-06,
+      "loss": 3.3789,
+      "step": 13699
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3528761061946904e-06,
+      "loss": 3.1833,
+      "step": 13700
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3528286978508217e-06,
+      "loss": 3.1375,
+      "step": 13701
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.352781289506953e-06,
+      "loss": 3.1118,
+      "step": 13702
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3527338811630847e-06,
+      "loss": 3.2696,
+      "step": 13703
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3526864728192164e-06,
+      "loss": 2.7692,
+      "step": 13704
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3526390644753477e-06,
+      "loss": 3.4108,
+      "step": 13705
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3525916561314794e-06,
+      "loss": 3.1162,
+      "step": 13706
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3525442477876106e-06,
+      "loss": 3.1445,
+      "step": 13707
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.352496839443742e-06,
+      "loss": 3.1951,
+      "step": 13708
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3524494310998736e-06,
+      "loss": 3.3381,
+      "step": 13709
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.3524020227560053e-06,
+      "loss": 3.004,
+      "step": 13710
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3523546144121366e-06,
+      "loss": 2.4019,
+      "step": 13711
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3523072060682683e-06,
+      "loss": 3.5472,
+      "step": 13712
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3522597977243996e-06,
+      "loss": 2.6608,
+      "step": 13713
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.352212389380531e-06,
+      "loss": 3.3256,
+      "step": 13714
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3521649810366626e-06,
+      "loss": 2.6151,
+      "step": 13715
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.352117572692794e-06,
+      "loss": 3.1871,
+      "step": 13716
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3520701643489256e-06,
+      "loss": 3.7854,
+      "step": 13717
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.352022756005057e-06,
+      "loss": 3.3662,
+      "step": 13718
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3519753476611885e-06,
+      "loss": 3.6842,
+      "step": 13719
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.35192793931732e-06,
+      "loss": 3.0462,
+      "step": 13720
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351880530973451e-06,
+      "loss": 3.6164,
+      "step": 13721
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351833122629583e-06,
+      "loss": 4.3597,
+      "step": 13722
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3517857142857145e-06,
+      "loss": 2.5592,
+      "step": 13723
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351738305941846e-06,
+      "loss": 3.0664,
+      "step": 13724
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3516908975979775e-06,
+      "loss": 3.1417,
+      "step": 13725
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3516434892541088e-06,
+      "loss": 3.5611,
+      "step": 13726
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.35159608091024e-06,
+      "loss": 2.5536,
+      "step": 13727
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3515486725663718e-06,
+      "loss": 3.1418,
+      "step": 13728
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351501264222503e-06,
+      "loss": 3.5676,
+      "step": 13729
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3514538558786348e-06,
+      "loss": 2.7248,
+      "step": 13730
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3514064475347665e-06,
+      "loss": 2.8953,
+      "step": 13731
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3513590391908977e-06,
+      "loss": 4.4675,
+      "step": 13732
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351311630847029e-06,
+      "loss": 2.6506,
+      "step": 13733
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3512642225031607e-06,
+      "loss": 2.2951,
+      "step": 13734
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351216814159292e-06,
+      "loss": 3.164,
+      "step": 13735
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3511694058154237e-06,
+      "loss": 4.1244,
+      "step": 13736
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3511219974715554e-06,
+      "loss": 3.2938,
+      "step": 13737
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3510745891276867e-06,
+      "loss": 3.2473,
+      "step": 13738
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.351027180783818e-06,
+      "loss": 2.2781,
+      "step": 13739
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3509797724399492e-06,
+      "loss": 4.0852,
+      "step": 13740
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350932364096081e-06,
+      "loss": 3.8783,
+      "step": 13741
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3508849557522122e-06,
+      "loss": 3.3245,
+      "step": 13742
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350837547408344e-06,
+      "loss": 2.2991,
+      "step": 13743
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3507901390644756e-06,
+      "loss": 3.4543,
+      "step": 13744
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350742730720607e-06,
+      "loss": 2.9105,
+      "step": 13745
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350695322376738e-06,
+      "loss": 3.26,
+      "step": 13746
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.35064791403287e-06,
+      "loss": 3.2643,
+      "step": 13747
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350600505689001e-06,
+      "loss": 3.3657,
+      "step": 13748
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350553097345133e-06,
+      "loss": 3.9522,
+      "step": 13749
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3505056890012646e-06,
+      "loss": 3.777,
+      "step": 13750
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350458280657396e-06,
+      "loss": 3.1331,
+      "step": 13751
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350410872313527e-06,
+      "loss": 2.5069,
+      "step": 13752
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350363463969659e-06,
+      "loss": 3.1738,
+      "step": 13753
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.35031605562579e-06,
+      "loss": 2.3526,
+      "step": 13754
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3502686472819214e-06,
+      "loss": 2.8189,
+      "step": 13755
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350221238938053e-06,
+      "loss": 3.6087,
+      "step": 13756
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350173830594185e-06,
+      "loss": 3.1121,
+      "step": 13757
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350126422250316e-06,
+      "loss": 2.4307,
+      "step": 13758
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350079013906448e-06,
+      "loss": 3.6393,
+      "step": 13759
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.350031605562579e-06,
+      "loss": 2.7413,
+      "step": 13760
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3499841972187104e-06,
+      "loss": 3.9707,
+      "step": 13761
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.349936788874842e-06,
+      "loss": 2.6097,
+      "step": 13762
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3498893805309738e-06,
+      "loss": 2.4192,
+      "step": 13763
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.349841972187105e-06,
+      "loss": 3.8125,
+      "step": 13764
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3497945638432363e-06,
+      "loss": 3.5011,
+      "step": 13765
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.349747155499368e-06,
+      "loss": 3.4413,
+      "step": 13766
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3496997471554993e-06,
+      "loss": 3.4706,
+      "step": 13767
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3496523388116306e-06,
+      "loss": 3.144,
+      "step": 13768
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3496049304677623e-06,
+      "loss": 3.0096,
+      "step": 13769
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.349557522123894e-06,
+      "loss": 3.2522,
+      "step": 13770
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3495101137800253e-06,
+      "loss": 2.67,
+      "step": 13771
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.349462705436157e-06,
+      "loss": 2.2596,
+      "step": 13772
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3494152970922883e-06,
+      "loss": 2.6869,
+      "step": 13773
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.3493678887484196e-06,
+      "loss": 3.0585,
+      "step": 13774
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3493204804045513e-06,
+      "loss": 3.3771,
+      "step": 13775
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.349273072060683e-06,
+      "loss": 3.8374,
+      "step": 13776
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3492256637168142e-06,
+      "loss": 2.9238,
+      "step": 13777
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.349178255372946e-06,
+      "loss": 3.1416,
+      "step": 13778
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3491308470290772e-06,
+      "loss": 2.9274,
+      "step": 13779
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3490834386852085e-06,
+      "loss": 2.9007,
+      "step": 13780
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3490360303413402e-06,
+      "loss": 2.5318,
+      "step": 13781
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3489886219974715e-06,
+      "loss": 2.4849,
+      "step": 13782
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.348941213653603e-06,
+      "loss": 2.8017,
+      "step": 13783
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3488938053097345e-06,
+      "loss": 3.5469,
+      "step": 13784
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.348846396965866e-06,
+      "loss": 3.5218,
+      "step": 13785
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3487989886219975e-06,
+      "loss": 2.7333,
+      "step": 13786
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3487515802781287e-06,
+      "loss": 4.1498,
+      "step": 13787
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3487041719342605e-06,
+      "loss": 2.9085,
+      "step": 13788
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3486567635903917e-06,
+      "loss": 3.0681,
+      "step": 13789
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3486093552465234e-06,
+      "loss": 3.4497,
+      "step": 13790
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.348561946902655e-06,
+      "loss": 3.0656,
+      "step": 13791
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3485145385587864e-06,
+      "loss": 4.1783,
+      "step": 13792
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3484671302149177e-06,
+      "loss": 3.9726,
+      "step": 13793
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3484197218710494e-06,
+      "loss": 2.1145,
+      "step": 13794
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3483723135271807e-06,
+      "loss": 3.2013,
+      "step": 13795
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3483249051833124e-06,
+      "loss": 3.2866,
+      "step": 13796
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.348277496839444e-06,
+      "loss": 3.4141,
+      "step": 13797
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3482300884955754e-06,
+      "loss": 3.0821,
+      "step": 13798
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3481826801517067e-06,
+      "loss": 2.7636,
+      "step": 13799
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3481352718078384e-06,
+      "loss": 4.1777,
+      "step": 13800
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3480878634639696e-06,
+      "loss": 2.9046,
+      "step": 13801
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.348040455120101e-06,
+      "loss": 3.3282,
+      "step": 13802
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3479930467762326e-06,
+      "loss": 2.5927,
+      "step": 13803
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3479456384323643e-06,
+      "loss": 2.6661,
+      "step": 13804
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3478982300884956e-06,
+      "loss": 3.3356,
+      "step": 13805
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.347850821744627e-06,
+      "loss": 3.9046,
+      "step": 13806
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3478034134007586e-06,
+      "loss": 3.5698,
+      "step": 13807
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.34775600505689e-06,
+      "loss": 3.875,
+      "step": 13808
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3477085967130216e-06,
+      "loss": 2.8389,
+      "step": 13809
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3476611883691533e-06,
+      "loss": 2.966,
+      "step": 13810
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3476137800252846e-06,
+      "loss": 2.3358,
+      "step": 13811
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.347566371681416e-06,
+      "loss": 3.4592,
+      "step": 13812
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3475189633375475e-06,
+      "loss": 3.5811,
+      "step": 13813
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.347471554993679e-06,
+      "loss": 3.4893,
+      "step": 13814
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.34742414664981e-06,
+      "loss": 3.3449,
+      "step": 13815
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.347376738305942e-06,
+      "loss": 3.775,
+      "step": 13816
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3473293299620735e-06,
+      "loss": 3.3229,
+      "step": 13817
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.347281921618205e-06,
+      "loss": 2.4386,
+      "step": 13818
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3472345132743365e-06,
+      "loss": 2.9677,
+      "step": 13819
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3471871049304678e-06,
+      "loss": 3.3623,
+      "step": 13820
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.347139696586599e-06,
+      "loss": 2.6283,
+      "step": 13821
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3470922882427308e-06,
+      "loss": 3.4543,
+      "step": 13822
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3470448798988625e-06,
+      "loss": 3.3236,
+      "step": 13823
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3469974715549937e-06,
+      "loss": 3.1139,
+      "step": 13824
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3469500632111254e-06,
+      "loss": 3.1686,
+      "step": 13825
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3469026548672567e-06,
+      "loss": 2.7516,
+      "step": 13826
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.346855246523388e-06,
+      "loss": 3.8925,
+      "step": 13827
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3468078381795193e-06,
+      "loss": 2.9825,
+      "step": 13828
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.346760429835651e-06,
+      "loss": 1.9096,
+      "step": 13829
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3467130214917827e-06,
+      "loss": 3.179,
+      "step": 13830
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.346665613147914e-06,
+      "loss": 2.906,
+      "step": 13831
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3466182048040457e-06,
+      "loss": 2.6777,
+      "step": 13832
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.346570796460177e-06,
+      "loss": 3.7382,
+      "step": 13833
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3465233881163082e-06,
+      "loss": 2.1165,
+      "step": 13834
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.34647597977244e-06,
+      "loss": 3.9969,
+      "step": 13835
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.3464285714285717e-06,
+      "loss": 3.1346,
+      "step": 13836
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.346381163084703e-06,
+      "loss": 3.936,
+      "step": 13837
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3463337547408346e-06,
+      "loss": 2.7839,
+      "step": 13838
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.346286346396966e-06,
+      "loss": 2.7482,
+      "step": 13839
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.346238938053097e-06,
+      "loss": 3.3307,
+      "step": 13840
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.346191529709229e-06,
+      "loss": 2.9325,
+      "step": 13841
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.34614412136536e-06,
+      "loss": 4.1802,
+      "step": 13842
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.346096713021492e-06,
+      "loss": 2.8641,
+      "step": 13843
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3460493046776236e-06,
+      "loss": 3.5427,
+      "step": 13844
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.346001896333755e-06,
+      "loss": 2.8628,
+      "step": 13845
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345954487989886e-06,
+      "loss": 3.5471,
+      "step": 13846
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345907079646018e-06,
+      "loss": 3.3673,
+      "step": 13847
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345859671302149e-06,
+      "loss": 2.7505,
+      "step": 13848
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345812262958281e-06,
+      "loss": 3.1573,
+      "step": 13849
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345764854614412e-06,
+      "loss": 2.4089,
+      "step": 13850
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345717446270544e-06,
+      "loss": 2.4745,
+      "step": 13851
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345670037926675e-06,
+      "loss": 4.0699,
+      "step": 13852
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3456226295828064e-06,
+      "loss": 4.2625,
+      "step": 13853
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345575221238938e-06,
+      "loss": 2.4562,
+      "step": 13854
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3455278128950694e-06,
+      "loss": 2.7777,
+      "step": 13855
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345480404551201e-06,
+      "loss": 3.2419,
+      "step": 13856
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3454329962073328e-06,
+      "loss": 3.0424,
+      "step": 13857
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345385587863464e-06,
+      "loss": 3.3405,
+      "step": 13858
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3453381795195953e-06,
+      "loss": 3.2163,
+      "step": 13859
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345290771175727e-06,
+      "loss": 2.6064,
+      "step": 13860
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3452433628318583e-06,
+      "loss": 3.7246,
+      "step": 13861
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.34519595448799e-06,
+      "loss": 3.6898,
+      "step": 13862
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3451485461441217e-06,
+      "loss": 2.4774,
+      "step": 13863
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345101137800253e-06,
+      "loss": 3.6611,
+      "step": 13864
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3450537294563843e-06,
+      "loss": 3.3494,
+      "step": 13865
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.345006321112516e-06,
+      "loss": 3.6896,
+      "step": 13866
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3449589127686473e-06,
+      "loss": 2.9584,
+      "step": 13867
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3449115044247786e-06,
+      "loss": 2.9614,
+      "step": 13868
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3448640960809103e-06,
+      "loss": 2.4083,
+      "step": 13869
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344816687737042e-06,
+      "loss": 2.8684,
+      "step": 13870
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3447692793931732e-06,
+      "loss": 3.4089,
+      "step": 13871
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3447218710493045e-06,
+      "loss": 2.9241,
+      "step": 13872
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3446744627054362e-06,
+      "loss": 2.9824,
+      "step": 13873
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3446270543615675e-06,
+      "loss": 2.997,
+      "step": 13874
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344579646017699e-06,
+      "loss": 1.9949,
+      "step": 13875
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344532237673831e-06,
+      "loss": 3.1906,
+      "step": 13876
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344484829329962e-06,
+      "loss": 4.0317,
+      "step": 13877
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3444374209860935e-06,
+      "loss": 2.6395,
+      "step": 13878
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344390012642225e-06,
+      "loss": 2.5634,
+      "step": 13879
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3443426042983565e-06,
+      "loss": 3.2323,
+      "step": 13880
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3442951959544877e-06,
+      "loss": 2.5667,
+      "step": 13881
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3442477876106194e-06,
+      "loss": 2.8127,
+      "step": 13882
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344200379266751e-06,
+      "loss": 2.9186,
+      "step": 13883
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3441529709228824e-06,
+      "loss": 2.9519,
+      "step": 13884
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.344105562579014e-06,
+      "loss": 3.9384,
+      "step": 13885
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3440581542351454e-06,
+      "loss": 3.1319,
+      "step": 13886
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3440107458912767e-06,
+      "loss": 3.2371,
+      "step": 13887
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3439633375474084e-06,
+      "loss": 2.4807,
+      "step": 13888
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.34391592920354e-06,
+      "loss": 2.296,
+      "step": 13889
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3438685208596714e-06,
+      "loss": 3.5107,
+      "step": 13890
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.343821112515803e-06,
+      "loss": 3.7744,
+      "step": 13891
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3437737041719344e-06,
+      "loss": 2.6076,
+      "step": 13892
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3437262958280656e-06,
+      "loss": 3.6382,
+      "step": 13893
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.343678887484197e-06,
+      "loss": 4.1477,
+      "step": 13894
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3436314791403286e-06,
+      "loss": 3.766,
+      "step": 13895
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3435840707964603e-06,
+      "loss": 3.3598,
+      "step": 13896
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3435366624525916e-06,
+      "loss": 3.5663,
+      "step": 13897
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3434892541087233e-06,
+      "loss": 3.3781,
+      "step": 13898
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.3434418457648546e-06,
+      "loss": 2.9165,
+      "step": 13899
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.343394437420986e-06,
+      "loss": 3.5518,
+      "step": 13900
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3433470290771176e-06,
+      "loss": 1.7351,
+      "step": 13901
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3432996207332493e-06,
+      "loss": 3.6201,
+      "step": 13902
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3432522123893806e-06,
+      "loss": 3.0148,
+      "step": 13903
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3432048040455123e-06,
+      "loss": 2.1653,
+      "step": 13904
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3431573957016436e-06,
+      "loss": 3.3843,
+      "step": 13905
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.343109987357775e-06,
+      "loss": 3.7849,
+      "step": 13906
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3430625790139065e-06,
+      "loss": 3.4528,
+      "step": 13907
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.343015170670038e-06,
+      "loss": 2.6314,
+      "step": 13908
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3429677623261695e-06,
+      "loss": 3.3454,
+      "step": 13909
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3429203539823012e-06,
+      "loss": 3.2825,
+      "step": 13910
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3428729456384325e-06,
+      "loss": 4.0657,
+      "step": 13911
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3428255372945638e-06,
+      "loss": 3.0026,
+      "step": 13912
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3427781289506955e-06,
+      "loss": 3.1031,
+      "step": 13913
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3427307206068268e-06,
+      "loss": 4.2092,
+      "step": 13914
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3426833122629585e-06,
+      "loss": 3.3748,
+      "step": 13915
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3426359039190898e-06,
+      "loss": 3.404,
+      "step": 13916
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3425884955752215e-06,
+      "loss": 3.6196,
+      "step": 13917
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3425410872313527e-06,
+      "loss": 2.763,
+      "step": 13918
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.342493678887484e-06,
+      "loss": 2.927,
+      "step": 13919
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3424462705436157e-06,
+      "loss": 3.7937,
+      "step": 13920
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.342398862199747e-06,
+      "loss": 3.2741,
+      "step": 13921
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3423514538558787e-06,
+      "loss": 2.7679,
+      "step": 13922
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3423040455120104e-06,
+      "loss": 2.7006,
+      "step": 13923
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3422566371681417e-06,
+      "loss": 2.7376,
+      "step": 13924
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.342209228824273e-06,
+      "loss": 3.1603,
+      "step": 13925
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3421618204804047e-06,
+      "loss": 3.9084,
+      "step": 13926
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.342114412136536e-06,
+      "loss": 3.915,
+      "step": 13927
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3420670037926677e-06,
+      "loss": 2.7933,
+      "step": 13928
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3420195954487994e-06,
+      "loss": 3.0107,
+      "step": 13929
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3419721871049306e-06,
+      "loss": 3.3415,
+      "step": 13930
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341924778761062e-06,
+      "loss": 2.6146,
+      "step": 13931
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3418773704171936e-06,
+      "loss": 3.0726,
+      "step": 13932
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341829962073325e-06,
+      "loss": 3.3883,
+      "step": 13933
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341782553729456e-06,
+      "loss": 3.0519,
+      "step": 13934
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341735145385588e-06,
+      "loss": 2.5967,
+      "step": 13935
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3416877370417196e-06,
+      "loss": 3.9815,
+      "step": 13936
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341640328697851e-06,
+      "loss": 2.862,
+      "step": 13937
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341592920353982e-06,
+      "loss": 3.0426,
+      "step": 13938
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341545512010114e-06,
+      "loss": 3.723,
+      "step": 13939
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341498103666245e-06,
+      "loss": 3.1291,
+      "step": 13940
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341450695322377e-06,
+      "loss": 3.5999,
+      "step": 13941
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3414032869785086e-06,
+      "loss": 3.0787,
+      "step": 13942
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.34135587863464e-06,
+      "loss": 3.4775,
+      "step": 13943
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341308470290771e-06,
+      "loss": 4.2907,
+      "step": 13944
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341261061946903e-06,
+      "loss": 4.6125,
+      "step": 13945
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341213653603034e-06,
+      "loss": 2.7061,
+      "step": 13946
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3411662452591654e-06,
+      "loss": 4.0239,
+      "step": 13947
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.341118836915297e-06,
+      "loss": 3.5048,
+      "step": 13948
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3410714285714288e-06,
+      "loss": 3.408,
+      "step": 13949
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.34102402022756e-06,
+      "loss": 2.7792,
+      "step": 13950
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3409766118836918e-06,
+      "loss": 3.2877,
+      "step": 13951
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.340929203539823e-06,
+      "loss": 3.4643,
+      "step": 13952
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3408817951959543e-06,
+      "loss": 3.4425,
+      "step": 13953
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.340834386852086e-06,
+      "loss": 2.2634,
+      "step": 13954
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3407869785082177e-06,
+      "loss": 2.9727,
+      "step": 13955
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.340739570164349e-06,
+      "loss": 3.4126,
+      "step": 13956
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3406921618204807e-06,
+      "loss": 3.4596,
+      "step": 13957
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.340644753476612e-06,
+      "loss": 2.9129,
+      "step": 13958
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3405973451327433e-06,
+      "loss": 3.1074,
+      "step": 13959
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3405499367888746e-06,
+      "loss": 3.0556,
+      "step": 13960
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3405025284450063e-06,
+      "loss": 3.8421,
+      "step": 13961
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.340455120101138e-06,
+      "loss": 3.4064,
+      "step": 13962
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.3404077117572693e-06,
+      "loss": 3.1181,
+      "step": 13963
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.340360303413401e-06,
+      "loss": 3.5276,
+      "step": 13964
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3403128950695322e-06,
+      "loss": 3.5097,
+      "step": 13965
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3402654867256635e-06,
+      "loss": 3.0259,
+      "step": 13966
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3402180783817952e-06,
+      "loss": 4.0022,
+      "step": 13967
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.340170670037927e-06,
+      "loss": 3.706,
+      "step": 13968
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.340123261694058e-06,
+      "loss": 3.6675,
+      "step": 13969
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.34007585335019e-06,
+      "loss": 2.2891,
+      "step": 13970
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.340028445006321e-06,
+      "loss": 3.5554,
+      "step": 13971
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3399810366624525e-06,
+      "loss": 2.9604,
+      "step": 13972
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339933628318584e-06,
+      "loss": 2.8357,
+      "step": 13973
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3398862199747155e-06,
+      "loss": 4.0618,
+      "step": 13974
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339838811630847e-06,
+      "loss": 3.139,
+      "step": 13975
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339791403286979e-06,
+      "loss": 2.7002,
+      "step": 13976
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.33974399494311e-06,
+      "loss": 4.1909,
+      "step": 13977
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3396965865992414e-06,
+      "loss": 2.6505,
+      "step": 13978
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339649178255373e-06,
+      "loss": 3.5835,
+      "step": 13979
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3396017699115044e-06,
+      "loss": 3.4588,
+      "step": 13980
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339554361567636e-06,
+      "loss": 3.8171,
+      "step": 13981
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3395069532237674e-06,
+      "loss": 2.9623,
+      "step": 13982
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339459544879899e-06,
+      "loss": 3.4946,
+      "step": 13983
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3394121365360304e-06,
+      "loss": 3.4366,
+      "step": 13984
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3393647281921617e-06,
+      "loss": 3.505,
+      "step": 13985
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3393173198482934e-06,
+      "loss": 2.6613,
+      "step": 13986
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3392699115044246e-06,
+      "loss": 3.5545,
+      "step": 13987
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3392225031605563e-06,
+      "loss": 2.4457,
+      "step": 13988
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.339175094816688e-06,
+      "loss": 3.6987,
+      "step": 13989
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3391276864728193e-06,
+      "loss": 3.6214,
+      "step": 13990
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3390802781289506e-06,
+      "loss": 4.0092,
+      "step": 13991
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3390328697850823e-06,
+      "loss": 2.6597,
+      "step": 13992
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3389854614412136e-06,
+      "loss": 3.8474,
+      "step": 13993
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3389380530973453e-06,
+      "loss": 3.4213,
+      "step": 13994
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.338890644753477e-06,
+      "loss": 3.147,
+      "step": 13995
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3388432364096083e-06,
+      "loss": 3.4589,
+      "step": 13996
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3387958280657396e-06,
+      "loss": 3.6174,
+      "step": 13997
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3387484197218713e-06,
+      "loss": 3.1221,
+      "step": 13998
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3387010113780025e-06,
+      "loss": 2.3474,
+      "step": 13999
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.338653603034134e-06,
+      "loss": 3.1019,
+      "step": 14000
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3386061946902655e-06,
+      "loss": 3.3943,
+      "step": 14001
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3385587863463972e-06,
+      "loss": 3.2591,
+      "step": 14002
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3385113780025285e-06,
+      "loss": 3.1852,
+      "step": 14003
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.33846396965866e-06,
+      "loss": 2.2591,
+      "step": 14004
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3384165613147915e-06,
+      "loss": 2.1083,
+      "step": 14005
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3383691529709228e-06,
+      "loss": 3.6875,
+      "step": 14006
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3383217446270545e-06,
+      "loss": 3.3833,
+      "step": 14007
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.338274336283186e-06,
+      "loss": 3.8488,
+      "step": 14008
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3382269279393175e-06,
+      "loss": 3.6374,
+      "step": 14009
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3381795195954487e-06,
+      "loss": 2.8294,
+      "step": 14010
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3381321112515805e-06,
+      "loss": 1.6659,
+      "step": 14011
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3380847029077117e-06,
+      "loss": 1.9699,
+      "step": 14012
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.338037294563843e-06,
+      "loss": 2.5661,
+      "step": 14013
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3379898862199747e-06,
+      "loss": 3.424,
+      "step": 14014
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3379424778761064e-06,
+      "loss": 2.7496,
+      "step": 14015
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3378950695322377e-06,
+      "loss": 3.894,
+      "step": 14016
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3378476611883694e-06,
+      "loss": 2.355,
+      "step": 14017
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3378002528445007e-06,
+      "loss": 3.6227,
+      "step": 14018
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.337752844500632e-06,
+      "loss": 3.8215,
+      "step": 14019
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3377054361567637e-06,
+      "loss": 2.587,
+      "step": 14020
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3376580278128954e-06,
+      "loss": 3.895,
+      "step": 14021
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3376106194690267e-06,
+      "loss": 2.2852,
+      "step": 14022
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3375632111251584e-06,
+      "loss": 3.8706,
+      "step": 14023
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.3375158027812896e-06,
+      "loss": 3.6188,
+      "step": 14024
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.337468394437421e-06,
+      "loss": 2.5765,
+      "step": 14025
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.337420986093552e-06,
+      "loss": 3.103,
+      "step": 14026
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.337373577749684e-06,
+      "loss": 3.2845,
+      "step": 14027
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3373261694058156e-06,
+      "loss": 4.1468,
+      "step": 14028
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.337278761061947e-06,
+      "loss": 2.3879,
+      "step": 14029
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3372313527180786e-06,
+      "loss": 3.4822,
+      "step": 14030
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.33718394437421e-06,
+      "loss": 3.7481,
+      "step": 14031
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.337136536030341e-06,
+      "loss": 2.431,
+      "step": 14032
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.337089127686473e-06,
+      "loss": 3.6609,
+      "step": 14033
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3370417193426046e-06,
+      "loss": 2.9249,
+      "step": 14034
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336994310998736e-06,
+      "loss": 2.8983,
+      "step": 14035
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3369469026548675e-06,
+      "loss": 2.2373,
+      "step": 14036
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336899494310999e-06,
+      "loss": 2.7172,
+      "step": 14037
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.33685208596713e-06,
+      "loss": 3.9902,
+      "step": 14038
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336804677623262e-06,
+      "loss": 3.6075,
+      "step": 14039
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336757269279393e-06,
+      "loss": 3.3351,
+      "step": 14040
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336709860935525e-06,
+      "loss": 3.5466,
+      "step": 14041
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3366624525916565e-06,
+      "loss": 3.1429,
+      "step": 14042
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3366150442477878e-06,
+      "loss": 2.8116,
+      "step": 14043
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336567635903919e-06,
+      "loss": 3.0826,
+      "step": 14044
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3365202275600508e-06,
+      "loss": 3.8775,
+      "step": 14045
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336472819216182e-06,
+      "loss": 4.1288,
+      "step": 14046
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3364254108723137e-06,
+      "loss": 3.5843,
+      "step": 14047
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336378002528445e-06,
+      "loss": 2.7325,
+      "step": 14048
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3363305941845767e-06,
+      "loss": 2.7068,
+      "step": 14049
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336283185840708e-06,
+      "loss": 1.8255,
+      "step": 14050
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3362357774968393e-06,
+      "loss": 3.5959,
+      "step": 14051
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336188369152971e-06,
+      "loss": 2.2143,
+      "step": 14052
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3361409608091023e-06,
+      "loss": 3.0523,
+      "step": 14053
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.336093552465234e-06,
+      "loss": 4.3355,
+      "step": 14054
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3360461441213657e-06,
+      "loss": 2.8357,
+      "step": 14055
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335998735777497e-06,
+      "loss": 2.3678,
+      "step": 14056
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3359513274336282e-06,
+      "loss": 3.3564,
+      "step": 14057
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.33590391908976e-06,
+      "loss": 3.8337,
+      "step": 14058
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3358565107458912e-06,
+      "loss": 3.6428,
+      "step": 14059
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3358091024020225e-06,
+      "loss": 3.4354,
+      "step": 14060
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3357616940581546e-06,
+      "loss": 3.0592,
+      "step": 14061
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335714285714286e-06,
+      "loss": 3.1505,
+      "step": 14062
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335666877370417e-06,
+      "loss": 3.1478,
+      "step": 14063
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335619469026549e-06,
+      "loss": 3.5548,
+      "step": 14064
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.33557206068268e-06,
+      "loss": 3.8602,
+      "step": 14065
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3355246523388115e-06,
+      "loss": 2.1307,
+      "step": 14066
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335477243994943e-06,
+      "loss": 3.0999,
+      "step": 14067
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335429835651075e-06,
+      "loss": 3.1004,
+      "step": 14068
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335382427307206e-06,
+      "loss": 2.3601,
+      "step": 14069
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3353350189633374e-06,
+      "loss": 3.1283,
+      "step": 14070
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335287610619469e-06,
+      "loss": 2.2572,
+      "step": 14071
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3352402022756004e-06,
+      "loss": 2.6419,
+      "step": 14072
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3351927939317317e-06,
+      "loss": 2.2815,
+      "step": 14073
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3351453855878634e-06,
+      "loss": 3.4655,
+      "step": 14074
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335097977243995e-06,
+      "loss": 4.1896,
+      "step": 14075
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3350505689001264e-06,
+      "loss": 2.7007,
+      "step": 14076
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.335003160556258e-06,
+      "loss": 2.1744,
+      "step": 14077
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3349557522123894e-06,
+      "loss": 2.8507,
+      "step": 14078
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3349083438685207e-06,
+      "loss": 2.9526,
+      "step": 14079
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3348609355246524e-06,
+      "loss": 3.2102,
+      "step": 14080
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.334813527180784e-06,
+      "loss": 3.6218,
+      "step": 14081
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3347661188369153e-06,
+      "loss": 3.3963,
+      "step": 14082
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.334718710493047e-06,
+      "loss": 2.9954,
+      "step": 14083
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3346713021491783e-06,
+      "loss": 3.717,
+      "step": 14084
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3346238938053096e-06,
+      "loss": 2.7143,
+      "step": 14085
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3345764854614413e-06,
+      "loss": 3.4893,
+      "step": 14086
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3345290771175726e-06,
+      "loss": 3.5408,
+      "step": 14087
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3344816687737043e-06,
+      "loss": 2.9883,
+      "step": 14088
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.334434260429836e-06,
+      "loss": 2.5906,
+      "step": 14089
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.3343868520859673e-06,
+      "loss": 2.5075,
+      "step": 14090
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3343394437420986e-06,
+      "loss": 2.9543,
+      "step": 14091
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.33429203539823e-06,
+      "loss": 2.8084,
+      "step": 14092
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3342446270543615e-06,
+      "loss": 3.4106,
+      "step": 14093
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3341972187104932e-06,
+      "loss": 3.48,
+      "step": 14094
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3341498103666245e-06,
+      "loss": 2.2521,
+      "step": 14095
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3341024020227562e-06,
+      "loss": 3.6269,
+      "step": 14096
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3340549936788875e-06,
+      "loss": 2.9389,
+      "step": 14097
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.334007585335019e-06,
+      "loss": 2.2093,
+      "step": 14098
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3339601769911505e-06,
+      "loss": 3.5302,
+      "step": 14099
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3339127686472818e-06,
+      "loss": 3.1471,
+      "step": 14100
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3338653603034135e-06,
+      "loss": 2.3637,
+      "step": 14101
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.333817951959545e-06,
+      "loss": 2.9016,
+      "step": 14102
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3337705436156765e-06,
+      "loss": 3.4377,
+      "step": 14103
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3337231352718077e-06,
+      "loss": 3.102,
+      "step": 14104
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3336757269279394e-06,
+      "loss": 2.7431,
+      "step": 14105
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3336283185840707e-06,
+      "loss": 3.4587,
+      "step": 14106
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3335809102402024e-06,
+      "loss": 2.1621,
+      "step": 14107
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.333533501896334e-06,
+      "loss": 3.8469,
+      "step": 14108
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3334860935524654e-06,
+      "loss": 3.5818,
+      "step": 14109
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3334386852085967e-06,
+      "loss": 2.6512,
+      "step": 14110
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3333912768647284e-06,
+      "loss": 3.0854,
+      "step": 14111
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3333438685208597e-06,
+      "loss": 3.2457,
+      "step": 14112
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.333296460176991e-06,
+      "loss": 3.2031,
+      "step": 14113
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3332490518331227e-06,
+      "loss": 2.7304,
+      "step": 14114
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3332016434892544e-06,
+      "loss": 3.937,
+      "step": 14115
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3331542351453856e-06,
+      "loss": 3.3487,
+      "step": 14116
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.333106826801517e-06,
+      "loss": 3.5558,
+      "step": 14117
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3330594184576486e-06,
+      "loss": 2.8641,
+      "step": 14118
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.33301201011378e-06,
+      "loss": 3.2485,
+      "step": 14119
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3329646017699116e-06,
+      "loss": 3.3366,
+      "step": 14120
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3329171934260433e-06,
+      "loss": 2.6488,
+      "step": 14121
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3328697850821746e-06,
+      "loss": 3.6497,
+      "step": 14122
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332822376738306e-06,
+      "loss": 3.4926,
+      "step": 14123
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3327749683944376e-06,
+      "loss": 3.5879,
+      "step": 14124
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332727560050569e-06,
+      "loss": 2.3085,
+      "step": 14125
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3326801517067e-06,
+      "loss": 3.2641,
+      "step": 14126
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332632743362832e-06,
+      "loss": 2.8764,
+      "step": 14127
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3325853350189636e-06,
+      "loss": 3.0952,
+      "step": 14128
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332537926675095e-06,
+      "loss": 3.4014,
+      "step": 14129
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3324905183312265e-06,
+      "loss": 4.5099,
+      "step": 14130
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332443109987358e-06,
+      "loss": 3.812,
+      "step": 14131
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332395701643489e-06,
+      "loss": 3.4922,
+      "step": 14132
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332348293299621e-06,
+      "loss": 3.2305,
+      "step": 14133
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3323008849557525e-06,
+      "loss": 3.0311,
+      "step": 14134
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332253476611884e-06,
+      "loss": 3.1891,
+      "step": 14135
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332206068268015e-06,
+      "loss": 3.4757,
+      "step": 14136
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3321586599241468e-06,
+      "loss": 2.1249,
+      "step": 14137
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332111251580278e-06,
+      "loss": 3.0402,
+      "step": 14138
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3320638432364093e-06,
+      "loss": 2.4534,
+      "step": 14139
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.332016434892541e-06,
+      "loss": 2.6828,
+      "step": 14140
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3319690265486727e-06,
+      "loss": 3.8177,
+      "step": 14141
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.331921618204804e-06,
+      "loss": 3.1332,
+      "step": 14142
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3318742098609357e-06,
+      "loss": 2.8266,
+      "step": 14143
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.331826801517067e-06,
+      "loss": 2.8335,
+      "step": 14144
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3317793931731983e-06,
+      "loss": 2.964,
+      "step": 14145
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.33173198482933e-06,
+      "loss": 3.0726,
+      "step": 14146
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3316845764854617e-06,
+      "loss": 3.6391,
+      "step": 14147
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.331637168141593e-06,
+      "loss": 2.7532,
+      "step": 14148
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3315897597977247e-06,
+      "loss": 3.3675,
+      "step": 14149
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.331542351453856e-06,
+      "loss": 3.0136,
+      "step": 14150
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3314949431099872e-06,
+      "loss": 3.4477,
+      "step": 14151
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.331447534766119e-06,
+      "loss": 3.1939,
+      "step": 14152
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.3314001264222502e-06,
+      "loss": 2.3903,
+      "step": 14153
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.331352718078382e-06,
+      "loss": 2.9772,
+      "step": 14154
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.331305309734513e-06,
+      "loss": 4.0192,
+      "step": 14155
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.331257901390645e-06,
+      "loss": 3.192,
+      "step": 14156
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.331210493046776e-06,
+      "loss": 3.2195,
+      "step": 14157
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3311630847029075e-06,
+      "loss": 2.7116,
+      "step": 14158
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.331115676359039e-06,
+      "loss": 2.3905,
+      "step": 14159
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.331068268015171e-06,
+      "loss": 3.7327,
+      "step": 14160
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.331020859671302e-06,
+      "loss": 3.0809,
+      "step": 14161
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330973451327434e-06,
+      "loss": 3.3178,
+      "step": 14162
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330926042983565e-06,
+      "loss": 3.2441,
+      "step": 14163
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3308786346396964e-06,
+      "loss": 3.3151,
+      "step": 14164
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330831226295828e-06,
+      "loss": 3.2415,
+      "step": 14165
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3307838179519594e-06,
+      "loss": 2.5152,
+      "step": 14166
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330736409608091e-06,
+      "loss": 3.6423,
+      "step": 14167
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330689001264223e-06,
+      "loss": 3.3513,
+      "step": 14168
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330641592920354e-06,
+      "loss": 3.4801,
+      "step": 14169
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3305941845764854e-06,
+      "loss": 3.5023,
+      "step": 14170
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330546776232617e-06,
+      "loss": 2.9276,
+      "step": 14171
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3304993678887484e-06,
+      "loss": 3.6895,
+      "step": 14172
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.33045195954488e-06,
+      "loss": 3.3664,
+      "step": 14173
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3304045512010118e-06,
+      "loss": 3.4487,
+      "step": 14174
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330357142857143e-06,
+      "loss": 2.5107,
+      "step": 14175
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3303097345132743e-06,
+      "loss": 3.2798,
+      "step": 14176
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3302623261694056e-06,
+      "loss": 3.1023,
+      "step": 14177
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3302149178255373e-06,
+      "loss": 3.694,
+      "step": 14178
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3301675094816686e-06,
+      "loss": 2.9203,
+      "step": 14179
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3301201011378003e-06,
+      "loss": 3.527,
+      "step": 14180
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.330072692793932e-06,
+      "loss": 3.0672,
+      "step": 14181
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3300252844500633e-06,
+      "loss": 2.2229,
+      "step": 14182
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3299778761061946e-06,
+      "loss": 3.8829,
+      "step": 14183
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3299304677623263e-06,
+      "loss": 2.7032,
+      "step": 14184
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3298830594184576e-06,
+      "loss": 3.2319,
+      "step": 14185
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3298356510745893e-06,
+      "loss": 2.6092,
+      "step": 14186
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.329788242730721e-06,
+      "loss": 2.8198,
+      "step": 14187
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3297408343868522e-06,
+      "loss": 2.6598,
+      "step": 14188
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3296934260429835e-06,
+      "loss": 2.8624,
+      "step": 14189
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3296460176991152e-06,
+      "loss": 2.6054,
+      "step": 14190
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3295986093552465e-06,
+      "loss": 1.8722,
+      "step": 14191
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3295512010113778e-06,
+      "loss": 1.7687,
+      "step": 14192
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3295037926675095e-06,
+      "loss": 2.4437,
+      "step": 14193
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.329456384323641e-06,
+      "loss": 3.1631,
+      "step": 14194
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3294089759797725e-06,
+      "loss": 3.7089,
+      "step": 14195
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.329361567635904e-06,
+      "loss": 2.5556,
+      "step": 14196
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3293141592920355e-06,
+      "loss": 2.4589,
+      "step": 14197
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3292667509481667e-06,
+      "loss": 3.2929,
+      "step": 14198
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3292193426042984e-06,
+      "loss": 3.4939,
+      "step": 14199
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.32917193426043e-06,
+      "loss": 4.0976,
+      "step": 14200
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3291245259165614e-06,
+      "loss": 3.3811,
+      "step": 14201
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3290771175726927e-06,
+      "loss": 2.5714,
+      "step": 14202
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3290297092288244e-06,
+      "loss": 2.7939,
+      "step": 14203
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3289823008849557e-06,
+      "loss": 3.0953,
+      "step": 14204
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.328934892541087e-06,
+      "loss": 4.3287,
+      "step": 14205
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3288874841972187e-06,
+      "loss": 3.3081,
+      "step": 14206
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3288400758533504e-06,
+      "loss": 1.7164,
+      "step": 14207
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3287926675094817e-06,
+      "loss": 3.417,
+      "step": 14208
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3287452591656134e-06,
+      "loss": 2.9856,
+      "step": 14209
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3286978508217446e-06,
+      "loss": 2.9204,
+      "step": 14210
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.328650442477876e-06,
+      "loss": 2.3482,
+      "step": 14211
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3286030341340076e-06,
+      "loss": 2.6318,
+      "step": 14212
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3285556257901393e-06,
+      "loss": 2.3815,
+      "step": 14213
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3285082174462706e-06,
+      "loss": 3.2211,
+      "step": 14214
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3284608091024023e-06,
+      "loss": 2.9596,
+      "step": 14215
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.3284134007585336e-06,
+      "loss": 3.5613,
+      "step": 14216
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.328365992414665e-06,
+      "loss": 2.9381,
+      "step": 14217
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3283185840707966e-06,
+      "loss": 2.9765,
+      "step": 14218
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.328271175726928e-06,
+      "loss": 3.3729,
+      "step": 14219
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3282237673830596e-06,
+      "loss": 1.8738,
+      "step": 14220
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.328176359039191e-06,
+      "loss": 4.0069,
+      "step": 14221
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3281289506953225e-06,
+      "loss": 3.3737,
+      "step": 14222
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.328081542351454e-06,
+      "loss": 2.6394,
+      "step": 14223
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.328034134007585e-06,
+      "loss": 3.2042,
+      "step": 14224
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327986725663717e-06,
+      "loss": 3.1522,
+      "step": 14225
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3279393173198485e-06,
+      "loss": 3.1093,
+      "step": 14226
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.32789190897598e-06,
+      "loss": 3.0726,
+      "step": 14227
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3278445006321115e-06,
+      "loss": 2.3444,
+      "step": 14228
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3277970922882428e-06,
+      "loss": 2.0632,
+      "step": 14229
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327749683944374e-06,
+      "loss": 2.8808,
+      "step": 14230
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3277022756005058e-06,
+      "loss": 3.9676,
+      "step": 14231
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327654867256637e-06,
+      "loss": 2.9439,
+      "step": 14232
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3276074589127688e-06,
+      "loss": 2.8996,
+      "step": 14233
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3275600505689005e-06,
+      "loss": 3.6559,
+      "step": 14234
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3275126422250317e-06,
+      "loss": 3.0694,
+      "step": 14235
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327465233881163e-06,
+      "loss": 3.0391,
+      "step": 14236
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3274178255372947e-06,
+      "loss": 1.9676,
+      "step": 14237
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327370417193426e-06,
+      "loss": 3.7279,
+      "step": 14238
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3273230088495577e-06,
+      "loss": 3.2385,
+      "step": 14239
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3272756005056894e-06,
+      "loss": 2.8774,
+      "step": 14240
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3272281921618207e-06,
+      "loss": 3.8695,
+      "step": 14241
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327180783817952e-06,
+      "loss": 2.9767,
+      "step": 14242
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3271333754740832e-06,
+      "loss": 2.6006,
+      "step": 14243
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.327085967130215e-06,
+      "loss": 3.4549,
+      "step": 14244
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3270385587863462e-06,
+      "loss": 2.9827,
+      "step": 14245
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326991150442478e-06,
+      "loss": 3.727,
+      "step": 14246
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3269437420986096e-06,
+      "loss": 2.8909,
+      "step": 14247
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326896333754741e-06,
+      "loss": 3.3223,
+      "step": 14248
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326848925410872e-06,
+      "loss": 3.3085,
+      "step": 14249
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326801517067004e-06,
+      "loss": 2.8946,
+      "step": 14250
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326754108723135e-06,
+      "loss": 3.813,
+      "step": 14251
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326706700379267e-06,
+      "loss": 2.7094,
+      "step": 14252
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3266592920353986e-06,
+      "loss": 3.9327,
+      "step": 14253
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.32661188369153e-06,
+      "loss": 2.3433,
+      "step": 14254
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326564475347661e-06,
+      "loss": 2.487,
+      "step": 14255
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326517067003793e-06,
+      "loss": 2.4451,
+      "step": 14256
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326469658659924e-06,
+      "loss": 2.9199,
+      "step": 14257
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3264222503160554e-06,
+      "loss": 3.2689,
+      "step": 14258
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326374841972187e-06,
+      "loss": 2.499,
+      "step": 14259
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326327433628319e-06,
+      "loss": 3.7382,
+      "step": 14260
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.32628002528445e-06,
+      "loss": 2.9822,
+      "step": 14261
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326232616940582e-06,
+      "loss": 4.1489,
+      "step": 14262
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326185208596713e-06,
+      "loss": 2.9525,
+      "step": 14263
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3261378002528444e-06,
+      "loss": 2.7331,
+      "step": 14264
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.326090391908976e-06,
+      "loss": 2.5365,
+      "step": 14265
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3260429835651078e-06,
+      "loss": 2.7609,
+      "step": 14266
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.325995575221239e-06,
+      "loss": 3.4939,
+      "step": 14267
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3259481668773703e-06,
+      "loss": 2.4513,
+      "step": 14268
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.325900758533502e-06,
+      "loss": 3.1897,
+      "step": 14269
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3258533501896333e-06,
+      "loss": 2.7957,
+      "step": 14270
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3258059418457646e-06,
+      "loss": 2.999,
+      "step": 14271
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3257585335018963e-06,
+      "loss": 4.1929,
+      "step": 14272
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.325711125158028e-06,
+      "loss": 4.0729,
+      "step": 14273
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3256637168141593e-06,
+      "loss": 3.1876,
+      "step": 14274
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.325616308470291e-06,
+      "loss": 3.2377,
+      "step": 14275
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3255689001264223e-06,
+      "loss": 2.3162,
+      "step": 14276
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3255214917825536e-06,
+      "loss": 3.0542,
+      "step": 14277
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3254740834386853e-06,
+      "loss": 3.2129,
+      "step": 14278
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.325426675094817e-06,
+      "loss": 3.1737,
+      "step": 14279
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.3253792667509482e-06,
+      "loss": 3.106,
+      "step": 14280
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.32533185840708e-06,
+      "loss": 2.1113,
+      "step": 14281
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3252844500632112e-06,
+      "loss": 3.3165,
+      "step": 14282
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3252370417193425e-06,
+      "loss": 3.715,
+      "step": 14283
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3251896333754742e-06,
+      "loss": 2.9511,
+      "step": 14284
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3251422250316055e-06,
+      "loss": 3.8905,
+      "step": 14285
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.325094816687737e-06,
+      "loss": 3.0366,
+      "step": 14286
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3250474083438685e-06,
+      "loss": 2.466,
+      "step": 14287
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.325e-06,
+      "loss": 4.0006,
+      "step": 14288
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3249525916561315e-06,
+      "loss": 3.8904,
+      "step": 14289
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3249051833122627e-06,
+      "loss": 3.6785,
+      "step": 14290
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3248577749683945e-06,
+      "loss": 3.3303,
+      "step": 14291
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.324810366624526e-06,
+      "loss": 3.5847,
+      "step": 14292
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3247629582806574e-06,
+      "loss": 3.1009,
+      "step": 14293
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.324715549936789e-06,
+      "loss": 2.4637,
+      "step": 14294
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3246681415929204e-06,
+      "loss": 2.8962,
+      "step": 14295
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3246207332490517e-06,
+      "loss": 2.5,
+      "step": 14296
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3245733249051834e-06,
+      "loss": 3.1446,
+      "step": 14297
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3245259165613147e-06,
+      "loss": 3.6018,
+      "step": 14298
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3244785082174464e-06,
+      "loss": 2.0642,
+      "step": 14299
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.324431099873578e-06,
+      "loss": 3.1149,
+      "step": 14300
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3243836915297094e-06,
+      "loss": 2.947,
+      "step": 14301
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3243362831858407e-06,
+      "loss": 3.1783,
+      "step": 14302
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3242888748419724e-06,
+      "loss": 3.347,
+      "step": 14303
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3242414664981036e-06,
+      "loss": 3.6313,
+      "step": 14304
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3241940581542353e-06,
+      "loss": 2.7416,
+      "step": 14305
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.324146649810367e-06,
+      "loss": 3.2749,
+      "step": 14306
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3240992414664983e-06,
+      "loss": 2.8922,
+      "step": 14307
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3240518331226296e-06,
+      "loss": 3.3958,
+      "step": 14308
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.324004424778761e-06,
+      "loss": 2.4935,
+      "step": 14309
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3239570164348926e-06,
+      "loss": 3.0104,
+      "step": 14310
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.323909608091024e-06,
+      "loss": 3.3112,
+      "step": 14311
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3238621997471556e-06,
+      "loss": 3.0725,
+      "step": 14312
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3238147914032873e-06,
+      "loss": 3.9799,
+      "step": 14313
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3237673830594186e-06,
+      "loss": 2.1924,
+      "step": 14314
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.32371997471555e-06,
+      "loss": 2.7686,
+      "step": 14315
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3236725663716815e-06,
+      "loss": 3.2253,
+      "step": 14316
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.323625158027813e-06,
+      "loss": 2.9219,
+      "step": 14317
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3235777496839445e-06,
+      "loss": 2.8812,
+      "step": 14318
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3235303413400762e-06,
+      "loss": 2.2895,
+      "step": 14319
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3234829329962075e-06,
+      "loss": 2.6247,
+      "step": 14320
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.323435524652339e-06,
+      "loss": 3.8437,
+      "step": 14321
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3233881163084705e-06,
+      "loss": 2.2567,
+      "step": 14322
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3233407079646018e-06,
+      "loss": 3.0759,
+      "step": 14323
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.323293299620733e-06,
+      "loss": 2.6575,
+      "step": 14324
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3232458912768648e-06,
+      "loss": 3.973,
+      "step": 14325
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3231984829329965e-06,
+      "loss": 2.8552,
+      "step": 14326
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3231510745891277e-06,
+      "loss": 3.0497,
+      "step": 14327
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3231036662452594e-06,
+      "loss": 2.9839,
+      "step": 14328
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3230562579013907e-06,
+      "loss": 2.9023,
+      "step": 14329
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.323008849557522e-06,
+      "loss": 3.4451,
+      "step": 14330
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3229614412136537e-06,
+      "loss": 3.1521,
+      "step": 14331
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3229140328697854e-06,
+      "loss": 2.8623,
+      "step": 14332
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3228666245259167e-06,
+      "loss": 3.8751,
+      "step": 14333
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.322819216182048e-06,
+      "loss": 2.6221,
+      "step": 14334
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3227718078381797e-06,
+      "loss": 3.3084,
+      "step": 14335
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.322724399494311e-06,
+      "loss": 3.3434,
+      "step": 14336
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3226769911504422e-06,
+      "loss": 2.6454,
+      "step": 14337
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.322629582806574e-06,
+      "loss": 3.042,
+      "step": 14338
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3225821744627057e-06,
+      "loss": 2.9796,
+      "step": 14339
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.322534766118837e-06,
+      "loss": 2.6072,
+      "step": 14340
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3224873577749686e-06,
+      "loss": 2.8432,
+      "step": 14341
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.3224399494311e-06,
+      "loss": 3.2699,
+      "step": 14342
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.322392541087231e-06,
+      "loss": 3.5492,
+      "step": 14343
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.322345132743363e-06,
+      "loss": 2.057,
+      "step": 14344
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3222977243994946e-06,
+      "loss": 3.7617,
+      "step": 14345
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.322250316055626e-06,
+      "loss": 2.4986,
+      "step": 14346
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3222029077117576e-06,
+      "loss": 2.8435,
+      "step": 14347
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.322155499367889e-06,
+      "loss": 2.522,
+      "step": 14348
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.32210809102402e-06,
+      "loss": 2.7219,
+      "step": 14349
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.322060682680152e-06,
+      "loss": 3.7314,
+      "step": 14350
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.322013274336283e-06,
+      "loss": 3.2619,
+      "step": 14351
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321965865992415e-06,
+      "loss": 2.7244,
+      "step": 14352
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321918457648546e-06,
+      "loss": 3.363,
+      "step": 14353
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321871049304678e-06,
+      "loss": 2.0346,
+      "step": 14354
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321823640960809e-06,
+      "loss": 3.1048,
+      "step": 14355
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3217762326169404e-06,
+      "loss": 2.1667,
+      "step": 14356
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321728824273072e-06,
+      "loss": 2.9793,
+      "step": 14357
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3216814159292034e-06,
+      "loss": 2.6792,
+      "step": 14358
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321634007585335e-06,
+      "loss": 3.4564,
+      "step": 14359
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3215865992414668e-06,
+      "loss": 3.5262,
+      "step": 14360
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321539190897598e-06,
+      "loss": 3.1578,
+      "step": 14361
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3214917825537293e-06,
+      "loss": 3.2229,
+      "step": 14362
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321444374209861e-06,
+      "loss": 2.7313,
+      "step": 14363
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3213969658659923e-06,
+      "loss": 2.6476,
+      "step": 14364
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321349557522124e-06,
+      "loss": 3.0293,
+      "step": 14365
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3213021491782557e-06,
+      "loss": 3.248,
+      "step": 14366
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.321254740834387e-06,
+      "loss": 3.037,
+      "step": 14367
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3212073324905183e-06,
+      "loss": 2.8768,
+      "step": 14368
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.32115992414665e-06,
+      "loss": 3.823,
+      "step": 14369
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3211125158027813e-06,
+      "loss": 2.3602,
+      "step": 14370
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3210651074589126e-06,
+      "loss": 4.3382,
+      "step": 14371
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3210176991150443e-06,
+      "loss": 2.958,
+      "step": 14372
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320970290771176e-06,
+      "loss": 2.5346,
+      "step": 14373
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3209228824273072e-06,
+      "loss": 2.7385,
+      "step": 14374
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3208754740834385e-06,
+      "loss": 2.7292,
+      "step": 14375
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3208280657395702e-06,
+      "loss": 3.1579,
+      "step": 14376
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3207806573957015e-06,
+      "loss": 3.6678,
+      "step": 14377
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320733249051833e-06,
+      "loss": 3.0018,
+      "step": 14378
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320685840707965e-06,
+      "loss": 2.9767,
+      "step": 14379
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320638432364096e-06,
+      "loss": 4.2464,
+      "step": 14380
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3205910240202275e-06,
+      "loss": 3.8947,
+      "step": 14381
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320543615676359e-06,
+      "loss": 2.6653,
+      "step": 14382
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3204962073324905e-06,
+      "loss": 2.3739,
+      "step": 14383
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3204487989886217e-06,
+      "loss": 3.9648,
+      "step": 14384
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3204013906447534e-06,
+      "loss": 3.0641,
+      "step": 14385
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320353982300885e-06,
+      "loss": 2.4817,
+      "step": 14386
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3203065739570164e-06,
+      "loss": 3.4287,
+      "step": 14387
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320259165613148e-06,
+      "loss": 2.8992,
+      "step": 14388
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3202117572692794e-06,
+      "loss": 3.5206,
+      "step": 14389
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3201643489254107e-06,
+      "loss": 2.7394,
+      "step": 14390
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3201169405815424e-06,
+      "loss": 2.6168,
+      "step": 14391
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.320069532237674e-06,
+      "loss": 2.8615,
+      "step": 14392
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3200221238938054e-06,
+      "loss": 2.5391,
+      "step": 14393
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.319974715549937e-06,
+      "loss": 2.6321,
+      "step": 14394
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3199273072060684e-06,
+      "loss": 3.5464,
+      "step": 14395
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3198798988621996e-06,
+      "loss": 3.1058,
+      "step": 14396
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.319832490518331e-06,
+      "loss": 3.8971,
+      "step": 14397
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3197850821744626e-06,
+      "loss": 2.6255,
+      "step": 14398
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3197376738305943e-06,
+      "loss": 2.6535,
+      "step": 14399
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3196902654867256e-06,
+      "loss": 4.0499,
+      "step": 14400
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3196428571428573e-06,
+      "loss": 3.5347,
+      "step": 14401
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3195954487989886e-06,
+      "loss": 3.1252,
+      "step": 14402
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.31954804045512e-06,
+      "loss": 1.9117,
+      "step": 14403
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3195006321112516e-06,
+      "loss": 2.5438,
+      "step": 14404
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3194532237673833e-06,
+      "loss": 3.5172,
+      "step": 14405
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3194058154235146e-06,
+      "loss": 2.8839,
+      "step": 14406
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.3193584070796463e-06,
+      "loss": 3.5044,
+      "step": 14407
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3193109987357776e-06,
+      "loss": 4.0196,
+      "step": 14408
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.319263590391909e-06,
+      "loss": 2.6511,
+      "step": 14409
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3192161820480405e-06,
+      "loss": 2.8328,
+      "step": 14410
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.319168773704172e-06,
+      "loss": 3.4219,
+      "step": 14411
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3191213653603035e-06,
+      "loss": 2.5685,
+      "step": 14412
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3190739570164352e-06,
+      "loss": 2.9582,
+      "step": 14413
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3190265486725665e-06,
+      "loss": 2.039,
+      "step": 14414
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3189791403286978e-06,
+      "loss": 3.5066,
+      "step": 14415
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3189317319848295e-06,
+      "loss": 2.8808,
+      "step": 14416
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3188843236409608e-06,
+      "loss": 3.3803,
+      "step": 14417
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3188369152970925e-06,
+      "loss": 2.8321,
+      "step": 14418
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3187895069532238e-06,
+      "loss": 2.3863,
+      "step": 14419
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3187420986093555e-06,
+      "loss": 4.2208,
+      "step": 14420
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3186946902654867e-06,
+      "loss": 3.3408,
+      "step": 14421
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.318647281921618e-06,
+      "loss": 2.6955,
+      "step": 14422
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3185998735777497e-06,
+      "loss": 3.906,
+      "step": 14423
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.318552465233881e-06,
+      "loss": 2.8941,
+      "step": 14424
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3185050568900127e-06,
+      "loss": 2.6302,
+      "step": 14425
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3184576485461444e-06,
+      "loss": 3.6856,
+      "step": 14426
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3184102402022757e-06,
+      "loss": 3.615,
+      "step": 14427
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.318362831858407e-06,
+      "loss": 3.5444,
+      "step": 14428
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3183154235145387e-06,
+      "loss": 3.2915,
+      "step": 14429
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.31826801517067e-06,
+      "loss": 2.338,
+      "step": 14430
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3182206068268017e-06,
+      "loss": 3.7886,
+      "step": 14431
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3181731984829334e-06,
+      "loss": 4.1194,
+      "step": 14432
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3181257901390646e-06,
+      "loss": 3.2111,
+      "step": 14433
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.318078381795196e-06,
+      "loss": 2.7946,
+      "step": 14434
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3180309734513276e-06,
+      "loss": 3.0777,
+      "step": 14435
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317983565107459e-06,
+      "loss": 2.4707,
+      "step": 14436
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.31793615676359e-06,
+      "loss": 3.1482,
+      "step": 14437
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317888748419722e-06,
+      "loss": 2.3395,
+      "step": 14438
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3178413400758536e-06,
+      "loss": 3.4486,
+      "step": 14439
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317793931731985e-06,
+      "loss": 2.3541,
+      "step": 14440
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317746523388116e-06,
+      "loss": 2.8218,
+      "step": 14441
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317699115044248e-06,
+      "loss": 2.6414,
+      "step": 14442
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317651706700379e-06,
+      "loss": 2.2972,
+      "step": 14443
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317604298356511e-06,
+      "loss": 2.1808,
+      "step": 14444
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3175568900126426e-06,
+      "loss": 3.2594,
+      "step": 14445
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317509481668774e-06,
+      "loss": 2.4838,
+      "step": 14446
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317462073324905e-06,
+      "loss": 2.8662,
+      "step": 14447
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317414664981037e-06,
+      "loss": 3.7288,
+      "step": 14448
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317367256637168e-06,
+      "loss": 2.9475,
+      "step": 14449
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3173198482932994e-06,
+      "loss": 3.9598,
+      "step": 14450
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317272439949431e-06,
+      "loss": 2.9889,
+      "step": 14451
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3172250316055628e-06,
+      "loss": 2.6238,
+      "step": 14452
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317177623261694e-06,
+      "loss": 2.7853,
+      "step": 14453
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3171302149178258e-06,
+      "loss": 3.5397,
+      "step": 14454
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.317082806573957e-06,
+      "loss": 3.654,
+      "step": 14455
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3170353982300883e-06,
+      "loss": 3.8609,
+      "step": 14456
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.31698798988622e-06,
+      "loss": 2.2997,
+      "step": 14457
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3169405815423517e-06,
+      "loss": 2.5274,
+      "step": 14458
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.316893173198483e-06,
+      "loss": 4.4561,
+      "step": 14459
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3168457648546147e-06,
+      "loss": 3.3289,
+      "step": 14460
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.316798356510746e-06,
+      "loss": 2.8913,
+      "step": 14461
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3167509481668773e-06,
+      "loss": 2.6644,
+      "step": 14462
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3167035398230086e-06,
+      "loss": 3.3854,
+      "step": 14463
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3166561314791403e-06,
+      "loss": 3.5609,
+      "step": 14464
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.316608723135272e-06,
+      "loss": 2.2724,
+      "step": 14465
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3165613147914033e-06,
+      "loss": 2.4998,
+      "step": 14466
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.316513906447535e-06,
+      "loss": 3.5942,
+      "step": 14467
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3164664981036662e-06,
+      "loss": 3.299,
+      "step": 14468
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3164190897597975e-06,
+      "loss": 2.1187,
+      "step": 14469
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.3163716814159292e-06,
+      "loss": 2.0815,
+      "step": 14470
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.316324273072061e-06,
+      "loss": 2.9501,
+      "step": 14471
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.316276864728192e-06,
+      "loss": 3.1012,
+      "step": 14472
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.316229456384324e-06,
+      "loss": 3.1172,
+      "step": 14473
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.316182048040455e-06,
+      "loss": 3.0707,
+      "step": 14474
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3161346396965865e-06,
+      "loss": 1.727,
+      "step": 14475
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.316087231352718e-06,
+      "loss": 3.4503,
+      "step": 14476
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3160398230088495e-06,
+      "loss": 3.8399,
+      "step": 14477
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315992414664981e-06,
+      "loss": 3.7807,
+      "step": 14478
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315945006321113e-06,
+      "loss": 3.0325,
+      "step": 14479
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315897597977244e-06,
+      "loss": 3.4798,
+      "step": 14480
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3158501896333754e-06,
+      "loss": 3.5083,
+      "step": 14481
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315802781289507e-06,
+      "loss": 2.7676,
+      "step": 14482
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3157553729456384e-06,
+      "loss": 2.765,
+      "step": 14483
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.31570796460177e-06,
+      "loss": 3.4944,
+      "step": 14484
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3156605562579014e-06,
+      "loss": 3.5758,
+      "step": 14485
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315613147914033e-06,
+      "loss": 2.5817,
+      "step": 14486
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3155657395701644e-06,
+      "loss": 3.6431,
+      "step": 14487
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3155183312262957e-06,
+      "loss": 3.1084,
+      "step": 14488
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3154709228824274e-06,
+      "loss": 4.3512,
+      "step": 14489
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3154235145385586e-06,
+      "loss": 4.0612,
+      "step": 14490
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3153761061946903e-06,
+      "loss": 2.7419,
+      "step": 14491
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315328697850822e-06,
+      "loss": 4.2121,
+      "step": 14492
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3152812895069533e-06,
+      "loss": 2.9011,
+      "step": 14493
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3152338811630846e-06,
+      "loss": 4.1365,
+      "step": 14494
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3151864728192163e-06,
+      "loss": 2.7922,
+      "step": 14495
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3151390644753476e-06,
+      "loss": 2.1594,
+      "step": 14496
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3150916561314793e-06,
+      "loss": 2.0722,
+      "step": 14497
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.315044247787611e-06,
+      "loss": 1.9459,
+      "step": 14498
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3149968394437423e-06,
+      "loss": 2.5059,
+      "step": 14499
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3149494310998736e-06,
+      "loss": 3.1832,
+      "step": 14500
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3149020227560053e-06,
+      "loss": 2.8759,
+      "step": 14501
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3148546144121365e-06,
+      "loss": 2.5866,
+      "step": 14502
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.314807206068268e-06,
+      "loss": 3.3056,
+      "step": 14503
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3147597977243995e-06,
+      "loss": 4.0041,
+      "step": 14504
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3147123893805312e-06,
+      "loss": 2.9908,
+      "step": 14505
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3146649810366625e-06,
+      "loss": 2.6509,
+      "step": 14506
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.314617572692794e-06,
+      "loss": 2.9238,
+      "step": 14507
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3145701643489255e-06,
+      "loss": 2.112,
+      "step": 14508
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3145227560050568e-06,
+      "loss": 3.8562,
+      "step": 14509
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3144753476611885e-06,
+      "loss": 3.8672,
+      "step": 14510
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.31442793931732e-06,
+      "loss": 2.9507,
+      "step": 14511
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3143805309734515e-06,
+      "loss": 2.9865,
+      "step": 14512
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3143331226295827e-06,
+      "loss": 3.9263,
+      "step": 14513
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3142857142857145e-06,
+      "loss": 2.4734,
+      "step": 14514
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3142383059418457e-06,
+      "loss": 2.6318,
+      "step": 14515
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.314190897597977e-06,
+      "loss": 3.4178,
+      "step": 14516
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3141434892541087e-06,
+      "loss": 3.0088,
+      "step": 14517
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3140960809102404e-06,
+      "loss": 3.7888,
+      "step": 14518
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3140486725663717e-06,
+      "loss": 2.3385,
+      "step": 14519
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3140012642225034e-06,
+      "loss": 3.4912,
+      "step": 14520
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3139538558786347e-06,
+      "loss": 3.4514,
+      "step": 14521
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.313906447534766e-06,
+      "loss": 3.2421,
+      "step": 14522
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3138590391908977e-06,
+      "loss": 3.2251,
+      "step": 14523
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3138116308470294e-06,
+      "loss": 2.7882,
+      "step": 14524
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3137642225031607e-06,
+      "loss": 3.3451,
+      "step": 14525
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3137168141592924e-06,
+      "loss": 2.4781,
+      "step": 14526
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3136694058154236e-06,
+      "loss": 2.8575,
+      "step": 14527
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.313621997471555e-06,
+      "loss": 3.3822,
+      "step": 14528
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.313574589127686e-06,
+      "loss": 3.0365,
+      "step": 14529
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.313527180783818e-06,
+      "loss": 3.1085,
+      "step": 14530
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3134797724399496e-06,
+      "loss": 3.4794,
+      "step": 14531
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.313432364096081e-06,
+      "loss": 3.1538,
+      "step": 14532
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.3133849557522126e-06,
+      "loss": 2.415,
+      "step": 14533
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.313337547408344e-06,
+      "loss": 2.7229,
+      "step": 14534
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.313290139064475e-06,
+      "loss": 2.6917,
+      "step": 14535
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.313242730720607e-06,
+      "loss": 2.0337,
+      "step": 14536
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3131953223767386e-06,
+      "loss": 2.9831,
+      "step": 14537
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.31314791403287e-06,
+      "loss": 3.5395,
+      "step": 14538
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3131005056890015e-06,
+      "loss": 3.5675,
+      "step": 14539
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.313053097345133e-06,
+      "loss": 3.6599,
+      "step": 14540
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.313005689001264e-06,
+      "loss": 2.5791,
+      "step": 14541
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312958280657396e-06,
+      "loss": 2.8613,
+      "step": 14542
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312910872313527e-06,
+      "loss": 2.6502,
+      "step": 14543
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312863463969659e-06,
+      "loss": 3.089,
+      "step": 14544
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3128160556257905e-06,
+      "loss": 3.4189,
+      "step": 14545
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3127686472819218e-06,
+      "loss": 3.0865,
+      "step": 14546
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312721238938053e-06,
+      "loss": 2.9876,
+      "step": 14547
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3126738305941848e-06,
+      "loss": 3.2692,
+      "step": 14548
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312626422250316e-06,
+      "loss": 3.6021,
+      "step": 14549
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3125790139064477e-06,
+      "loss": 2.3087,
+      "step": 14550
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312531605562579e-06,
+      "loss": 2.0707,
+      "step": 14551
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3124841972187107e-06,
+      "loss": 3.3667,
+      "step": 14552
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312436788874842e-06,
+      "loss": 3.2373,
+      "step": 14553
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3123893805309733e-06,
+      "loss": 3.1768,
+      "step": 14554
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312341972187105e-06,
+      "loss": 3.1439,
+      "step": 14555
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3122945638432363e-06,
+      "loss": 3.9382,
+      "step": 14556
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312247155499368e-06,
+      "loss": 3.2781,
+      "step": 14557
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3121997471554997e-06,
+      "loss": 3.3297,
+      "step": 14558
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312152338811631e-06,
+      "loss": 2.5314,
+      "step": 14559
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3121049304677622e-06,
+      "loss": 3.3548,
+      "step": 14560
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.312057522123894e-06,
+      "loss": 3.5552,
+      "step": 14561
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3120101137800252e-06,
+      "loss": 2.8126,
+      "step": 14562
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311962705436157e-06,
+      "loss": 4.4511,
+      "step": 14563
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3119152970922886e-06,
+      "loss": 3.4623,
+      "step": 14564
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.31186788874842e-06,
+      "loss": 4.0433,
+      "step": 14565
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311820480404551e-06,
+      "loss": 2.932,
+      "step": 14566
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311773072060683e-06,
+      "loss": 3.8751,
+      "step": 14567
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311725663716814e-06,
+      "loss": 3.2722,
+      "step": 14568
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3116782553729455e-06,
+      "loss": 2.5745,
+      "step": 14569
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311630847029077e-06,
+      "loss": 4.022,
+      "step": 14570
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311583438685209e-06,
+      "loss": 3.1183,
+      "step": 14571
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.31153603034134e-06,
+      "loss": 2.7063,
+      "step": 14572
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3114886219974714e-06,
+      "loss": 3.0642,
+      "step": 14573
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311441213653603e-06,
+      "loss": 2.9688,
+      "step": 14574
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3113938053097344e-06,
+      "loss": 2.8293,
+      "step": 14575
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311346396965866e-06,
+      "loss": 4.321,
+      "step": 14576
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311298988621998e-06,
+      "loss": 2.3763,
+      "step": 14577
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311251580278129e-06,
+      "loss": 1.9207,
+      "step": 14578
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3112041719342604e-06,
+      "loss": 2.8043,
+      "step": 14579
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.311156763590392e-06,
+      "loss": 2.6134,
+      "step": 14580
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3111093552465234e-06,
+      "loss": 3.0305,
+      "step": 14581
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3110619469026547e-06,
+      "loss": 2.8987,
+      "step": 14582
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3110145385587864e-06,
+      "loss": 2.0993,
+      "step": 14583
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.310967130214918e-06,
+      "loss": 3.319,
+      "step": 14584
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3109197218710493e-06,
+      "loss": 3.572,
+      "step": 14585
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.310872313527181e-06,
+      "loss": 3.0642,
+      "step": 14586
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3108249051833123e-06,
+      "loss": 3.2117,
+      "step": 14587
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3107774968394436e-06,
+      "loss": 2.6886,
+      "step": 14588
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3107300884955753e-06,
+      "loss": 2.8213,
+      "step": 14589
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.310682680151707e-06,
+      "loss": 3.1663,
+      "step": 14590
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3106352718078383e-06,
+      "loss": 2.7883,
+      "step": 14591
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3105878634639696e-06,
+      "loss": 2.4962,
+      "step": 14592
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3105404551201013e-06,
+      "loss": 2.8266,
+      "step": 14593
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3104930467762326e-06,
+      "loss": 3.2233,
+      "step": 14594
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.310445638432364e-06,
+      "loss": 3.6497,
+      "step": 14595
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3103982300884955e-06,
+      "loss": 3.0231,
+      "step": 14596
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.3103508217446272e-06,
+      "loss": 3.9642,
+      "step": 14597
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3103034134007585e-06,
+      "loss": 3.2351,
+      "step": 14598
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3102560050568902e-06,
+      "loss": 2.6555,
+      "step": 14599
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3102085967130215e-06,
+      "loss": 3.0812,
+      "step": 14600
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.310161188369153e-06,
+      "loss": 2.0726,
+      "step": 14601
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3101137800252845e-06,
+      "loss": 3.0378,
+      "step": 14602
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.310066371681416e-06,
+      "loss": 2.9233,
+      "step": 14603
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3100189633375475e-06,
+      "loss": 3.553,
+      "step": 14604
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.309971554993679e-06,
+      "loss": 2.2615,
+      "step": 14605
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3099241466498105e-06,
+      "loss": 3.0746,
+      "step": 14606
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3098767383059417e-06,
+      "loss": 2.7594,
+      "step": 14607
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3098293299620734e-06,
+      "loss": 2.5478,
+      "step": 14608
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3097819216182047e-06,
+      "loss": 3.442,
+      "step": 14609
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3097345132743364e-06,
+      "loss": 3.9486,
+      "step": 14610
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.309687104930468e-06,
+      "loss": 2.8874,
+      "step": 14611
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3096396965865994e-06,
+      "loss": 2.3611,
+      "step": 14612
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3095922882427307e-06,
+      "loss": 2.7926,
+      "step": 14613
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.309544879898862e-06,
+      "loss": 1.9373,
+      "step": 14614
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3094974715549937e-06,
+      "loss": 2.5189,
+      "step": 14615
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3094500632111254e-06,
+      "loss": 3.5876,
+      "step": 14616
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3094026548672567e-06,
+      "loss": 2.4404,
+      "step": 14617
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3093552465233884e-06,
+      "loss": 3.423,
+      "step": 14618
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3093078381795196e-06,
+      "loss": 2.2004,
+      "step": 14619
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.309260429835651e-06,
+      "loss": 3.1256,
+      "step": 14620
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3092130214917826e-06,
+      "loss": 2.5832,
+      "step": 14621
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.309165613147914e-06,
+      "loss": 2.6244,
+      "step": 14622
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3091182048040456e-06,
+      "loss": 3.4415,
+      "step": 14623
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3090707964601773e-06,
+      "loss": 3.508,
+      "step": 14624
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3090233881163086e-06,
+      "loss": 3.0929,
+      "step": 14625
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.30897597977244e-06,
+      "loss": 2.5086,
+      "step": 14626
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3089285714285716e-06,
+      "loss": 1.8333,
+      "step": 14627
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308881163084703e-06,
+      "loss": 2.637,
+      "step": 14628
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308833754740834e-06,
+      "loss": 3.6129,
+      "step": 14629
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3087863463969663e-06,
+      "loss": 2.2014,
+      "step": 14630
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3087389380530976e-06,
+      "loss": 2.4363,
+      "step": 14631
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308691529709229e-06,
+      "loss": 2.5014,
+      "step": 14632
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3086441213653605e-06,
+      "loss": 3.6692,
+      "step": 14633
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308596713021492e-06,
+      "loss": 3.0315,
+      "step": 14634
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308549304677623e-06,
+      "loss": 3.1687,
+      "step": 14635
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308501896333755e-06,
+      "loss": 2.4718,
+      "step": 14636
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3084544879898865e-06,
+      "loss": 2.6709,
+      "step": 14637
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308407079646018e-06,
+      "loss": 4.5035,
+      "step": 14638
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308359671302149e-06,
+      "loss": 3.5111,
+      "step": 14639
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3083122629582808e-06,
+      "loss": 3.0027,
+      "step": 14640
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308264854614412e-06,
+      "loss": 3.1135,
+      "step": 14641
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3082174462705433e-06,
+      "loss": 2.9025,
+      "step": 14642
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308170037926675e-06,
+      "loss": 2.7126,
+      "step": 14643
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3081226295828067e-06,
+      "loss": 3.4594,
+      "step": 14644
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.308075221238938e-06,
+      "loss": 3.7544,
+      "step": 14645
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3080278128950697e-06,
+      "loss": 3.1507,
+      "step": 14646
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307980404551201e-06,
+      "loss": 3.8648,
+      "step": 14647
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3079329962073323e-06,
+      "loss": 2.381,
+      "step": 14648
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307885587863464e-06,
+      "loss": 3.7428,
+      "step": 14649
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3078381795195957e-06,
+      "loss": 2.8478,
+      "step": 14650
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307790771175727e-06,
+      "loss": 2.3962,
+      "step": 14651
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3077433628318587e-06,
+      "loss": 3.4408,
+      "step": 14652
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.30769595448799e-06,
+      "loss": 3.7937,
+      "step": 14653
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3076485461441212e-06,
+      "loss": 3.0399,
+      "step": 14654
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307601137800253e-06,
+      "loss": 3.2071,
+      "step": 14655
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.3075537294563842e-06,
+      "loss": 2.7021,
+      "step": 14656
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307506321112516e-06,
+      "loss": 3.524,
+      "step": 14657
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307458912768647e-06,
+      "loss": 3.3182,
+      "step": 14658
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.307411504424779e-06,
+      "loss": 2.6525,
+      "step": 14659
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.30736409608091e-06,
+      "loss": 4.1655,
+      "step": 14660
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3073166877370415e-06,
+      "loss": 3.8306,
+      "step": 14661
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.307269279393173e-06,
+      "loss": 2.5307,
+      "step": 14662
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.307221871049305e-06,
+      "loss": 3.0827,
+      "step": 14663
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.307174462705436e-06,
+      "loss": 3.5573,
+      "step": 14664
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.307127054361568e-06,
+      "loss": 2.1489,
+      "step": 14665
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.307079646017699e-06,
+      "loss": 3.0156,
+      "step": 14666
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3070322376738304e-06,
+      "loss": 2.0604,
+      "step": 14667
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306984829329962e-06,
+      "loss": 3.2426,
+      "step": 14668
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3069374209860934e-06,
+      "loss": 3.4529,
+      "step": 14669
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306890012642225e-06,
+      "loss": 3.6692,
+      "step": 14670
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306842604298357e-06,
+      "loss": 3.303,
+      "step": 14671
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306795195954488e-06,
+      "loss": 3.3909,
+      "step": 14672
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3067477876106194e-06,
+      "loss": 2.6012,
+      "step": 14673
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306700379266751e-06,
+      "loss": 3.3815,
+      "step": 14674
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3066529709228824e-06,
+      "loss": 3.8947,
+      "step": 14675
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306605562579014e-06,
+      "loss": 3.2925,
+      "step": 14676
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3065581542351458e-06,
+      "loss": 3.4479,
+      "step": 14677
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306510745891277e-06,
+      "loss": 2.1808,
+      "step": 14678
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3064633375474083e-06,
+      "loss": 2.3843,
+      "step": 14679
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3064159292035396e-06,
+      "loss": 3.5355,
+      "step": 14680
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3063685208596713e-06,
+      "loss": 3.3071,
+      "step": 14681
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3063211125158026e-06,
+      "loss": 3.8655,
+      "step": 14682
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3062737041719343e-06,
+      "loss": 3.0532,
+      "step": 14683
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.306226295828066e-06,
+      "loss": 3.1346,
+      "step": 14684
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3061788874841973e-06,
+      "loss": 3.9227,
+      "step": 14685
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3061314791403286e-06,
+      "loss": 3.4824,
+      "step": 14686
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3060840707964603e-06,
+      "loss": 3.6396,
+      "step": 14687
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3060366624525916e-06,
+      "loss": 2.653,
+      "step": 14688
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3059892541087233e-06,
+      "loss": 2.1739,
+      "step": 14689
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.305941845764855e-06,
+      "loss": 3.8462,
+      "step": 14690
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3058944374209862e-06,
+      "loss": 3.0787,
+      "step": 14691
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3058470290771175e-06,
+      "loss": 3.8124,
+      "step": 14692
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3057996207332492e-06,
+      "loss": 2.9329,
+      "step": 14693
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3057522123893805e-06,
+      "loss": 2.8206,
+      "step": 14694
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3057048040455118e-06,
+      "loss": 3.0082,
+      "step": 14695
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3056573957016435e-06,
+      "loss": 2.6692,
+      "step": 14696
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.305609987357775e-06,
+      "loss": 3.3037,
+      "step": 14697
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3055625790139065e-06,
+      "loss": 3.7964,
+      "step": 14698
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.305515170670038e-06,
+      "loss": 3.6529,
+      "step": 14699
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3054677623261695e-06,
+      "loss": 3.844,
+      "step": 14700
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3054203539823007e-06,
+      "loss": 2.3372,
+      "step": 14701
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3053729456384324e-06,
+      "loss": 3.033,
+      "step": 14702
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.305325537294564e-06,
+      "loss": 3.0026,
+      "step": 14703
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3052781289506954e-06,
+      "loss": 3.0054,
+      "step": 14704
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3052307206068267e-06,
+      "loss": 3.103,
+      "step": 14705
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3051833122629584e-06,
+      "loss": 2.9733,
+      "step": 14706
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3051359039190897e-06,
+      "loss": 2.6808,
+      "step": 14707
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.305088495575221e-06,
+      "loss": 3.1293,
+      "step": 14708
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3050410872313527e-06,
+      "loss": 3.6535,
+      "step": 14709
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3049936788874844e-06,
+      "loss": 2.5501,
+      "step": 14710
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3049462705436157e-06,
+      "loss": 2.5767,
+      "step": 14711
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3048988621997474e-06,
+      "loss": 3.9394,
+      "step": 14712
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3048514538558786e-06,
+      "loss": 2.1471,
+      "step": 14713
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.30480404551201e-06,
+      "loss": 3.3289,
+      "step": 14714
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3047566371681416e-06,
+      "loss": 3.0996,
+      "step": 14715
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3047092288242733e-06,
+      "loss": 2.8841,
+      "step": 14716
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3046618204804046e-06,
+      "loss": 3.134,
+      "step": 14717
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3046144121365363e-06,
+      "loss": 2.775,
+      "step": 14718
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3045670037926676e-06,
+      "loss": 2.9133,
+      "step": 14719
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.304519595448799e-06,
+      "loss": 2.1983,
+      "step": 14720
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3044721871049306e-06,
+      "loss": 3.7444,
+      "step": 14721
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.304424778761062e-06,
+      "loss": 2.9882,
+      "step": 14722
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.3043773704171936e-06,
+      "loss": 2.5267,
+      "step": 14723
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.304329962073325e-06,
+      "loss": 2.5438,
+      "step": 14724
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3042825537294565e-06,
+      "loss": 3.5331,
+      "step": 14725
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.304235145385588e-06,
+      "loss": 3.2826,
+      "step": 14726
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.304187737041719e-06,
+      "loss": 4.362,
+      "step": 14727
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.304140328697851e-06,
+      "loss": 2.2496,
+      "step": 14728
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3040929203539825e-06,
+      "loss": 2.9151,
+      "step": 14729
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.304045512010114e-06,
+      "loss": 4.0945,
+      "step": 14730
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3039981036662455e-06,
+      "loss": 2.7172,
+      "step": 14731
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3039506953223768e-06,
+      "loss": 3.8551,
+      "step": 14732
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303903286978508e-06,
+      "loss": 3.5451,
+      "step": 14733
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3038558786346398e-06,
+      "loss": 2.6482,
+      "step": 14734
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303808470290771e-06,
+      "loss": 2.7519,
+      "step": 14735
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3037610619469028e-06,
+      "loss": 3.5231,
+      "step": 14736
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3037136536030345e-06,
+      "loss": 2.9417,
+      "step": 14737
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3036662452591657e-06,
+      "loss": 2.9841,
+      "step": 14738
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303618836915297e-06,
+      "loss": 3.4838,
+      "step": 14739
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3035714285714287e-06,
+      "loss": 3.749,
+      "step": 14740
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.30352402022756e-06,
+      "loss": 3.5729,
+      "step": 14741
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3034766118836917e-06,
+      "loss": 2.7625,
+      "step": 14742
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3034292035398234e-06,
+      "loss": 2.3528,
+      "step": 14743
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3033817951959547e-06,
+      "loss": 2.0663,
+      "step": 14744
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303334386852086e-06,
+      "loss": 1.7435,
+      "step": 14745
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3032869785082173e-06,
+      "loss": 3.1619,
+      "step": 14746
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303239570164349e-06,
+      "loss": 3.3683,
+      "step": 14747
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3031921618204802e-06,
+      "loss": 2.4402,
+      "step": 14748
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303144753476612e-06,
+      "loss": 2.9723,
+      "step": 14749
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3030973451327436e-06,
+      "loss": 3.5482,
+      "step": 14750
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303049936788875e-06,
+      "loss": 3.0852,
+      "step": 14751
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.303002528445006e-06,
+      "loss": 3.5662,
+      "step": 14752
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302955120101138e-06,
+      "loss": 2.8791,
+      "step": 14753
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302907711757269e-06,
+      "loss": 3.3795,
+      "step": 14754
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302860303413401e-06,
+      "loss": 2.8064,
+      "step": 14755
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3028128950695326e-06,
+      "loss": 3.7562,
+      "step": 14756
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302765486725664e-06,
+      "loss": 2.8129,
+      "step": 14757
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302718078381795e-06,
+      "loss": 1.6998,
+      "step": 14758
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302670670037927e-06,
+      "loss": 3.8858,
+      "step": 14759
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302623261694058e-06,
+      "loss": 3.0468,
+      "step": 14760
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3025758533501894e-06,
+      "loss": 2.0202,
+      "step": 14761
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302528445006321e-06,
+      "loss": 2.7219,
+      "step": 14762
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302481036662453e-06,
+      "loss": 2.5478,
+      "step": 14763
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302433628318584e-06,
+      "loss": 2.6262,
+      "step": 14764
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302386219974716e-06,
+      "loss": 4.0503,
+      "step": 14765
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302338811630847e-06,
+      "loss": 3.9478,
+      "step": 14766
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3022914032869784e-06,
+      "loss": 3.0899,
+      "step": 14767
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.30224399494311e-06,
+      "loss": 2.6252,
+      "step": 14768
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3021965865992418e-06,
+      "loss": 3.2647,
+      "step": 14769
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302149178255373e-06,
+      "loss": 3.567,
+      "step": 14770
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3021017699115043e-06,
+      "loss": 2.9531,
+      "step": 14771
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.302054361567636e-06,
+      "loss": 4.0694,
+      "step": 14772
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3020069532237673e-06,
+      "loss": 3.9423,
+      "step": 14773
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3019595448798986e-06,
+      "loss": 3.5669,
+      "step": 14774
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3019121365360303e-06,
+      "loss": 2.9605,
+      "step": 14775
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.301864728192162e-06,
+      "loss": 2.5189,
+      "step": 14776
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3018173198482933e-06,
+      "loss": 3.1482,
+      "step": 14777
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.301769911504425e-06,
+      "loss": 3.1183,
+      "step": 14778
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3017225031605563e-06,
+      "loss": 3.2973,
+      "step": 14779
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3016750948166876e-06,
+      "loss": 3.8761,
+      "step": 14780
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3016276864728193e-06,
+      "loss": 2.0234,
+      "step": 14781
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.301580278128951e-06,
+      "loss": 1.8536,
+      "step": 14782
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3015328697850822e-06,
+      "loss": 2.5236,
+      "step": 14783
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.301485461441214e-06,
+      "loss": 4.0828,
+      "step": 14784
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3014380530973452e-06,
+      "loss": 2.9562,
+      "step": 14785
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3013906447534765e-06,
+      "loss": 2.952,
+      "step": 14786
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.3013432364096082e-06,
+      "loss": 3.3075,
+      "step": 14787
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3012958280657395e-06,
+      "loss": 3.4284,
+      "step": 14788
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.301248419721871e-06,
+      "loss": 3.4288,
+      "step": 14789
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3012010113780025e-06,
+      "loss": 3.2752,
+      "step": 14790
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.301153603034134e-06,
+      "loss": 3.0235,
+      "step": 14791
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3011061946902655e-06,
+      "loss": 3.9692,
+      "step": 14792
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3010587863463967e-06,
+      "loss": 2.494,
+      "step": 14793
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3010113780025285e-06,
+      "loss": 3.7625,
+      "step": 14794
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.30096396965866e-06,
+      "loss": 3.8646,
+      "step": 14795
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3009165613147914e-06,
+      "loss": 3.8003,
+      "step": 14796
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.300869152970923e-06,
+      "loss": 3.9827,
+      "step": 14797
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3008217446270544e-06,
+      "loss": 3.6676,
+      "step": 14798
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3007743362831857e-06,
+      "loss": 3.2198,
+      "step": 14799
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3007269279393174e-06,
+      "loss": 2.9616,
+      "step": 14800
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3006795195954487e-06,
+      "loss": 2.9409,
+      "step": 14801
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3006321112515804e-06,
+      "loss": 3.5711,
+      "step": 14802
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.300584702907712e-06,
+      "loss": 3.186,
+      "step": 14803
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3005372945638434e-06,
+      "loss": 2.9401,
+      "step": 14804
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3004898862199747e-06,
+      "loss": 3.1441,
+      "step": 14805
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3004424778761064e-06,
+      "loss": 2.3736,
+      "step": 14806
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3003950695322376e-06,
+      "loss": 2.9347,
+      "step": 14807
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3003476611883693e-06,
+      "loss": 2.779,
+      "step": 14808
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.300300252844501e-06,
+      "loss": 3.5719,
+      "step": 14809
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3002528445006323e-06,
+      "loss": 2.3469,
+      "step": 14810
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3002054361567636e-06,
+      "loss": 3.7767,
+      "step": 14811
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.300158027812895e-06,
+      "loss": 2.5276,
+      "step": 14812
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3001106194690266e-06,
+      "loss": 3.9712,
+      "step": 14813
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.300063211125158e-06,
+      "loss": 2.1788,
+      "step": 14814
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.3000158027812896e-06,
+      "loss": 2.7562,
+      "step": 14815
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2999683944374213e-06,
+      "loss": 2.3843,
+      "step": 14816
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2999209860935526e-06,
+      "loss": 3.3935,
+      "step": 14817
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.299873577749684e-06,
+      "loss": 1.9686,
+      "step": 14818
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2998261694058155e-06,
+      "loss": 2.5375,
+      "step": 14819
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.299778761061947e-06,
+      "loss": 3.5072,
+      "step": 14820
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2997313527180785e-06,
+      "loss": 2.6132,
+      "step": 14821
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2996839443742102e-06,
+      "loss": 2.9051,
+      "step": 14822
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2996365360303415e-06,
+      "loss": 3.3278,
+      "step": 14823
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.299589127686473e-06,
+      "loss": 3.252,
+      "step": 14824
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2995417193426045e-06,
+      "loss": 3.3362,
+      "step": 14825
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2994943109987358e-06,
+      "loss": 1.7443,
+      "step": 14826
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.299446902654867e-06,
+      "loss": 3.5799,
+      "step": 14827
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2993994943109988e-06,
+      "loss": 3.7473,
+      "step": 14828
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2993520859671305e-06,
+      "loss": 3.5267,
+      "step": 14829
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2993046776232617e-06,
+      "loss": 3.8225,
+      "step": 14830
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2992572692793934e-06,
+      "loss": 2.8096,
+      "step": 14831
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2992098609355247e-06,
+      "loss": 2.9768,
+      "step": 14832
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.299162452591656e-06,
+      "loss": 4.4713,
+      "step": 14833
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2991150442477877e-06,
+      "loss": 2.5872,
+      "step": 14834
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2990676359039194e-06,
+      "loss": 3.3049,
+      "step": 14835
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2990202275600507e-06,
+      "loss": 3.3198,
+      "step": 14836
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298972819216182e-06,
+      "loss": 3.4461,
+      "step": 14837
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2989254108723137e-06,
+      "loss": 4.0182,
+      "step": 14838
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298878002528445e-06,
+      "loss": 3.7586,
+      "step": 14839
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2988305941845762e-06,
+      "loss": 2.2752,
+      "step": 14840
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298783185840708e-06,
+      "loss": 2.6627,
+      "step": 14841
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2987357774968397e-06,
+      "loss": 3.2422,
+      "step": 14842
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298688369152971e-06,
+      "loss": 3.0786,
+      "step": 14843
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2986409608091026e-06,
+      "loss": 2.9479,
+      "step": 14844
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298593552465234e-06,
+      "loss": 3.7183,
+      "step": 14845
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298546144121365e-06,
+      "loss": 3.138,
+      "step": 14846
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.298498735777497e-06,
+      "loss": 3.605,
+      "step": 14847
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2984513274336286e-06,
+      "loss": 3.064,
+      "step": 14848
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.29840391908976e-06,
+      "loss": 2.2499,
+      "step": 14849
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.2983565107458916e-06,
+      "loss": 2.7266,
+      "step": 14850
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.298309102402023e-06,
+      "loss": 4.2654,
+      "step": 14851
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.298261694058154e-06,
+      "loss": 3.0976,
+      "step": 14852
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.298214285714286e-06,
+      "loss": 3.6753,
+      "step": 14853
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.298166877370417e-06,
+      "loss": 3.0235,
+      "step": 14854
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.298119469026549e-06,
+      "loss": 3.0912,
+      "step": 14855
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.29807206068268e-06,
+      "loss": 2.9554,
+      "step": 14856
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.298024652338812e-06,
+      "loss": 2.9696,
+      "step": 14857
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297977243994943e-06,
+      "loss": 2.9541,
+      "step": 14858
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2979298356510744e-06,
+      "loss": 3.2764,
+      "step": 14859
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297882427307206e-06,
+      "loss": 3.5272,
+      "step": 14860
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297835018963338e-06,
+      "loss": 2.8463,
+      "step": 14861
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297787610619469e-06,
+      "loss": 3.243,
+      "step": 14862
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2977402022756008e-06,
+      "loss": 2.9999,
+      "step": 14863
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297692793931732e-06,
+      "loss": 2.6666,
+      "step": 14864
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2976453855878633e-06,
+      "loss": 3.2269,
+      "step": 14865
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297597977243995e-06,
+      "loss": 4.0773,
+      "step": 14866
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2975505689001263e-06,
+      "loss": 2.6757,
+      "step": 14867
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297503160556258e-06,
+      "loss": 2.6601,
+      "step": 14868
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2974557522123897e-06,
+      "loss": 3.8383,
+      "step": 14869
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297408343868521e-06,
+      "loss": 3.4817,
+      "step": 14870
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2973609355246523e-06,
+      "loss": 2.9121,
+      "step": 14871
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297313527180784e-06,
+      "loss": 2.7365,
+      "step": 14872
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2972661188369153e-06,
+      "loss": 3.5812,
+      "step": 14873
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.297218710493047e-06,
+      "loss": 3.295,
+      "step": 14874
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2971713021491787e-06,
+      "loss": 1.634,
+      "step": 14875
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.29712389380531e-06,
+      "loss": 2.5974,
+      "step": 14876
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2970764854614412e-06,
+      "loss": 3.178,
+      "step": 14877
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2970290771175725e-06,
+      "loss": 3.504,
+      "step": 14878
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2969816687737042e-06,
+      "loss": 3.7644,
+      "step": 14879
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2969342604298355e-06,
+      "loss": 3.7429,
+      "step": 14880
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296886852085967e-06,
+      "loss": 3.53,
+      "step": 14881
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296839443742099e-06,
+      "loss": 2.7039,
+      "step": 14882
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.29679203539823e-06,
+      "loss": 2.1375,
+      "step": 14883
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2967446270543615e-06,
+      "loss": 2.2753,
+      "step": 14884
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296697218710493e-06,
+      "loss": 2.6302,
+      "step": 14885
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2966498103666245e-06,
+      "loss": 2.7848,
+      "step": 14886
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296602402022756e-06,
+      "loss": 3.0396,
+      "step": 14887
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296554993678888e-06,
+      "loss": 3.4832,
+      "step": 14888
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296507585335019e-06,
+      "loss": 2.6798,
+      "step": 14889
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2964601769911504e-06,
+      "loss": 1.8483,
+      "step": 14890
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296412768647282e-06,
+      "loss": 2.8597,
+      "step": 14891
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2963653603034134e-06,
+      "loss": 3.7288,
+      "step": 14892
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2963179519595447e-06,
+      "loss": 2.7618,
+      "step": 14893
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2962705436156764e-06,
+      "loss": 3.1572,
+      "step": 14894
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296223135271808e-06,
+      "loss": 3.0529,
+      "step": 14895
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2961757269279394e-06,
+      "loss": 2.6481,
+      "step": 14896
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.296128318584071e-06,
+      "loss": 3.6326,
+      "step": 14897
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2960809102402024e-06,
+      "loss": 2.1884,
+      "step": 14898
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2960335018963336e-06,
+      "loss": 3.6707,
+      "step": 14899
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.295986093552465e-06,
+      "loss": 3.5196,
+      "step": 14900
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.295938685208597e-06,
+      "loss": 2.3094,
+      "step": 14901
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2958912768647283e-06,
+      "loss": 3.4509,
+      "step": 14902
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2958438685208596e-06,
+      "loss": 3.1785,
+      "step": 14903
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2957964601769913e-06,
+      "loss": 2.9086,
+      "step": 14904
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2957490518331226e-06,
+      "loss": 2.7635,
+      "step": 14905
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.295701643489254e-06,
+      "loss": 2.6333,
+      "step": 14906
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2956542351453856e-06,
+      "loss": 2.2083,
+      "step": 14907
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2956068268015173e-06,
+      "loss": 3.749,
+      "step": 14908
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2955594184576486e-06,
+      "loss": 3.1942,
+      "step": 14909
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2955120101137803e-06,
+      "loss": 3.7372,
+      "step": 14910
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2954646017699116e-06,
+      "loss": 2.9847,
+      "step": 14911
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.295417193426043e-06,
+      "loss": 4.1238,
+      "step": 14912
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.2953697850821745e-06,
+      "loss": 2.8328,
+      "step": 14913
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.295322376738306e-06,
+      "loss": 3.2069,
+      "step": 14914
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2952749683944375e-06,
+      "loss": 3.1233,
+      "step": 14915
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2952275600505692e-06,
+      "loss": 2.5868,
+      "step": 14916
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2951801517067005e-06,
+      "loss": 3.5264,
+      "step": 14917
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2951327433628318e-06,
+      "loss": 4.0667,
+      "step": 14918
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2950853350189635e-06,
+      "loss": 3.6829,
+      "step": 14919
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2950379266750948e-06,
+      "loss": 2.7398,
+      "step": 14920
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2949905183312265e-06,
+      "loss": 3.5585,
+      "step": 14921
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2949431099873578e-06,
+      "loss": 3.0312,
+      "step": 14922
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2948957016434895e-06,
+      "loss": 3.2701,
+      "step": 14923
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2948482932996207e-06,
+      "loss": 2.7345,
+      "step": 14924
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294800884955752e-06,
+      "loss": 2.5772,
+      "step": 14925
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2947534766118837e-06,
+      "loss": 3.1937,
+      "step": 14926
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294706068268015e-06,
+      "loss": 2.5372,
+      "step": 14927
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2946586599241467e-06,
+      "loss": 2.5536,
+      "step": 14928
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2946112515802784e-06,
+      "loss": 2.6229,
+      "step": 14929
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2945638432364097e-06,
+      "loss": 2.5465,
+      "step": 14930
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294516434892541e-06,
+      "loss": 3.8376,
+      "step": 14931
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2944690265486727e-06,
+      "loss": 2.1663,
+      "step": 14932
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294421618204804e-06,
+      "loss": 3.0824,
+      "step": 14933
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2943742098609357e-06,
+      "loss": 3.332,
+      "step": 14934
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2943268015170674e-06,
+      "loss": 2.5256,
+      "step": 14935
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2942793931731986e-06,
+      "loss": 2.6709,
+      "step": 14936
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.29423198482933e-06,
+      "loss": 2.9539,
+      "step": 14937
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2941845764854616e-06,
+      "loss": 2.9325,
+      "step": 14938
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294137168141593e-06,
+      "loss": 3.0513,
+      "step": 14939
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294089759797724e-06,
+      "loss": 2.5417,
+      "step": 14940
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.294042351453856e-06,
+      "loss": 1.9543,
+      "step": 14941
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2939949431099876e-06,
+      "loss": 3.122,
+      "step": 14942
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293947534766119e-06,
+      "loss": 2.9714,
+      "step": 14943
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.29390012642225e-06,
+      "loss": 3.2903,
+      "step": 14944
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293852718078382e-06,
+      "loss": 3.1161,
+      "step": 14945
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293805309734513e-06,
+      "loss": 2.6628,
+      "step": 14946
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293757901390645e-06,
+      "loss": 2.2699,
+      "step": 14947
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2937104930467766e-06,
+      "loss": 2.8213,
+      "step": 14948
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293663084702908e-06,
+      "loss": 3.1995,
+      "step": 14949
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293615676359039e-06,
+      "loss": 2.7792,
+      "step": 14950
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293568268015171e-06,
+      "loss": 3.7566,
+      "step": 14951
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293520859671302e-06,
+      "loss": 3.0427,
+      "step": 14952
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2934734513274334e-06,
+      "loss": 2.0906,
+      "step": 14953
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293426042983565e-06,
+      "loss": 3.0292,
+      "step": 14954
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2933786346396968e-06,
+      "loss": 2.8337,
+      "step": 14955
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293331226295828e-06,
+      "loss": 1.7524,
+      "step": 14956
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2932838179519598e-06,
+      "loss": 2.5166,
+      "step": 14957
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293236409608091e-06,
+      "loss": 3.3865,
+      "step": 14958
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2931890012642223e-06,
+      "loss": 3.363,
+      "step": 14959
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293141592920354e-06,
+      "loss": 3.2192,
+      "step": 14960
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2930941845764857e-06,
+      "loss": 3.115,
+      "step": 14961
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.293046776232617e-06,
+      "loss": 2.7692,
+      "step": 14962
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2929993678887487e-06,
+      "loss": 3.3319,
+      "step": 14963
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.29295195954488e-06,
+      "loss": 3.4977,
+      "step": 14964
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2929045512010113e-06,
+      "loss": 3.1367,
+      "step": 14965
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2928571428571426e-06,
+      "loss": 2.5533,
+      "step": 14966
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2928097345132743e-06,
+      "loss": 2.5429,
+      "step": 14967
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.292762326169406e-06,
+      "loss": 3.4357,
+      "step": 14968
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2927149178255373e-06,
+      "loss": 3.3713,
+      "step": 14969
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.292667509481669e-06,
+      "loss": 2.9007,
+      "step": 14970
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2926201011378002e-06,
+      "loss": 3.2809,
+      "step": 14971
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2925726927939315e-06,
+      "loss": 3.3849,
+      "step": 14972
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.2925252844500632e-06,
+      "loss": 2.5641,
+      "step": 14973
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.292477876106195e-06,
+      "loss": 3.3033,
+      "step": 14974
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.292430467762326e-06,
+      "loss": 3.1006,
+      "step": 14975
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.292383059418458e-06,
+      "loss": 3.087,
+      "step": 14976
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.292335651074589e-06,
+      "loss": 3.7397,
+      "step": 14977
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2922882427307205e-06,
+      "loss": 3.009,
+      "step": 14978
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.292240834386852e-06,
+      "loss": 2.503,
+      "step": 14979
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2921934260429835e-06,
+      "loss": 2.6196,
+      "step": 14980
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.292146017699115e-06,
+      "loss": 2.4303,
+      "step": 14981
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.292098609355247e-06,
+      "loss": 2.6034,
+      "step": 14982
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.292051201011378e-06,
+      "loss": 3.2094,
+      "step": 14983
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2920037926675094e-06,
+      "loss": 3.4224,
+      "step": 14984
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.291956384323641e-06,
+      "loss": 2.8915,
+      "step": 14985
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2919089759797724e-06,
+      "loss": 3.2071,
+      "step": 14986
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.291861567635904e-06,
+      "loss": 3.2336,
+      "step": 14987
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2918141592920354e-06,
+      "loss": 1.7905,
+      "step": 14988
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.291766750948167e-06,
+      "loss": 2.5608,
+      "step": 14989
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2917193426042984e-06,
+      "loss": 2.3296,
+      "step": 14990
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2916719342604297e-06,
+      "loss": 3.7927,
+      "step": 14991
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2916245259165614e-06,
+      "loss": 2.6249,
+      "step": 14992
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2915771175726926e-06,
+      "loss": 4.1332,
+      "step": 14993
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2915297092288243e-06,
+      "loss": 3.5756,
+      "step": 14994
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.291482300884956e-06,
+      "loss": 3.7797,
+      "step": 14995
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2914348925410873e-06,
+      "loss": 3.1003,
+      "step": 14996
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2913874841972186e-06,
+      "loss": 3.5016,
+      "step": 14997
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2913400758533503e-06,
+      "loss": 3.4436,
+      "step": 14998
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2912926675094816e-06,
+      "loss": 3.0829,
+      "step": 14999
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2912452591656133e-06,
+      "loss": 1.9356,
+      "step": 15000
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.291197850821745e-06,
+      "loss": 3.2946,
+      "step": 15001
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2911504424778763e-06,
+      "loss": 3.519,
+      "step": 15002
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2911030341340076e-06,
+      "loss": 2.8943,
+      "step": 15003
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2910556257901393e-06,
+      "loss": 3.7677,
+      "step": 15004
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2910082174462705e-06,
+      "loss": 3.2355,
+      "step": 15005
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.290960809102402e-06,
+      "loss": 3.5274,
+      "step": 15006
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2909134007585335e-06,
+      "loss": 3.6789,
+      "step": 15007
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2908659924146652e-06,
+      "loss": 3.274,
+      "step": 15008
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2908185840707965e-06,
+      "loss": 3.2214,
+      "step": 15009
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.290771175726928e-06,
+      "loss": 2.8337,
+      "step": 15010
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2907237673830595e-06,
+      "loss": 3.0702,
+      "step": 15011
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2906763590391908e-06,
+      "loss": 3.0841,
+      "step": 15012
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2906289506953225e-06,
+      "loss": 2.9428,
+      "step": 15013
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.290581542351454e-06,
+      "loss": 2.8285,
+      "step": 15014
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2905341340075855e-06,
+      "loss": 2.9814,
+      "step": 15015
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2904867256637167e-06,
+      "loss": 3.7281,
+      "step": 15016
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2904393173198485e-06,
+      "loss": 3.2371,
+      "step": 15017
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2903919089759797e-06,
+      "loss": 2.7428,
+      "step": 15018
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.290344500632111e-06,
+      "loss": 1.7458,
+      "step": 15019
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2902970922882427e-06,
+      "loss": 2.5795,
+      "step": 15020
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2902496839443744e-06,
+      "loss": 1.9498,
+      "step": 15021
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2902022756005057e-06,
+      "loss": 1.7161,
+      "step": 15022
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2901548672566374e-06,
+      "loss": 3.3548,
+      "step": 15023
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2901074589127687e-06,
+      "loss": 3.9684,
+      "step": 15024
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2900600505689e-06,
+      "loss": 3.31,
+      "step": 15025
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2900126422250317e-06,
+      "loss": 3.4205,
+      "step": 15026
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2899652338811634e-06,
+      "loss": 3.4555,
+      "step": 15027
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2899178255372947e-06,
+      "loss": 3.6835,
+      "step": 15028
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289870417193426e-06,
+      "loss": 4.2507,
+      "step": 15029
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2898230088495576e-06,
+      "loss": 3.3824,
+      "step": 15030
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289775600505689e-06,
+      "loss": 3.435,
+      "step": 15031
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.28972819216182e-06,
+      "loss": 2.8076,
+      "step": 15032
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289680783817952e-06,
+      "loss": 3.0697,
+      "step": 15033
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2896333754740836e-06,
+      "loss": 2.9639,
+      "step": 15034
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289585967130215e-06,
+      "loss": 2.8318,
+      "step": 15035
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2895385587863466e-06,
+      "loss": 3.8081,
+      "step": 15036
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289491150442478e-06,
+      "loss": 3.1048,
+      "step": 15037
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289443742098609e-06,
+      "loss": 3.1192,
+      "step": 15038
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.289396333754741e-06,
+      "loss": 3.0089,
+      "step": 15039
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.2893489254108726e-06,
+      "loss": 3.0617,
+      "step": 15040
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.289301517067004e-06,
+      "loss": 1.8096,
+      "step": 15041
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2892541087231355e-06,
+      "loss": 3.0154,
+      "step": 15042
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.289206700379267e-06,
+      "loss": 3.5163,
+      "step": 15043
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.289159292035398e-06,
+      "loss": 2.3858,
+      "step": 15044
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.28911188369153e-06,
+      "loss": 3.51,
+      "step": 15045
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.289064475347661e-06,
+      "loss": 3.0101,
+      "step": 15046
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.289017067003793e-06,
+      "loss": 3.4421,
+      "step": 15047
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2889696586599245e-06,
+      "loss": 2.9062,
+      "step": 15048
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2889222503160558e-06,
+      "loss": 3.4926,
+      "step": 15049
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288874841972187e-06,
+      "loss": 3.2775,
+      "step": 15050
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2888274336283183e-06,
+      "loss": 3.073,
+      "step": 15051
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.28878002528445e-06,
+      "loss": 2.4844,
+      "step": 15052
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2887326169405817e-06,
+      "loss": 3.8776,
+      "step": 15053
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288685208596713e-06,
+      "loss": 3.1075,
+      "step": 15054
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2886378002528447e-06,
+      "loss": 2.9466,
+      "step": 15055
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288590391908976e-06,
+      "loss": 4.3981,
+      "step": 15056
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2885429835651073e-06,
+      "loss": 3.296,
+      "step": 15057
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288495575221239e-06,
+      "loss": 3.6667,
+      "step": 15058
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2884481668773703e-06,
+      "loss": 2.8292,
+      "step": 15059
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288400758533502e-06,
+      "loss": 2.8138,
+      "step": 15060
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2883533501896337e-06,
+      "loss": 4.082,
+      "step": 15061
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288305941845765e-06,
+      "loss": 3.9826,
+      "step": 15062
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2882585335018962e-06,
+      "loss": 2.6107,
+      "step": 15063
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288211125158028e-06,
+      "loss": 2.6586,
+      "step": 15064
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2881637168141592e-06,
+      "loss": 2.6857,
+      "step": 15065
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288116308470291e-06,
+      "loss": 3.3069,
+      "step": 15066
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2880689001264226e-06,
+      "loss": 2.3855,
+      "step": 15067
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.288021491782554e-06,
+      "loss": 3.3407,
+      "step": 15068
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287974083438685e-06,
+      "loss": 2.3211,
+      "step": 15069
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287926675094817e-06,
+      "loss": 4.0716,
+      "step": 15070
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287879266750948e-06,
+      "loss": 2.6313,
+      "step": 15071
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2878318584070795e-06,
+      "loss": 3.7436,
+      "step": 15072
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287784450063211e-06,
+      "loss": 3.2996,
+      "step": 15073
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287737041719343e-06,
+      "loss": 4.0118,
+      "step": 15074
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287689633375474e-06,
+      "loss": 2.8015,
+      "step": 15075
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2876422250316054e-06,
+      "loss": 2.3228,
+      "step": 15076
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287594816687737e-06,
+      "loss": 3.411,
+      "step": 15077
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2875474083438684e-06,
+      "loss": 2.8385,
+      "step": 15078
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2875e-06,
+      "loss": 3.5149,
+      "step": 15079
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287452591656132e-06,
+      "loss": 3.2511,
+      "step": 15080
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287405183312263e-06,
+      "loss": 2.9681,
+      "step": 15081
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2873577749683944e-06,
+      "loss": 2.2532,
+      "step": 15082
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287310366624526e-06,
+      "loss": 3.1376,
+      "step": 15083
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2872629582806574e-06,
+      "loss": 2.192,
+      "step": 15084
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2872155499367887e-06,
+      "loss": 2.8474,
+      "step": 15085
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2871681415929204e-06,
+      "loss": 2.4499,
+      "step": 15086
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287120733249052e-06,
+      "loss": 4.0694,
+      "step": 15087
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2870733249051833e-06,
+      "loss": 3.3031,
+      "step": 15088
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.287025916561315e-06,
+      "loss": 3.6385,
+      "step": 15089
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2869785082174463e-06,
+      "loss": 2.3324,
+      "step": 15090
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2869310998735776e-06,
+      "loss": 2.4179,
+      "step": 15091
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2868836915297093e-06,
+      "loss": 2.4073,
+      "step": 15092
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.286836283185841e-06,
+      "loss": 2.3772,
+      "step": 15093
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2867888748419723e-06,
+      "loss": 2.8972,
+      "step": 15094
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2867414664981036e-06,
+      "loss": 3.0685,
+      "step": 15095
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2866940581542353e-06,
+      "loss": 2.7345,
+      "step": 15096
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2866466498103666e-06,
+      "loss": 2.923,
+      "step": 15097
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.286599241466498e-06,
+      "loss": 2.4548,
+      "step": 15098
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2865518331226295e-06,
+      "loss": 2.8363,
+      "step": 15099
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2865044247787612e-06,
+      "loss": 2.9583,
+      "step": 15100
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2864570164348925e-06,
+      "loss": 3.4785,
+      "step": 15101
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2864096080910242e-06,
+      "loss": 3.7083,
+      "step": 15102
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.2863621997471555e-06,
+      "loss": 3.7836,
+      "step": 15103
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.286314791403287e-06,
+      "loss": 4.0257,
+      "step": 15104
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2862673830594185e-06,
+      "loss": 3.2361,
+      "step": 15105
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.28621997471555e-06,
+      "loss": 3.7572,
+      "step": 15106
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2861725663716815e-06,
+      "loss": 3.6887,
+      "step": 15107
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.286125158027813e-06,
+      "loss": 3.8801,
+      "step": 15108
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2860777496839445e-06,
+      "loss": 4.1577,
+      "step": 15109
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2860303413400757e-06,
+      "loss": 2.7338,
+      "step": 15110
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2859829329962074e-06,
+      "loss": 3.2494,
+      "step": 15111
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2859355246523387e-06,
+      "loss": 3.3572,
+      "step": 15112
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2858881163084704e-06,
+      "loss": 4.1551,
+      "step": 15113
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.285840707964602e-06,
+      "loss": 2.7841,
+      "step": 15114
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2857932996207334e-06,
+      "loss": 2.4256,
+      "step": 15115
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2857458912768647e-06,
+      "loss": 2.8,
+      "step": 15116
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.285698482932996e-06,
+      "loss": 3.0174,
+      "step": 15117
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2856510745891277e-06,
+      "loss": 3.0464,
+      "step": 15118
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2856036662452594e-06,
+      "loss": 2.8786,
+      "step": 15119
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2855562579013907e-06,
+      "loss": 2.4165,
+      "step": 15120
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2855088495575224e-06,
+      "loss": 3.7562,
+      "step": 15121
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2854614412136536e-06,
+      "loss": 2.8965,
+      "step": 15122
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.285414032869785e-06,
+      "loss": 2.9284,
+      "step": 15123
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2853666245259166e-06,
+      "loss": 2.9341,
+      "step": 15124
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.285319216182048e-06,
+      "loss": 2.7448,
+      "step": 15125
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2852718078381796e-06,
+      "loss": 3.3973,
+      "step": 15126
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2852243994943113e-06,
+      "loss": 3.6619,
+      "step": 15127
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2851769911504426e-06,
+      "loss": 4.0567,
+      "step": 15128
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.285129582806574e-06,
+      "loss": 2.0301,
+      "step": 15129
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2850821744627056e-06,
+      "loss": 3.3636,
+      "step": 15130
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.285034766118837e-06,
+      "loss": 3.5908,
+      "step": 15131
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2849873577749686e-06,
+      "loss": 3.808,
+      "step": 15132
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2849399494311003e-06,
+      "loss": 3.2182,
+      "step": 15133
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2848925410872316e-06,
+      "loss": 3.5973,
+      "step": 15134
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284845132743363e-06,
+      "loss": 3.0808,
+      "step": 15135
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2847977243994945e-06,
+      "loss": 3.0531,
+      "step": 15136
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284750316055626e-06,
+      "loss": 3.1496,
+      "step": 15137
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284702907711757e-06,
+      "loss": 2.904,
+      "step": 15138
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284655499367889e-06,
+      "loss": 3.6596,
+      "step": 15139
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2846080910240205e-06,
+      "loss": 2.1036,
+      "step": 15140
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284560682680152e-06,
+      "loss": 3.3858,
+      "step": 15141
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284513274336283e-06,
+      "loss": 3.6046,
+      "step": 15142
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2844658659924148e-06,
+      "loss": 3.4174,
+      "step": 15143
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284418457648546e-06,
+      "loss": 2.7247,
+      "step": 15144
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2843710493046778e-06,
+      "loss": 2.8895,
+      "step": 15145
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2843236409608095e-06,
+      "loss": 3.4641,
+      "step": 15146
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2842762326169407e-06,
+      "loss": 3.8516,
+      "step": 15147
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284228824273072e-06,
+      "loss": 4.0787,
+      "step": 15148
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2841814159292037e-06,
+      "loss": 2.31,
+      "step": 15149
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284134007585335e-06,
+      "loss": 3.1179,
+      "step": 15150
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2840865992414663e-06,
+      "loss": 3.2057,
+      "step": 15151
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.284039190897598e-06,
+      "loss": 3.4955,
+      "step": 15152
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2839917825537297e-06,
+      "loss": 3.5738,
+      "step": 15153
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283944374209861e-06,
+      "loss": 3.0359,
+      "step": 15154
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2838969658659927e-06,
+      "loss": 2.4797,
+      "step": 15155
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283849557522124e-06,
+      "loss": 2.411,
+      "step": 15156
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2838021491782552e-06,
+      "loss": 3.48,
+      "step": 15157
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283754740834387e-06,
+      "loss": 3.4407,
+      "step": 15158
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2837073324905186e-06,
+      "loss": 3.868,
+      "step": 15159
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.28365992414665e-06,
+      "loss": 3.1962,
+      "step": 15160
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283612515802781e-06,
+      "loss": 2.9758,
+      "step": 15161
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283565107458913e-06,
+      "loss": 3.4847,
+      "step": 15162
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283517699115044e-06,
+      "loss": 2.6718,
+      "step": 15163
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.2834702907711755e-06,
+      "loss": 2.3725,
+      "step": 15164
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283422882427307e-06,
+      "loss": 1.4835,
+      "step": 15165
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.283375474083439e-06,
+      "loss": 3.1038,
+      "step": 15166
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.28332806573957e-06,
+      "loss": 1.8962,
+      "step": 15167
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.283280657395702e-06,
+      "loss": 3.0391,
+      "step": 15168
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.283233249051833e-06,
+      "loss": 3.3643,
+      "step": 15169
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2831858407079644e-06,
+      "loss": 2.5785,
+      "step": 15170
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.283138432364096e-06,
+      "loss": 3.1017,
+      "step": 15171
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.283091024020228e-06,
+      "loss": 3.2423,
+      "step": 15172
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.283043615676359e-06,
+      "loss": 3.217,
+      "step": 15173
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.282996207332491e-06,
+      "loss": 3.0826,
+      "step": 15174
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.282948798988622e-06,
+      "loss": 3.7395,
+      "step": 15175
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2829013906447534e-06,
+      "loss": 2.3838,
+      "step": 15176
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.282853982300885e-06,
+      "loss": 2.1777,
+      "step": 15177
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2828065739570164e-06,
+      "loss": 3.0469,
+      "step": 15178
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.282759165613148e-06,
+      "loss": 3.2796,
+      "step": 15179
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2827117572692798e-06,
+      "loss": 3.361,
+      "step": 15180
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.282664348925411e-06,
+      "loss": 1.8511,
+      "step": 15181
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2826169405815423e-06,
+      "loss": 2.3365,
+      "step": 15182
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2825695322376736e-06,
+      "loss": 3.4061,
+      "step": 15183
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2825221238938053e-06,
+      "loss": 2.8952,
+      "step": 15184
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2824747155499366e-06,
+      "loss": 2.8229,
+      "step": 15185
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2824273072060683e-06,
+      "loss": 2.6522,
+      "step": 15186
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2823798988622e-06,
+      "loss": 4.255,
+      "step": 15187
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2823324905183313e-06,
+      "loss": 3.4879,
+      "step": 15188
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2822850821744626e-06,
+      "loss": 2.3214,
+      "step": 15189
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2822376738305943e-06,
+      "loss": 3.44,
+      "step": 15190
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2821902654867256e-06,
+      "loss": 3.1606,
+      "step": 15191
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2821428571428573e-06,
+      "loss": 2.888,
+      "step": 15192
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.282095448798989e-06,
+      "loss": 2.3969,
+      "step": 15193
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2820480404551202e-06,
+      "loss": 3.2334,
+      "step": 15194
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2820006321112515e-06,
+      "loss": 2.9834,
+      "step": 15195
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2819532237673832e-06,
+      "loss": 3.1317,
+      "step": 15196
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2819058154235145e-06,
+      "loss": 3.4456,
+      "step": 15197
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2818584070796458e-06,
+      "loss": 3.763,
+      "step": 15198
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2818109987357775e-06,
+      "loss": 2.6785,
+      "step": 15199
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.281763590391909e-06,
+      "loss": 3.9904,
+      "step": 15200
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2817161820480405e-06,
+      "loss": 3.8481,
+      "step": 15201
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.281668773704172e-06,
+      "loss": 3.4291,
+      "step": 15202
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2816213653603035e-06,
+      "loss": 2.9891,
+      "step": 15203
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2815739570164347e-06,
+      "loss": 3.4772,
+      "step": 15204
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2815265486725664e-06,
+      "loss": 3.6593,
+      "step": 15205
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.281479140328698e-06,
+      "loss": 3.8499,
+      "step": 15206
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2814317319848294e-06,
+      "loss": 2.4947,
+      "step": 15207
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2813843236409607e-06,
+      "loss": 3.8747,
+      "step": 15208
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2813369152970924e-06,
+      "loss": 3.4358,
+      "step": 15209
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2812895069532237e-06,
+      "loss": 3.4799,
+      "step": 15210
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.281242098609355e-06,
+      "loss": 3.6642,
+      "step": 15211
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2811946902654867e-06,
+      "loss": 2.5822,
+      "step": 15212
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2811472819216184e-06,
+      "loss": 2.949,
+      "step": 15213
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2810998735777497e-06,
+      "loss": 2.8017,
+      "step": 15214
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2810524652338814e-06,
+      "loss": 3.8006,
+      "step": 15215
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2810050568900126e-06,
+      "loss": 2.4016,
+      "step": 15216
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.280957648546144e-06,
+      "loss": 2.4934,
+      "step": 15217
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2809102402022756e-06,
+      "loss": 3.401,
+      "step": 15218
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2808628318584073e-06,
+      "loss": 2.9603,
+      "step": 15219
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2808154235145386e-06,
+      "loss": 3.2033,
+      "step": 15220
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2807680151706703e-06,
+      "loss": 2.7446,
+      "step": 15221
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2807206068268016e-06,
+      "loss": 2.5956,
+      "step": 15222
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.280673198482933e-06,
+      "loss": 4.2421,
+      "step": 15223
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2806257901390646e-06,
+      "loss": 3.0863,
+      "step": 15224
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.280578381795196e-06,
+      "loss": 2.4499,
+      "step": 15225
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2805309734513276e-06,
+      "loss": 3.771,
+      "step": 15226
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.280483565107459e-06,
+      "loss": 2.953,
+      "step": 15227
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.2804361567635905e-06,
+      "loss": 2.2105,
+      "step": 15228
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.280388748419722e-06,
+      "loss": 2.6765,
+      "step": 15229
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2.280341340075853e-06,
+      "loss": 3.6966,
+      "step": 15230
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.280293931731985e-06,
+      "loss": 3.7907,
+      "step": 15231
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2802465233881165e-06,
+      "loss": 3.3467,
+      "step": 15232
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.280199115044248e-06,
+      "loss": 3.2642,
+      "step": 15233
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2801517067003795e-06,
+      "loss": 3.2735,
+      "step": 15234
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2801042983565108e-06,
+      "loss": 2.7897,
+      "step": 15235
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.280056890012642e-06,
+      "loss": 3.6646,
+      "step": 15236
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2800094816687738e-06,
+      "loss": 3.6631,
+      "step": 15237
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279962073324905e-06,
+      "loss": 3.3757,
+      "step": 15238
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2799146649810368e-06,
+      "loss": 3.5374,
+      "step": 15239
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2798672566371685e-06,
+      "loss": 2.7796,
+      "step": 15240
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2798198482932997e-06,
+      "loss": 2.9583,
+      "step": 15241
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279772439949431e-06,
+      "loss": 3.014,
+      "step": 15242
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2797250316055627e-06,
+      "loss": 2.9121,
+      "step": 15243
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279677623261694e-06,
+      "loss": 2.6285,
+      "step": 15244
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2796302149178257e-06,
+      "loss": 3.5873,
+      "step": 15245
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2795828065739574e-06,
+      "loss": 4.0265,
+      "step": 15246
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2795353982300887e-06,
+      "loss": 3.3892,
+      "step": 15247
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.27948798988622e-06,
+      "loss": 3.8179,
+      "step": 15248
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2794405815423513e-06,
+      "loss": 3.1173,
+      "step": 15249
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279393173198483e-06,
+      "loss": 3.5281,
+      "step": 15250
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2793457648546142e-06,
+      "loss": 2.5604,
+      "step": 15251
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279298356510746e-06,
+      "loss": 2.7055,
+      "step": 15252
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2792509481668776e-06,
+      "loss": 2.5118,
+      "step": 15253
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279203539823009e-06,
+      "loss": 3.6995,
+      "step": 15254
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.27915613147914e-06,
+      "loss": 3.2503,
+      "step": 15255
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279108723135272e-06,
+      "loss": 3.4666,
+      "step": 15256
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279061314791403e-06,
+      "loss": 2.9606,
+      "step": 15257
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.279013906447535e-06,
+      "loss": 2.9516,
+      "step": 15258
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2789664981036666e-06,
+      "loss": 1.8152,
+      "step": 15259
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278919089759798e-06,
+      "loss": 3.1953,
+      "step": 15260
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278871681415929e-06,
+      "loss": 3.2498,
+      "step": 15261
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278824273072061e-06,
+      "loss": 4.0897,
+      "step": 15262
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278776864728192e-06,
+      "loss": 2.4297,
+      "step": 15263
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2787294563843234e-06,
+      "loss": 2.2429,
+      "step": 15264
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278682048040455e-06,
+      "loss": 3.1256,
+      "step": 15265
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278634639696587e-06,
+      "loss": 2.6466,
+      "step": 15266
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278587231352718e-06,
+      "loss": 3.0947,
+      "step": 15267
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.27853982300885e-06,
+      "loss": 3.2833,
+      "step": 15268
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278492414664981e-06,
+      "loss": 2.6944,
+      "step": 15269
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2784450063211124e-06,
+      "loss": 2.081,
+      "step": 15270
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278397597977244e-06,
+      "loss": 2.7333,
+      "step": 15271
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2783501896333758e-06,
+      "loss": 2.1874,
+      "step": 15272
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278302781289507e-06,
+      "loss": 2.5881,
+      "step": 15273
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2782553729456383e-06,
+      "loss": 2.7222,
+      "step": 15274
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.27820796460177e-06,
+      "loss": 2.8253,
+      "step": 15275
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2781605562579013e-06,
+      "loss": 2.1704,
+      "step": 15276
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2781131479140326e-06,
+      "loss": 3.345,
+      "step": 15277
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2780657395701643e-06,
+      "loss": 3.1911,
+      "step": 15278
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.278018331226296e-06,
+      "loss": 3.1795,
+      "step": 15279
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2779709228824273e-06,
+      "loss": 3.9438,
+      "step": 15280
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.277923514538559e-06,
+      "loss": 2.3932,
+      "step": 15281
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2778761061946903e-06,
+      "loss": 2.8879,
+      "step": 15282
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2778286978508216e-06,
+      "loss": 1.9686,
+      "step": 15283
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2777812895069533e-06,
+      "loss": 3.5491,
+      "step": 15284
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.277733881163085e-06,
+      "loss": 3.2578,
+      "step": 15285
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2776864728192162e-06,
+      "loss": 2.7713,
+      "step": 15286
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.277639064475348e-06,
+      "loss": 2.9307,
+      "step": 15287
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2775916561314792e-06,
+      "loss": 3.3113,
+      "step": 15288
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2775442477876105e-06,
+      "loss": 3.2993,
+      "step": 15289
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2774968394437422e-06,
+      "loss": 2.5853,
+      "step": 15290
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2774494310998735e-06,
+      "loss": 2.16,
+      "step": 15291
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.277402022756005e-06,
+      "loss": 3.8772,
+      "step": 15292
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.2773546144121365e-06,
+      "loss": 2.6357,
+      "step": 15293
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.277307206068268e-06,
+      "loss": 2.8665,
+      "step": 15294
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2772597977243995e-06,
+      "loss": 3.7724,
+      "step": 15295
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2772123893805307e-06,
+      "loss": 2.4379,
+      "step": 15296
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2771649810366625e-06,
+      "loss": 3.2496,
+      "step": 15297
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.277117572692794e-06,
+      "loss": 3.4166,
+      "step": 15298
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2770701643489254e-06,
+      "loss": 4.0529,
+      "step": 15299
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.277022756005057e-06,
+      "loss": 2.8194,
+      "step": 15300
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2769753476611884e-06,
+      "loss": 2.6718,
+      "step": 15301
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2769279393173197e-06,
+      "loss": 3.1279,
+      "step": 15302
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2768805309734514e-06,
+      "loss": 2.9246,
+      "step": 15303
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2768331226295827e-06,
+      "loss": 3.4004,
+      "step": 15304
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2767857142857144e-06,
+      "loss": 2.5253,
+      "step": 15305
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.276738305941846e-06,
+      "loss": 4.1252,
+      "step": 15306
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2766908975979774e-06,
+      "loss": 2.8013,
+      "step": 15307
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2766434892541087e-06,
+      "loss": 3.0258,
+      "step": 15308
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2765960809102404e-06,
+      "loss": 2.4671,
+      "step": 15309
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2765486725663716e-06,
+      "loss": 3.7307,
+      "step": 15310
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2765012642225033e-06,
+      "loss": 2.5898,
+      "step": 15311
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.276453855878635e-06,
+      "loss": 3.3511,
+      "step": 15312
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2764064475347663e-06,
+      "loss": 1.9669,
+      "step": 15313
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2763590391908976e-06,
+      "loss": 2.067,
+      "step": 15314
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.276311630847029e-06,
+      "loss": 4.1964,
+      "step": 15315
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2762642225031606e-06,
+      "loss": 3.3287,
+      "step": 15316
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.276216814159292e-06,
+      "loss": 2.235,
+      "step": 15317
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2761694058154236e-06,
+      "loss": 3.7236,
+      "step": 15318
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2761219974715553e-06,
+      "loss": 3.9151,
+      "step": 15319
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2760745891276866e-06,
+      "loss": 3.1249,
+      "step": 15320
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.276027180783818e-06,
+      "loss": 2.3558,
+      "step": 15321
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2759797724399495e-06,
+      "loss": 2.8539,
+      "step": 15322
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.275932364096081e-06,
+      "loss": 2.7584,
+      "step": 15323
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2758849557522125e-06,
+      "loss": 2.4232,
+      "step": 15324
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2758375474083442e-06,
+      "loss": 2.6619,
+      "step": 15325
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2757901390644755e-06,
+      "loss": 2.8432,
+      "step": 15326
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.275742730720607e-06,
+      "loss": 3.3802,
+      "step": 15327
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2756953223767385e-06,
+      "loss": 2.9073,
+      "step": 15328
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2756479140328698e-06,
+      "loss": 2.6603,
+      "step": 15329
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.275600505689001e-06,
+      "loss": 2.8675,
+      "step": 15330
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2755530973451328e-06,
+      "loss": 2.6683,
+      "step": 15331
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2755056890012645e-06,
+      "loss": 2.4768,
+      "step": 15332
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2754582806573957e-06,
+      "loss": 3.9955,
+      "step": 15333
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2754108723135275e-06,
+      "loss": 2.5246,
+      "step": 15334
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2753634639696587e-06,
+      "loss": 2.9563,
+      "step": 15335
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.27531605562579e-06,
+      "loss": 2.0908,
+      "step": 15336
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2752686472819217e-06,
+      "loss": 3.0874,
+      "step": 15337
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2752212389380534e-06,
+      "loss": 2.9805,
+      "step": 15338
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2751738305941847e-06,
+      "loss": 3.6524,
+      "step": 15339
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.275126422250316e-06,
+      "loss": 2.2904,
+      "step": 15340
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2750790139064477e-06,
+      "loss": 2.4841,
+      "step": 15341
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.275031605562579e-06,
+      "loss": 3.2693,
+      "step": 15342
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2749841972187102e-06,
+      "loss": 3.0065,
+      "step": 15343
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274936788874842e-06,
+      "loss": 3.0671,
+      "step": 15344
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2748893805309737e-06,
+      "loss": 3.6876,
+      "step": 15345
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274841972187105e-06,
+      "loss": 2.7679,
+      "step": 15346
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2747945638432366e-06,
+      "loss": 3.093,
+      "step": 15347
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274747155499368e-06,
+      "loss": 2.7673,
+      "step": 15348
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274699747155499e-06,
+      "loss": 2.9517,
+      "step": 15349
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274652338811631e-06,
+      "loss": 2.4841,
+      "step": 15350
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2746049304677626e-06,
+      "loss": 3.9055,
+      "step": 15351
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274557522123894e-06,
+      "loss": 3.5047,
+      "step": 15352
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.2745101137800256e-06,
+      "loss": 2.6375,
+      "step": 15353
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274462705436157e-06,
+      "loss": 2.6387,
+      "step": 15354
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274415297092288e-06,
+      "loss": 3.0305,
+      "step": 15355
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.27436788874842e-06,
+      "loss": 3.5112,
+      "step": 15356
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2.274320480404551e-06,
+      "loss": 2.9481,
+      "step": 15357
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.274273072060683e-06,
+      "loss": 3.4365,
+      "step": 15358
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.274225663716814e-06,
+      "loss": 3.8645,
+      "step": 15359
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.274178255372946e-06,
+      "loss": 2.4989,
+      "step": 15360
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.274130847029077e-06,
+      "loss": 3.1991,
+      "step": 15361
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2740834386852084e-06,
+      "loss": 3.6247,
+      "step": 15362
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.27403603034134e-06,
+      "loss": 2.4202,
+      "step": 15363
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273988621997472e-06,
+      "loss": 4.0469,
+      "step": 15364
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273941213653603e-06,
+      "loss": 2.7646,
+      "step": 15365
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2738938053097348e-06,
+      "loss": 3.71,
+      "step": 15366
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273846396965866e-06,
+      "loss": 3.6399,
+      "step": 15367
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2737989886219973e-06,
+      "loss": 2.7445,
+      "step": 15368
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273751580278129e-06,
+      "loss": 3.0329,
+      "step": 15369
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2737041719342603e-06,
+      "loss": 3.0851,
+      "step": 15370
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273656763590392e-06,
+      "loss": 3.4379,
+      "step": 15371
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2736093552465237e-06,
+      "loss": 3.4525,
+      "step": 15372
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273561946902655e-06,
+      "loss": 3.8633,
+      "step": 15373
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2735145385587863e-06,
+      "loss": 2.2671,
+      "step": 15374
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273467130214918e-06,
+      "loss": 3.3245,
+      "step": 15375
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2734197218710493e-06,
+      "loss": 3.5805,
+      "step": 15376
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273372313527181e-06,
+      "loss": 3.6879,
+      "step": 15377
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2733249051833127e-06,
+      "loss": 2.9214,
+      "step": 15378
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.273277496839444e-06,
+      "loss": 3.4853,
+      "step": 15379
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2732300884955752e-06,
+      "loss": 3.642,
+      "step": 15380
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2731826801517065e-06,
+      "loss": 4.2148,
+      "step": 15381
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2731352718078382e-06,
+      "loss": 2.0284,
+      "step": 15382
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2730878634639695e-06,
+      "loss": 2.9489,
+      "step": 15383
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2730404551201012e-06,
+      "loss": 2.3164,
+      "step": 15384
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272993046776233e-06,
+      "loss": 1.8381,
+      "step": 15385
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272945638432364e-06,
+      "loss": 3.132,
+      "step": 15386
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2728982300884955e-06,
+      "loss": 3.7983,
+      "step": 15387
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272850821744627e-06,
+      "loss": 2.9482,
+      "step": 15388
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2728034134007585e-06,
+      "loss": 3.6945,
+      "step": 15389
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.27275600505689e-06,
+      "loss": 3.767,
+      "step": 15390
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272708596713022e-06,
+      "loss": 3.0548,
+      "step": 15391
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272661188369153e-06,
+      "loss": 2.8131,
+      "step": 15392
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2726137800252844e-06,
+      "loss": 3.5493,
+      "step": 15393
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272566371681416e-06,
+      "loss": 3.4338,
+      "step": 15394
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2725189633375474e-06,
+      "loss": 2.9848,
+      "step": 15395
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2724715549936787e-06,
+      "loss": 3.0252,
+      "step": 15396
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2724241466498104e-06,
+      "loss": 3.6545,
+      "step": 15397
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272376738305942e-06,
+      "loss": 3.3396,
+      "step": 15398
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2723293299620734e-06,
+      "loss": 3.1314,
+      "step": 15399
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272281921618205e-06,
+      "loss": 2.7314,
+      "step": 15400
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2722345132743364e-06,
+      "loss": 2.1419,
+      "step": 15401
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2721871049304676e-06,
+      "loss": 2.7143,
+      "step": 15402
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2721396965865994e-06,
+      "loss": 3.2751,
+      "step": 15403
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.272092288242731e-06,
+      "loss": 2.3759,
+      "step": 15404
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2720448798988623e-06,
+      "loss": 3.3815,
+      "step": 15405
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2719974715549936e-06,
+      "loss": 3.8358,
+      "step": 15406
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2719500632111253e-06,
+      "loss": 2.3887,
+      "step": 15407
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2719026548672566e-06,
+      "loss": 2.8024,
+      "step": 15408
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.271855246523388e-06,
+      "loss": 3.0428,
+      "step": 15409
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2718078381795196e-06,
+      "loss": 2.0706,
+      "step": 15410
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2717604298356513e-06,
+      "loss": 2.8377,
+      "step": 15411
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2717130214917826e-06,
+      "loss": 2.4626,
+      "step": 15412
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2716656131479143e-06,
+      "loss": 3.1159,
+      "step": 15413
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2716182048040456e-06,
+      "loss": 2.6566,
+      "step": 15414
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.271570796460177e-06,
+      "loss": 2.9434,
+      "step": 15415
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2715233881163085e-06,
+      "loss": 3.3591,
+      "step": 15416
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2714759797724402e-06,
+      "loss": 2.5836,
+      "step": 15417
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2714285714285715e-06,
+      "loss": 2.292,
+      "step": 15418
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2713811630847032e-06,
+      "loss": 2.8444,
+      "step": 15419
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.2713337547408345e-06,
+      "loss": 1.9028,
+      "step": 15420
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2712863463969658e-06,
+      "loss": 2.3932,
+      "step": 15421
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2712389380530975e-06,
+      "loss": 3.2439,
+      "step": 15422
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2711915297092288e-06,
+      "loss": 3.0834,
+      "step": 15423
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2711441213653605e-06,
+      "loss": 3.265,
+      "step": 15424
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2710967130214918e-06,
+      "loss": 3.136,
+      "step": 15425
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2710493046776235e-06,
+      "loss": 3.6135,
+      "step": 15426
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2710018963337547e-06,
+      "loss": 3.0598,
+      "step": 15427
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270954487989886e-06,
+      "loss": 3.6645,
+      "step": 15428
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2709070796460177e-06,
+      "loss": 2.6271,
+      "step": 15429
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2708596713021494e-06,
+      "loss": 2.6678,
+      "step": 15430
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2708122629582807e-06,
+      "loss": 3.5459,
+      "step": 15431
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2707648546144124e-06,
+      "loss": 2.7419,
+      "step": 15432
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2707174462705437e-06,
+      "loss": 2.6447,
+      "step": 15433
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270670037926675e-06,
+      "loss": 2.7557,
+      "step": 15434
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2706226295828067e-06,
+      "loss": 2.7246,
+      "step": 15435
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270575221238938e-06,
+      "loss": 2.9647,
+      "step": 15436
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2705278128950697e-06,
+      "loss": 3.53,
+      "step": 15437
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2704804045512014e-06,
+      "loss": 2.7244,
+      "step": 15438
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2704329962073326e-06,
+      "loss": 2.8718,
+      "step": 15439
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270385587863464e-06,
+      "loss": 3.5079,
+      "step": 15440
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2703381795195956e-06,
+      "loss": 2.0822,
+      "step": 15441
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270290771175727e-06,
+      "loss": 3.5673,
+      "step": 15442
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2702433628318586e-06,
+      "loss": 3.0539,
+      "step": 15443
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.27019595448799e-06,
+      "loss": 3.5895,
+      "step": 15444
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2701485461441216e-06,
+      "loss": 3.8638,
+      "step": 15445
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270101137800253e-06,
+      "loss": 3.111,
+      "step": 15446
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270053729456384e-06,
+      "loss": 3.4274,
+      "step": 15447
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.270006321112516e-06,
+      "loss": 1.8023,
+      "step": 15448
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269958912768647e-06,
+      "loss": 2.7903,
+      "step": 15449
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269911504424779e-06,
+      "loss": 3.3661,
+      "step": 15450
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2698640960809106e-06,
+      "loss": 3.3463,
+      "step": 15451
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269816687737042e-06,
+      "loss": 2.628,
+      "step": 15452
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269769279393173e-06,
+      "loss": 3.2203,
+      "step": 15453
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269721871049305e-06,
+      "loss": 3.5779,
+      "step": 15454
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269674462705436e-06,
+      "loss": 3.1745,
+      "step": 15455
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2696270543615674e-06,
+      "loss": 2.3272,
+      "step": 15456
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2695796460176995e-06,
+      "loss": 3.0221,
+      "step": 15457
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2695322376738308e-06,
+      "loss": 2.4386,
+      "step": 15458
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269484829329962e-06,
+      "loss": 3.609,
+      "step": 15459
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2694374209860938e-06,
+      "loss": 3.0797,
+      "step": 15460
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269390012642225e-06,
+      "loss": 2.5042,
+      "step": 15461
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2693426042983563e-06,
+      "loss": 2.7061,
+      "step": 15462
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269295195954488e-06,
+      "loss": 3.1419,
+      "step": 15463
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2692477876106197e-06,
+      "loss": 3.0952,
+      "step": 15464
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269200379266751e-06,
+      "loss": 3.6036,
+      "step": 15465
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2691529709228823e-06,
+      "loss": 3.3909,
+      "step": 15466
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.269105562579014e-06,
+      "loss": 3.6186,
+      "step": 15467
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2690581542351453e-06,
+      "loss": 2.7891,
+      "step": 15468
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2690107458912766e-06,
+      "loss": 4.0175,
+      "step": 15469
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2689633375474083e-06,
+      "loss": 3.2864,
+      "step": 15470
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.26891592920354e-06,
+      "loss": 3.2469,
+      "step": 15471
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2688685208596713e-06,
+      "loss": 3.7893,
+      "step": 15472
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.268821112515803e-06,
+      "loss": 2.5625,
+      "step": 15473
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2687737041719342e-06,
+      "loss": 3.741,
+      "step": 15474
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2687262958280655e-06,
+      "loss": 1.9835,
+      "step": 15475
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2686788874841972e-06,
+      "loss": 3.176,
+      "step": 15476
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.268631479140329e-06,
+      "loss": 3.1104,
+      "step": 15477
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.26858407079646e-06,
+      "loss": 3.801,
+      "step": 15478
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.268536662452592e-06,
+      "loss": 1.896,
+      "step": 15479
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.268489254108723e-06,
+      "loss": 2.6252,
+      "step": 15480
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2684418457648545e-06,
+      "loss": 3.1544,
+      "step": 15481
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.268394437420986e-06,
+      "loss": 3.5125,
+      "step": 15482
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.2683470290771175e-06,
+      "loss": 2.2552,
+      "step": 15483
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2.268299620733249e-06,
+      "loss": 2.8689,
+      "step": 15484
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.268252212389381e-06,
+      "loss": 3.5109,
+      "step": 15485
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.268204804045512e-06,
+      "loss": 2.7621,
+      "step": 15486
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2681573957016434e-06,
+      "loss": 3.6344,
+      "step": 15487
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2681099873577747e-06,
+      "loss": 2.8345,
+      "step": 15488
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2680625790139064e-06,
+      "loss": 3.2266,
+      "step": 15489
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.268015170670038e-06,
+      "loss": 4.1848,
+      "step": 15490
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2679677623261694e-06,
+      "loss": 3.3223,
+      "step": 15491
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.267920353982301e-06,
+      "loss": 2.4758,
+      "step": 15492
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2678729456384324e-06,
+      "loss": 2.8542,
+      "step": 15493
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2678255372945637e-06,
+      "loss": 2.578,
+      "step": 15494
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2677781289506954e-06,
+      "loss": 2.5252,
+      "step": 15495
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2677307206068266e-06,
+      "loss": 3.1788,
+      "step": 15496
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2676833122629583e-06,
+      "loss": 3.5512,
+      "step": 15497
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.26763590391909e-06,
+      "loss": 3.4307,
+      "step": 15498
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2675884955752213e-06,
+      "loss": 2.6461,
+      "step": 15499
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2675410872313526e-06,
+      "loss": 3.3894,
+      "step": 15500
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2674936788874843e-06,
+      "loss": 3.2867,
+      "step": 15501
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2674462705436156e-06,
+      "loss": 4.0155,
+      "step": 15502
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2673988621997473e-06,
+      "loss": 2.7724,
+      "step": 15503
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.267351453855879e-06,
+      "loss": 3.2744,
+      "step": 15504
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2673040455120103e-06,
+      "loss": 2.9683,
+      "step": 15505
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2672566371681416e-06,
+      "loss": 3.0705,
+      "step": 15506
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2672092288242733e-06,
+      "loss": 3.1433,
+      "step": 15507
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2671618204804045e-06,
+      "loss": 2.8417,
+      "step": 15508
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.267114412136536e-06,
+      "loss": 2.9498,
+      "step": 15509
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2670670037926675e-06,
+      "loss": 2.5761,
+      "step": 15510
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2670195954487992e-06,
+      "loss": 2.2686,
+      "step": 15511
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2669721871049305e-06,
+      "loss": 1.7334,
+      "step": 15512
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.266924778761062e-06,
+      "loss": 3.1286,
+      "step": 15513
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2668773704171935e-06,
+      "loss": 2.4873,
+      "step": 15514
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2668299620733248e-06,
+      "loss": 3.1058,
+      "step": 15515
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2667825537294565e-06,
+      "loss": 2.6809,
+      "step": 15516
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.266735145385588e-06,
+      "loss": 2.7314,
+      "step": 15517
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2666877370417195e-06,
+      "loss": 2.7241,
+      "step": 15518
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2666403286978507e-06,
+      "loss": 2.7153,
+      "step": 15519
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2665929203539825e-06,
+      "loss": 3.3786,
+      "step": 15520
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2665455120101137e-06,
+      "loss": 3.5702,
+      "step": 15521
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.266498103666245e-06,
+      "loss": 3.643,
+      "step": 15522
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2664506953223767e-06,
+      "loss": 3.206,
+      "step": 15523
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2664032869785084e-06,
+      "loss": 2.9423,
+      "step": 15524
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2663558786346397e-06,
+      "loss": 3.823,
+      "step": 15525
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2663084702907714e-06,
+      "loss": 3.2205,
+      "step": 15526
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2662610619469027e-06,
+      "loss": 2.2319,
+      "step": 15527
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.266213653603034e-06,
+      "loss": 3.3495,
+      "step": 15528
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2661662452591657e-06,
+      "loss": 2.7582,
+      "step": 15529
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2661188369152974e-06,
+      "loss": 3.0563,
+      "step": 15530
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2660714285714287e-06,
+      "loss": 2.9366,
+      "step": 15531
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.26602402022756e-06,
+      "loss": 3.3955,
+      "step": 15532
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2659766118836916e-06,
+      "loss": 2.4825,
+      "step": 15533
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265929203539823e-06,
+      "loss": 1.9691,
+      "step": 15534
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265881795195954e-06,
+      "loss": 3.3274,
+      "step": 15535
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265834386852086e-06,
+      "loss": 2.4462,
+      "step": 15536
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2657869785082176e-06,
+      "loss": 2.7959,
+      "step": 15537
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265739570164349e-06,
+      "loss": 2.983,
+      "step": 15538
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2656921618204806e-06,
+      "loss": 2.8436,
+      "step": 15539
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265644753476612e-06,
+      "loss": 3.3431,
+      "step": 15540
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265597345132743e-06,
+      "loss": 3.1034,
+      "step": 15541
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265549936788875e-06,
+      "loss": 2.9688,
+      "step": 15542
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2655025284450066e-06,
+      "loss": 3.5542,
+      "step": 15543
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265455120101138e-06,
+      "loss": 2.5087,
+      "step": 15544
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.2654077117572695e-06,
+      "loss": 3.7861,
+      "step": 15545
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265360303413401e-06,
+      "loss": 1.7361,
+      "step": 15546
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.265312895069532e-06,
+      "loss": 3.031,
+      "step": 15547
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.265265486725664e-06,
+      "loss": 2.0314,
+      "step": 15548
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.265218078381795e-06,
+      "loss": 2.6112,
+      "step": 15549
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.265170670037927e-06,
+      "loss": 3.1108,
+      "step": 15550
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2651232616940585e-06,
+      "loss": 3.7041,
+      "step": 15551
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2650758533501898e-06,
+      "loss": 3.259,
+      "step": 15552
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.265028445006321e-06,
+      "loss": 4.1568,
+      "step": 15553
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2649810366624523e-06,
+      "loss": 4.1261,
+      "step": 15554
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264933628318584e-06,
+      "loss": 2.1503,
+      "step": 15555
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2648862199747157e-06,
+      "loss": 2.822,
+      "step": 15556
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264838811630847e-06,
+      "loss": 2.5579,
+      "step": 15557
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2647914032869787e-06,
+      "loss": 3.3245,
+      "step": 15558
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.26474399494311e-06,
+      "loss": 3.8179,
+      "step": 15559
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2646965865992413e-06,
+      "loss": 2.1687,
+      "step": 15560
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264649178255373e-06,
+      "loss": 2.5516,
+      "step": 15561
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2646017699115043e-06,
+      "loss": 2.0083,
+      "step": 15562
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264554361567636e-06,
+      "loss": 4.0712,
+      "step": 15563
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2645069532237677e-06,
+      "loss": 3.5622,
+      "step": 15564
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264459544879899e-06,
+      "loss": 3.8533,
+      "step": 15565
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2644121365360302e-06,
+      "loss": 3.6692,
+      "step": 15566
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264364728192162e-06,
+      "loss": 2.2471,
+      "step": 15567
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2643173198482932e-06,
+      "loss": 2.212,
+      "step": 15568
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264269911504425e-06,
+      "loss": 3.356,
+      "step": 15569
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2642225031605566e-06,
+      "loss": 3.386,
+      "step": 15570
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264175094816688e-06,
+      "loss": 3.5604,
+      "step": 15571
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264127686472819e-06,
+      "loss": 2.2655,
+      "step": 15572
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264080278128951e-06,
+      "loss": 2.6806,
+      "step": 15573
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.264032869785082e-06,
+      "loss": 3.9572,
+      "step": 15574
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2639854614412135e-06,
+      "loss": 4.0718,
+      "step": 15575
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263938053097345e-06,
+      "loss": 3.0602,
+      "step": 15576
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263890644753477e-06,
+      "loss": 3.8491,
+      "step": 15577
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263843236409608e-06,
+      "loss": 3.7498,
+      "step": 15578
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2637958280657394e-06,
+      "loss": 2.783,
+      "step": 15579
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263748419721871e-06,
+      "loss": 3.8584,
+      "step": 15580
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2637010113780024e-06,
+      "loss": 2.2603,
+      "step": 15581
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263653603034134e-06,
+      "loss": 3.9399,
+      "step": 15582
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263606194690266e-06,
+      "loss": 2.9715,
+      "step": 15583
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263558786346397e-06,
+      "loss": 2.3026,
+      "step": 15584
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2635113780025284e-06,
+      "loss": 3.0249,
+      "step": 15585
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.26346396965866e-06,
+      "loss": 2.8444,
+      "step": 15586
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2634165613147914e-06,
+      "loss": 3.2155,
+      "step": 15587
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2633691529709227e-06,
+      "loss": 2.5,
+      "step": 15588
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2633217446270544e-06,
+      "loss": 2.7661,
+      "step": 15589
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263274336283186e-06,
+      "loss": 2.9586,
+      "step": 15590
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2632269279393173e-06,
+      "loss": 3.3985,
+      "step": 15591
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.263179519595449e-06,
+      "loss": 3.0746,
+      "step": 15592
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2631321112515803e-06,
+      "loss": 3.8227,
+      "step": 15593
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2630847029077116e-06,
+      "loss": 2.7035,
+      "step": 15594
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2630372945638433e-06,
+      "loss": 2.2667,
+      "step": 15595
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.262989886219975e-06,
+      "loss": 3.3313,
+      "step": 15596
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2629424778761063e-06,
+      "loss": 2.0255,
+      "step": 15597
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2628950695322376e-06,
+      "loss": 2.9179,
+      "step": 15598
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2628476611883693e-06,
+      "loss": 3.1369,
+      "step": 15599
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2628002528445006e-06,
+      "loss": 3.3406,
+      "step": 15600
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.262752844500632e-06,
+      "loss": 3.3749,
+      "step": 15601
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2627054361567635e-06,
+      "loss": 1.9332,
+      "step": 15602
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2626580278128952e-06,
+      "loss": 4.0062,
+      "step": 15603
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2626106194690265e-06,
+      "loss": 1.352,
+      "step": 15604
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2625632111251582e-06,
+      "loss": 2.1237,
+      "step": 15605
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2625158027812895e-06,
+      "loss": 2.4007,
+      "step": 15606
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.262468394437421e-06,
+      "loss": 1.8725,
+      "step": 15607
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2624209860935525e-06,
+      "loss": 2.7935,
+      "step": 15608
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.262373577749684e-06,
+      "loss": 2.342,
+      "step": 15609
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.2623261694058155e-06,
+      "loss": 2.5347,
+      "step": 15610
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.262278761061947e-06,
+      "loss": 2.2282,
+      "step": 15611
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2622313527180785e-06,
+      "loss": 3.1164,
+      "step": 15612
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2621839443742097e-06,
+      "loss": 3.1342,
+      "step": 15613
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2621365360303414e-06,
+      "loss": 2.4348,
+      "step": 15614
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2620891276864727e-06,
+      "loss": 2.874,
+      "step": 15615
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2620417193426044e-06,
+      "loss": 3.1239,
+      "step": 15616
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.261994310998736e-06,
+      "loss": 2.7681,
+      "step": 15617
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2619469026548674e-06,
+      "loss": 3.0714,
+      "step": 15618
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2618994943109987e-06,
+      "loss": 2.6284,
+      "step": 15619
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.26185208596713e-06,
+      "loss": 3.627,
+      "step": 15620
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2618046776232617e-06,
+      "loss": 3.3405,
+      "step": 15621
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2617572692793934e-06,
+      "loss": 2.4215,
+      "step": 15622
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2617098609355247e-06,
+      "loss": 3.3327,
+      "step": 15623
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2616624525916564e-06,
+      "loss": 4.485,
+      "step": 15624
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2616150442477877e-06,
+      "loss": 3.335,
+      "step": 15625
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.261567635903919e-06,
+      "loss": 3.0647,
+      "step": 15626
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2615202275600506e-06,
+      "loss": 3.2761,
+      "step": 15627
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.261472819216182e-06,
+      "loss": 3.9363,
+      "step": 15628
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2614254108723136e-06,
+      "loss": 3.2947,
+      "step": 15629
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2613780025284453e-06,
+      "loss": 3.4678,
+      "step": 15630
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2613305941845766e-06,
+      "loss": 2.9462,
+      "step": 15631
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.261283185840708e-06,
+      "loss": 3.0109,
+      "step": 15632
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2612357774968396e-06,
+      "loss": 1.8757,
+      "step": 15633
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.261188369152971e-06,
+      "loss": 3.2805,
+      "step": 15634
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2611409608091026e-06,
+      "loss": 2.7987,
+      "step": 15635
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2610935524652343e-06,
+      "loss": 2.672,
+      "step": 15636
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2610461441213656e-06,
+      "loss": 2.4884,
+      "step": 15637
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260998735777497e-06,
+      "loss": 3.9003,
+      "step": 15638
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2609513274336285e-06,
+      "loss": 2.437,
+      "step": 15639
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.26090391908976e-06,
+      "loss": 2.6003,
+      "step": 15640
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260856510745891e-06,
+      "loss": 3.6855,
+      "step": 15641
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260809102402023e-06,
+      "loss": 2.032,
+      "step": 15642
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2607616940581545e-06,
+      "loss": 2.6538,
+      "step": 15643
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260714285714286e-06,
+      "loss": 2.909,
+      "step": 15644
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260666877370417e-06,
+      "loss": 2.4169,
+      "step": 15645
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2606194690265488e-06,
+      "loss": 3.191,
+      "step": 15646
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.26057206068268e-06,
+      "loss": 3.376,
+      "step": 15647
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2605246523388118e-06,
+      "loss": 3.091,
+      "step": 15648
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2604772439949435e-06,
+      "loss": 3.4831,
+      "step": 15649
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2604298356510747e-06,
+      "loss": 3.231,
+      "step": 15650
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260382427307206e-06,
+      "loss": 2.964,
+      "step": 15651
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2603350189633377e-06,
+      "loss": 2.5483,
+      "step": 15652
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260287610619469e-06,
+      "loss": 2.2744,
+      "step": 15653
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2602402022756003e-06,
+      "loss": 2.6411,
+      "step": 15654
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260192793931732e-06,
+      "loss": 2.2322,
+      "step": 15655
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2601453855878637e-06,
+      "loss": 4.0515,
+      "step": 15656
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260097977243995e-06,
+      "loss": 4.0053,
+      "step": 15657
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2600505689001267e-06,
+      "loss": 2.8059,
+      "step": 15658
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.260003160556258e-06,
+      "loss": 2.9016,
+      "step": 15659
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2599557522123892e-06,
+      "loss": 2.5286,
+      "step": 15660
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259908343868521e-06,
+      "loss": 3.6154,
+      "step": 15661
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2598609355246526e-06,
+      "loss": 3.634,
+      "step": 15662
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259813527180784e-06,
+      "loss": 3.1869,
+      "step": 15663
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259766118836915e-06,
+      "loss": 3.8924,
+      "step": 15664
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259718710493047e-06,
+      "loss": 2.0464,
+      "step": 15665
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259671302149178e-06,
+      "loss": 2.3611,
+      "step": 15666
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2596238938053095e-06,
+      "loss": 3.4847,
+      "step": 15667
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259576485461441e-06,
+      "loss": 2.415,
+      "step": 15668
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259529077117573e-06,
+      "loss": 3.074,
+      "step": 15669
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259481668773704e-06,
+      "loss": 2.6463,
+      "step": 15670
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259434260429836e-06,
+      "loss": 3.1075,
+      "step": 15671
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.259386852085967e-06,
+      "loss": 2.5283,
+      "step": 15672
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.2593394437420984e-06,
+      "loss": 3.6732,
+      "step": 15673
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2.25929203539823e-06,
+      "loss": 3.0217,
+      "step": 15674
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.259244627054362e-06,
+      "loss": 3.8553,
+      "step": 15675
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.259197218710493e-06,
+      "loss": 3.2071,
+      "step": 15676
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.259149810366625e-06,
+      "loss": 3.6233,
+      "step": 15677
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.259102402022756e-06,
+      "loss": 2.7004,
+      "step": 15678
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2590549936788874e-06,
+      "loss": 3.8947,
+      "step": 15679
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.259007585335019e-06,
+      "loss": 3.4175,
+      "step": 15680
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2589601769911504e-06,
+      "loss": 3.3323,
+      "step": 15681
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.258912768647282e-06,
+      "loss": 3.7263,
+      "step": 15682
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2588653603034138e-06,
+      "loss": 3.8288,
+      "step": 15683
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.258817951959545e-06,
+      "loss": 2.6576,
+      "step": 15684
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2587705436156763e-06,
+      "loss": 3.7679,
+      "step": 15685
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2587231352718076e-06,
+      "loss": 2.703,
+      "step": 15686
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2586757269279393e-06,
+      "loss": 2.6392,
+      "step": 15687
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.258628318584071e-06,
+      "loss": 3.0278,
+      "step": 15688
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2585809102402023e-06,
+      "loss": 3.0951,
+      "step": 15689
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.258533501896334e-06,
+      "loss": 2.9735,
+      "step": 15690
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2584860935524653e-06,
+      "loss": 3.7246,
+      "step": 15691
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2584386852085966e-06,
+      "loss": 2.9498,
+      "step": 15692
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2583912768647283e-06,
+      "loss": 3.189,
+      "step": 15693
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2583438685208596e-06,
+      "loss": 2.3656,
+      "step": 15694
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2582964601769913e-06,
+      "loss": 2.3026,
+      "step": 15695
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.258249051833123e-06,
+      "loss": 3.001,
+      "step": 15696
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2582016434892542e-06,
+      "loss": 3.4641,
+      "step": 15697
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2581542351453855e-06,
+      "loss": 3.8859,
+      "step": 15698
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2581068268015172e-06,
+      "loss": 3.051,
+      "step": 15699
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2580594184576485e-06,
+      "loss": 3.6034,
+      "step": 15700
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.25801201011378e-06,
+      "loss": 1.8505,
+      "step": 15701
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.257964601769912e-06,
+      "loss": 3.2493,
+      "step": 15702
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.257917193426043e-06,
+      "loss": 3.4793,
+      "step": 15703
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2578697850821745e-06,
+      "loss": 2.456,
+      "step": 15704
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.257822376738306e-06,
+      "loss": 2.6896,
+      "step": 15705
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2577749683944375e-06,
+      "loss": 3.4105,
+      "step": 15706
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2577275600505687e-06,
+      "loss": 2.742,
+      "step": 15707
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2576801517067004e-06,
+      "loss": 4.3545,
+      "step": 15708
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.257632743362832e-06,
+      "loss": 3.2568,
+      "step": 15709
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2575853350189634e-06,
+      "loss": 2.9898,
+      "step": 15710
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2575379266750947e-06,
+      "loss": 2.9442,
+      "step": 15711
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2574905183312264e-06,
+      "loss": 2.6624,
+      "step": 15712
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2574431099873577e-06,
+      "loss": 3.5091,
+      "step": 15713
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2573957016434894e-06,
+      "loss": 2.953,
+      "step": 15714
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.257348293299621e-06,
+      "loss": 3.6905,
+      "step": 15715
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2573008849557524e-06,
+      "loss": 2.9615,
+      "step": 15716
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2572534766118837e-06,
+      "loss": 3.4665,
+      "step": 15717
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2572060682680154e-06,
+      "loss": 1.8619,
+      "step": 15718
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2571586599241466e-06,
+      "loss": 1.7445,
+      "step": 15719
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.257111251580278e-06,
+      "loss": 2.787,
+      "step": 15720
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2570638432364096e-06,
+      "loss": 2.5504,
+      "step": 15721
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2570164348925413e-06,
+      "loss": 3.0677,
+      "step": 15722
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2569690265486726e-06,
+      "loss": 3.0799,
+      "step": 15723
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2569216182048043e-06,
+      "loss": 2.3946,
+      "step": 15724
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2568742098609356e-06,
+      "loss": 3.8839,
+      "step": 15725
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.256826801517067e-06,
+      "loss": 1.7216,
+      "step": 15726
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2567793931731986e-06,
+      "loss": 3.1188,
+      "step": 15727
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2567319848293303e-06,
+      "loss": 4.1986,
+      "step": 15728
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2566845764854616e-06,
+      "loss": 3.1142,
+      "step": 15729
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.256637168141593e-06,
+      "loss": 2.666,
+      "step": 15730
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2565897597977246e-06,
+      "loss": 3.6488,
+      "step": 15731
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.256542351453856e-06,
+      "loss": 2.0955,
+      "step": 15732
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.256494943109987e-06,
+      "loss": 3.581,
+      "step": 15733
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.256447534766119e-06,
+      "loss": 2.9297,
+      "step": 15734
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2564001264222505e-06,
+      "loss": 2.5766,
+      "step": 15735
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.256352718078382e-06,
+      "loss": 3.0777,
+      "step": 15736
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.2563053097345135e-06,
+      "loss": 3.1711,
+      "step": 15737
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2562579013906448e-06,
+      "loss": 3.3046,
+      "step": 15738
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.256210493046776e-06,
+      "loss": 2.6508,
+      "step": 15739
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2561630847029078e-06,
+      "loss": 2.5489,
+      "step": 15740
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.256115676359039e-06,
+      "loss": 2.2303,
+      "step": 15741
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2560682680151708e-06,
+      "loss": 2.9018,
+      "step": 15742
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2560208596713025e-06,
+      "loss": 2.5065,
+      "step": 15743
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2559734513274337e-06,
+      "loss": 2.4576,
+      "step": 15744
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255926042983565e-06,
+      "loss": 2.301,
+      "step": 15745
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2558786346396967e-06,
+      "loss": 2.6494,
+      "step": 15746
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255831226295828e-06,
+      "loss": 2.5652,
+      "step": 15747
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2557838179519597e-06,
+      "loss": 2.9279,
+      "step": 15748
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2557364096080914e-06,
+      "loss": 2.7267,
+      "step": 15749
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2556890012642227e-06,
+      "loss": 2.9879,
+      "step": 15750
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255641592920354e-06,
+      "loss": 3.1012,
+      "step": 15751
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2555941845764853e-06,
+      "loss": 2.2893,
+      "step": 15752
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255546776232617e-06,
+      "loss": 2.5815,
+      "step": 15753
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2554993678887482e-06,
+      "loss": 3.2079,
+      "step": 15754
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.25545195954488e-06,
+      "loss": 3.0422,
+      "step": 15755
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2554045512010116e-06,
+      "loss": 3.1949,
+      "step": 15756
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255357142857143e-06,
+      "loss": 4.6333,
+      "step": 15757
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255309734513274e-06,
+      "loss": 3.6125,
+      "step": 15758
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255262326169406e-06,
+      "loss": 2.3409,
+      "step": 15759
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255214917825537e-06,
+      "loss": 2.311,
+      "step": 15760
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255167509481669e-06,
+      "loss": 3.2839,
+      "step": 15761
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2551201011378006e-06,
+      "loss": 3.7581,
+      "step": 15762
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255072692793932e-06,
+      "loss": 2.9306,
+      "step": 15763
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.255025284450063e-06,
+      "loss": 3.3151,
+      "step": 15764
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254977876106195e-06,
+      "loss": 2.8666,
+      "step": 15765
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254930467762326e-06,
+      "loss": 2.9925,
+      "step": 15766
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2548830594184574e-06,
+      "loss": 2.3506,
+      "step": 15767
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254835651074589e-06,
+      "loss": 3.2745,
+      "step": 15768
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254788242730721e-06,
+      "loss": 2.8019,
+      "step": 15769
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254740834386852e-06,
+      "loss": 3.7145,
+      "step": 15770
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254693426042984e-06,
+      "loss": 3.0914,
+      "step": 15771
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254646017699115e-06,
+      "loss": 3.9848,
+      "step": 15772
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2545986093552464e-06,
+      "loss": 2.017,
+      "step": 15773
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254551201011378e-06,
+      "loss": 2.867,
+      "step": 15774
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2545037926675098e-06,
+      "loss": 3.3579,
+      "step": 15775
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254456384323641e-06,
+      "loss": 2.6275,
+      "step": 15776
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2544089759797723e-06,
+      "loss": 2.3819,
+      "step": 15777
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254361567635904e-06,
+      "loss": 2.4663,
+      "step": 15778
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2543141592920353e-06,
+      "loss": 2.7509,
+      "step": 15779
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2542667509481666e-06,
+      "loss": 2.6983,
+      "step": 15780
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2542193426042983e-06,
+      "loss": 1.9417,
+      "step": 15781
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.25417193426043e-06,
+      "loss": 2.5915,
+      "step": 15782
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2541245259165613e-06,
+      "loss": 3.2509,
+      "step": 15783
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.254077117572693e-06,
+      "loss": 3.0423,
+      "step": 15784
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2540297092288243e-06,
+      "loss": 3.275,
+      "step": 15785
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2539823008849556e-06,
+      "loss": 3.2971,
+      "step": 15786
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2539348925410873e-06,
+      "loss": 2.2463,
+      "step": 15787
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.253887484197219e-06,
+      "loss": 2.9606,
+      "step": 15788
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2538400758533502e-06,
+      "loss": 1.8282,
+      "step": 15789
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.253792667509482e-06,
+      "loss": 3.0264,
+      "step": 15790
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2537452591656132e-06,
+      "loss": 2.0547,
+      "step": 15791
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2536978508217445e-06,
+      "loss": 2.032,
+      "step": 15792
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2536504424778762e-06,
+      "loss": 3.3849,
+      "step": 15793
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2536030341340075e-06,
+      "loss": 2.8036,
+      "step": 15794
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.253555625790139e-06,
+      "loss": 3.8671,
+      "step": 15795
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2535082174462705e-06,
+      "loss": 3.7412,
+      "step": 15796
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.253460809102402e-06,
+      "loss": 3.1451,
+      "step": 15797
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2534134007585335e-06,
+      "loss": 3.0903,
+      "step": 15798
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2533659924146647e-06,
+      "loss": 2.2651,
+      "step": 15799
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2.2533185840707965e-06,
+      "loss": 3.371,
+      "step": 15800
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.253271175726928e-06,
+      "loss": 2.7237,
+      "step": 15801
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2532237673830594e-06,
+      "loss": 3.683,
+      "step": 15802
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.253176359039191e-06,
+      "loss": 3.3385,
+      "step": 15803
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2531289506953224e-06,
+      "loss": 3.6288,
+      "step": 15804
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2530815423514537e-06,
+      "loss": 3.1287,
+      "step": 15805
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2530341340075854e-06,
+      "loss": 4.09,
+      "step": 15806
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2529867256637167e-06,
+      "loss": 2.8453,
+      "step": 15807
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2529393173198484e-06,
+      "loss": 3.4375,
+      "step": 15808
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.25289190897598e-06,
+      "loss": 2.8198,
+      "step": 15809
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2528445006321114e-06,
+      "loss": 2.1629,
+      "step": 15810
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2527970922882427e-06,
+      "loss": 3.2316,
+      "step": 15811
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2527496839443744e-06,
+      "loss": 3.5703,
+      "step": 15812
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2527022756005056e-06,
+      "loss": 1.9712,
+      "step": 15813
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2526548672566373e-06,
+      "loss": 3.1572,
+      "step": 15814
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.252607458912769e-06,
+      "loss": 2.6821,
+      "step": 15815
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2525600505689003e-06,
+      "loss": 3.7065,
+      "step": 15816
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2525126422250316e-06,
+      "loss": 2.0325,
+      "step": 15817
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.252465233881163e-06,
+      "loss": 3.3858,
+      "step": 15818
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2524178255372946e-06,
+      "loss": 3.121,
+      "step": 15819
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.252370417193426e-06,
+      "loss": 3.3964,
+      "step": 15820
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2523230088495576e-06,
+      "loss": 3.2233,
+      "step": 15821
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2522756005056893e-06,
+      "loss": 2.8017,
+      "step": 15822
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2522281921618206e-06,
+      "loss": 2.8227,
+      "step": 15823
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.252180783817952e-06,
+      "loss": 3.5851,
+      "step": 15824
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2521333754740835e-06,
+      "loss": 2.8181,
+      "step": 15825
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.252085967130215e-06,
+      "loss": 3.6335,
+      "step": 15826
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2520385587863465e-06,
+      "loss": 4.1773,
+      "step": 15827
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2519911504424782e-06,
+      "loss": 3.6328,
+      "step": 15828
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2519437420986095e-06,
+      "loss": 3.7849,
+      "step": 15829
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.251896333754741e-06,
+      "loss": 3.4819,
+      "step": 15830
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2518489254108725e-06,
+      "loss": 2.9978,
+      "step": 15831
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2518015170670038e-06,
+      "loss": 2.7003,
+      "step": 15832
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.251754108723135e-06,
+      "loss": 2.8388,
+      "step": 15833
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2517067003792668e-06,
+      "loss": 2.8925,
+      "step": 15834
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2516592920353985e-06,
+      "loss": 2.0133,
+      "step": 15835
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2516118836915297e-06,
+      "loss": 3.2333,
+      "step": 15836
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2515644753476615e-06,
+      "loss": 3.2838,
+      "step": 15837
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2515170670037927e-06,
+      "loss": 3.7488,
+      "step": 15838
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.251469658659924e-06,
+      "loss": 2.3427,
+      "step": 15839
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2514222503160557e-06,
+      "loss": 3.0053,
+      "step": 15840
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2513748419721874e-06,
+      "loss": 2.0333,
+      "step": 15841
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2513274336283187e-06,
+      "loss": 3.2852,
+      "step": 15842
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.25128002528445e-06,
+      "loss": 2.1224,
+      "step": 15843
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2512326169405817e-06,
+      "loss": 3.9367,
+      "step": 15844
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.251185208596713e-06,
+      "loss": 2.6858,
+      "step": 15845
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2511378002528442e-06,
+      "loss": 2.4876,
+      "step": 15846
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.251090391908976e-06,
+      "loss": 2.5576,
+      "step": 15847
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2510429835651077e-06,
+      "loss": 3.3522,
+      "step": 15848
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250995575221239e-06,
+      "loss": 2.5136,
+      "step": 15849
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2509481668773706e-06,
+      "loss": 2.5556,
+      "step": 15850
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250900758533502e-06,
+      "loss": 2.3448,
+      "step": 15851
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250853350189633e-06,
+      "loss": 3.2075,
+      "step": 15852
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250805941845765e-06,
+      "loss": 2.5526,
+      "step": 15853
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2507585335018966e-06,
+      "loss": 3.0775,
+      "step": 15854
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250711125158028e-06,
+      "loss": 2.263,
+      "step": 15855
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.2506637168141596e-06,
+      "loss": 2.9909,
+      "step": 15856
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250616308470291e-06,
+      "loss": 2.6314,
+      "step": 15857
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250568900126422e-06,
+      "loss": 3.3279,
+      "step": 15858
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250521491782554e-06,
+      "loss": 3.2926,
+      "step": 15859
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250474083438685e-06,
+      "loss": 3.684,
+      "step": 15860
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250426675094817e-06,
+      "loss": 2.8321,
+      "step": 15861
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250379266750948e-06,
+      "loss": 3.6619,
+      "step": 15862
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.25033185840708e-06,
+      "loss": 3.3737,
+      "step": 15863
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.250284450063211e-06,
+      "loss": 2.6751,
+      "step": 15864
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2502370417193424e-06,
+      "loss": 2.9489,
+      "step": 15865
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.250189633375474e-06,
+      "loss": 3.4323,
+      "step": 15866
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.250142225031606e-06,
+      "loss": 3.32,
+      "step": 15867
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.250094816687737e-06,
+      "loss": 2.2984,
+      "step": 15868
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2500474083438688e-06,
+      "loss": 2.5479,
+      "step": 15869
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.25e-06,
+      "loss": 3.3379,
+      "step": 15870
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2499525916561313e-06,
+      "loss": 2.891,
+      "step": 15871
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249905183312263e-06,
+      "loss": 3.292,
+      "step": 15872
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2498577749683943e-06,
+      "loss": 2.9631,
+      "step": 15873
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249810366624526e-06,
+      "loss": 2.2464,
+      "step": 15874
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2497629582806577e-06,
+      "loss": 3.402,
+      "step": 15875
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249715549936789e-06,
+      "loss": 3.9381,
+      "step": 15876
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2496681415929203e-06,
+      "loss": 3.4288,
+      "step": 15877
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249620733249052e-06,
+      "loss": 3.114,
+      "step": 15878
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2495733249051833e-06,
+      "loss": 3.2632,
+      "step": 15879
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249525916561315e-06,
+      "loss": 3.1313,
+      "step": 15880
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2494785082174463e-06,
+      "loss": 3.2462,
+      "step": 15881
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249431099873578e-06,
+      "loss": 3.148,
+      "step": 15882
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2493836915297092e-06,
+      "loss": 2.499,
+      "step": 15883
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2493362831858405e-06,
+      "loss": 3.0764,
+      "step": 15884
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2492888748419722e-06,
+      "loss": 3.5295,
+      "step": 15885
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2492414664981035e-06,
+      "loss": 1.9754,
+      "step": 15886
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2491940581542352e-06,
+      "loss": 1.8189,
+      "step": 15887
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249146649810367e-06,
+      "loss": 3.5268,
+      "step": 15888
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249099241466498e-06,
+      "loss": 3.2958,
+      "step": 15889
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2490518331226295e-06,
+      "loss": 4.2452,
+      "step": 15890
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.249004424778761e-06,
+      "loss": 2.7524,
+      "step": 15891
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2489570164348925e-06,
+      "loss": 3.4303,
+      "step": 15892
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.248909608091024e-06,
+      "loss": 2.7046,
+      "step": 15893
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.248862199747156e-06,
+      "loss": 2.8756,
+      "step": 15894
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.248814791403287e-06,
+      "loss": 3.6089,
+      "step": 15895
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2487673830594184e-06,
+      "loss": 3.5165,
+      "step": 15896
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.24871997471555e-06,
+      "loss": 2.247,
+      "step": 15897
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2486725663716814e-06,
+      "loss": 2.745,
+      "step": 15898
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2486251580278127e-06,
+      "loss": 2.4809,
+      "step": 15899
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2485777496839444e-06,
+      "loss": 1.8228,
+      "step": 15900
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.248530341340076e-06,
+      "loss": 2.3282,
+      "step": 15901
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2484829329962074e-06,
+      "loss": 3.8536,
+      "step": 15902
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2484355246523387e-06,
+      "loss": 4.0007,
+      "step": 15903
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2483881163084704e-06,
+      "loss": 2.7267,
+      "step": 15904
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2483407079646016e-06,
+      "loss": 2.5368,
+      "step": 15905
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2482932996207334e-06,
+      "loss": 3.244,
+      "step": 15906
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.248245891276865e-06,
+      "loss": 3.3157,
+      "step": 15907
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2481984829329963e-06,
+      "loss": 3.6843,
+      "step": 15908
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2481510745891276e-06,
+      "loss": 3.4078,
+      "step": 15909
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2481036662452593e-06,
+      "loss": 2.8093,
+      "step": 15910
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2480562579013906e-06,
+      "loss": 2.8396,
+      "step": 15911
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.248008849557522e-06,
+      "loss": 2.6591,
+      "step": 15912
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2479614412136536e-06,
+      "loss": 4.0509,
+      "step": 15913
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2479140328697853e-06,
+      "loss": 3.6171,
+      "step": 15914
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2478666245259166e-06,
+      "loss": 1.7377,
+      "step": 15915
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2478192161820483e-06,
+      "loss": 3.7018,
+      "step": 15916
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2477718078381796e-06,
+      "loss": 3.8187,
+      "step": 15917
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.247724399494311e-06,
+      "loss": 3.4317,
+      "step": 15918
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2476769911504425e-06,
+      "loss": 2.7777,
+      "step": 15919
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2476295828065742e-06,
+      "loss": 4.1379,
+      "step": 15920
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2475821744627055e-06,
+      "loss": 3.3758,
+      "step": 15921
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2475347661188372e-06,
+      "loss": 2.7661,
+      "step": 15922
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2474873577749685e-06,
+      "loss": 3.192,
+      "step": 15923
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2474399494310998e-06,
+      "loss": 3.5106,
+      "step": 15924
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.247392541087231e-06,
+      "loss": 2.8507,
+      "step": 15925
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2473451327433628e-06,
+      "loss": 1.9235,
+      "step": 15926
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2.2472977243994945e-06,
+      "loss": 3.7565,
+      "step": 15927
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2472503160556258e-06,
+      "loss": 3.2082,
+      "step": 15928
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2472029077117575e-06,
+      "loss": 2.4286,
+      "step": 15929
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2471554993678887e-06,
+      "loss": 3.6653,
+      "step": 15930
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24710809102402e-06,
+      "loss": 2.3341,
+      "step": 15931
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2470606826801517e-06,
+      "loss": 3.6644,
+      "step": 15932
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2470132743362834e-06,
+      "loss": 3.297,
+      "step": 15933
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2469658659924147e-06,
+      "loss": 2.4819,
+      "step": 15934
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2469184576485464e-06,
+      "loss": 3.2563,
+      "step": 15935
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2468710493046777e-06,
+      "loss": 3.0253,
+      "step": 15936
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246823640960809e-06,
+      "loss": 3.5364,
+      "step": 15937
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2467762326169407e-06,
+      "loss": 2.8957,
+      "step": 15938
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246728824273072e-06,
+      "loss": 3.6729,
+      "step": 15939
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2466814159292037e-06,
+      "loss": 3.5493,
+      "step": 15940
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2466340075853354e-06,
+      "loss": 3.1154,
+      "step": 15941
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2465865992414666e-06,
+      "loss": 3.3384,
+      "step": 15942
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246539190897598e-06,
+      "loss": 2.2131,
+      "step": 15943
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2464917825537296e-06,
+      "loss": 3.2005,
+      "step": 15944
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246444374209861e-06,
+      "loss": 3.0525,
+      "step": 15945
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2463969658659926e-06,
+      "loss": 2.8967,
+      "step": 15946
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246349557522124e-06,
+      "loss": 3.5905,
+      "step": 15947
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2463021491782556e-06,
+      "loss": 4.1187,
+      "step": 15948
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246254740834387e-06,
+      "loss": 2.3023,
+      "step": 15949
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246207332490518e-06,
+      "loss": 2.9146,
+      "step": 15950
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24615992414665e-06,
+      "loss": 3.0306,
+      "step": 15951
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246112515802781e-06,
+      "loss": 1.9947,
+      "step": 15952
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.246065107458913e-06,
+      "loss": 3.457,
+      "step": 15953
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2460176991150446e-06,
+      "loss": 3.1006,
+      "step": 15954
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245970290771176e-06,
+      "loss": 2.3234,
+      "step": 15955
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245922882427307e-06,
+      "loss": 2.7544,
+      "step": 15956
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245875474083439e-06,
+      "loss": 2.8237,
+      "step": 15957
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24582806573957e-06,
+      "loss": 4.2006,
+      "step": 15958
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245780657395702e-06,
+      "loss": 2.5537,
+      "step": 15959
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2457332490518335e-06,
+      "loss": 3.7264,
+      "step": 15960
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2456858407079648e-06,
+      "loss": 3.3136,
+      "step": 15961
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245638432364096e-06,
+      "loss": 3.0226,
+      "step": 15962
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2455910240202278e-06,
+      "loss": 3.3306,
+      "step": 15963
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245543615676359e-06,
+      "loss": 3.0865,
+      "step": 15964
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2454962073324903e-06,
+      "loss": 3.9576,
+      "step": 15965
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245448798988622e-06,
+      "loss": 2.7834,
+      "step": 15966
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2454013906447537e-06,
+      "loss": 3.009,
+      "step": 15967
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245353982300885e-06,
+      "loss": 2.8035,
+      "step": 15968
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2453065739570163e-06,
+      "loss": 2.4964,
+      "step": 15969
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245259165613148e-06,
+      "loss": 2.6025,
+      "step": 15970
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2452117572692793e-06,
+      "loss": 3.5246,
+      "step": 15971
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245164348925411e-06,
+      "loss": 3.1802,
+      "step": 15972
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2451169405815427e-06,
+      "loss": 2.7031,
+      "step": 15973
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.245069532237674e-06,
+      "loss": 2.8044,
+      "step": 15974
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2450221238938053e-06,
+      "loss": 3.1497,
+      "step": 15975
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244974715549937e-06,
+      "loss": 2.1629,
+      "step": 15976
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2449273072060682e-06,
+      "loss": 3.882,
+      "step": 15977
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2448798988621995e-06,
+      "loss": 3.8866,
+      "step": 15978
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2448324905183312e-06,
+      "loss": 2.1806,
+      "step": 15979
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244785082174463e-06,
+      "loss": 2.5111,
+      "step": 15980
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244737673830594e-06,
+      "loss": 3.8039,
+      "step": 15981
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244690265486726e-06,
+      "loss": 2.6869,
+      "step": 15982
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244642857142857e-06,
+      "loss": 2.1502,
+      "step": 15983
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2445954487989885e-06,
+      "loss": 3.3443,
+      "step": 15984
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.24454804045512e-06,
+      "loss": 3.33,
+      "step": 15985
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244500632111252e-06,
+      "loss": 2.4406,
+      "step": 15986
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244453223767383e-06,
+      "loss": 2.4204,
+      "step": 15987
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244405815423515e-06,
+      "loss": 2.7528,
+      "step": 15988
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.244358407079646e-06,
+      "loss": 1.8591,
+      "step": 15989
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.2443109987357774e-06,
+      "loss": 3.2104,
+      "step": 15990
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2442635903919087e-06,
+      "loss": 2.4303,
+      "step": 15991
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2442161820480404e-06,
+      "loss": 2.6652,
+      "step": 15992
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.244168773704172e-06,
+      "loss": 3.0226,
+      "step": 15993
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2441213653603034e-06,
+      "loss": 3.1448,
+      "step": 15994
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.244073957016435e-06,
+      "loss": 2.2079,
+      "step": 15995
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2440265486725664e-06,
+      "loss": 2.7275,
+      "step": 15996
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2439791403286977e-06,
+      "loss": 2.8514,
+      "step": 15997
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2439317319848294e-06,
+      "loss": 2.2823,
+      "step": 15998
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.243884323640961e-06,
+      "loss": 2.6575,
+      "step": 15999
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2438369152970923e-06,
+      "loss": 2.3401,
+      "step": 16000
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.243789506953224e-06,
+      "loss": 3.1079,
+      "step": 16001
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2437420986093553e-06,
+      "loss": 2.2878,
+      "step": 16002
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2436946902654866e-06,
+      "loss": 3.6793,
+      "step": 16003
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2436472819216183e-06,
+      "loss": 3.0897,
+      "step": 16004
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2435998735777496e-06,
+      "loss": 3.0834,
+      "step": 16005
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2435524652338813e-06,
+      "loss": 3.2663,
+      "step": 16006
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.243505056890013e-06,
+      "loss": 3.565,
+      "step": 16007
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2434576485461443e-06,
+      "loss": 3.6645,
+      "step": 16008
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2434102402022756e-06,
+      "loss": 2.9068,
+      "step": 16009
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2433628318584073e-06,
+      "loss": 4.0962,
+      "step": 16010
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2433154235145385e-06,
+      "loss": 3.6338,
+      "step": 16011
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.24326801517067e-06,
+      "loss": 3.7873,
+      "step": 16012
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2432206068268015e-06,
+      "loss": 2.796,
+      "step": 16013
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2431731984829332e-06,
+      "loss": 2.8653,
+      "step": 16014
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2431257901390645e-06,
+      "loss": 2.4835,
+      "step": 16015
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.243078381795196e-06,
+      "loss": 1.8243,
+      "step": 16016
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2430309734513275e-06,
+      "loss": 2.7894,
+      "step": 16017
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2429835651074588e-06,
+      "loss": 3.5075,
+      "step": 16018
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2429361567635905e-06,
+      "loss": 2.5245,
+      "step": 16019
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.242888748419722e-06,
+      "loss": 2.2709,
+      "step": 16020
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2428413400758535e-06,
+      "loss": 2.0243,
+      "step": 16021
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2427939317319848e-06,
+      "loss": 2.6966,
+      "step": 16022
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2427465233881165e-06,
+      "loss": 2.7322,
+      "step": 16023
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2426991150442477e-06,
+      "loss": 2.799,
+      "step": 16024
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.242651706700379e-06,
+      "loss": 3.4887,
+      "step": 16025
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2426042983565107e-06,
+      "loss": 3.1182,
+      "step": 16026
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2425568900126424e-06,
+      "loss": 3.8874,
+      "step": 16027
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2425094816687737e-06,
+      "loss": 3.0386,
+      "step": 16028
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2424620733249054e-06,
+      "loss": 3.1232,
+      "step": 16029
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2424146649810367e-06,
+      "loss": 2.4451,
+      "step": 16030
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.242367256637168e-06,
+      "loss": 3.1595,
+      "step": 16031
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2423198482932997e-06,
+      "loss": 2.3414,
+      "step": 16032
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2422724399494314e-06,
+      "loss": 3.1701,
+      "step": 16033
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2422250316055627e-06,
+      "loss": 2.8662,
+      "step": 16034
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.242177623261694e-06,
+      "loss": 3.1503,
+      "step": 16035
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2421302149178256e-06,
+      "loss": 3.5574,
+      "step": 16036
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.242082806573957e-06,
+      "loss": 3.0524,
+      "step": 16037
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.242035398230088e-06,
+      "loss": 3.027,
+      "step": 16038
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.24198798988622e-06,
+      "loss": 3.1403,
+      "step": 16039
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2419405815423516e-06,
+      "loss": 3.1653,
+      "step": 16040
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241893173198483e-06,
+      "loss": 3.7317,
+      "step": 16041
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2418457648546146e-06,
+      "loss": 2.6823,
+      "step": 16042
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241798356510746e-06,
+      "loss": 2.8697,
+      "step": 16043
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241750948166877e-06,
+      "loss": 2.6148,
+      "step": 16044
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241703539823009e-06,
+      "loss": 3.8784,
+      "step": 16045
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2416561314791406e-06,
+      "loss": 3.3867,
+      "step": 16046
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241608723135272e-06,
+      "loss": 2.3316,
+      "step": 16047
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2415613147914035e-06,
+      "loss": 3.9848,
+      "step": 16048
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241513906447535e-06,
+      "loss": 2.673,
+      "step": 16049
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241466498103666e-06,
+      "loss": 2.1644,
+      "step": 16050
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241419089759798e-06,
+      "loss": 3.2829,
+      "step": 16051
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241371681415929e-06,
+      "loss": 3.3552,
+      "step": 16052
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.241324273072061e-06,
+      "loss": 2.4557,
+      "step": 16053
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2.2412768647281925e-06,
+      "loss": 3.688,
+      "step": 16054
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2412294563843238e-06,
+      "loss": 2.3384,
+      "step": 16055
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.241182048040455e-06,
+      "loss": 3.2405,
+      "step": 16056
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2411346396965863e-06,
+      "loss": 3.2588,
+      "step": 16057
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.241087231352718e-06,
+      "loss": 4.2513,
+      "step": 16058
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2410398230088497e-06,
+      "loss": 3.5564,
+      "step": 16059
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240992414664981e-06,
+      "loss": 2.1985,
+      "step": 16060
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2409450063211127e-06,
+      "loss": 3.2589,
+      "step": 16061
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240897597977244e-06,
+      "loss": 2.9272,
+      "step": 16062
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2408501896333753e-06,
+      "loss": 3.3563,
+      "step": 16063
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240802781289507e-06,
+      "loss": 3.3413,
+      "step": 16064
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2407553729456383e-06,
+      "loss": 3.377,
+      "step": 16065
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.24070796460177e-06,
+      "loss": 2.3364,
+      "step": 16066
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2406605562579017e-06,
+      "loss": 3.2794,
+      "step": 16067
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240613147914033e-06,
+      "loss": 3.5227,
+      "step": 16068
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2405657395701642e-06,
+      "loss": 2.8942,
+      "step": 16069
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240518331226296e-06,
+      "loss": 3.6582,
+      "step": 16070
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2404709228824272e-06,
+      "loss": 3.3033,
+      "step": 16071
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240423514538559e-06,
+      "loss": 3.2341,
+      "step": 16072
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2403761061946906e-06,
+      "loss": 3.4629,
+      "step": 16073
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240328697850822e-06,
+      "loss": 3.8342,
+      "step": 16074
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240281289506953e-06,
+      "loss": 3.0495,
+      "step": 16075
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240233881163085e-06,
+      "loss": 2.658,
+      "step": 16076
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240186472819216e-06,
+      "loss": 2.1053,
+      "step": 16077
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2401390644753475e-06,
+      "loss": 3.8312,
+      "step": 16078
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240091656131479e-06,
+      "loss": 3.334,
+      "step": 16079
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.240044247787611e-06,
+      "loss": 2.955,
+      "step": 16080
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239996839443742e-06,
+      "loss": 3.6396,
+      "step": 16081
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2399494310998734e-06,
+      "loss": 3.018,
+      "step": 16082
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239902022756005e-06,
+      "loss": 3.3466,
+      "step": 16083
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2398546144121364e-06,
+      "loss": 2.0809,
+      "step": 16084
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239807206068268e-06,
+      "loss": 3.1821,
+      "step": 16085
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2397597977244e-06,
+      "loss": 3.6124,
+      "step": 16086
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239712389380531e-06,
+      "loss": 2.9246,
+      "step": 16087
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2396649810366624e-06,
+      "loss": 2.4553,
+      "step": 16088
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239617572692794e-06,
+      "loss": 2.9154,
+      "step": 16089
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2395701643489254e-06,
+      "loss": 3.3517,
+      "step": 16090
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2395227560050567e-06,
+      "loss": 3.7389,
+      "step": 16091
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2394753476611884e-06,
+      "loss": 3.637,
+      "step": 16092
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.23942793931732e-06,
+      "loss": 2.3234,
+      "step": 16093
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2393805309734513e-06,
+      "loss": 3.2429,
+      "step": 16094
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239333122629583e-06,
+      "loss": 2.0092,
+      "step": 16095
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2392857142857143e-06,
+      "loss": 3.4393,
+      "step": 16096
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2392383059418456e-06,
+      "loss": 2.7692,
+      "step": 16097
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2391908975979773e-06,
+      "loss": 3.1684,
+      "step": 16098
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.239143489254109e-06,
+      "loss": 3.0334,
+      "step": 16099
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2390960809102403e-06,
+      "loss": 2.4026,
+      "step": 16100
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2390486725663716e-06,
+      "loss": 3.8778,
+      "step": 16101
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2390012642225033e-06,
+      "loss": 3.5648,
+      "step": 16102
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2389538558786346e-06,
+      "loss": 3.8032,
+      "step": 16103
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.238906447534766e-06,
+      "loss": 3.0505,
+      "step": 16104
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2388590391908975e-06,
+      "loss": 2.3038,
+      "step": 16105
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2388116308470292e-06,
+      "loss": 2.2803,
+      "step": 16106
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2387642225031605e-06,
+      "loss": 2.4981,
+      "step": 16107
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2387168141592922e-06,
+      "loss": 2.0449,
+      "step": 16108
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2386694058154235e-06,
+      "loss": 2.6751,
+      "step": 16109
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.238621997471555e-06,
+      "loss": 2.8716,
+      "step": 16110
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2385745891276865e-06,
+      "loss": 3.247,
+      "step": 16111
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.238527180783818e-06,
+      "loss": 1.7409,
+      "step": 16112
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2384797724399495e-06,
+      "loss": 2.554,
+      "step": 16113
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.238432364096081e-06,
+      "loss": 2.3377,
+      "step": 16114
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2383849557522125e-06,
+      "loss": 2.7554,
+      "step": 16115
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2383375474083437e-06,
+      "loss": 3.1631,
+      "step": 16116
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.2382901390644754e-06,
+      "loss": 3.4975,
+      "step": 16117
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2382427307206067e-06,
+      "loss": 3.7375,
+      "step": 16118
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2381953223767384e-06,
+      "loss": 3.6501,
+      "step": 16119
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.23814791403287e-06,
+      "loss": 2.9981,
+      "step": 16120
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2381005056890014e-06,
+      "loss": 2.6708,
+      "step": 16121
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2380530973451327e-06,
+      "loss": 2.9983,
+      "step": 16122
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.238005689001264e-06,
+      "loss": 3.3469,
+      "step": 16123
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2379582806573957e-06,
+      "loss": 3.0858,
+      "step": 16124
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2379108723135274e-06,
+      "loss": 2.2301,
+      "step": 16125
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2378634639696587e-06,
+      "loss": 2.9683,
+      "step": 16126
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2378160556257904e-06,
+      "loss": 2.8124,
+      "step": 16127
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2377686472819217e-06,
+      "loss": 2.5599,
+      "step": 16128
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237721238938053e-06,
+      "loss": 3.3944,
+      "step": 16129
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2376738305941846e-06,
+      "loss": 3.48,
+      "step": 16130
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237626422250316e-06,
+      "loss": 3.0964,
+      "step": 16131
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2375790139064476e-06,
+      "loss": 3.2788,
+      "step": 16132
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2375316055625793e-06,
+      "loss": 1.8663,
+      "step": 16133
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2374841972187106e-06,
+      "loss": 2.2801,
+      "step": 16134
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237436788874842e-06,
+      "loss": 2.8542,
+      "step": 16135
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2373893805309736e-06,
+      "loss": 2.6906,
+      "step": 16136
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237341972187105e-06,
+      "loss": 3.4648,
+      "step": 16137
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2372945638432366e-06,
+      "loss": 1.9771,
+      "step": 16138
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2372471554993683e-06,
+      "loss": 2.2291,
+      "step": 16139
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2371997471554996e-06,
+      "loss": 3.7248,
+      "step": 16140
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237152338811631e-06,
+      "loss": 2.2901,
+      "step": 16141
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2371049304677625e-06,
+      "loss": 2.3072,
+      "step": 16142
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237057522123894e-06,
+      "loss": 2.3977,
+      "step": 16143
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.237010113780025e-06,
+      "loss": 3.0918,
+      "step": 16144
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236962705436157e-06,
+      "loss": 3.0577,
+      "step": 16145
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2369152970922885e-06,
+      "loss": 2.9507,
+      "step": 16146
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.23686788874842e-06,
+      "loss": 3.8816,
+      "step": 16147
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236820480404551e-06,
+      "loss": 3.5743,
+      "step": 16148
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2367730720606828e-06,
+      "loss": 2.1137,
+      "step": 16149
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236725663716814e-06,
+      "loss": 3.3506,
+      "step": 16150
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2366782553729458e-06,
+      "loss": 2.256,
+      "step": 16151
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2366308470290775e-06,
+      "loss": 2.449,
+      "step": 16152
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2365834386852087e-06,
+      "loss": 2.5605,
+      "step": 16153
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.23653603034134e-06,
+      "loss": 3.6466,
+      "step": 16154
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2364886219974717e-06,
+      "loss": 1.8252,
+      "step": 16155
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236441213653603e-06,
+      "loss": 2.8575,
+      "step": 16156
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2363938053097343e-06,
+      "loss": 1.7914,
+      "step": 16157
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236346396965866e-06,
+      "loss": 2.7466,
+      "step": 16158
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2362989886219977e-06,
+      "loss": 3.3005,
+      "step": 16159
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236251580278129e-06,
+      "loss": 3.4797,
+      "step": 16160
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2362041719342607e-06,
+      "loss": 2.9074,
+      "step": 16161
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236156763590392e-06,
+      "loss": 2.7778,
+      "step": 16162
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2361093552465232e-06,
+      "loss": 4.1527,
+      "step": 16163
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.236061946902655e-06,
+      "loss": 2.4014,
+      "step": 16164
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2360145385587866e-06,
+      "loss": 3.5402,
+      "step": 16165
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235967130214918e-06,
+      "loss": 4.3306,
+      "step": 16166
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235919721871049e-06,
+      "loss": 3.1199,
+      "step": 16167
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235872313527181e-06,
+      "loss": 3.7328,
+      "step": 16168
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235824905183312e-06,
+      "loss": 2.9608,
+      "step": 16169
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2357774968394435e-06,
+      "loss": 3.048,
+      "step": 16170
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235730088495575e-06,
+      "loss": 2.638,
+      "step": 16171
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235682680151707e-06,
+      "loss": 3.5525,
+      "step": 16172
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235635271807838e-06,
+      "loss": 4.5043,
+      "step": 16173
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.23558786346397e-06,
+      "loss": 3.1068,
+      "step": 16174
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235540455120101e-06,
+      "loss": 3.4052,
+      "step": 16175
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2354930467762324e-06,
+      "loss": 2.5825,
+      "step": 16176
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235445638432364e-06,
+      "loss": 3.5987,
+      "step": 16177
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235398230088496e-06,
+      "loss": 3.1539,
+      "step": 16178
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235350821744627e-06,
+      "loss": 2.4996,
+      "step": 16179
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.235303413400759e-06,
+      "loss": 2.9083,
+      "step": 16180
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.23525600505689e-06,
+      "loss": 3.4754,
+      "step": 16181
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2352085967130214e-06,
+      "loss": 2.9823,
+      "step": 16182
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.235161188369153e-06,
+      "loss": 3.3031,
+      "step": 16183
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2351137800252844e-06,
+      "loss": 2.8985,
+      "step": 16184
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.235066371681416e-06,
+      "loss": 2.7624,
+      "step": 16185
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2350189633375478e-06,
+      "loss": 2.6084,
+      "step": 16186
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234971554993679e-06,
+      "loss": 2.4893,
+      "step": 16187
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2349241466498103e-06,
+      "loss": 2.6351,
+      "step": 16188
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2348767383059416e-06,
+      "loss": 3.8288,
+      "step": 16189
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2348293299620733e-06,
+      "loss": 3.6232,
+      "step": 16190
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234781921618205e-06,
+      "loss": 3.1395,
+      "step": 16191
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2347345132743363e-06,
+      "loss": 2.7398,
+      "step": 16192
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234687104930468e-06,
+      "loss": 2.0031,
+      "step": 16193
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2346396965865993e-06,
+      "loss": 3.8461,
+      "step": 16194
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2345922882427306e-06,
+      "loss": 2.6567,
+      "step": 16195
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2345448798988623e-06,
+      "loss": 2.233,
+      "step": 16196
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2344974715549936e-06,
+      "loss": 3.2553,
+      "step": 16197
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2344500632111253e-06,
+      "loss": 3.0856,
+      "step": 16198
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234402654867257e-06,
+      "loss": 3.3481,
+      "step": 16199
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2343552465233882e-06,
+      "loss": 3.2704,
+      "step": 16200
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2343078381795195e-06,
+      "loss": 2.4435,
+      "step": 16201
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2342604298356512e-06,
+      "loss": 2.3469,
+      "step": 16202
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2342130214917825e-06,
+      "loss": 2.7302,
+      "step": 16203
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234165613147914e-06,
+      "loss": 3.3973,
+      "step": 16204
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234118204804046e-06,
+      "loss": 3.2406,
+      "step": 16205
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.234070796460177e-06,
+      "loss": 3.8539,
+      "step": 16206
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2340233881163085e-06,
+      "loss": 2.1152,
+      "step": 16207
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.23397597977244e-06,
+      "loss": 3.0315,
+      "step": 16208
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2339285714285715e-06,
+      "loss": 1.9296,
+      "step": 16209
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2338811630847027e-06,
+      "loss": 2.939,
+      "step": 16210
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2338337547408344e-06,
+      "loss": 3.4609,
+      "step": 16211
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.233786346396966e-06,
+      "loss": 3.1959,
+      "step": 16212
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2337389380530974e-06,
+      "loss": 2.9811,
+      "step": 16213
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2336915297092287e-06,
+      "loss": 2.7748,
+      "step": 16214
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2336441213653604e-06,
+      "loss": 2.8043,
+      "step": 16215
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2335967130214917e-06,
+      "loss": 1.7542,
+      "step": 16216
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2335493046776234e-06,
+      "loss": 1.9129,
+      "step": 16217
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.233501896333755e-06,
+      "loss": 4.0357,
+      "step": 16218
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2334544879898864e-06,
+      "loss": 3.9828,
+      "step": 16219
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2334070796460177e-06,
+      "loss": 3.1736,
+      "step": 16220
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2333596713021494e-06,
+      "loss": 3.1082,
+      "step": 16221
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2333122629582806e-06,
+      "loss": 2.7774,
+      "step": 16222
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.233264854614412e-06,
+      "loss": 3.334,
+      "step": 16223
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2332174462705436e-06,
+      "loss": 3.759,
+      "step": 16224
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2331700379266753e-06,
+      "loss": 2.1797,
+      "step": 16225
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2331226295828066e-06,
+      "loss": 2.1332,
+      "step": 16226
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2330752212389383e-06,
+      "loss": 2.8574,
+      "step": 16227
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2330278128950696e-06,
+      "loss": 2.698,
+      "step": 16228
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.232980404551201e-06,
+      "loss": 4.0269,
+      "step": 16229
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2329329962073326e-06,
+      "loss": 3.2394,
+      "step": 16230
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2328855878634643e-06,
+      "loss": 2.3482,
+      "step": 16231
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2328381795195956e-06,
+      "loss": 2.5288,
+      "step": 16232
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.232790771175727e-06,
+      "loss": 3.2173,
+      "step": 16233
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2327433628318586e-06,
+      "loss": 2.682,
+      "step": 16234
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.23269595448799e-06,
+      "loss": 2.5613,
+      "step": 16235
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.232648546144121e-06,
+      "loss": 3.2674,
+      "step": 16236
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.232601137800253e-06,
+      "loss": 3.2385,
+      "step": 16237
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2325537294563845e-06,
+      "loss": 2.874,
+      "step": 16238
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.232506321112516e-06,
+      "loss": 1.8302,
+      "step": 16239
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2324589127686475e-06,
+      "loss": 3.6344,
+      "step": 16240
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2324115044247788e-06,
+      "loss": 3.3539,
+      "step": 16241
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.23236409608091e-06,
+      "loss": 3.2119,
+      "step": 16242
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2323166877370418e-06,
+      "loss": 2.084,
+      "step": 16243
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.2322692793931735e-06,
+      "loss": 2.3414,
+      "step": 16244
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2322218710493048e-06,
+      "loss": 3.5891,
+      "step": 16245
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2321744627054365e-06,
+      "loss": 2.7854,
+      "step": 16246
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2321270543615677e-06,
+      "loss": 3.9012,
+      "step": 16247
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.232079646017699e-06,
+      "loss": 2.1455,
+      "step": 16248
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2320322376738307e-06,
+      "loss": 3.1582,
+      "step": 16249
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231984829329962e-06,
+      "loss": 4.1825,
+      "step": 16250
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2319374209860937e-06,
+      "loss": 2.4785,
+      "step": 16251
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2318900126422254e-06,
+      "loss": 2.4551,
+      "step": 16252
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2318426042983567e-06,
+      "loss": 2.9093,
+      "step": 16253
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231795195954488e-06,
+      "loss": 2.459,
+      "step": 16254
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2317477876106193e-06,
+      "loss": 2.9661,
+      "step": 16255
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231700379266751e-06,
+      "loss": 4.2045,
+      "step": 16256
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2316529709228827e-06,
+      "loss": 2.3728,
+      "step": 16257
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231605562579014e-06,
+      "loss": 2.452,
+      "step": 16258
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2315581542351456e-06,
+      "loss": 3.3921,
+      "step": 16259
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231510745891277e-06,
+      "loss": 3.3446,
+      "step": 16260
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231463337547408e-06,
+      "loss": 2.9305,
+      "step": 16261
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.23141592920354e-06,
+      "loss": 1.8328,
+      "step": 16262
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231368520859671e-06,
+      "loss": 3.843,
+      "step": 16263
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231321112515803e-06,
+      "loss": 3.1701,
+      "step": 16264
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2312737041719346e-06,
+      "loss": 3.1372,
+      "step": 16265
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231226295828066e-06,
+      "loss": 2.4801,
+      "step": 16266
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231178887484197e-06,
+      "loss": 2.5287,
+      "step": 16267
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231131479140329e-06,
+      "loss": 3.897,
+      "step": 16268
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.23108407079646e-06,
+      "loss": 1.995,
+      "step": 16269
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.231036662452592e-06,
+      "loss": 2.2482,
+      "step": 16270
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2309892541087235e-06,
+      "loss": 2.2546,
+      "step": 16271
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230941845764855e-06,
+      "loss": 3.6316,
+      "step": 16272
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230894437420986e-06,
+      "loss": 2.573,
+      "step": 16273
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230847029077118e-06,
+      "loss": 4.0704,
+      "step": 16274
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230799620733249e-06,
+      "loss": 4.3022,
+      "step": 16275
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2307522123893804e-06,
+      "loss": 2.6244,
+      "step": 16276
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230704804045512e-06,
+      "loss": 2.6685,
+      "step": 16277
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2306573957016438e-06,
+      "loss": 3.1596,
+      "step": 16278
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230609987357775e-06,
+      "loss": 2.5214,
+      "step": 16279
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2305625790139063e-06,
+      "loss": 2.9019,
+      "step": 16280
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230515170670038e-06,
+      "loss": 3.1091,
+      "step": 16281
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2304677623261693e-06,
+      "loss": 2.4511,
+      "step": 16282
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2304203539823006e-06,
+      "loss": 2.6478,
+      "step": 16283
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2303729456384327e-06,
+      "loss": 2.735,
+      "step": 16284
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230325537294564e-06,
+      "loss": 2.927,
+      "step": 16285
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2302781289506953e-06,
+      "loss": 2.8818,
+      "step": 16286
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230230720606827e-06,
+      "loss": 3.5894,
+      "step": 16287
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2301833122629583e-06,
+      "loss": 1.7368,
+      "step": 16288
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2301359039190896e-06,
+      "loss": 2.9823,
+      "step": 16289
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2300884955752213e-06,
+      "loss": 3.5891,
+      "step": 16290
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.230041087231353e-06,
+      "loss": 3.3111,
+      "step": 16291
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2299936788874842e-06,
+      "loss": 3.0996,
+      "step": 16292
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.229946270543616e-06,
+      "loss": 3.0026,
+      "step": 16293
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2298988621997472e-06,
+      "loss": 2.4672,
+      "step": 16294
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2298514538558785e-06,
+      "loss": 3.8428,
+      "step": 16295
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2298040455120102e-06,
+      "loss": 3.1699,
+      "step": 16296
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2297566371681415e-06,
+      "loss": 4.2619,
+      "step": 16297
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.229709228824273e-06,
+      "loss": 4.0297,
+      "step": 16298
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2296618204804045e-06,
+      "loss": 2.7548,
+      "step": 16299
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.229614412136536e-06,
+      "loss": 2.5913,
+      "step": 16300
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2295670037926675e-06,
+      "loss": 3.7046,
+      "step": 16301
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2295195954487987e-06,
+      "loss": 2.3647,
+      "step": 16302
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2294721871049305e-06,
+      "loss": 3.3516,
+      "step": 16303
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.229424778761062e-06,
+      "loss": 2.4003,
+      "step": 16304
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2293773704171934e-06,
+      "loss": 3.0297,
+      "step": 16305
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.229329962073325e-06,
+      "loss": 3.3084,
+      "step": 16306
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2292825537294564e-06,
+      "loss": 3.0029,
+      "step": 16307
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2292351453855877e-06,
+      "loss": 1.4478,
+      "step": 16308
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2291877370417194e-06,
+      "loss": 3.6106,
+      "step": 16309
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2291403286978507e-06,
+      "loss": 2.9711,
+      "step": 16310
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2290929203539824e-06,
+      "loss": 3.3288,
+      "step": 16311
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.229045512010114e-06,
+      "loss": 3.7897,
+      "step": 16312
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2289981036662454e-06,
+      "loss": 3.2976,
+      "step": 16313
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2289506953223767e-06,
+      "loss": 3.6507,
+      "step": 16314
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2289032869785084e-06,
+      "loss": 3.0422,
+      "step": 16315
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2288558786346396e-06,
+      "loss": 3.1161,
+      "step": 16316
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2288084702907713e-06,
+      "loss": 2.2314,
+      "step": 16317
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2287610619469026e-06,
+      "loss": 2.9054,
+      "step": 16318
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2287136536030343e-06,
+      "loss": 2.8971,
+      "step": 16319
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2286662452591656e-06,
+      "loss": 3.1474,
+      "step": 16320
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.228618836915297e-06,
+      "loss": 2.0203,
+      "step": 16321
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2285714285714286e-06,
+      "loss": 3.898,
+      "step": 16322
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.22852402022756e-06,
+      "loss": 3.2012,
+      "step": 16323
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2284766118836916e-06,
+      "loss": 3.454,
+      "step": 16324
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2284292035398233e-06,
+      "loss": 3.1726,
+      "step": 16325
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2283817951959546e-06,
+      "loss": 3.2094,
+      "step": 16326
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.228334386852086e-06,
+      "loss": 2.5099,
+      "step": 16327
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2282869785082175e-06,
+      "loss": 2.8798,
+      "step": 16328
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.228239570164349e-06,
+      "loss": 2.8813,
+      "step": 16329
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2281921618204805e-06,
+      "loss": 3.5943,
+      "step": 16330
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2281447534766122e-06,
+      "loss": 2.8167,
+      "step": 16331
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2280973451327435e-06,
+      "loss": 1.3343,
+      "step": 16332
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.228049936788875e-06,
+      "loss": 3.8036,
+      "step": 16333
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2280025284450065e-06,
+      "loss": 3.4679,
+      "step": 16334
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2279551201011378e-06,
+      "loss": 2.7402,
+      "step": 16335
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227907711757269e-06,
+      "loss": 3.0733,
+      "step": 16336
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2278603034134008e-06,
+      "loss": 3.2967,
+      "step": 16337
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2278128950695325e-06,
+      "loss": 2.343,
+      "step": 16338
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2277654867256637e-06,
+      "loss": 1.8829,
+      "step": 16339
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227718078381795e-06,
+      "loss": 3.3349,
+      "step": 16340
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2276706700379267e-06,
+      "loss": 2.4344,
+      "step": 16341
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227623261694058e-06,
+      "loss": 3.2014,
+      "step": 16342
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2275758533501897e-06,
+      "loss": 2.3564,
+      "step": 16343
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2275284450063214e-06,
+      "loss": 2.3261,
+      "step": 16344
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2274810366624527e-06,
+      "loss": 2.6968,
+      "step": 16345
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227433628318584e-06,
+      "loss": 3.2756,
+      "step": 16346
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2273862199747157e-06,
+      "loss": 3.8287,
+      "step": 16347
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227338811630847e-06,
+      "loss": 2.3241,
+      "step": 16348
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2272914032869782e-06,
+      "loss": 2.7066,
+      "step": 16349
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.22724399494311e-06,
+      "loss": 2.5386,
+      "step": 16350
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2271965865992417e-06,
+      "loss": 2.7761,
+      "step": 16351
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227149178255373e-06,
+      "loss": 2.9177,
+      "step": 16352
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2271017699115046e-06,
+      "loss": 2.7207,
+      "step": 16353
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227054361567636e-06,
+      "loss": 3.6475,
+      "step": 16354
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.227006953223767e-06,
+      "loss": 2.1787,
+      "step": 16355
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226959544879899e-06,
+      "loss": 3.0041,
+      "step": 16356
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2269121365360306e-06,
+      "loss": 2.5234,
+      "step": 16357
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226864728192162e-06,
+      "loss": 3.1977,
+      "step": 16358
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2268173198482936e-06,
+      "loss": 2.3452,
+      "step": 16359
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226769911504425e-06,
+      "loss": 3.2462,
+      "step": 16360
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226722503160556e-06,
+      "loss": 4.6153,
+      "step": 16361
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2266750948166874e-06,
+      "loss": 3.2383,
+      "step": 16362
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226627686472819e-06,
+      "loss": 2.0173,
+      "step": 16363
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226580278128951e-06,
+      "loss": 3.0019,
+      "step": 16364
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226532869785082e-06,
+      "loss": 3.2589,
+      "step": 16365
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226485461441214e-06,
+      "loss": 2.9188,
+      "step": 16366
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226438053097345e-06,
+      "loss": 3.4225,
+      "step": 16367
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.2263906447534764e-06,
+      "loss": 2.8824,
+      "step": 16368
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.226343236409608e-06,
+      "loss": 3.1487,
+      "step": 16369
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.22629582806574e-06,
+      "loss": 3.0046,
+      "step": 16370
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.226248419721871e-06,
+      "loss": 2.8919,
+      "step": 16371
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2262010113780028e-06,
+      "loss": 2.0743,
+      "step": 16372
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.226153603034134e-06,
+      "loss": 3.4749,
+      "step": 16373
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2261061946902653e-06,
+      "loss": 1.7862,
+      "step": 16374
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.226058786346397e-06,
+      "loss": 2.9139,
+      "step": 16375
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2260113780025283e-06,
+      "loss": 3.0064,
+      "step": 16376
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.22596396965866e-06,
+      "loss": 3.6828,
+      "step": 16377
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2259165613147917e-06,
+      "loss": 3.1464,
+      "step": 16378
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225869152970923e-06,
+      "loss": 2.4669,
+      "step": 16379
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2258217446270543e-06,
+      "loss": 2.4104,
+      "step": 16380
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225774336283186e-06,
+      "loss": 2.8083,
+      "step": 16381
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2257269279393173e-06,
+      "loss": 3.3526,
+      "step": 16382
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225679519595449e-06,
+      "loss": 2.2387,
+      "step": 16383
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2256321112515803e-06,
+      "loss": 2.3535,
+      "step": 16384
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225584702907712e-06,
+      "loss": 3.805,
+      "step": 16385
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2255372945638432e-06,
+      "loss": 3.9513,
+      "step": 16386
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2254898862199745e-06,
+      "loss": 3.3006,
+      "step": 16387
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2254424778761062e-06,
+      "loss": 1.9914,
+      "step": 16388
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2253950695322375e-06,
+      "loss": 2.6111,
+      "step": 16389
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2253476611883692e-06,
+      "loss": 2.4097,
+      "step": 16390
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225300252844501e-06,
+      "loss": 2.0581,
+      "step": 16391
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225252844500632e-06,
+      "loss": 3.6991,
+      "step": 16392
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2252054361567635e-06,
+      "loss": 3.4709,
+      "step": 16393
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225158027812895e-06,
+      "loss": 3.407,
+      "step": 16394
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2251106194690265e-06,
+      "loss": 2.8164,
+      "step": 16395
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.225063211125158e-06,
+      "loss": 3.3504,
+      "step": 16396
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.22501580278129e-06,
+      "loss": 3.0276,
+      "step": 16397
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.224968394437421e-06,
+      "loss": 1.5463,
+      "step": 16398
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2249209860935524e-06,
+      "loss": 3.1679,
+      "step": 16399
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.224873577749684e-06,
+      "loss": 3.3472,
+      "step": 16400
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2248261694058154e-06,
+      "loss": 3.4605,
+      "step": 16401
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2247787610619467e-06,
+      "loss": 3.046,
+      "step": 16402
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2247313527180784e-06,
+      "loss": 4.4397,
+      "step": 16403
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.22468394437421e-06,
+      "loss": 4.6288,
+      "step": 16404
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2246365360303414e-06,
+      "loss": 2.2403,
+      "step": 16405
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2245891276864727e-06,
+      "loss": 2.5905,
+      "step": 16406
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2245417193426044e-06,
+      "loss": 3.9221,
+      "step": 16407
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2244943109987356e-06,
+      "loss": 2.0415,
+      "step": 16408
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2244469026548674e-06,
+      "loss": 2.638,
+      "step": 16409
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.224399494310999e-06,
+      "loss": 3.5217,
+      "step": 16410
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2243520859671303e-06,
+      "loss": 2.0628,
+      "step": 16411
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2243046776232616e-06,
+      "loss": 2.4524,
+      "step": 16412
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2242572692793933e-06,
+      "loss": 3.11,
+      "step": 16413
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2242098609355246e-06,
+      "loss": 3.1145,
+      "step": 16414
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.224162452591656e-06,
+      "loss": 3.3119,
+      "step": 16415
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2241150442477876e-06,
+      "loss": 2.8694,
+      "step": 16416
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2240676359039193e-06,
+      "loss": 3.0693,
+      "step": 16417
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2240202275600506e-06,
+      "loss": 2.2082,
+      "step": 16418
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2239728192161823e-06,
+      "loss": 3.2677,
+      "step": 16419
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2239254108723136e-06,
+      "loss": 2.8883,
+      "step": 16420
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.223878002528445e-06,
+      "loss": 1.8645,
+      "step": 16421
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2238305941845765e-06,
+      "loss": 3.9541,
+      "step": 16422
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2237831858407082e-06,
+      "loss": 3.1762,
+      "step": 16423
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2237357774968395e-06,
+      "loss": 2.6075,
+      "step": 16424
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2236883691529712e-06,
+      "loss": 3.0652,
+      "step": 16425
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2236409608091025e-06,
+      "loss": 1.8368,
+      "step": 16426
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2235935524652338e-06,
+      "loss": 3.1886,
+      "step": 16427
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.223546144121365e-06,
+      "loss": 3.366,
+      "step": 16428
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2234987357774968e-06,
+      "loss": 2.6595,
+      "step": 16429
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2234513274336285e-06,
+      "loss": 3.763,
+      "step": 16430
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2234039190897598e-06,
+      "loss": 2.2184,
+      "step": 16431
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2233565107458915e-06,
+      "loss": 3.3565,
+      "step": 16432
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2233091024020227e-06,
+      "loss": 3.0867,
+      "step": 16433
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.223261694058154e-06,
+      "loss": 2.1593,
+      "step": 16434
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2232142857142857e-06,
+      "loss": 2.5815,
+      "step": 16435
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2231668773704174e-06,
+      "loss": 3.6669,
+      "step": 16436
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2231194690265487e-06,
+      "loss": 2.5252,
+      "step": 16437
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2230720606826804e-06,
+      "loss": 2.3122,
+      "step": 16438
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2230246523388117e-06,
+      "loss": 2.2643,
+      "step": 16439
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222977243994943e-06,
+      "loss": 3.7809,
+      "step": 16440
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2229298356510747e-06,
+      "loss": 4.3542,
+      "step": 16441
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222882427307206e-06,
+      "loss": 2.8155,
+      "step": 16442
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2228350189633377e-06,
+      "loss": 3.2833,
+      "step": 16443
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2227876106194694e-06,
+      "loss": 3.3613,
+      "step": 16444
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2227402022756006e-06,
+      "loss": 3.1089,
+      "step": 16445
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222692793931732e-06,
+      "loss": 3.0906,
+      "step": 16446
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2226453855878636e-06,
+      "loss": 3.6031,
+      "step": 16447
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222597977243995e-06,
+      "loss": 2.1004,
+      "step": 16448
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2225505689001266e-06,
+      "loss": 4.3252,
+      "step": 16449
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222503160556258e-06,
+      "loss": 3.3371,
+      "step": 16450
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2224557522123896e-06,
+      "loss": 2.6509,
+      "step": 16451
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222408343868521e-06,
+      "loss": 2.2008,
+      "step": 16452
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222360935524652e-06,
+      "loss": 3.3422,
+      "step": 16453
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222313527180784e-06,
+      "loss": 2.9801,
+      "step": 16454
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222266118836915e-06,
+      "loss": 3.3584,
+      "step": 16455
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222218710493047e-06,
+      "loss": 3.2545,
+      "step": 16456
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2221713021491786e-06,
+      "loss": 3.6651,
+      "step": 16457
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.22212389380531e-06,
+      "loss": 2.3788,
+      "step": 16458
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222076485461441e-06,
+      "loss": 2.8357,
+      "step": 16459
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.222029077117573e-06,
+      "loss": 2.4194,
+      "step": 16460
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221981668773704e-06,
+      "loss": 3.1034,
+      "step": 16461
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221934260429836e-06,
+      "loss": 3.2262,
+      "step": 16462
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2218868520859675e-06,
+      "loss": 3.4635,
+      "step": 16463
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2218394437420988e-06,
+      "loss": 3.6849,
+      "step": 16464
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.22179203539823e-06,
+      "loss": 2.6751,
+      "step": 16465
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2217446270543618e-06,
+      "loss": 2.5533,
+      "step": 16466
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221697218710493e-06,
+      "loss": 2.6048,
+      "step": 16467
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2216498103666243e-06,
+      "loss": 2.8594,
+      "step": 16468
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221602402022756e-06,
+      "loss": 3.4081,
+      "step": 16469
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2215549936788877e-06,
+      "loss": 3.2846,
+      "step": 16470
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221507585335019e-06,
+      "loss": 2.7629,
+      "step": 16471
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2214601769911503e-06,
+      "loss": 4.0966,
+      "step": 16472
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221412768647282e-06,
+      "loss": 2.3363,
+      "step": 16473
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2213653603034133e-06,
+      "loss": 1.5825,
+      "step": 16474
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221317951959545e-06,
+      "loss": 2.9063,
+      "step": 16475
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2212705436156767e-06,
+      "loss": 2.8704,
+      "step": 16476
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221223135271808e-06,
+      "loss": 2.4089,
+      "step": 16477
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2211757269279393e-06,
+      "loss": 3.4816,
+      "step": 16478
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.221128318584071e-06,
+      "loss": 2.6322,
+      "step": 16479
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2210809102402022e-06,
+      "loss": 2.5726,
+      "step": 16480
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2210335018963335e-06,
+      "loss": 2.2082,
+      "step": 16481
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2209860935524652e-06,
+      "loss": 4.128,
+      "step": 16482
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220938685208597e-06,
+      "loss": 3.0266,
+      "step": 16483
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220891276864728e-06,
+      "loss": 3.7298,
+      "step": 16484
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.22084386852086e-06,
+      "loss": 3.6233,
+      "step": 16485
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220796460176991e-06,
+      "loss": 3.5034,
+      "step": 16486
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2207490518331225e-06,
+      "loss": 3.919,
+      "step": 16487
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220701643489254e-06,
+      "loss": 1.8252,
+      "step": 16488
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220654235145386e-06,
+      "loss": 2.1773,
+      "step": 16489
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220606826801517e-06,
+      "loss": 3.0398,
+      "step": 16490
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220559418457649e-06,
+      "loss": 3.0167,
+      "step": 16491
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.22051201011378e-06,
+      "loss": 2.3348,
+      "step": 16492
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2204646017699114e-06,
+      "loss": 2.9616,
+      "step": 16493
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2204171934260427e-06,
+      "loss": 2.4946,
+      "step": 16494
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2203697850821744e-06,
+      "loss": 4.0833,
+      "step": 16495
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.220322376738306e-06,
+      "loss": 2.1627,
+      "step": 16496
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.2202749683944374e-06,
+      "loss": 1.9597,
+      "step": 16497
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.220227560050569e-06,
+      "loss": 3.1233,
+      "step": 16498
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2201801517067004e-06,
+      "loss": 3.3337,
+      "step": 16499
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2201327433628317e-06,
+      "loss": 2.5933,
+      "step": 16500
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2200853350189634e-06,
+      "loss": 2.6606,
+      "step": 16501
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.220037926675095e-06,
+      "loss": 2.7675,
+      "step": 16502
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2199905183312263e-06,
+      "loss": 3.3797,
+      "step": 16503
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.219943109987358e-06,
+      "loss": 2.8294,
+      "step": 16504
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2198957016434893e-06,
+      "loss": 3.3436,
+      "step": 16505
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2198482932996206e-06,
+      "loss": 3.3697,
+      "step": 16506
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2198008849557523e-06,
+      "loss": 3.7983,
+      "step": 16507
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2197534766118836e-06,
+      "loss": 2.5647,
+      "step": 16508
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2197060682680153e-06,
+      "loss": 3.2569,
+      "step": 16509
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.219658659924147e-06,
+      "loss": 3.6258,
+      "step": 16510
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2196112515802783e-06,
+      "loss": 2.9017,
+      "step": 16511
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2195638432364096e-06,
+      "loss": 3.8177,
+      "step": 16512
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2195164348925413e-06,
+      "loss": 2.8779,
+      "step": 16513
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2194690265486725e-06,
+      "loss": 2.4882,
+      "step": 16514
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2194216182048043e-06,
+      "loss": 3.227,
+      "step": 16515
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2193742098609355e-06,
+      "loss": 2.5204,
+      "step": 16516
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2193268015170672e-06,
+      "loss": 2.8291,
+      "step": 16517
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2192793931731985e-06,
+      "loss": 2.2151,
+      "step": 16518
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.21923198482933e-06,
+      "loss": 2.2002,
+      "step": 16519
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2191845764854615e-06,
+      "loss": 2.761,
+      "step": 16520
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2191371681415928e-06,
+      "loss": 2.1341,
+      "step": 16521
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2190897597977245e-06,
+      "loss": 2.7824,
+      "step": 16522
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.219042351453856e-06,
+      "loss": 1.3326,
+      "step": 16523
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2189949431099875e-06,
+      "loss": 3.3095,
+      "step": 16524
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2189475347661188e-06,
+      "loss": 2.3754,
+      "step": 16525
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2189001264222505e-06,
+      "loss": 3.178,
+      "step": 16526
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2188527180783817e-06,
+      "loss": 2.687,
+      "step": 16527
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2188053097345134e-06,
+      "loss": 3.6799,
+      "step": 16528
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.218757901390645e-06,
+      "loss": 2.7063,
+      "step": 16529
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2187104930467764e-06,
+      "loss": 2.8745,
+      "step": 16530
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2186630847029077e-06,
+      "loss": 3.3764,
+      "step": 16531
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2186156763590394e-06,
+      "loss": 2.2965,
+      "step": 16532
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2185682680151707e-06,
+      "loss": 3.8533,
+      "step": 16533
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.218520859671302e-06,
+      "loss": 4.0854,
+      "step": 16534
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2184734513274337e-06,
+      "loss": 3.3095,
+      "step": 16535
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2184260429835654e-06,
+      "loss": 2.9435,
+      "step": 16536
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2183786346396967e-06,
+      "loss": 2.734,
+      "step": 16537
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.218331226295828e-06,
+      "loss": 3.3455,
+      "step": 16538
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2182838179519596e-06,
+      "loss": 1.6159,
+      "step": 16539
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.218236409608091e-06,
+      "loss": 2.9399,
+      "step": 16540
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2181890012642226e-06,
+      "loss": 3.8385,
+      "step": 16541
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2181415929203543e-06,
+      "loss": 2.5285,
+      "step": 16542
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2180941845764856e-06,
+      "loss": 3.4775,
+      "step": 16543
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.218046776232617e-06,
+      "loss": 2.4878,
+      "step": 16544
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2179993678887486e-06,
+      "loss": 2.9371,
+      "step": 16545
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.21795195954488e-06,
+      "loss": 3.8527,
+      "step": 16546
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217904551201011e-06,
+      "loss": 3.9683,
+      "step": 16547
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217857142857143e-06,
+      "loss": 3.1107,
+      "step": 16548
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2178097345132746e-06,
+      "loss": 2.8098,
+      "step": 16549
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217762326169406e-06,
+      "loss": 3.1534,
+      "step": 16550
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2177149178255375e-06,
+      "loss": 3.48,
+      "step": 16551
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217667509481669e-06,
+      "loss": 2.5532,
+      "step": 16552
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2176201011378e-06,
+      "loss": 3.6296,
+      "step": 16553
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217572692793932e-06,
+      "loss": 2.1237,
+      "step": 16554
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2175252844500635e-06,
+      "loss": 3.1764,
+      "step": 16555
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217477876106195e-06,
+      "loss": 2.1231,
+      "step": 16556
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2174304677623265e-06,
+      "loss": 3.0935,
+      "step": 16557
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2173830594184578e-06,
+      "loss": 2.6129,
+      "step": 16558
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.217335651074589e-06,
+      "loss": 2.4973,
+      "step": 16559
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.2172882427307203e-06,
+      "loss": 3.5363,
+      "step": 16560
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.217240834386852e-06,
+      "loss": 2.9671,
+      "step": 16561
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2171934260429837e-06,
+      "loss": 3.5508,
+      "step": 16562
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.217146017699115e-06,
+      "loss": 2.4893,
+      "step": 16563
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2170986093552467e-06,
+      "loss": 2.5878,
+      "step": 16564
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.217051201011378e-06,
+      "loss": 1.575,
+      "step": 16565
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2170037926675093e-06,
+      "loss": 3.163,
+      "step": 16566
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216956384323641e-06,
+      "loss": 1.9263,
+      "step": 16567
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2169089759797723e-06,
+      "loss": 2.4929,
+      "step": 16568
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216861567635904e-06,
+      "loss": 3.4503,
+      "step": 16569
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2168141592920357e-06,
+      "loss": 4.7,
+      "step": 16570
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216766750948167e-06,
+      "loss": 2.7817,
+      "step": 16571
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2167193426042982e-06,
+      "loss": 2.8776,
+      "step": 16572
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.21667193426043e-06,
+      "loss": 3.0827,
+      "step": 16573
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2166245259165612e-06,
+      "loss": 3.4069,
+      "step": 16574
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216577117572693e-06,
+      "loss": 2.6709,
+      "step": 16575
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2165297092288246e-06,
+      "loss": 3.2091,
+      "step": 16576
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216482300884956e-06,
+      "loss": 3.4123,
+      "step": 16577
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216434892541087e-06,
+      "loss": 1.7546,
+      "step": 16578
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216387484197219e-06,
+      "loss": 3.5355,
+      "step": 16579
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.21634007585335e-06,
+      "loss": 2.7967,
+      "step": 16580
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2162926675094815e-06,
+      "loss": 3.1637,
+      "step": 16581
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216245259165613e-06,
+      "loss": 2.5176,
+      "step": 16582
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216197850821745e-06,
+      "loss": 4.1887,
+      "step": 16583
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216150442477876e-06,
+      "loss": 3.608,
+      "step": 16584
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2161030341340074e-06,
+      "loss": 2.185,
+      "step": 16585
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.216055625790139e-06,
+      "loss": 3.4094,
+      "step": 16586
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2160082174462704e-06,
+      "loss": 2.26,
+      "step": 16587
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215960809102402e-06,
+      "loss": 2.6054,
+      "step": 16588
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215913400758534e-06,
+      "loss": 3.8384,
+      "step": 16589
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215865992414665e-06,
+      "loss": 2.8758,
+      "step": 16590
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2158185840707964e-06,
+      "loss": 2.5693,
+      "step": 16591
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215771175726928e-06,
+      "loss": 2.3334,
+      "step": 16592
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2157237673830594e-06,
+      "loss": 2.4075,
+      "step": 16593
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2156763590391907e-06,
+      "loss": 3.8052,
+      "step": 16594
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2156289506953224e-06,
+      "loss": 3.5111,
+      "step": 16595
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215581542351454e-06,
+      "loss": 3.2182,
+      "step": 16596
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2155341340075853e-06,
+      "loss": 3.1892,
+      "step": 16597
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215486725663717e-06,
+      "loss": 2.9248,
+      "step": 16598
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2154393173198483e-06,
+      "loss": 3.1159,
+      "step": 16599
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2153919089759796e-06,
+      "loss": 2.5822,
+      "step": 16600
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2153445006321113e-06,
+      "loss": 3.4696,
+      "step": 16601
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.215297092288243e-06,
+      "loss": 3.0184,
+      "step": 16602
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2152496839443743e-06,
+      "loss": 4.4215,
+      "step": 16603
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2152022756005056e-06,
+      "loss": 3.891,
+      "step": 16604
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2151548672566373e-06,
+      "loss": 3.7095,
+      "step": 16605
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2151074589127686e-06,
+      "loss": 3.1047,
+      "step": 16606
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2150600505689e-06,
+      "loss": 2.5711,
+      "step": 16607
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2150126422250315e-06,
+      "loss": 2.0992,
+      "step": 16608
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2149652338811632e-06,
+      "loss": 2.3588,
+      "step": 16609
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2149178255372945e-06,
+      "loss": 3.4441,
+      "step": 16610
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2148704171934262e-06,
+      "loss": 3.1102,
+      "step": 16611
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2148230088495575e-06,
+      "loss": 3.5414,
+      "step": 16612
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.214775600505689e-06,
+      "loss": 2.9481,
+      "step": 16613
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2147281921618205e-06,
+      "loss": 3.6654,
+      "step": 16614
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.214680783817952e-06,
+      "loss": 3.3707,
+      "step": 16615
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2146333754740835e-06,
+      "loss": 3.0648,
+      "step": 16616
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.214585967130215e-06,
+      "loss": 1.5836,
+      "step": 16617
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2145385587863465e-06,
+      "loss": 3.5945,
+      "step": 16618
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2144911504424777e-06,
+      "loss": 2.1109,
+      "step": 16619
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2144437420986094e-06,
+      "loss": 2.5948,
+      "step": 16620
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2143963337547407e-06,
+      "loss": 2.9032,
+      "step": 16621
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2143489254108724e-06,
+      "loss": 2.4284,
+      "step": 16622
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.214301517067004e-06,
+      "loss": 3.5716,
+      "step": 16623
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.2142541087231354e-06,
+      "loss": 2.902,
+      "step": 16624
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2142067003792667e-06,
+      "loss": 2.8678,
+      "step": 16625
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.214159292035398e-06,
+      "loss": 2.4335,
+      "step": 16626
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2141118836915297e-06,
+      "loss": 3.5662,
+      "step": 16627
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2140644753476614e-06,
+      "loss": 3.1955,
+      "step": 16628
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2140170670037927e-06,
+      "loss": 3.9614,
+      "step": 16629
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2139696586599244e-06,
+      "loss": 2.861,
+      "step": 16630
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2139222503160557e-06,
+      "loss": 2.2488,
+      "step": 16631
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213874841972187e-06,
+      "loss": 2.4088,
+      "step": 16632
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2138274336283186e-06,
+      "loss": 1.7351,
+      "step": 16633
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.21378002528445e-06,
+      "loss": 2.1996,
+      "step": 16634
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2137326169405816e-06,
+      "loss": 2.5447,
+      "step": 16635
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2136852085967133e-06,
+      "loss": 2.5105,
+      "step": 16636
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2136378002528446e-06,
+      "loss": 1.9991,
+      "step": 16637
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213590391908976e-06,
+      "loss": 2.9591,
+      "step": 16638
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2135429835651076e-06,
+      "loss": 2.1172,
+      "step": 16639
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213495575221239e-06,
+      "loss": 3.2041,
+      "step": 16640
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2134481668773706e-06,
+      "loss": 2.3912,
+      "step": 16641
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2134007585335023e-06,
+      "loss": 3.7915,
+      "step": 16642
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2133533501896336e-06,
+      "loss": 2.5076,
+      "step": 16643
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213305941845765e-06,
+      "loss": 1.9945,
+      "step": 16644
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2132585335018965e-06,
+      "loss": 2.3129,
+      "step": 16645
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213211125158028e-06,
+      "loss": 2.1898,
+      "step": 16646
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213163716814159e-06,
+      "loss": 2.9718,
+      "step": 16647
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213116308470291e-06,
+      "loss": 3.9684,
+      "step": 16648
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2130689001264225e-06,
+      "loss": 2.9489,
+      "step": 16649
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.213021491782554e-06,
+      "loss": 3.129,
+      "step": 16650
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212974083438685e-06,
+      "loss": 3.048,
+      "step": 16651
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2129266750948168e-06,
+      "loss": 2.8458,
+      "step": 16652
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212879266750948e-06,
+      "loss": 2.9409,
+      "step": 16653
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2128318584070798e-06,
+      "loss": 3.1571,
+      "step": 16654
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2127844500632115e-06,
+      "loss": 3.2885,
+      "step": 16655
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2127370417193427e-06,
+      "loss": 2.5925,
+      "step": 16656
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212689633375474e-06,
+      "loss": 3.4634,
+      "step": 16657
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2126422250316057e-06,
+      "loss": 2.1827,
+      "step": 16658
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212594816687737e-06,
+      "loss": 2.7877,
+      "step": 16659
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2125474083438683e-06,
+      "loss": 3.9512,
+      "step": 16660
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2125e-06,
+      "loss": 2.123,
+      "step": 16661
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2124525916561317e-06,
+      "loss": 3.0441,
+      "step": 16662
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212405183312263e-06,
+      "loss": 3.5419,
+      "step": 16663
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2123577749683947e-06,
+      "loss": 2.4551,
+      "step": 16664
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212310366624526e-06,
+      "loss": 3.3859,
+      "step": 16665
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2122629582806572e-06,
+      "loss": 2.9073,
+      "step": 16666
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212215549936789e-06,
+      "loss": 3.2254,
+      "step": 16667
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2121681415929206e-06,
+      "loss": 4.2385,
+      "step": 16668
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212120733249052e-06,
+      "loss": 4.4159,
+      "step": 16669
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212073324905183e-06,
+      "loss": 2.4884,
+      "step": 16670
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.212025916561315e-06,
+      "loss": 3.7039,
+      "step": 16671
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211978508217446e-06,
+      "loss": 1.8395,
+      "step": 16672
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2119310998735775e-06,
+      "loss": 2.8203,
+      "step": 16673
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211883691529709e-06,
+      "loss": 3.1103,
+      "step": 16674
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211836283185841e-06,
+      "loss": 3.366,
+      "step": 16675
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211788874841972e-06,
+      "loss": 3.4869,
+      "step": 16676
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211741466498104e-06,
+      "loss": 3.564,
+      "step": 16677
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211694058154235e-06,
+      "loss": 2.5026,
+      "step": 16678
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2116466498103664e-06,
+      "loss": 2.1523,
+      "step": 16679
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211599241466498e-06,
+      "loss": 2.5983,
+      "step": 16680
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.21155183312263e-06,
+      "loss": 2.9662,
+      "step": 16681
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211504424778761e-06,
+      "loss": 2.307,
+      "step": 16682
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211457016434893e-06,
+      "loss": 3.6395,
+      "step": 16683
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211409608091024e-06,
+      "loss": 4.4461,
+      "step": 16684
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2113621997471554e-06,
+      "loss": 2.7888,
+      "step": 16685
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.211314791403287e-06,
+      "loss": 3.2094,
+      "step": 16686
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.2112673830594184e-06,
+      "loss": 3.7997,
+      "step": 16687
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.21121997471555e-06,
+      "loss": 2.3041,
+      "step": 16688
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2111725663716818e-06,
+      "loss": 2.7575,
+      "step": 16689
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.211125158027813e-06,
+      "loss": 2.706,
+      "step": 16690
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2110777496839443e-06,
+      "loss": 2.0358,
+      "step": 16691
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2110303413400756e-06,
+      "loss": 2.6309,
+      "step": 16692
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2109829329962073e-06,
+      "loss": 3.0211,
+      "step": 16693
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.210935524652339e-06,
+      "loss": 3.4779,
+      "step": 16694
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2108881163084703e-06,
+      "loss": 2.6865,
+      "step": 16695
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.210840707964602e-06,
+      "loss": 2.814,
+      "step": 16696
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2107932996207333e-06,
+      "loss": 3.2103,
+      "step": 16697
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2107458912768646e-06,
+      "loss": 2.8863,
+      "step": 16698
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2106984829329963e-06,
+      "loss": 1.929,
+      "step": 16699
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2106510745891276e-06,
+      "loss": 2.8653,
+      "step": 16700
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2106036662452593e-06,
+      "loss": 3.2897,
+      "step": 16701
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.210556257901391e-06,
+      "loss": 1.662,
+      "step": 16702
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2105088495575222e-06,
+      "loss": 2.8728,
+      "step": 16703
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2104614412136535e-06,
+      "loss": 3.5261,
+      "step": 16704
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2104140328697852e-06,
+      "loss": 3.7676,
+      "step": 16705
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2103666245259165e-06,
+      "loss": 3.7356,
+      "step": 16706
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.210319216182048e-06,
+      "loss": 3.0503,
+      "step": 16707
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.21027180783818e-06,
+      "loss": 2.9676,
+      "step": 16708
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.210224399494311e-06,
+      "loss": 2.6443,
+      "step": 16709
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2101769911504425e-06,
+      "loss": 3.178,
+      "step": 16710
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.210129582806574e-06,
+      "loss": 3.6197,
+      "step": 16711
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2100821744627055e-06,
+      "loss": 3.4041,
+      "step": 16712
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2100347661188367e-06,
+      "loss": 3.4355,
+      "step": 16713
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2099873577749684e-06,
+      "loss": 2.9649,
+      "step": 16714
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2099399494311e-06,
+      "loss": 1.9219,
+      "step": 16715
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2098925410872314e-06,
+      "loss": 3.4084,
+      "step": 16716
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2098451327433627e-06,
+      "loss": 3.4388,
+      "step": 16717
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2097977243994944e-06,
+      "loss": 2.856,
+      "step": 16718
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2097503160556257e-06,
+      "loss": 2.4165,
+      "step": 16719
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2097029077117574e-06,
+      "loss": 4.442,
+      "step": 16720
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.209655499367889e-06,
+      "loss": 2.2323,
+      "step": 16721
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2096080910240204e-06,
+      "loss": 3.2361,
+      "step": 16722
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2095606826801517e-06,
+      "loss": 3.5981,
+      "step": 16723
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2095132743362834e-06,
+      "loss": 3.1041,
+      "step": 16724
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2094658659924146e-06,
+      "loss": 2.8099,
+      "step": 16725
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.209418457648546e-06,
+      "loss": 3.2272,
+      "step": 16726
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2093710493046776e-06,
+      "loss": 3.6189,
+      "step": 16727
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2093236409608093e-06,
+      "loss": 3.0615,
+      "step": 16728
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2092762326169406e-06,
+      "loss": 2.8511,
+      "step": 16729
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2092288242730723e-06,
+      "loss": 3.3637,
+      "step": 16730
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2091814159292036e-06,
+      "loss": 2.5953,
+      "step": 16731
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.209134007585335e-06,
+      "loss": 5.0262,
+      "step": 16732
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2090865992414666e-06,
+      "loss": 3.2556,
+      "step": 16733
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2090391908975983e-06,
+      "loss": 2.0834,
+      "step": 16734
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2089917825537296e-06,
+      "loss": 3.5641,
+      "step": 16735
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.208944374209861e-06,
+      "loss": 2.9299,
+      "step": 16736
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2088969658659926e-06,
+      "loss": 2.1188,
+      "step": 16737
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.208849557522124e-06,
+      "loss": 2.8499,
+      "step": 16738
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.208802149178255e-06,
+      "loss": 3.9749,
+      "step": 16739
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.208754740834387e-06,
+      "loss": 2.4406,
+      "step": 16740
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2087073324905185e-06,
+      "loss": 2.5493,
+      "step": 16741
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.20865992414665e-06,
+      "loss": 2.4231,
+      "step": 16742
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2086125158027815e-06,
+      "loss": 1.4818,
+      "step": 16743
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2085651074589128e-06,
+      "loss": 4.0888,
+      "step": 16744
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.208517699115044e-06,
+      "loss": 2.0543,
+      "step": 16745
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2084702907711758e-06,
+      "loss": 2.9069,
+      "step": 16746
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2084228824273075e-06,
+      "loss": 2.5558,
+      "step": 16747
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2083754740834388e-06,
+      "loss": 3.8935,
+      "step": 16748
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2083280657395705e-06,
+      "loss": 3.0924,
+      "step": 16749
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2082806573957017e-06,
+      "loss": 2.6174,
+      "step": 16750
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.208233249051833e-06,
+      "loss": 2.3552,
+      "step": 16751
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2081858407079647e-06,
+      "loss": 2.5143,
+      "step": 16752
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.208138432364096e-06,
+      "loss": 3.0606,
+      "step": 16753
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2080910240202277e-06,
+      "loss": 3.902,
+      "step": 16754
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.208043615676359e-06,
+      "loss": 2.7183,
+      "step": 16755
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2079962073324907e-06,
+      "loss": 4.007,
+      "step": 16756
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207948798988622e-06,
+      "loss": 2.8888,
+      "step": 16757
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2079013906447533e-06,
+      "loss": 2.6662,
+      "step": 16758
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207853982300885e-06,
+      "loss": 2.0488,
+      "step": 16759
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2078065739570167e-06,
+      "loss": 2.7859,
+      "step": 16760
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207759165613148e-06,
+      "loss": 2.6012,
+      "step": 16761
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2077117572692796e-06,
+      "loss": 2.8015,
+      "step": 16762
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207664348925411e-06,
+      "loss": 3.2766,
+      "step": 16763
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207616940581542e-06,
+      "loss": 2.6149,
+      "step": 16764
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207569532237674e-06,
+      "loss": 2.721,
+      "step": 16765
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207522123893805e-06,
+      "loss": 2.9197,
+      "step": 16766
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207474715549937e-06,
+      "loss": 2.4961,
+      "step": 16767
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2074273072060686e-06,
+      "loss": 3.741,
+      "step": 16768
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2073798988622e-06,
+      "loss": 3.3099,
+      "step": 16769
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207332490518331e-06,
+      "loss": 3.5611,
+      "step": 16770
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207285082174463e-06,
+      "loss": 4.0697,
+      "step": 16771
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207237673830594e-06,
+      "loss": 2.6846,
+      "step": 16772
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207190265486726e-06,
+      "loss": 2.4166,
+      "step": 16773
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2071428571428575e-06,
+      "loss": 3.4638,
+      "step": 16774
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.207095448798989e-06,
+      "loss": 2.5359,
+      "step": 16775
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.20704804045512e-06,
+      "loss": 2.6923,
+      "step": 16776
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2070006321112514e-06,
+      "loss": 2.3983,
+      "step": 16777
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206953223767383e-06,
+      "loss": 2.3737,
+      "step": 16778
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2069058154235144e-06,
+      "loss": 3.6737,
+      "step": 16779
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206858407079646e-06,
+      "loss": 3.2131,
+      "step": 16780
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2068109987357778e-06,
+      "loss": 3.1574,
+      "step": 16781
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206763590391909e-06,
+      "loss": 3.2296,
+      "step": 16782
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2067161820480403e-06,
+      "loss": 3.0246,
+      "step": 16783
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206668773704172e-06,
+      "loss": 3.3376,
+      "step": 16784
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2066213653603033e-06,
+      "loss": 2.8723,
+      "step": 16785
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206573957016435e-06,
+      "loss": 2.8662,
+      "step": 16786
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2065265486725667e-06,
+      "loss": 3.5355,
+      "step": 16787
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206479140328698e-06,
+      "loss": 2.5579,
+      "step": 16788
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2064317319848293e-06,
+      "loss": 3.3209,
+      "step": 16789
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206384323640961e-06,
+      "loss": 2.1599,
+      "step": 16790
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2063369152970923e-06,
+      "loss": 2.0096,
+      "step": 16791
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2062895069532236e-06,
+      "loss": 2.0108,
+      "step": 16792
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2062420986093553e-06,
+      "loss": 2.8527,
+      "step": 16793
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.206194690265487e-06,
+      "loss": 3.2997,
+      "step": 16794
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2061472819216183e-06,
+      "loss": 2.6134,
+      "step": 16795
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.20609987357775e-06,
+      "loss": 3.1557,
+      "step": 16796
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2060524652338812e-06,
+      "loss": 2.8527,
+      "step": 16797
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2060050568900125e-06,
+      "loss": 4.267,
+      "step": 16798
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2059576485461442e-06,
+      "loss": 3.2051,
+      "step": 16799
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.205910240202276e-06,
+      "loss": 3.1226,
+      "step": 16800
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.205862831858407e-06,
+      "loss": 1.6747,
+      "step": 16801
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2058154235145385e-06,
+      "loss": 3.0788,
+      "step": 16802
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.20576801517067e-06,
+      "loss": 2.7747,
+      "step": 16803
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2057206068268015e-06,
+      "loss": 2.3876,
+      "step": 16804
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2056731984829327e-06,
+      "loss": 3.6866,
+      "step": 16805
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2056257901390645e-06,
+      "loss": 3.7238,
+      "step": 16806
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.205578381795196e-06,
+      "loss": 2.7459,
+      "step": 16807
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2055309734513274e-06,
+      "loss": 2.9963,
+      "step": 16808
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.205483565107459e-06,
+      "loss": 3.4498,
+      "step": 16809
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2054361567635904e-06,
+      "loss": 2.4716,
+      "step": 16810
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2053887484197217e-06,
+      "loss": 4.1794,
+      "step": 16811
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2053413400758534e-06,
+      "loss": 3.5481,
+      "step": 16812
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.205293931731985e-06,
+      "loss": 1.8895,
+      "step": 16813
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.2052465233881164e-06,
+      "loss": 3.6344,
+      "step": 16814
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.205199115044248e-06,
+      "loss": 3.4964,
+      "step": 16815
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2051517067003794e-06,
+      "loss": 2.4453,
+      "step": 16816
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2051042983565107e-06,
+      "loss": 3.768,
+      "step": 16817
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2050568900126424e-06,
+      "loss": 1.8621,
+      "step": 16818
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2050094816687736e-06,
+      "loss": 2.8294,
+      "step": 16819
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2049620733249053e-06,
+      "loss": 2.9606,
+      "step": 16820
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2049146649810366e-06,
+      "loss": 2.8198,
+      "step": 16821
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2048672566371683e-06,
+      "loss": 3.9262,
+      "step": 16822
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2048198482932996e-06,
+      "loss": 3.1885,
+      "step": 16823
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.204772439949431e-06,
+      "loss": 2.4088,
+      "step": 16824
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2047250316055626e-06,
+      "loss": 1.6996,
+      "step": 16825
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2046776232616943e-06,
+      "loss": 3.1436,
+      "step": 16826
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2046302149178256e-06,
+      "loss": 2.8448,
+      "step": 16827
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2045828065739573e-06,
+      "loss": 2.9108,
+      "step": 16828
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2045353982300886e-06,
+      "loss": 2.9931,
+      "step": 16829
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.20448798988622e-06,
+      "loss": 2.504,
+      "step": 16830
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2044405815423515e-06,
+      "loss": 3.7081,
+      "step": 16831
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.204393173198483e-06,
+      "loss": 2.7137,
+      "step": 16832
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2043457648546145e-06,
+      "loss": 2.8823,
+      "step": 16833
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2042983565107462e-06,
+      "loss": 4.2791,
+      "step": 16834
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2042509481668775e-06,
+      "loss": 2.6221,
+      "step": 16835
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.204203539823009e-06,
+      "loss": 2.2468,
+      "step": 16836
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2041561314791405e-06,
+      "loss": 2.4596,
+      "step": 16837
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2041087231352718e-06,
+      "loss": 2.3172,
+      "step": 16838
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.204061314791403e-06,
+      "loss": 3.7087,
+      "step": 16839
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.204013906447535e-06,
+      "loss": 3.4882,
+      "step": 16840
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2039664981036665e-06,
+      "loss": 2.5683,
+      "step": 16841
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2039190897597977e-06,
+      "loss": 2.9707,
+      "step": 16842
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203871681415929e-06,
+      "loss": 3.1416,
+      "step": 16843
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2038242730720607e-06,
+      "loss": 3.6408,
+      "step": 16844
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203776864728192e-06,
+      "loss": 3.3838,
+      "step": 16845
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2037294563843237e-06,
+      "loss": 3.48,
+      "step": 16846
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2036820480404554e-06,
+      "loss": 3.2772,
+      "step": 16847
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2036346396965867e-06,
+      "loss": 2.7957,
+      "step": 16848
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203587231352718e-06,
+      "loss": 3.0137,
+      "step": 16849
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2035398230088497e-06,
+      "loss": 3.3262,
+      "step": 16850
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203492414664981e-06,
+      "loss": 3.6644,
+      "step": 16851
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2034450063211122e-06,
+      "loss": 3.0261,
+      "step": 16852
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203397597977244e-06,
+      "loss": 2.1907,
+      "step": 16853
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2033501896333757e-06,
+      "loss": 2.4726,
+      "step": 16854
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203302781289507e-06,
+      "loss": 3.0473,
+      "step": 16855
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2032553729456386e-06,
+      "loss": 2.5779,
+      "step": 16856
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.20320796460177e-06,
+      "loss": 3.4015,
+      "step": 16857
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203160556257901e-06,
+      "loss": 3.9474,
+      "step": 16858
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203113147914033e-06,
+      "loss": 2.5047,
+      "step": 16859
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2030657395701646e-06,
+      "loss": 2.8131,
+      "step": 16860
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.203018331226296e-06,
+      "loss": 2.0969,
+      "step": 16861
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2029709228824276e-06,
+      "loss": 3.2794,
+      "step": 16862
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202923514538559e-06,
+      "loss": 3.0927,
+      "step": 16863
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.20287610619469e-06,
+      "loss": 3.3987,
+      "step": 16864
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2028286978508214e-06,
+      "loss": 2.8513,
+      "step": 16865
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202781289506953e-06,
+      "loss": 2.8909,
+      "step": 16866
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202733881163085e-06,
+      "loss": 3.7443,
+      "step": 16867
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202686472819216e-06,
+      "loss": 2.809,
+      "step": 16868
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202639064475348e-06,
+      "loss": 4.1156,
+      "step": 16869
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202591656131479e-06,
+      "loss": 2.1024,
+      "step": 16870
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2025442477876104e-06,
+      "loss": 2.0495,
+      "step": 16871
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202496839443742e-06,
+      "loss": 1.9331,
+      "step": 16872
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202449431099874e-06,
+      "loss": 2.629,
+      "step": 16873
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202402022756005e-06,
+      "loss": 2.8598,
+      "step": 16874
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2023546144121368e-06,
+      "loss": 2.257,
+      "step": 16875
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.202307206068268e-06,
+      "loss": 2.1085,
+      "step": 16876
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2022597977243993e-06,
+      "loss": 2.5273,
+      "step": 16877
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.202212389380531e-06,
+      "loss": 2.9135,
+      "step": 16878
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2021649810366623e-06,
+      "loss": 3.6756,
+      "step": 16879
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.202117572692794e-06,
+      "loss": 3.9371,
+      "step": 16880
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2020701643489257e-06,
+      "loss": 3.696,
+      "step": 16881
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.202022756005057e-06,
+      "loss": 3.5508,
+      "step": 16882
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2019753476611883e-06,
+      "loss": 3.6452,
+      "step": 16883
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.20192793931732e-06,
+      "loss": 3.2649,
+      "step": 16884
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2018805309734513e-06,
+      "loss": 3.2576,
+      "step": 16885
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201833122629583e-06,
+      "loss": 3.0343,
+      "step": 16886
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2017857142857143e-06,
+      "loss": 2.6656,
+      "step": 16887
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201738305941846e-06,
+      "loss": 3.1983,
+      "step": 16888
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2016908975979772e-06,
+      "loss": 2.907,
+      "step": 16889
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2016434892541085e-06,
+      "loss": 2.9864,
+      "step": 16890
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2015960809102402e-06,
+      "loss": 2.8003,
+      "step": 16891
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2015486725663715e-06,
+      "loss": 2.1244,
+      "step": 16892
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2015012642225032e-06,
+      "loss": 3.3328,
+      "step": 16893
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201453855878635e-06,
+      "loss": 2.5735,
+      "step": 16894
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201406447534766e-06,
+      "loss": 2.4609,
+      "step": 16895
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2013590391908975e-06,
+      "loss": 2.5755,
+      "step": 16896
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201311630847029e-06,
+      "loss": 3.5403,
+      "step": 16897
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2012642225031605e-06,
+      "loss": 2.9688,
+      "step": 16898
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201216814159292e-06,
+      "loss": 3.3751,
+      "step": 16899
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201169405815424e-06,
+      "loss": 2.8795,
+      "step": 16900
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201121997471555e-06,
+      "loss": 2.7159,
+      "step": 16901
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2010745891276864e-06,
+      "loss": 2.8521,
+      "step": 16902
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201027180783818e-06,
+      "loss": 3.5,
+      "step": 16903
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2009797724399494e-06,
+      "loss": 3.3059,
+      "step": 16904
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2009323640960807e-06,
+      "loss": 2.6511,
+      "step": 16905
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2008849557522124e-06,
+      "loss": 2.6747,
+      "step": 16906
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.200837547408344e-06,
+      "loss": 2.324,
+      "step": 16907
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2007901390644754e-06,
+      "loss": 2.4805,
+      "step": 16908
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2007427307206067e-06,
+      "loss": 3.6383,
+      "step": 16909
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2006953223767384e-06,
+      "loss": 2.2034,
+      "step": 16910
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2006479140328696e-06,
+      "loss": 2.5096,
+      "step": 16911
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2006005056890014e-06,
+      "loss": 3.725,
+      "step": 16912
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.200553097345133e-06,
+      "loss": 2.7373,
+      "step": 16913
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2005056890012643e-06,
+      "loss": 2.5884,
+      "step": 16914
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2004582806573956e-06,
+      "loss": 3.6744,
+      "step": 16915
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2004108723135273e-06,
+      "loss": 2.3631,
+      "step": 16916
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2003634639696586e-06,
+      "loss": 1.8986,
+      "step": 16917
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.20031605562579e-06,
+      "loss": 2.6338,
+      "step": 16918
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2002686472819216e-06,
+      "loss": 3.3234,
+      "step": 16919
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2002212389380533e-06,
+      "loss": 1.991,
+      "step": 16920
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2001738305941846e-06,
+      "loss": 2.6443,
+      "step": 16921
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2001264222503163e-06,
+      "loss": 3.2091,
+      "step": 16922
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2000790139064476e-06,
+      "loss": 3.2856,
+      "step": 16923
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.200031605562579e-06,
+      "loss": 2.5378,
+      "step": 16924
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1999841972187105e-06,
+      "loss": 3.1925,
+      "step": 16925
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1999367888748422e-06,
+      "loss": 2.7829,
+      "step": 16926
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1998893805309735e-06,
+      "loss": 3.6091,
+      "step": 16927
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1998419721871052e-06,
+      "loss": 2.7377,
+      "step": 16928
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1997945638432365e-06,
+      "loss": 3.2585,
+      "step": 16929
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1997471554993678e-06,
+      "loss": 3.2456,
+      "step": 16930
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.199699747155499e-06,
+      "loss": 3.2191,
+      "step": 16931
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1996523388116308e-06,
+      "loss": 4.2222,
+      "step": 16932
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1996049304677625e-06,
+      "loss": 3.6259,
+      "step": 16933
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1995575221238938e-06,
+      "loss": 2.5959,
+      "step": 16934
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1995101137800255e-06,
+      "loss": 3.0776,
+      "step": 16935
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1994627054361567e-06,
+      "loss": 2.1283,
+      "step": 16936
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.199415297092288e-06,
+      "loss": 3.6177,
+      "step": 16937
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1993678887484197e-06,
+      "loss": 3.8416,
+      "step": 16938
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1993204804045514e-06,
+      "loss": 2.1373,
+      "step": 16939
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1992730720606827e-06,
+      "loss": 3.1264,
+      "step": 16940
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1992256637168144e-06,
+      "loss": 3.1806,
+      "step": 16941
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1991782553729457e-06,
+      "loss": 1.6125,
+      "step": 16942
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.199130847029077e-06,
+      "loss": 2.4234,
+      "step": 16943
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1990834386852087e-06,
+      "loss": 2.4751,
+      "step": 16944
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.19903603034134e-06,
+      "loss": 3.9706,
+      "step": 16945
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1989886219974717e-06,
+      "loss": 3.7717,
+      "step": 16946
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1989412136536034e-06,
+      "loss": 3.0885,
+      "step": 16947
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1988938053097346e-06,
+      "loss": 3.1219,
+      "step": 16948
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198846396965866e-06,
+      "loss": 3.4659,
+      "step": 16949
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1987989886219976e-06,
+      "loss": 2.7553,
+      "step": 16950
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198751580278129e-06,
+      "loss": 2.4176,
+      "step": 16951
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1987041719342606e-06,
+      "loss": 2.9671,
+      "step": 16952
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198656763590392e-06,
+      "loss": 3.4531,
+      "step": 16953
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1986093552465236e-06,
+      "loss": 3.3584,
+      "step": 16954
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198561946902655e-06,
+      "loss": 2.7549,
+      "step": 16955
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198514538558786e-06,
+      "loss": 2.2339,
+      "step": 16956
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198467130214918e-06,
+      "loss": 3.9402,
+      "step": 16957
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198419721871049e-06,
+      "loss": 3.2659,
+      "step": 16958
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198372313527181e-06,
+      "loss": 1.8948,
+      "step": 16959
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1983249051833126e-06,
+      "loss": 3.4287,
+      "step": 16960
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198277496839444e-06,
+      "loss": 3.2087,
+      "step": 16961
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198230088495575e-06,
+      "loss": 2.2166,
+      "step": 16962
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198182680151707e-06,
+      "loss": 3.1036,
+      "step": 16963
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.198135271807838e-06,
+      "loss": 2.0188,
+      "step": 16964
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.19808786346397e-06,
+      "loss": 2.7242,
+      "step": 16965
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1980404551201015e-06,
+      "loss": 2.445,
+      "step": 16966
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1979930467762328e-06,
+      "loss": 3.93,
+      "step": 16967
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197945638432364e-06,
+      "loss": 1.5982,
+      "step": 16968
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1978982300884958e-06,
+      "loss": 2.1297,
+      "step": 16969
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197850821744627e-06,
+      "loss": 1.616,
+      "step": 16970
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1978034134007583e-06,
+      "loss": 3.3335,
+      "step": 16971
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.19775600505689e-06,
+      "loss": 2.0471,
+      "step": 16972
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1977085967130217e-06,
+      "loss": 2.0412,
+      "step": 16973
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197661188369153e-06,
+      "loss": 3.2569,
+      "step": 16974
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1976137800252843e-06,
+      "loss": 2.3757,
+      "step": 16975
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197566371681416e-06,
+      "loss": 2.1621,
+      "step": 16976
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1975189633375473e-06,
+      "loss": 3.103,
+      "step": 16977
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197471554993679e-06,
+      "loss": 1.7157,
+      "step": 16978
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1974241466498107e-06,
+      "loss": 3.2168,
+      "step": 16979
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197376738305942e-06,
+      "loss": 2.9913,
+      "step": 16980
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1973293299620733e-06,
+      "loss": 3.161,
+      "step": 16981
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197281921618205e-06,
+      "loss": 3.2175,
+      "step": 16982
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1972345132743362e-06,
+      "loss": 3.0357,
+      "step": 16983
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1971871049304675e-06,
+      "loss": 3.5332,
+      "step": 16984
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1971396965865992e-06,
+      "loss": 3.6375,
+      "step": 16985
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197092288242731e-06,
+      "loss": 2.4794,
+      "step": 16986
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.197044879898862e-06,
+      "loss": 3.6024,
+      "step": 16987
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196997471554994e-06,
+      "loss": 2.1439,
+      "step": 16988
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196950063211125e-06,
+      "loss": 3.2921,
+      "step": 16989
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1969026548672565e-06,
+      "loss": 3.7465,
+      "step": 16990
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196855246523388e-06,
+      "loss": 2.7056,
+      "step": 16991
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.19680783817952e-06,
+      "loss": 3.0766,
+      "step": 16992
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196760429835651e-06,
+      "loss": 3.3029,
+      "step": 16993
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196713021491783e-06,
+      "loss": 3.1976,
+      "step": 16994
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196665613147914e-06,
+      "loss": 3.1492,
+      "step": 16995
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1966182048040454e-06,
+      "loss": 4.1558,
+      "step": 16996
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1965707964601767e-06,
+      "loss": 2.1964,
+      "step": 16997
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1965233881163084e-06,
+      "loss": 2.6166,
+      "step": 16998
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.19647597977244e-06,
+      "loss": 3.5089,
+      "step": 16999
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1964285714285714e-06,
+      "loss": 3.3988,
+      "step": 17000
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.196381163084703e-06,
+      "loss": 3.2937,
+      "step": 17001
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1963337547408344e-06,
+      "loss": 1.6754,
+      "step": 17002
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1962863463969657e-06,
+      "loss": 3.377,
+      "step": 17003
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1962389380530974e-06,
+      "loss": 3.4339,
+      "step": 17004
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.196191529709229e-06,
+      "loss": 2.8954,
+      "step": 17005
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1961441213653603e-06,
+      "loss": 3.7175,
+      "step": 17006
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.196096713021492e-06,
+      "loss": 3.0315,
+      "step": 17007
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1960493046776233e-06,
+      "loss": 3.853,
+      "step": 17008
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1960018963337546e-06,
+      "loss": 2.5125,
+      "step": 17009
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1959544879898863e-06,
+      "loss": 2.1511,
+      "step": 17010
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1959070796460176e-06,
+      "loss": 2.2031,
+      "step": 17011
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1958596713021493e-06,
+      "loss": 3.4452,
+      "step": 17012
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.195812262958281e-06,
+      "loss": 2.6528,
+      "step": 17013
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1957648546144123e-06,
+      "loss": 2.1222,
+      "step": 17014
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1957174462705436e-06,
+      "loss": 2.3278,
+      "step": 17015
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1956700379266753e-06,
+      "loss": 2.52,
+      "step": 17016
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1956226295828065e-06,
+      "loss": 4.0863,
+      "step": 17017
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1955752212389383e-06,
+      "loss": 2.8115,
+      "step": 17018
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1955278128950695e-06,
+      "loss": 3.1683,
+      "step": 17019
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1954804045512012e-06,
+      "loss": 4.3729,
+      "step": 17020
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1954329962073325e-06,
+      "loss": 2.9554,
+      "step": 17021
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.195385587863464e-06,
+      "loss": 3.1763,
+      "step": 17022
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1953381795195955e-06,
+      "loss": 3.5134,
+      "step": 17023
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1952907711757268e-06,
+      "loss": 2.8284,
+      "step": 17024
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1952433628318585e-06,
+      "loss": 3.3156,
+      "step": 17025
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.19519595448799e-06,
+      "loss": 3.3316,
+      "step": 17026
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1951485461441215e-06,
+      "loss": 2.071,
+      "step": 17027
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1951011378002528e-06,
+      "loss": 3.1883,
+      "step": 17028
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1950537294563845e-06,
+      "loss": 2.9445,
+      "step": 17029
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1950063211125157e-06,
+      "loss": 2.6757,
+      "step": 17030
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1949589127686474e-06,
+      "loss": 1.9703,
+      "step": 17031
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194911504424779e-06,
+      "loss": 2.0264,
+      "step": 17032
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1948640960809104e-06,
+      "loss": 2.7608,
+      "step": 17033
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1948166877370417e-06,
+      "loss": 3.5357,
+      "step": 17034
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1947692793931734e-06,
+      "loss": 2.6453,
+      "step": 17035
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1947218710493047e-06,
+      "loss": 2.6479,
+      "step": 17036
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194674462705436e-06,
+      "loss": 2.3714,
+      "step": 17037
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1946270543615677e-06,
+      "loss": 3.1156,
+      "step": 17038
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1945796460176994e-06,
+      "loss": 2.7261,
+      "step": 17039
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1945322376738307e-06,
+      "loss": 2.7238,
+      "step": 17040
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194484829329962e-06,
+      "loss": 3.2092,
+      "step": 17041
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1944374209860936e-06,
+      "loss": 3.4541,
+      "step": 17042
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194390012642225e-06,
+      "loss": 1.8902,
+      "step": 17043
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1943426042983566e-06,
+      "loss": 3.0685,
+      "step": 17044
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1942951959544883e-06,
+      "loss": 3.5489,
+      "step": 17045
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1942477876106196e-06,
+      "loss": 3.1093,
+      "step": 17046
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194200379266751e-06,
+      "loss": 3.0822,
+      "step": 17047
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1941529709228826e-06,
+      "loss": 3.3337,
+      "step": 17048
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194105562579014e-06,
+      "loss": 2.9016,
+      "step": 17049
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194058154235145e-06,
+      "loss": 3.1945,
+      "step": 17050
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.194010745891277e-06,
+      "loss": 2.9615,
+      "step": 17051
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1939633375474086e-06,
+      "loss": 2.7624,
+      "step": 17052
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.19391592920354e-06,
+      "loss": 2.9046,
+      "step": 17053
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1938685208596715e-06,
+      "loss": 3.3076,
+      "step": 17054
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193821112515803e-06,
+      "loss": 2.3074,
+      "step": 17055
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193773704171934e-06,
+      "loss": 3.0874,
+      "step": 17056
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193726295828066e-06,
+      "loss": 2.555,
+      "step": 17057
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1936788874841975e-06,
+      "loss": 1.7144,
+      "step": 17058
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193631479140329e-06,
+      "loss": 3.8733,
+      "step": 17059
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1935840707964605e-06,
+      "loss": 2.3158,
+      "step": 17060
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1935366624525918e-06,
+      "loss": 2.5347,
+      "step": 17061
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193489254108723e-06,
+      "loss": 2.884,
+      "step": 17062
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1934418457648543e-06,
+      "loss": 2.1317,
+      "step": 17063
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193394437420986e-06,
+      "loss": 3.4539,
+      "step": 17064
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1933470290771177e-06,
+      "loss": 2.7626,
+      "step": 17065
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.193299620733249e-06,
+      "loss": 3.0809,
+      "step": 17066
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.1932522123893807e-06,
+      "loss": 2.6919,
+      "step": 17067
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.193204804045512e-06,
+      "loss": 2.58,
+      "step": 17068
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1931573957016433e-06,
+      "loss": 3.9275,
+      "step": 17069
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.193109987357775e-06,
+      "loss": 2.9405,
+      "step": 17070
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1930625790139067e-06,
+      "loss": 2.7917,
+      "step": 17071
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.193015170670038e-06,
+      "loss": 3.7589,
+      "step": 17072
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1929677623261697e-06,
+      "loss": 2.7766,
+      "step": 17073
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192920353982301e-06,
+      "loss": 2.6348,
+      "step": 17074
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1928729456384322e-06,
+      "loss": 3.1015,
+      "step": 17075
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192825537294564e-06,
+      "loss": 3.565,
+      "step": 17076
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1927781289506952e-06,
+      "loss": 3.4352,
+      "step": 17077
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192730720606827e-06,
+      "loss": 3.7549,
+      "step": 17078
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1926833122629586e-06,
+      "loss": 1.9084,
+      "step": 17079
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.19263590391909e-06,
+      "loss": 3.2649,
+      "step": 17080
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192588495575221e-06,
+      "loss": 3.3725,
+      "step": 17081
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192541087231353e-06,
+      "loss": 2.8564,
+      "step": 17082
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192493678887484e-06,
+      "loss": 3.7539,
+      "step": 17083
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192446270543616e-06,
+      "loss": 3.2875,
+      "step": 17084
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192398862199747e-06,
+      "loss": 2.0707,
+      "step": 17085
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192351453855879e-06,
+      "loss": 3.7773,
+      "step": 17086
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.19230404551201e-06,
+      "loss": 2.0835,
+      "step": 17087
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1922566371681414e-06,
+      "loss": 3.1179,
+      "step": 17088
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192209228824273e-06,
+      "loss": 2.4436,
+      "step": 17089
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1921618204804044e-06,
+      "loss": 4.1442,
+      "step": 17090
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192114412136536e-06,
+      "loss": 3.1225,
+      "step": 17091
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192067003792668e-06,
+      "loss": 2.7765,
+      "step": 17092
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.192019595448799e-06,
+      "loss": 3.1499,
+      "step": 17093
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1919721871049304e-06,
+      "loss": 3.5853,
+      "step": 17094
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191924778761062e-06,
+      "loss": 2.4113,
+      "step": 17095
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1918773704171934e-06,
+      "loss": 3.4177,
+      "step": 17096
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191829962073325e-06,
+      "loss": 2.6402,
+      "step": 17097
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1917825537294568e-06,
+      "loss": 2.4314,
+      "step": 17098
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191735145385588e-06,
+      "loss": 3.1153,
+      "step": 17099
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1916877370417193e-06,
+      "loss": 2.2818,
+      "step": 17100
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191640328697851e-06,
+      "loss": 2.019,
+      "step": 17101
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1915929203539823e-06,
+      "loss": 2.1665,
+      "step": 17102
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1915455120101136e-06,
+      "loss": 3.2026,
+      "step": 17103
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1914981036662453e-06,
+      "loss": 4.1626,
+      "step": 17104
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191450695322377e-06,
+      "loss": 3.1932,
+      "step": 17105
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1914032869785083e-06,
+      "loss": 2.8769,
+      "step": 17106
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1913558786346396e-06,
+      "loss": 3.6216,
+      "step": 17107
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1913084702907713e-06,
+      "loss": 2.3819,
+      "step": 17108
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1912610619469026e-06,
+      "loss": 3.0617,
+      "step": 17109
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191213653603034e-06,
+      "loss": 2.6766,
+      "step": 17110
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.191166245259166e-06,
+      "loss": 2.9173,
+      "step": 17111
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1911188369152972e-06,
+      "loss": 2.3461,
+      "step": 17112
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1910714285714285e-06,
+      "loss": 3.0381,
+      "step": 17113
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1910240202275602e-06,
+      "loss": 2.6675,
+      "step": 17114
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1909766118836915e-06,
+      "loss": 3.2477,
+      "step": 17115
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.190929203539823e-06,
+      "loss": 4.0481,
+      "step": 17116
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1908817951959545e-06,
+      "loss": 3.2248,
+      "step": 17117
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.190834386852086e-06,
+      "loss": 3.6898,
+      "step": 17118
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1907869785082175e-06,
+      "loss": 1.4562,
+      "step": 17119
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.190739570164349e-06,
+      "loss": 2.4419,
+      "step": 17120
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1906921618204805e-06,
+      "loss": 2.7219,
+      "step": 17121
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1906447534766117e-06,
+      "loss": 1.832,
+      "step": 17122
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1905973451327434e-06,
+      "loss": 3.3255,
+      "step": 17123
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1905499367888747e-06,
+      "loss": 3.2917,
+      "step": 17124
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1905025284450064e-06,
+      "loss": 2.3352,
+      "step": 17125
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.190455120101138e-06,
+      "loss": 1.7684,
+      "step": 17126
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1904077117572694e-06,
+      "loss": 3.4198,
+      "step": 17127
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1903603034134007e-06,
+      "loss": 3.3313,
+      "step": 17128
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.190312895069532e-06,
+      "loss": 2.738,
+      "step": 17129
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.1902654867256637e-06,
+      "loss": 3.1421,
+      "step": 17130
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1902180783817954e-06,
+      "loss": 3.4572,
+      "step": 17131
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1901706700379267e-06,
+      "loss": 3.5926,
+      "step": 17132
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1901232616940584e-06,
+      "loss": 2.6525,
+      "step": 17133
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1900758533501897e-06,
+      "loss": 2.2012,
+      "step": 17134
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.190028445006321e-06,
+      "loss": 2.7397,
+      "step": 17135
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1899810366624526e-06,
+      "loss": 2.9579,
+      "step": 17136
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189933628318584e-06,
+      "loss": 2.5173,
+      "step": 17137
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1898862199747156e-06,
+      "loss": 2.2123,
+      "step": 17138
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1898388116308473e-06,
+      "loss": 2.8088,
+      "step": 17139
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1897914032869786e-06,
+      "loss": 3.3906,
+      "step": 17140
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.18974399494311e-06,
+      "loss": 3.6733,
+      "step": 17141
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1896965865992416e-06,
+      "loss": 2.6622,
+      "step": 17142
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189649178255373e-06,
+      "loss": 4.2416,
+      "step": 17143
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1896017699115046e-06,
+      "loss": 4.0412,
+      "step": 17144
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1895543615676363e-06,
+      "loss": 3.9807,
+      "step": 17145
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1895069532237676e-06,
+      "loss": 3.6443,
+      "step": 17146
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189459544879899e-06,
+      "loss": 2.1695,
+      "step": 17147
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1894121365360305e-06,
+      "loss": 3.0183,
+      "step": 17148
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189364728192162e-06,
+      "loss": 3.5613,
+      "step": 17149
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189317319848293e-06,
+      "loss": 3.4942,
+      "step": 17150
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189269911504425e-06,
+      "loss": 2.2239,
+      "step": 17151
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1892225031605565e-06,
+      "loss": 2.0904,
+      "step": 17152
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189175094816688e-06,
+      "loss": 3.0136,
+      "step": 17153
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189127686472819e-06,
+      "loss": 2.2334,
+      "step": 17154
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1890802781289508e-06,
+      "loss": 3.3551,
+      "step": 17155
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.189032869785082e-06,
+      "loss": 3.5528,
+      "step": 17156
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1889854614412138e-06,
+      "loss": 2.4165,
+      "step": 17157
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1889380530973455e-06,
+      "loss": 3.7738,
+      "step": 17158
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1888906447534767e-06,
+      "loss": 2.9523,
+      "step": 17159
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188843236409608e-06,
+      "loss": 2.3352,
+      "step": 17160
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1887958280657397e-06,
+      "loss": 3.0759,
+      "step": 17161
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188748419721871e-06,
+      "loss": 2.9211,
+      "step": 17162
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1887010113780023e-06,
+      "loss": 2.214,
+      "step": 17163
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188653603034134e-06,
+      "loss": 3.0898,
+      "step": 17164
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1886061946902657e-06,
+      "loss": 3.9664,
+      "step": 17165
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188558786346397e-06,
+      "loss": 3.9534,
+      "step": 17166
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1885113780025287e-06,
+      "loss": 3.5765,
+      "step": 17167
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.18846396965866e-06,
+      "loss": 2.1107,
+      "step": 17168
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1884165613147912e-06,
+      "loss": 2.8556,
+      "step": 17169
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188369152970923e-06,
+      "loss": 2.8536,
+      "step": 17170
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1883217446270546e-06,
+      "loss": 3.3064,
+      "step": 17171
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188274336283186e-06,
+      "loss": 2.0003,
+      "step": 17172
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188226927939317e-06,
+      "loss": 3.902,
+      "step": 17173
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188179519595449e-06,
+      "loss": 2.2075,
+      "step": 17174
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.18813211125158e-06,
+      "loss": 2.7965,
+      "step": 17175
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1880847029077115e-06,
+      "loss": 1.9966,
+      "step": 17176
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.188037294563843e-06,
+      "loss": 3.5149,
+      "step": 17177
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187989886219975e-06,
+      "loss": 3.2814,
+      "step": 17178
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187942477876106e-06,
+      "loss": 3.7431,
+      "step": 17179
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187895069532238e-06,
+      "loss": 2.4098,
+      "step": 17180
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187847661188369e-06,
+      "loss": 3.5204,
+      "step": 17181
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1878002528445004e-06,
+      "loss": 1.795,
+      "step": 17182
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187752844500632e-06,
+      "loss": 2.4666,
+      "step": 17183
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187705436156764e-06,
+      "loss": 3.0736,
+      "step": 17184
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187658027812895e-06,
+      "loss": 3.3605,
+      "step": 17185
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187610619469027e-06,
+      "loss": 3.186,
+      "step": 17186
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187563211125158e-06,
+      "loss": 4.2471,
+      "step": 17187
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1875158027812894e-06,
+      "loss": 2.579,
+      "step": 17188
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187468394437421e-06,
+      "loss": 3.2051,
+      "step": 17189
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1874209860935524e-06,
+      "loss": 2.4137,
+      "step": 17190
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187373577749684e-06,
+      "loss": 3.4322,
+      "step": 17191
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1873261694058154e-06,
+      "loss": 2.4664,
+      "step": 17192
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.187278761061947e-06,
+      "loss": 4.1628,
+      "step": 17193
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.1872313527180783e-06,
+      "loss": 3.3998,
+      "step": 17194
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1871839443742096e-06,
+      "loss": 3.195,
+      "step": 17195
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1871365360303413e-06,
+      "loss": 3.1588,
+      "step": 17196
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.187089127686473e-06,
+      "loss": 3.8553,
+      "step": 17197
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1870417193426043e-06,
+      "loss": 2.7524,
+      "step": 17198
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.186994310998736e-06,
+      "loss": 3.3971,
+      "step": 17199
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1869469026548673e-06,
+      "loss": 2.726,
+      "step": 17200
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1868994943109986e-06,
+      "loss": 3.1012,
+      "step": 17201
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1868520859671303e-06,
+      "loss": 3.6202,
+      "step": 17202
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1868046776232616e-06,
+      "loss": 3.155,
+      "step": 17203
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1867572692793933e-06,
+      "loss": 2.3981,
+      "step": 17204
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.186709860935525e-06,
+      "loss": 3.217,
+      "step": 17205
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1866624525916562e-06,
+      "loss": 2.5631,
+      "step": 17206
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1866150442477875e-06,
+      "loss": 2.8889,
+      "step": 17207
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1865676359039192e-06,
+      "loss": 3.1008,
+      "step": 17208
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1865202275600505e-06,
+      "loss": 2.8285,
+      "step": 17209
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.186472819216182e-06,
+      "loss": 2.7674,
+      "step": 17210
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.186425410872314e-06,
+      "loss": 3.1576,
+      "step": 17211
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.186378002528445e-06,
+      "loss": 2.446,
+      "step": 17212
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1863305941845765e-06,
+      "loss": 2.7107,
+      "step": 17213
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1862831858407078e-06,
+      "loss": 3.5195,
+      "step": 17214
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1862357774968395e-06,
+      "loss": 2.8774,
+      "step": 17215
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1861883691529707e-06,
+      "loss": 2.5992,
+      "step": 17216
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1861409608091024e-06,
+      "loss": 3.7011,
+      "step": 17217
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.186093552465234e-06,
+      "loss": 3.3908,
+      "step": 17218
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1860461441213654e-06,
+      "loss": 2.5155,
+      "step": 17219
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1859987357774967e-06,
+      "loss": 3.5807,
+      "step": 17220
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1859513274336284e-06,
+      "loss": 3.4545,
+      "step": 17221
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1859039190897597e-06,
+      "loss": 4.1411,
+      "step": 17222
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1858565107458914e-06,
+      "loss": 3.204,
+      "step": 17223
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.185809102402023e-06,
+      "loss": 1.9626,
+      "step": 17224
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1857616940581544e-06,
+      "loss": 3.1415,
+      "step": 17225
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1857142857142857e-06,
+      "loss": 2.8573,
+      "step": 17226
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1856668773704174e-06,
+      "loss": 3.0036,
+      "step": 17227
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1856194690265486e-06,
+      "loss": 2.3673,
+      "step": 17228
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.18557206068268e-06,
+      "loss": 2.0813,
+      "step": 17229
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1855246523388116e-06,
+      "loss": 3.3669,
+      "step": 17230
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1854772439949433e-06,
+      "loss": 3.1875,
+      "step": 17231
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1854298356510746e-06,
+      "loss": 1.962,
+      "step": 17232
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1853824273072063e-06,
+      "loss": 3.6798,
+      "step": 17233
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1853350189633376e-06,
+      "loss": 3.0507,
+      "step": 17234
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.185287610619469e-06,
+      "loss": 2.9947,
+      "step": 17235
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1852402022756006e-06,
+      "loss": 1.9683,
+      "step": 17236
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1851927939317323e-06,
+      "loss": 3.9197,
+      "step": 17237
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1851453855878636e-06,
+      "loss": 2.7713,
+      "step": 17238
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.185097977243995e-06,
+      "loss": 2.5857,
+      "step": 17239
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1850505689001266e-06,
+      "loss": 2.932,
+      "step": 17240
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.185003160556258e-06,
+      "loss": 3.4193,
+      "step": 17241
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.184955752212389e-06,
+      "loss": 2.8231,
+      "step": 17242
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.184908343868521e-06,
+      "loss": 3.0444,
+      "step": 17243
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1848609355246525e-06,
+      "loss": 2.477,
+      "step": 17244
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.184813527180784e-06,
+      "loss": 3.5032,
+      "step": 17245
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1847661188369155e-06,
+      "loss": 2.9042,
+      "step": 17246
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1847187104930468e-06,
+      "loss": 3.4034,
+      "step": 17247
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.184671302149178e-06,
+      "loss": 3.1132,
+      "step": 17248
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1846238938053098e-06,
+      "loss": 3.2537,
+      "step": 17249
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1845764854614415e-06,
+      "loss": 2.6236,
+      "step": 17250
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1845290771175728e-06,
+      "loss": 3.3195,
+      "step": 17251
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1844816687737045e-06,
+      "loss": 3.8175,
+      "step": 17252
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1844342604298357e-06,
+      "loss": 2.4843,
+      "step": 17253
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.184386852085967e-06,
+      "loss": 2.4135,
+      "step": 17254
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1843394437420987e-06,
+      "loss": 3.544,
+      "step": 17255
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.18429203539823e-06,
+      "loss": 3.9918,
+      "step": 17256
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2.1842446270543617e-06,
+      "loss": 2.7668,
+      "step": 17257
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.184197218710493e-06,
+      "loss": 3.7328,
+      "step": 17258
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1841498103666247e-06,
+      "loss": 2.977,
+      "step": 17259
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.184102402022756e-06,
+      "loss": 1.7165,
+      "step": 17260
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1840549936788873e-06,
+      "loss": 2.411,
+      "step": 17261
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.184007585335019e-06,
+      "loss": 3.2432,
+      "step": 17262
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1839601769911507e-06,
+      "loss": 2.6205,
+      "step": 17263
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183912768647282e-06,
+      "loss": 3.0009,
+      "step": 17264
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1838653603034136e-06,
+      "loss": 2.2263,
+      "step": 17265
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183817951959545e-06,
+      "loss": 3.0778,
+      "step": 17266
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183770543615676e-06,
+      "loss": 2.5025,
+      "step": 17267
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183723135271808e-06,
+      "loss": 1.8105,
+      "step": 17268
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183675726927939e-06,
+      "loss": 4.1122,
+      "step": 17269
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183628318584071e-06,
+      "loss": 3.0425,
+      "step": 17270
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1835809102402026e-06,
+      "loss": 3.0084,
+      "step": 17271
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183533501896334e-06,
+      "loss": 2.1408,
+      "step": 17272
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183486093552465e-06,
+      "loss": 3.1847,
+      "step": 17273
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183438685208597e-06,
+      "loss": 2.5217,
+      "step": 17274
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183391276864728e-06,
+      "loss": 3.6228,
+      "step": 17275
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.18334386852086e-06,
+      "loss": 3.3134,
+      "step": 17276
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1832964601769916e-06,
+      "loss": 2.5325,
+      "step": 17277
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183249051833123e-06,
+      "loss": 3.1336,
+      "step": 17278
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183201643489254e-06,
+      "loss": 3.0993,
+      "step": 17279
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1831542351453854e-06,
+      "loss": 1.9514,
+      "step": 17280
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.183106826801517e-06,
+      "loss": 2.8725,
+      "step": 17281
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1830594184576484e-06,
+      "loss": 2.8299,
+      "step": 17282
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.18301201011378e-06,
+      "loss": 2.8316,
+      "step": 17283
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1829646017699118e-06,
+      "loss": 1.4384,
+      "step": 17284
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182917193426043e-06,
+      "loss": 2.5783,
+      "step": 17285
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1828697850821743e-06,
+      "loss": 3.092,
+      "step": 17286
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182822376738306e-06,
+      "loss": 2.9922,
+      "step": 17287
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1827749683944373e-06,
+      "loss": 2.016,
+      "step": 17288
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182727560050569e-06,
+      "loss": 3.4196,
+      "step": 17289
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1826801517067007e-06,
+      "loss": 2.9164,
+      "step": 17290
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182632743362832e-06,
+      "loss": 2.909,
+      "step": 17291
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1825853350189633e-06,
+      "loss": 3.0191,
+      "step": 17292
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182537926675095e-06,
+      "loss": 2.0266,
+      "step": 17293
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1824905183312263e-06,
+      "loss": 3.4375,
+      "step": 17294
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1824431099873576e-06,
+      "loss": 3.1963,
+      "step": 17295
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1823957016434893e-06,
+      "loss": 3.4211,
+      "step": 17296
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182348293299621e-06,
+      "loss": 3.2846,
+      "step": 17297
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1823008849557523e-06,
+      "loss": 3.2806,
+      "step": 17298
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182253476611884e-06,
+      "loss": 2.3892,
+      "step": 17299
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1822060682680152e-06,
+      "loss": 3.3601,
+      "step": 17300
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1821586599241465e-06,
+      "loss": 2.7975,
+      "step": 17301
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1821112515802782e-06,
+      "loss": 3.5686,
+      "step": 17302
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.18206384323641e-06,
+      "loss": 3.1571,
+      "step": 17303
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.182016434892541e-06,
+      "loss": 2.6921,
+      "step": 17304
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1819690265486725e-06,
+      "loss": 3.4772,
+      "step": 17305
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.181921618204804e-06,
+      "loss": 2.9414,
+      "step": 17306
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1818742098609355e-06,
+      "loss": 3.1432,
+      "step": 17307
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1818268015170667e-06,
+      "loss": 3.3947,
+      "step": 17308
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1817793931731985e-06,
+      "loss": 2.4997,
+      "step": 17309
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.18173198482933e-06,
+      "loss": 3.6251,
+      "step": 17310
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1816845764854614e-06,
+      "loss": 3.6201,
+      "step": 17311
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.181637168141593e-06,
+      "loss": 3.6985,
+      "step": 17312
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1815897597977244e-06,
+      "loss": 3.136,
+      "step": 17313
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1815423514538557e-06,
+      "loss": 2.7635,
+      "step": 17314
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1814949431099874e-06,
+      "loss": 3.2397,
+      "step": 17315
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.181447534766119e-06,
+      "loss": 2.4001,
+      "step": 17316
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1814001264222504e-06,
+      "loss": 3.9963,
+      "step": 17317
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.181352718078382e-06,
+      "loss": 3.7399,
+      "step": 17318
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1813053097345134e-06,
+      "loss": 2.3458,
+      "step": 17319
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.1812579013906447e-06,
+      "loss": 3.2116,
+      "step": 17320
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1812104930467764e-06,
+      "loss": 2.1035,
+      "step": 17321
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1811630847029076e-06,
+      "loss": 2.4175,
+      "step": 17322
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1811156763590393e-06,
+      "loss": 2.6738,
+      "step": 17323
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1810682680151706e-06,
+      "loss": 3.5024,
+      "step": 17324
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1810208596713023e-06,
+      "loss": 1.9151,
+      "step": 17325
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1809734513274336e-06,
+      "loss": 2.6131,
+      "step": 17326
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.180926042983565e-06,
+      "loss": 1.5434,
+      "step": 17327
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1808786346396966e-06,
+      "loss": 4.0095,
+      "step": 17328
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1808312262958283e-06,
+      "loss": 3.3743,
+      "step": 17329
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1807838179519596e-06,
+      "loss": 3.4937,
+      "step": 17330
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1807364096080913e-06,
+      "loss": 3.4373,
+      "step": 17331
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1806890012642226e-06,
+      "loss": 3.3499,
+      "step": 17332
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.180641592920354e-06,
+      "loss": 2.713,
+      "step": 17333
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1805941845764855e-06,
+      "loss": 3.2903,
+      "step": 17334
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.180546776232617e-06,
+      "loss": 3.5042,
+      "step": 17335
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1804993678887485e-06,
+      "loss": 3.0268,
+      "step": 17336
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1804519595448802e-06,
+      "loss": 3.4072,
+      "step": 17337
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1804045512010115e-06,
+      "loss": 3.3575,
+      "step": 17338
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.180357142857143e-06,
+      "loss": 3.3677,
+      "step": 17339
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1803097345132745e-06,
+      "loss": 3.8206,
+      "step": 17340
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1802623261694058e-06,
+      "loss": 2.4113,
+      "step": 17341
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1802149178255375e-06,
+      "loss": 2.9921,
+      "step": 17342
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.180167509481669e-06,
+      "loss": 2.9803,
+      "step": 17343
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1801201011378005e-06,
+      "loss": 4.1892,
+      "step": 17344
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1800726927939317e-06,
+      "loss": 2.9091,
+      "step": 17345
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.180025284450063e-06,
+      "loss": 2.6969,
+      "step": 17346
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1799778761061947e-06,
+      "loss": 3.0363,
+      "step": 17347
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179930467762326e-06,
+      "loss": 2.4617,
+      "step": 17348
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1798830594184577e-06,
+      "loss": 3.6571,
+      "step": 17349
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1798356510745894e-06,
+      "loss": 2.7763,
+      "step": 17350
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1797882427307207e-06,
+      "loss": 1.9846,
+      "step": 17351
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179740834386852e-06,
+      "loss": 3.5249,
+      "step": 17352
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1796934260429837e-06,
+      "loss": 3.1593,
+      "step": 17353
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179646017699115e-06,
+      "loss": 2.2797,
+      "step": 17354
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1795986093552467e-06,
+      "loss": 2.5318,
+      "step": 17355
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1795512010113784e-06,
+      "loss": 2.5546,
+      "step": 17356
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1795037926675097e-06,
+      "loss": 3.0241,
+      "step": 17357
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179456384323641e-06,
+      "loss": 2.8479,
+      "step": 17358
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1794089759797726e-06,
+      "loss": 3.8535,
+      "step": 17359
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179361567635904e-06,
+      "loss": 3.665,
+      "step": 17360
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179314159292035e-06,
+      "loss": 2.3431,
+      "step": 17361
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179266750948167e-06,
+      "loss": 3.1262,
+      "step": 17362
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1792193426042986e-06,
+      "loss": 2.8964,
+      "step": 17363
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.17917193426043e-06,
+      "loss": 2.697,
+      "step": 17364
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1791245259165616e-06,
+      "loss": 2.7662,
+      "step": 17365
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179077117572693e-06,
+      "loss": 3.2239,
+      "step": 17366
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.179029709228824e-06,
+      "loss": 2.1457,
+      "step": 17367
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178982300884956e-06,
+      "loss": 2.4705,
+      "step": 17368
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1789348925410876e-06,
+      "loss": 3.0594,
+      "step": 17369
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178887484197219e-06,
+      "loss": 1.6616,
+      "step": 17370
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.17884007585335e-06,
+      "loss": 2.2105,
+      "step": 17371
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178792667509482e-06,
+      "loss": 3.3994,
+      "step": 17372
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178745259165613e-06,
+      "loss": 2.2753,
+      "step": 17373
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1786978508217444e-06,
+      "loss": 2.7964,
+      "step": 17374
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178650442477876e-06,
+      "loss": 3.3365,
+      "step": 17375
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178603034134008e-06,
+      "loss": 3.0222,
+      "step": 17376
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178555625790139e-06,
+      "loss": 3.4612,
+      "step": 17377
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1785082174462708e-06,
+      "loss": 2.4288,
+      "step": 17378
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178460809102402e-06,
+      "loss": 2.7669,
+      "step": 17379
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1784134007585333e-06,
+      "loss": 3.6132,
+      "step": 17380
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178365992414665e-06,
+      "loss": 2.2957,
+      "step": 17381
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1783185840707967e-06,
+      "loss": 2.787,
+      "step": 17382
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.178271175726928e-06,
+      "loss": 3.612,
+      "step": 17383
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2.1782237673830597e-06,
+      "loss": 2.2,
+      "step": 17384
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.178176359039191e-06,
+      "loss": 3.2503,
+      "step": 17385
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1781289506953223e-06,
+      "loss": 2.928,
+      "step": 17386
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.178081542351454e-06,
+      "loss": 3.4035,
+      "step": 17387
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1780341340075853e-06,
+      "loss": 3.3492,
+      "step": 17388
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177986725663717e-06,
+      "loss": 1.8791,
+      "step": 17389
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1779393173198483e-06,
+      "loss": 2.6257,
+      "step": 17390
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.17789190897598e-06,
+      "loss": 3.0263,
+      "step": 17391
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1778445006321112e-06,
+      "loss": 3.6849,
+      "step": 17392
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1777970922882425e-06,
+      "loss": 2.2085,
+      "step": 17393
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1777496839443742e-06,
+      "loss": 3.3849,
+      "step": 17394
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1777022756005055e-06,
+      "loss": 3.3695,
+      "step": 17395
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1776548672566372e-06,
+      "loss": 3.3381,
+      "step": 17396
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177607458912769e-06,
+      "loss": 2.1522,
+      "step": 17397
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1775600505689e-06,
+      "loss": 2.7767,
+      "step": 17398
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1775126422250315e-06,
+      "loss": 2.3117,
+      "step": 17399
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177465233881163e-06,
+      "loss": 2.7944,
+      "step": 17400
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1774178255372945e-06,
+      "loss": 3.5533,
+      "step": 17401
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177370417193426e-06,
+      "loss": 1.7213,
+      "step": 17402
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177323008849558e-06,
+      "loss": 3.631,
+      "step": 17403
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177275600505689e-06,
+      "loss": 2.7854,
+      "step": 17404
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1772281921618204e-06,
+      "loss": 2.6327,
+      "step": 17405
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.177180783817952e-06,
+      "loss": 2.6511,
+      "step": 17406
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1771333754740834e-06,
+      "loss": 1.7919,
+      "step": 17407
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1770859671302147e-06,
+      "loss": 2.9052,
+      "step": 17408
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1770385587863464e-06,
+      "loss": 3.4576,
+      "step": 17409
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.176991150442478e-06,
+      "loss": 2.8608,
+      "step": 17410
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1769437420986094e-06,
+      "loss": 3.3824,
+      "step": 17411
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1768963337547407e-06,
+      "loss": 2.9465,
+      "step": 17412
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1768489254108724e-06,
+      "loss": 2.1881,
+      "step": 17413
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1768015170670036e-06,
+      "loss": 2.6031,
+      "step": 17414
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1767541087231354e-06,
+      "loss": 2.5961,
+      "step": 17415
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.176706700379267e-06,
+      "loss": 3.2222,
+      "step": 17416
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1766592920353983e-06,
+      "loss": 2.338,
+      "step": 17417
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1766118836915296e-06,
+      "loss": 3.3092,
+      "step": 17418
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1765644753476613e-06,
+      "loss": 2.2423,
+      "step": 17419
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1765170670037926e-06,
+      "loss": 1.5459,
+      "step": 17420
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.176469658659924e-06,
+      "loss": 3.1401,
+      "step": 17421
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1764222503160556e-06,
+      "loss": 3.621,
+      "step": 17422
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1763748419721873e-06,
+      "loss": 3.0594,
+      "step": 17423
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1763274336283186e-06,
+      "loss": 3.5535,
+      "step": 17424
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1762800252844503e-06,
+      "loss": 2.8181,
+      "step": 17425
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1762326169405816e-06,
+      "loss": 2.8089,
+      "step": 17426
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.176185208596713e-06,
+      "loss": 2.3349,
+      "step": 17427
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1761378002528445e-06,
+      "loss": 2.3162,
+      "step": 17428
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1760903919089762e-06,
+      "loss": 2.7742,
+      "step": 17429
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1760429835651075e-06,
+      "loss": 3.3248,
+      "step": 17430
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1759955752212392e-06,
+      "loss": 2.1569,
+      "step": 17431
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1759481668773705e-06,
+      "loss": 2.8727,
+      "step": 17432
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.175900758533502e-06,
+      "loss": 2.4782,
+      "step": 17433
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.175853350189633e-06,
+      "loss": 2.6699,
+      "step": 17434
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1758059418457648e-06,
+      "loss": 3.7469,
+      "step": 17435
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1757585335018965e-06,
+      "loss": 3.1769,
+      "step": 17436
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1757111251580278e-06,
+      "loss": 2.8093,
+      "step": 17437
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1756637168141595e-06,
+      "loss": 3.4706,
+      "step": 17438
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1756163084702907e-06,
+      "loss": 2.3962,
+      "step": 17439
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.175568900126422e-06,
+      "loss": 2.7334,
+      "step": 17440
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1755214917825537e-06,
+      "loss": 3.1798,
+      "step": 17441
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1754740834386854e-06,
+      "loss": 3.1153,
+      "step": 17442
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1754266750948167e-06,
+      "loss": 3.184,
+      "step": 17443
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1753792667509484e-06,
+      "loss": 3.2953,
+      "step": 17444
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1753318584070797e-06,
+      "loss": 3.0026,
+      "step": 17445
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.175284450063211e-06,
+      "loss": 2.803,
+      "step": 17446
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2.1752370417193427e-06,
+      "loss": 3.2236,
+      "step": 17447
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.175189633375474e-06,
+      "loss": 3.4438,
+      "step": 17448
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1751422250316057e-06,
+      "loss": 2.5159,
+      "step": 17449
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1750948166877374e-06,
+      "loss": 2.8376,
+      "step": 17450
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1750474083438686e-06,
+      "loss": 2.3901,
+      "step": 17451
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.175e-06,
+      "loss": 3.3545,
+      "step": 17452
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1749525916561316e-06,
+      "loss": 2.8799,
+      "step": 17453
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174905183312263e-06,
+      "loss": 3.5657,
+      "step": 17454
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1748577749683946e-06,
+      "loss": 2.9817,
+      "step": 17455
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174810366624526e-06,
+      "loss": 4.2094,
+      "step": 17456
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1747629582806576e-06,
+      "loss": 2.8349,
+      "step": 17457
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174715549936789e-06,
+      "loss": 1.9731,
+      "step": 17458
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.17466814159292e-06,
+      "loss": 3.5806,
+      "step": 17459
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174620733249052e-06,
+      "loss": 3.4907,
+      "step": 17460
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174573324905183e-06,
+      "loss": 3.6819,
+      "step": 17461
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174525916561315e-06,
+      "loss": 3.5522,
+      "step": 17462
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1744785082174466e-06,
+      "loss": 2.1922,
+      "step": 17463
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174431099873578e-06,
+      "loss": 2.2892,
+      "step": 17464
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174383691529709e-06,
+      "loss": 2.6841,
+      "step": 17465
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174336283185841e-06,
+      "loss": 3.9331,
+      "step": 17466
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174288874841972e-06,
+      "loss": 2.6562,
+      "step": 17467
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174241466498104e-06,
+      "loss": 3.0925,
+      "step": 17468
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1741940581542355e-06,
+      "loss": 2.6784,
+      "step": 17469
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1741466498103668e-06,
+      "loss": 3.6015,
+      "step": 17470
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174099241466498e-06,
+      "loss": 3.4479,
+      "step": 17471
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1740518331226298e-06,
+      "loss": 3.4738,
+      "step": 17472
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.174004424778761e-06,
+      "loss": 2.5102,
+      "step": 17473
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1739570164348923e-06,
+      "loss": 3.1797,
+      "step": 17474
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173909608091024e-06,
+      "loss": 3.6167,
+      "step": 17475
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1738621997471557e-06,
+      "loss": 3.5416,
+      "step": 17476
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173814791403287e-06,
+      "loss": 3.8367,
+      "step": 17477
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1737673830594183e-06,
+      "loss": 2.8574,
+      "step": 17478
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.17371997471555e-06,
+      "loss": 2.7724,
+      "step": 17479
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1736725663716813e-06,
+      "loss": 3.5407,
+      "step": 17480
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173625158027813e-06,
+      "loss": 1.6861,
+      "step": 17481
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1735777496839447e-06,
+      "loss": 2.504,
+      "step": 17482
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173530341340076e-06,
+      "loss": 2.8543,
+      "step": 17483
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1734829329962073e-06,
+      "loss": 1.9551,
+      "step": 17484
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173435524652339e-06,
+      "loss": 2.1317,
+      "step": 17485
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1733881163084702e-06,
+      "loss": 2.6924,
+      "step": 17486
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1733407079646015e-06,
+      "loss": 1.9156,
+      "step": 17487
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1732932996207332e-06,
+      "loss": 1.7774,
+      "step": 17488
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173245891276865e-06,
+      "loss": 3.2664,
+      "step": 17489
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173198482932996e-06,
+      "loss": 3.1161,
+      "step": 17490
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173151074589128e-06,
+      "loss": 3.5724,
+      "step": 17491
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173103666245259e-06,
+      "loss": 2.3649,
+      "step": 17492
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1730562579013905e-06,
+      "loss": 2.485,
+      "step": 17493
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.173008849557522e-06,
+      "loss": 2.7052,
+      "step": 17494
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172961441213654e-06,
+      "loss": 3.7086,
+      "step": 17495
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172914032869785e-06,
+      "loss": 3.3756,
+      "step": 17496
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172866624525917e-06,
+      "loss": 3.3186,
+      "step": 17497
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172819216182048e-06,
+      "loss": 4.2126,
+      "step": 17498
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1727718078381794e-06,
+      "loss": 3.1505,
+      "step": 17499
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1727243994943107e-06,
+      "loss": 2.2574,
+      "step": 17500
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1726769911504424e-06,
+      "loss": 2.7733,
+      "step": 17501
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172629582806574e-06,
+      "loss": 3.0933,
+      "step": 17502
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1725821744627054e-06,
+      "loss": 1.9619,
+      "step": 17503
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172534766118837e-06,
+      "loss": 2.7186,
+      "step": 17504
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1724873577749684e-06,
+      "loss": 3.4245,
+      "step": 17505
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1724399494310997e-06,
+      "loss": 3.6037,
+      "step": 17506
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1723925410872314e-06,
+      "loss": 2.99,
+      "step": 17507
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172345132743363e-06,
+      "loss": 3.1822,
+      "step": 17508
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.1722977243994943e-06,
+      "loss": 3.2905,
+      "step": 17509
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2.172250316055626e-06,
+      "loss": 2.951,
+      "step": 17510
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1722029077117573e-06,
+      "loss": 2.4253,
+      "step": 17511
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1721554993678886e-06,
+      "loss": 2.0644,
+      "step": 17512
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1721080910240203e-06,
+      "loss": 2.1437,
+      "step": 17513
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1720606826801516e-06,
+      "loss": 2.9438,
+      "step": 17514
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1720132743362833e-06,
+      "loss": 3.7359,
+      "step": 17515
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.171965865992415e-06,
+      "loss": 3.6206,
+      "step": 17516
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1719184576485463e-06,
+      "loss": 4.196,
+      "step": 17517
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1718710493046776e-06,
+      "loss": 3.2531,
+      "step": 17518
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1718236409608093e-06,
+      "loss": 3.9608,
+      "step": 17519
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1717762326169405e-06,
+      "loss": 3.161,
+      "step": 17520
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1717288242730723e-06,
+      "loss": 3.1563,
+      "step": 17521
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1716814159292035e-06,
+      "loss": 3.0038,
+      "step": 17522
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1716340075853352e-06,
+      "loss": 2.94,
+      "step": 17523
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1715865992414665e-06,
+      "loss": 2.1477,
+      "step": 17524
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.171539190897598e-06,
+      "loss": 4.1406,
+      "step": 17525
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1714917825537295e-06,
+      "loss": 3.1286,
+      "step": 17526
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1714443742098608e-06,
+      "loss": 2.5817,
+      "step": 17527
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1713969658659925e-06,
+      "loss": 2.8537,
+      "step": 17528
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.171349557522124e-06,
+      "loss": 2.1624,
+      "step": 17529
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1713021491782555e-06,
+      "loss": 3.7573,
+      "step": 17530
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1712547408343868e-06,
+      "loss": 3.0899,
+      "step": 17531
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1712073324905185e-06,
+      "loss": 3.3357,
+      "step": 17532
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1711599241466497e-06,
+      "loss": 3.3656,
+      "step": 17533
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1711125158027814e-06,
+      "loss": 3.4314,
+      "step": 17534
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.171065107458913e-06,
+      "loss": 3.0735,
+      "step": 17535
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1710176991150444e-06,
+      "loss": 2.0254,
+      "step": 17536
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1709702907711757e-06,
+      "loss": 3.4087,
+      "step": 17537
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1709228824273074e-06,
+      "loss": 3.6054,
+      "step": 17538
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1708754740834387e-06,
+      "loss": 2.9693,
+      "step": 17539
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.17082806573957e-06,
+      "loss": 3.3953,
+      "step": 17540
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1707806573957017e-06,
+      "loss": 3.2163,
+      "step": 17541
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1707332490518334e-06,
+      "loss": 2.9811,
+      "step": 17542
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1706858407079647e-06,
+      "loss": 3.6403,
+      "step": 17543
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170638432364096e-06,
+      "loss": 2.7344,
+      "step": 17544
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1705910240202276e-06,
+      "loss": 2.5759,
+      "step": 17545
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170543615676359e-06,
+      "loss": 2.7487,
+      "step": 17546
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1704962073324906e-06,
+      "loss": 3.7823,
+      "step": 17547
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1704487989886223e-06,
+      "loss": 2.5393,
+      "step": 17548
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1704013906447536e-06,
+      "loss": 3.6827,
+      "step": 17549
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170353982300885e-06,
+      "loss": 3.0325,
+      "step": 17550
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1703065739570166e-06,
+      "loss": 3.0754,
+      "step": 17551
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170259165613148e-06,
+      "loss": 3.7137,
+      "step": 17552
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170211757269279e-06,
+      "loss": 2.9239,
+      "step": 17553
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170164348925411e-06,
+      "loss": 2.6671,
+      "step": 17554
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1701169405815426e-06,
+      "loss": 3.1256,
+      "step": 17555
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.170069532237674e-06,
+      "loss": 2.5139,
+      "step": 17556
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1700221238938055e-06,
+      "loss": 2.0002,
+      "step": 17557
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169974715549937e-06,
+      "loss": 3.0036,
+      "step": 17558
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169927307206068e-06,
+      "loss": 2.8177,
+      "step": 17559
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1698798988622e-06,
+      "loss": 2.8114,
+      "step": 17560
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1698324905183315e-06,
+      "loss": 2.8597,
+      "step": 17561
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169785082174463e-06,
+      "loss": 3.0025,
+      "step": 17562
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1697376738305945e-06,
+      "loss": 2.6667,
+      "step": 17563
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1696902654867258e-06,
+      "loss": 2.719,
+      "step": 17564
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169642857142857e-06,
+      "loss": 3.1736,
+      "step": 17565
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1695954487989883e-06,
+      "loss": 2.3626,
+      "step": 17566
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.16954804045512e-06,
+      "loss": 3.6364,
+      "step": 17567
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1695006321112517e-06,
+      "loss": 2.9436,
+      "step": 17568
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169453223767383e-06,
+      "loss": 4.0122,
+      "step": 17569
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1694058154235147e-06,
+      "loss": 2.2876,
+      "step": 17570
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169358407079646e-06,
+      "loss": 2.5421,
+      "step": 17571
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1693109987357773e-06,
+      "loss": 3.1791,
+      "step": 17572
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.169263590391909e-06,
+      "loss": 3.1648,
+      "step": 17573
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2.1692161820480407e-06,
+      "loss": 2.9131,
+      "step": 17574
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.169168773704172e-06,
+      "loss": 3.4286,
+      "step": 17575
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1691213653603037e-06,
+      "loss": 2.1423,
+      "step": 17576
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.169073957016435e-06,
+      "loss": 2.9271,
+      "step": 17577
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1690265486725662e-06,
+      "loss": 3.6036,
+      "step": 17578
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168979140328698e-06,
+      "loss": 2.4388,
+      "step": 17579
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1689317319848292e-06,
+      "loss": 2.5592,
+      "step": 17580
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168884323640961e-06,
+      "loss": 2.8706,
+      "step": 17581
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1688369152970926e-06,
+      "loss": 2.5396,
+      "step": 17582
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168789506953224e-06,
+      "loss": 2.3091,
+      "step": 17583
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168742098609355e-06,
+      "loss": 2.6532,
+      "step": 17584
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168694690265487e-06,
+      "loss": 3.732,
+      "step": 17585
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168647281921618e-06,
+      "loss": 3.4122,
+      "step": 17586
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.16859987357775e-06,
+      "loss": 3.6301,
+      "step": 17587
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168552465233881e-06,
+      "loss": 2.8823,
+      "step": 17588
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168505056890013e-06,
+      "loss": 2.5808,
+      "step": 17589
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168457648546144e-06,
+      "loss": 3.3339,
+      "step": 17590
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1684102402022754e-06,
+      "loss": 2.5941,
+      "step": 17591
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168362831858407e-06,
+      "loss": 3.7085,
+      "step": 17592
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1683154235145384e-06,
+      "loss": 2.5918,
+      "step": 17593
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.16826801517067e-06,
+      "loss": 2.8607,
+      "step": 17594
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168220606826802e-06,
+      "loss": 2.3315,
+      "step": 17595
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168173198482933e-06,
+      "loss": 3.3495,
+      "step": 17596
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1681257901390644e-06,
+      "loss": 2.6504,
+      "step": 17597
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.168078381795196e-06,
+      "loss": 3.4268,
+      "step": 17598
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1680309734513274e-06,
+      "loss": 2.2764,
+      "step": 17599
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.167983565107459e-06,
+      "loss": 2.9694,
+      "step": 17600
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1679361567635908e-06,
+      "loss": 3.6435,
+      "step": 17601
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.167888748419722e-06,
+      "loss": 2.7797,
+      "step": 17602
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1678413400758533e-06,
+      "loss": 2.5367,
+      "step": 17603
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.167793931731985e-06,
+      "loss": 2.3946,
+      "step": 17604
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1677465233881163e-06,
+      "loss": 3.1753,
+      "step": 17605
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1676991150442476e-06,
+      "loss": 3.9347,
+      "step": 17606
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1676517067003793e-06,
+      "loss": 3.2133,
+      "step": 17607
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.167604298356511e-06,
+      "loss": 2.4609,
+      "step": 17608
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1675568900126423e-06,
+      "loss": 2.9235,
+      "step": 17609
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1675094816687736e-06,
+      "loss": 3.2816,
+      "step": 17610
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1674620733249053e-06,
+      "loss": 2.9317,
+      "step": 17611
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1674146649810366e-06,
+      "loss": 2.7339,
+      "step": 17612
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1673672566371683e-06,
+      "loss": 2.5567,
+      "step": 17613
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1673198482933e-06,
+      "loss": 2.1067,
+      "step": 17614
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1672724399494312e-06,
+      "loss": 4.2549,
+      "step": 17615
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1672250316055625e-06,
+      "loss": 2.416,
+      "step": 17616
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1671776232616942e-06,
+      "loss": 3.2607,
+      "step": 17617
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1671302149178255e-06,
+      "loss": 1.9218,
+      "step": 17618
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.167082806573957e-06,
+      "loss": 2.4125,
+      "step": 17619
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1670353982300885e-06,
+      "loss": 3.6993,
+      "step": 17620
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.16698798988622e-06,
+      "loss": 2.5156,
+      "step": 17621
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1669405815423515e-06,
+      "loss": 1.9771,
+      "step": 17622
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.166893173198483e-06,
+      "loss": 3.668,
+      "step": 17623
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1668457648546145e-06,
+      "loss": 2.4639,
+      "step": 17624
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1667983565107457e-06,
+      "loss": 1.8631,
+      "step": 17625
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1667509481668774e-06,
+      "loss": 3.5114,
+      "step": 17626
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.166703539823009e-06,
+      "loss": 3.281,
+      "step": 17627
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1666561314791404e-06,
+      "loss": 1.8821,
+      "step": 17628
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1666087231352717e-06,
+      "loss": 3.4748,
+      "step": 17629
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1665613147914034e-06,
+      "loss": 3.5714,
+      "step": 17630
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1665139064475347e-06,
+      "loss": 3.0335,
+      "step": 17631
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.166466498103666e-06,
+      "loss": 2.0033,
+      "step": 17632
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1664190897597977e-06,
+      "loss": 3.9489,
+      "step": 17633
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1663716814159294e-06,
+      "loss": 2.4891,
+      "step": 17634
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1663242730720607e-06,
+      "loss": 1.9628,
+      "step": 17635
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1662768647281924e-06,
+      "loss": 3.1137,
+      "step": 17636
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2.1662294563843237e-06,
+      "loss": 3.3048,
+      "step": 17637
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.166182048040455e-06,
+      "loss": 3.5769,
+      "step": 17638
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1661346396965866e-06,
+      "loss": 2.8611,
+      "step": 17639
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1660872313527183e-06,
+      "loss": 3.1049,
+      "step": 17640
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1660398230088496e-06,
+      "loss": 3.5791,
+      "step": 17641
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1659924146649813e-06,
+      "loss": 2.2106,
+      "step": 17642
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1659450063211126e-06,
+      "loss": 3.6205,
+      "step": 17643
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165897597977244e-06,
+      "loss": 2.8136,
+      "step": 17644
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1658501896333756e-06,
+      "loss": 3.2209,
+      "step": 17645
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165802781289507e-06,
+      "loss": 3.1692,
+      "step": 17646
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1657553729456386e-06,
+      "loss": 2.4318,
+      "step": 17647
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1657079646017703e-06,
+      "loss": 1.4993,
+      "step": 17648
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1656605562579016e-06,
+      "loss": 3.0862,
+      "step": 17649
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165613147914033e-06,
+      "loss": 3.6272,
+      "step": 17650
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165565739570164e-06,
+      "loss": 2.9317,
+      "step": 17651
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165518331226296e-06,
+      "loss": 2.8193,
+      "step": 17652
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1654709228824275e-06,
+      "loss": 3.1222,
+      "step": 17653
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165423514538559e-06,
+      "loss": 2.3187,
+      "step": 17654
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1653761061946905e-06,
+      "loss": 2.9616,
+      "step": 17655
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165328697850822e-06,
+      "loss": 2.8913,
+      "step": 17656
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165281289506953e-06,
+      "loss": 2.6195,
+      "step": 17657
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1652338811630848e-06,
+      "loss": 2.3429,
+      "step": 17658
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.165186472819216e-06,
+      "loss": 2.7673,
+      "step": 17659
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1651390644753478e-06,
+      "loss": 2.6032,
+      "step": 17660
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1650916561314795e-06,
+      "loss": 3.8031,
+      "step": 17661
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1650442477876107e-06,
+      "loss": 2.8745,
+      "step": 17662
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164996839443742e-06,
+      "loss": 2.3261,
+      "step": 17663
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1649494310998737e-06,
+      "loss": 3.0646,
+      "step": 17664
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164902022756005e-06,
+      "loss": 2.339,
+      "step": 17665
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1648546144121367e-06,
+      "loss": 2.3429,
+      "step": 17666
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1648072060682684e-06,
+      "loss": 3.5045,
+      "step": 17667
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1647597977243997e-06,
+      "loss": 1.809,
+      "step": 17668
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164712389380531e-06,
+      "loss": 2.7848,
+      "step": 17669
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1646649810366627e-06,
+      "loss": 2.1758,
+      "step": 17670
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164617572692794e-06,
+      "loss": 3.6,
+      "step": 17671
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1645701643489252e-06,
+      "loss": 2.9053,
+      "step": 17672
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164522756005057e-06,
+      "loss": 2.8298,
+      "step": 17673
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1644753476611887e-06,
+      "loss": 2.8401,
+      "step": 17674
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.16442793931732e-06,
+      "loss": 3.7261,
+      "step": 17675
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164380530973451e-06,
+      "loss": 2.8795,
+      "step": 17676
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164333122629583e-06,
+      "loss": 3.4904,
+      "step": 17677
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164285714285714e-06,
+      "loss": 3.1461,
+      "step": 17678
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1642383059418455e-06,
+      "loss": 2.733,
+      "step": 17679
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1641908975979776e-06,
+      "loss": 3.2243,
+      "step": 17680
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164143489254109e-06,
+      "loss": 3.4778,
+      "step": 17681
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.16409608091024e-06,
+      "loss": 3.3171,
+      "step": 17682
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164048672566372e-06,
+      "loss": 3.2681,
+      "step": 17683
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.164001264222503e-06,
+      "loss": 1.4964,
+      "step": 17684
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1639538558786344e-06,
+      "loss": 3.7447,
+      "step": 17685
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163906447534766e-06,
+      "loss": 3.7616,
+      "step": 17686
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163859039190898e-06,
+      "loss": 2.9159,
+      "step": 17687
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163811630847029e-06,
+      "loss": 3.4764,
+      "step": 17688
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163764222503161e-06,
+      "loss": 3.0974,
+      "step": 17689
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163716814159292e-06,
+      "loss": 3.023,
+      "step": 17690
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1636694058154234e-06,
+      "loss": 4.1317,
+      "step": 17691
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163621997471555e-06,
+      "loss": 2.7351,
+      "step": 17692
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1635745891276864e-06,
+      "loss": 2.5728,
+      "step": 17693
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163527180783818e-06,
+      "loss": 3.1854,
+      "step": 17694
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1634797724399494e-06,
+      "loss": 3.9332,
+      "step": 17695
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163432364096081e-06,
+      "loss": 3.8178,
+      "step": 17696
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1633849557522123e-06,
+      "loss": 3.6979,
+      "step": 17697
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1633375474083436e-06,
+      "loss": 3.0563,
+      "step": 17698
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.1632901390644753e-06,
+      "loss": 2.3001,
+      "step": 17699
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2.163242730720607e-06,
+      "loss": 2.1745,
+      "step": 17700
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1631953223767383e-06,
+      "loss": 3.1625,
+      "step": 17701
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.16314791403287e-06,
+      "loss": 4.114,
+      "step": 17702
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1631005056890013e-06,
+      "loss": 2.4898,
+      "step": 17703
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1630530973451326e-06,
+      "loss": 3.1603,
+      "step": 17704
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1630056890012643e-06,
+      "loss": 3.4846,
+      "step": 17705
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1629582806573956e-06,
+      "loss": 2.9802,
+      "step": 17706
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1629108723135273e-06,
+      "loss": 2.9287,
+      "step": 17707
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.162863463969659e-06,
+      "loss": 3.1702,
+      "step": 17708
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1628160556257902e-06,
+      "loss": 3.1225,
+      "step": 17709
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1627686472819215e-06,
+      "loss": 2.2542,
+      "step": 17710
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1627212389380532e-06,
+      "loss": 2.8247,
+      "step": 17711
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1626738305941845e-06,
+      "loss": 4.0743,
+      "step": 17712
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.162626422250316e-06,
+      "loss": 3.0232,
+      "step": 17713
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.162579013906448e-06,
+      "loss": 3.1199,
+      "step": 17714
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.162531605562579e-06,
+      "loss": 2.8983,
+      "step": 17715
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1624841972187105e-06,
+      "loss": 1.789,
+      "step": 17716
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1624367888748418e-06,
+      "loss": 3.6714,
+      "step": 17717
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1623893805309735e-06,
+      "loss": 3.8078,
+      "step": 17718
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1623419721871047e-06,
+      "loss": 2.9642,
+      "step": 17719
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1622945638432364e-06,
+      "loss": 3.3473,
+      "step": 17720
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.162247155499368e-06,
+      "loss": 3.1486,
+      "step": 17721
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1621997471554994e-06,
+      "loss": 2.5559,
+      "step": 17722
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1621523388116307e-06,
+      "loss": 2.3831,
+      "step": 17723
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1621049304677624e-06,
+      "loss": 2.933,
+      "step": 17724
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1620575221238937e-06,
+      "loss": 3.409,
+      "step": 17725
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1620101137800254e-06,
+      "loss": 3.125,
+      "step": 17726
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161962705436157e-06,
+      "loss": 2.024,
+      "step": 17727
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1619152970922884e-06,
+      "loss": 3.4842,
+      "step": 17728
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1618678887484197e-06,
+      "loss": 2.9952,
+      "step": 17729
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1618204804045514e-06,
+      "loss": 3.1679,
+      "step": 17730
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1617730720606826e-06,
+      "loss": 3.8743,
+      "step": 17731
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161725663716814e-06,
+      "loss": 2.6272,
+      "step": 17732
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1616782553729456e-06,
+      "loss": 3.101,
+      "step": 17733
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1616308470290773e-06,
+      "loss": 3.0217,
+      "step": 17734
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1615834386852086e-06,
+      "loss": 2.7915,
+      "step": 17735
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1615360303413403e-06,
+      "loss": 3.0197,
+      "step": 17736
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1614886219974716e-06,
+      "loss": 3.4765,
+      "step": 17737
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161441213653603e-06,
+      "loss": 3.3993,
+      "step": 17738
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1613938053097346e-06,
+      "loss": 2.5478,
+      "step": 17739
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1613463969658663e-06,
+      "loss": 2.1903,
+      "step": 17740
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1612989886219976e-06,
+      "loss": 2.9235,
+      "step": 17741
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161251580278129e-06,
+      "loss": 2.5468,
+      "step": 17742
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1612041719342606e-06,
+      "loss": 3.3029,
+      "step": 17743
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161156763590392e-06,
+      "loss": 2.6459,
+      "step": 17744
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161109355246523e-06,
+      "loss": 2.8568,
+      "step": 17745
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.161061946902655e-06,
+      "loss": 3.4801,
+      "step": 17746
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1610145385587865e-06,
+      "loss": 2.4183,
+      "step": 17747
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.160967130214918e-06,
+      "loss": 3.2023,
+      "step": 17748
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1609197218710495e-06,
+      "loss": 3.1753,
+      "step": 17749
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1608723135271808e-06,
+      "loss": 2.6094,
+      "step": 17750
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.160824905183312e-06,
+      "loss": 3.3801,
+      "step": 17751
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1607774968394438e-06,
+      "loss": 2.3742,
+      "step": 17752
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1607300884955755e-06,
+      "loss": 3.2911,
+      "step": 17753
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1606826801517068e-06,
+      "loss": 2.971,
+      "step": 17754
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1606352718078385e-06,
+      "loss": 3.6159,
+      "step": 17755
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1605878634639697e-06,
+      "loss": 3.4653,
+      "step": 17756
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.160540455120101e-06,
+      "loss": 2.4369,
+      "step": 17757
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1604930467762327e-06,
+      "loss": 2.425,
+      "step": 17758
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.160445638432364e-06,
+      "loss": 3.0527,
+      "step": 17759
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1603982300884957e-06,
+      "loss": 3.7018,
+      "step": 17760
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.160350821744627e-06,
+      "loss": 3.2815,
+      "step": 17761
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1603034134007587e-06,
+      "loss": 3.268,
+      "step": 17762
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.16025600505689e-06,
+      "loss": 2.4862,
+      "step": 17763
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2.1602085967130213e-06,
+      "loss": 2.6444,
+      "step": 17764
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.160161188369153e-06,
+      "loss": 3.1275,
+      "step": 17765
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1601137800252847e-06,
+      "loss": 2.9594,
+      "step": 17766
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.160066371681416e-06,
+      "loss": 2.1908,
+      "step": 17767
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1600189633375476e-06,
+      "loss": 3.5402,
+      "step": 17768
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159971554993679e-06,
+      "loss": 2.9801,
+      "step": 17769
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.15992414664981e-06,
+      "loss": 2.3499,
+      "step": 17770
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159876738305942e-06,
+      "loss": 2.617,
+      "step": 17771
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159829329962073e-06,
+      "loss": 2.6303,
+      "step": 17772
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159781921618205e-06,
+      "loss": 2.9831,
+      "step": 17773
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1597345132743366e-06,
+      "loss": 2.6719,
+      "step": 17774
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159687104930468e-06,
+      "loss": 2.9854,
+      "step": 17775
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159639696586599e-06,
+      "loss": 3.789,
+      "step": 17776
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159592288242731e-06,
+      "loss": 2.9289,
+      "step": 17777
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159544879898862e-06,
+      "loss": 3.8788,
+      "step": 17778
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159497471554994e-06,
+      "loss": 3.2673,
+      "step": 17779
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1594500632111256e-06,
+      "loss": 2.823,
+      "step": 17780
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159402654867257e-06,
+      "loss": 2.139,
+      "step": 17781
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159355246523388e-06,
+      "loss": 3.1252,
+      "step": 17782
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1593078381795194e-06,
+      "loss": 3.2035,
+      "step": 17783
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159260429835651e-06,
+      "loss": 3.0864,
+      "step": 17784
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1592130214917824e-06,
+      "loss": 2.3904,
+      "step": 17785
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159165613147914e-06,
+      "loss": 3.1709,
+      "step": 17786
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1591182048040458e-06,
+      "loss": 2.1675,
+      "step": 17787
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.159070796460177e-06,
+      "loss": 2.9606,
+      "step": 17788
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1590233881163083e-06,
+      "loss": 2.8233,
+      "step": 17789
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.15897597977244e-06,
+      "loss": 3.2179,
+      "step": 17790
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1589285714285713e-06,
+      "loss": 2.7357,
+      "step": 17791
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158881163084703e-06,
+      "loss": 2.0937,
+      "step": 17792
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1588337547408347e-06,
+      "loss": 2.4738,
+      "step": 17793
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158786346396966e-06,
+      "loss": 3.2293,
+      "step": 17794
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1587389380530973e-06,
+      "loss": 1.9345,
+      "step": 17795
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158691529709229e-06,
+      "loss": 2.9995,
+      "step": 17796
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1586441213653603e-06,
+      "loss": 2.3618,
+      "step": 17797
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1585967130214916e-06,
+      "loss": 2.051,
+      "step": 17798
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1585493046776233e-06,
+      "loss": 3.0204,
+      "step": 17799
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158501896333755e-06,
+      "loss": 3.2399,
+      "step": 17800
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1584544879898863e-06,
+      "loss": 3.5221,
+      "step": 17801
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158407079646018e-06,
+      "loss": 3.2102,
+      "step": 17802
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1583596713021492e-06,
+      "loss": 1.7976,
+      "step": 17803
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1583122629582805e-06,
+      "loss": 2.6968,
+      "step": 17804
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1582648546144122e-06,
+      "loss": 3.9758,
+      "step": 17805
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158217446270544e-06,
+      "loss": 2.9813,
+      "step": 17806
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158170037926675e-06,
+      "loss": 2.2979,
+      "step": 17807
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1581226295828065e-06,
+      "loss": 2.2148,
+      "step": 17808
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.158075221238938e-06,
+      "loss": 3.1235,
+      "step": 17809
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1580278128950695e-06,
+      "loss": 2.3821,
+      "step": 17810
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1579804045512007e-06,
+      "loss": 2.7209,
+      "step": 17811
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1579329962073325e-06,
+      "loss": 2.552,
+      "step": 17812
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.157885587863464e-06,
+      "loss": 3.0188,
+      "step": 17813
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1578381795195954e-06,
+      "loss": 1.6986,
+      "step": 17814
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.157790771175727e-06,
+      "loss": 1.5539,
+      "step": 17815
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1577433628318584e-06,
+      "loss": 1.6451,
+      "step": 17816
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1576959544879897e-06,
+      "loss": 2.34,
+      "step": 17817
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1576485461441214e-06,
+      "loss": 3.5893,
+      "step": 17818
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.157601137800253e-06,
+      "loss": 2.6697,
+      "step": 17819
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1575537294563844e-06,
+      "loss": 2.8869,
+      "step": 17820
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.157506321112516e-06,
+      "loss": 2.1727,
+      "step": 17821
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1574589127686474e-06,
+      "loss": 1.9022,
+      "step": 17822
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1574115044247787e-06,
+      "loss": 3.0558,
+      "step": 17823
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1573640960809104e-06,
+      "loss": 3.9998,
+      "step": 17824
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1573166877370416e-06,
+      "loss": 3.2224,
+      "step": 17825
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1572692793931733e-06,
+      "loss": 1.7216,
+      "step": 17826
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2.1572218710493046e-06,
+      "loss": 2.0085,
+      "step": 17827
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1571744627054363e-06,
+      "loss": 2.0976,
+      "step": 17828
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1571270543615676e-06,
+      "loss": 3.6709,
+      "step": 17829
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.157079646017699e-06,
+      "loss": 2.4058,
+      "step": 17830
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1570322376738306e-06,
+      "loss": 3.3178,
+      "step": 17831
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1569848293299623e-06,
+      "loss": 2.5407,
+      "step": 17832
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1569374209860936e-06,
+      "loss": 2.6857,
+      "step": 17833
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1568900126422253e-06,
+      "loss": 2.1311,
+      "step": 17834
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1568426042983566e-06,
+      "loss": 3.5717,
+      "step": 17835
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.156795195954488e-06,
+      "loss": 2.9502,
+      "step": 17836
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1567477876106195e-06,
+      "loss": 2.6895,
+      "step": 17837
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.156700379266751e-06,
+      "loss": 3.7818,
+      "step": 17838
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1566529709228825e-06,
+      "loss": 2.5983,
+      "step": 17839
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1566055625790142e-06,
+      "loss": 1.5028,
+      "step": 17840
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1565581542351455e-06,
+      "loss": 3.0847,
+      "step": 17841
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.156510745891277e-06,
+      "loss": 2.4514,
+      "step": 17842
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1564633375474085e-06,
+      "loss": 2.5462,
+      "step": 17843
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1564159292035398e-06,
+      "loss": 2.6267,
+      "step": 17844
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1563685208596715e-06,
+      "loss": 3.2527,
+      "step": 17845
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.156321112515803e-06,
+      "loss": 1.3856,
+      "step": 17846
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1562737041719345e-06,
+      "loss": 3.1278,
+      "step": 17847
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1562262958280657e-06,
+      "loss": 4.3209,
+      "step": 17848
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.156178887484197e-06,
+      "loss": 3.9017,
+      "step": 17849
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1561314791403287e-06,
+      "loss": 3.0623,
+      "step": 17850
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.15608407079646e-06,
+      "loss": 2.4035,
+      "step": 17851
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1560366624525917e-06,
+      "loss": 4.0629,
+      "step": 17852
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1559892541087234e-06,
+      "loss": 2.5814,
+      "step": 17853
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1559418457648547e-06,
+      "loss": 3.0006,
+      "step": 17854
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155894437420986e-06,
+      "loss": 1.7289,
+      "step": 17855
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1558470290771177e-06,
+      "loss": 2.5895,
+      "step": 17856
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155799620733249e-06,
+      "loss": 2.4979,
+      "step": 17857
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1557522123893807e-06,
+      "loss": 3.0874,
+      "step": 17858
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1557048040455124e-06,
+      "loss": 2.3775,
+      "step": 17859
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1556573957016437e-06,
+      "loss": 2.2901,
+      "step": 17860
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155609987357775e-06,
+      "loss": 2.2769,
+      "step": 17861
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1555625790139066e-06,
+      "loss": 3.263,
+      "step": 17862
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155515170670038e-06,
+      "loss": 3.0814,
+      "step": 17863
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155467762326169e-06,
+      "loss": 3.9326,
+      "step": 17864
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155420353982301e-06,
+      "loss": 3.332,
+      "step": 17865
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1553729456384326e-06,
+      "loss": 2.9113,
+      "step": 17866
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155325537294564e-06,
+      "loss": 2.8754,
+      "step": 17867
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1552781289506956e-06,
+      "loss": 2.167,
+      "step": 17868
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155230720606827e-06,
+      "loss": 3.1169,
+      "step": 17869
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155183312262958e-06,
+      "loss": 3.5632,
+      "step": 17870
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.15513590391909e-06,
+      "loss": 2.3168,
+      "step": 17871
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1550884955752216e-06,
+      "loss": 2.317,
+      "step": 17872
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.155041087231353e-06,
+      "loss": 3.1699,
+      "step": 17873
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154993678887484e-06,
+      "loss": 2.2634,
+      "step": 17874
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154946270543616e-06,
+      "loss": 1.445,
+      "step": 17875
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154898862199747e-06,
+      "loss": 2.7026,
+      "step": 17876
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1548514538558784e-06,
+      "loss": 3.3991,
+      "step": 17877
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.15480404551201e-06,
+      "loss": 3.7853,
+      "step": 17878
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154756637168142e-06,
+      "loss": 3.6013,
+      "step": 17879
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154709228824273e-06,
+      "loss": 2.5538,
+      "step": 17880
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1546618204804048e-06,
+      "loss": 2.8299,
+      "step": 17881
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154614412136536e-06,
+      "loss": 3.3642,
+      "step": 17882
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1545670037926673e-06,
+      "loss": 3.1021,
+      "step": 17883
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154519595448799e-06,
+      "loss": 2.5458,
+      "step": 17884
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1544721871049307e-06,
+      "loss": 2.5322,
+      "step": 17885
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154424778761062e-06,
+      "loss": 4.191,
+      "step": 17886
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1543773704171937e-06,
+      "loss": 3.3353,
+      "step": 17887
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154329962073325e-06,
+      "loss": 3.8835,
+      "step": 17888
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.1542825537294563e-06,
+      "loss": 3.4439,
+      "step": 17889
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2.154235145385588e-06,
+      "loss": 4.0422,
+      "step": 17890
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1541877370417193e-06,
+      "loss": 3.0737,
+      "step": 17891
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.154140328697851e-06,
+      "loss": 2.5613,
+      "step": 17892
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1540929203539823e-06,
+      "loss": 2.3246,
+      "step": 17893
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.154045512010114e-06,
+      "loss": 2.6934,
+      "step": 17894
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1539981036662452e-06,
+      "loss": 2.7427,
+      "step": 17895
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1539506953223765e-06,
+      "loss": 3.4342,
+      "step": 17896
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1539032869785082e-06,
+      "loss": 3.1195,
+      "step": 17897
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.15385587863464e-06,
+      "loss": 2.9702,
+      "step": 17898
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1538084702907712e-06,
+      "loss": 3.4999,
+      "step": 17899
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153761061946903e-06,
+      "loss": 3.1131,
+      "step": 17900
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153713653603034e-06,
+      "loss": 3.12,
+      "step": 17901
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1536662452591655e-06,
+      "loss": 2.1845,
+      "step": 17902
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153618836915297e-06,
+      "loss": 1.6579,
+      "step": 17903
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1535714285714285e-06,
+      "loss": 3.2492,
+      "step": 17904
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.15352402022756e-06,
+      "loss": 3.4029,
+      "step": 17905
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153476611883692e-06,
+      "loss": 4.0296,
+      "step": 17906
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153429203539823e-06,
+      "loss": 3.5884,
+      "step": 17907
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1533817951959544e-06,
+      "loss": 2.6211,
+      "step": 17908
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153334386852086e-06,
+      "loss": 3.4358,
+      "step": 17909
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1532869785082174e-06,
+      "loss": 2.8525,
+      "step": 17910
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153239570164349e-06,
+      "loss": 3.5807,
+      "step": 17911
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153192161820481e-06,
+      "loss": 2.9851,
+      "step": 17912
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.153144753476612e-06,
+      "loss": 2.1651,
+      "step": 17913
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1530973451327434e-06,
+      "loss": 4.0449,
+      "step": 17914
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1530499367888747e-06,
+      "loss": 1.9776,
+      "step": 17915
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1530025284450064e-06,
+      "loss": 1.7941,
+      "step": 17916
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1529551201011376e-06,
+      "loss": 2.659,
+      "step": 17917
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1529077117572694e-06,
+      "loss": 2.6175,
+      "step": 17918
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.152860303413401e-06,
+      "loss": 2.2406,
+      "step": 17919
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1528128950695323e-06,
+      "loss": 3.3195,
+      "step": 17920
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1527654867256636e-06,
+      "loss": 3.0029,
+      "step": 17921
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1527180783817953e-06,
+      "loss": 3.5115,
+      "step": 17922
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1526706700379266e-06,
+      "loss": 2.6511,
+      "step": 17923
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1526232616940583e-06,
+      "loss": 3.124,
+      "step": 17924
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.15257585335019e-06,
+      "loss": 2.4834,
+      "step": 17925
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1525284450063213e-06,
+      "loss": 2.4475,
+      "step": 17926
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1524810366624526e-06,
+      "loss": 2.8198,
+      "step": 17927
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1524336283185843e-06,
+      "loss": 1.6502,
+      "step": 17928
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1523862199747156e-06,
+      "loss": 2.898,
+      "step": 17929
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.152338811630847e-06,
+      "loss": 2.3836,
+      "step": 17930
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1522914032869785e-06,
+      "loss": 3.8695,
+      "step": 17931
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1522439949431102e-06,
+      "loss": 2.3429,
+      "step": 17932
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1521965865992415e-06,
+      "loss": 2.4362,
+      "step": 17933
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1521491782553732e-06,
+      "loss": 3.472,
+      "step": 17934
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1521017699115045e-06,
+      "loss": 2.5173,
+      "step": 17935
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.152054361567636e-06,
+      "loss": 2.5224,
+      "step": 17936
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1520069532237675e-06,
+      "loss": 3.1566,
+      "step": 17937
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.151959544879899e-06,
+      "loss": 2.712,
+      "step": 17938
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1519121365360305e-06,
+      "loss": 3.4842,
+      "step": 17939
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1518647281921618e-06,
+      "loss": 3.0076,
+      "step": 17940
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1518173198482935e-06,
+      "loss": 2.5353,
+      "step": 17941
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1517699115044247e-06,
+      "loss": 2.056,
+      "step": 17942
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.151722503160556e-06,
+      "loss": 4.1658,
+      "step": 17943
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1516750948166877e-06,
+      "loss": 3.8996,
+      "step": 17944
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1516276864728194e-06,
+      "loss": 2.9618,
+      "step": 17945
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1515802781289507e-06,
+      "loss": 2.6802,
+      "step": 17946
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1515328697850824e-06,
+      "loss": 3.8205,
+      "step": 17947
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1514854614412137e-06,
+      "loss": 3.6362,
+      "step": 17948
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.151438053097345e-06,
+      "loss": 2.4916,
+      "step": 17949
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1513906447534767e-06,
+      "loss": 3.75,
+      "step": 17950
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1513432364096084e-06,
+      "loss": 1.7759,
+      "step": 17951
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1512958280657397e-06,
+      "loss": 2.3102,
+      "step": 17952
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1512484197218714e-06,
+      "loss": 2.9378,
+      "step": 17953
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2.1512010113780026e-06,
+      "loss": 3.8195,
+      "step": 17954
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.151153603034134e-06,
+      "loss": 2.0263,
+      "step": 17955
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1511061946902656e-06,
+      "loss": 2.0763,
+      "step": 17956
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.151058786346397e-06,
+      "loss": 2.3035,
+      "step": 17957
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1510113780025286e-06,
+      "loss": 2.9377,
+      "step": 17958
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.15096396965866e-06,
+      "loss": 3.5542,
+      "step": 17959
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1509165613147916e-06,
+      "loss": 3.0683,
+      "step": 17960
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150869152970923e-06,
+      "loss": 1.5775,
+      "step": 17961
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150821744627054e-06,
+      "loss": 3.0329,
+      "step": 17962
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150774336283186e-06,
+      "loss": 2.8969,
+      "step": 17963
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150726927939317e-06,
+      "loss": 3.404,
+      "step": 17964
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150679519595449e-06,
+      "loss": 2.9805,
+      "step": 17965
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1506321112515806e-06,
+      "loss": 2.2632,
+      "step": 17966
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150584702907712e-06,
+      "loss": 2.8459,
+      "step": 17967
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150537294563843e-06,
+      "loss": 3.3624,
+      "step": 17968
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150489886219975e-06,
+      "loss": 2.8596,
+      "step": 17969
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150442477876106e-06,
+      "loss": 2.709,
+      "step": 17970
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150395069532238e-06,
+      "loss": 3.1531,
+      "step": 17971
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1503476611883695e-06,
+      "loss": 2.4606,
+      "step": 17972
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1503002528445008e-06,
+      "loss": 2.3766,
+      "step": 17973
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150252844500632e-06,
+      "loss": 2.5369,
+      "step": 17974
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1502054361567638e-06,
+      "loss": 3.6994,
+      "step": 17975
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150158027812895e-06,
+      "loss": 2.9007,
+      "step": 17976
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1501106194690263e-06,
+      "loss": 2.9987,
+      "step": 17977
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.150063211125158e-06,
+      "loss": 2.0158,
+      "step": 17978
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1500158027812897e-06,
+      "loss": 3.6167,
+      "step": 17979
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149968394437421e-06,
+      "loss": 3.3723,
+      "step": 17980
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1499209860935523e-06,
+      "loss": 3.0474,
+      "step": 17981
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149873577749684e-06,
+      "loss": 2.5557,
+      "step": 17982
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1498261694058153e-06,
+      "loss": 2.8559,
+      "step": 17983
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149778761061947e-06,
+      "loss": 1.8482,
+      "step": 17984
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1497313527180787e-06,
+      "loss": 2.9152,
+      "step": 17985
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.14968394437421e-06,
+      "loss": 1.7761,
+      "step": 17986
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1496365360303413e-06,
+      "loss": 3.6693,
+      "step": 17987
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149589127686473e-06,
+      "loss": 4.1795,
+      "step": 17988
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1495417193426042e-06,
+      "loss": 3.5382,
+      "step": 17989
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1494943109987355e-06,
+      "loss": 2.9337,
+      "step": 17990
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1494469026548672e-06,
+      "loss": 2.7126,
+      "step": 17991
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149399494310999e-06,
+      "loss": 2.5572,
+      "step": 17992
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.14935208596713e-06,
+      "loss": 3.3818,
+      "step": 17993
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149304677623262e-06,
+      "loss": 2.0129,
+      "step": 17994
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149257269279393e-06,
+      "loss": 3.1784,
+      "step": 17995
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1492098609355245e-06,
+      "loss": 3.9507,
+      "step": 17996
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149162452591656e-06,
+      "loss": 2.6761,
+      "step": 17997
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149115044247788e-06,
+      "loss": 2.8183,
+      "step": 17998
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149067635903919e-06,
+      "loss": 2.574,
+      "step": 17999
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.149020227560051e-06,
+      "loss": 2.7439,
+      "step": 18000
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.148972819216182e-06,
+      "loss": 3.393,
+      "step": 18001
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1489254108723134e-06,
+      "loss": 2.5316,
+      "step": 18002
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1488780025284447e-06,
+      "loss": 2.1006,
+      "step": 18003
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1488305941845764e-06,
+      "loss": 3.4939,
+      "step": 18004
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.148783185840708e-06,
+      "loss": 2.1228,
+      "step": 18005
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1487357774968394e-06,
+      "loss": 3.1798,
+      "step": 18006
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.148688369152971e-06,
+      "loss": 2.5046,
+      "step": 18007
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1486409608091024e-06,
+      "loss": 1.7784,
+      "step": 18008
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1485935524652337e-06,
+      "loss": 2.8995,
+      "step": 18009
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1485461441213654e-06,
+      "loss": 2.155,
+      "step": 18010
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.148498735777497e-06,
+      "loss": 3.5844,
+      "step": 18011
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1484513274336283e-06,
+      "loss": 3.1029,
+      "step": 18012
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.14840391908976e-06,
+      "loss": 2.0765,
+      "step": 18013
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1483565107458913e-06,
+      "loss": 1.5275,
+      "step": 18014
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1483091024020226e-06,
+      "loss": 3.524,
+      "step": 18015
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1482616940581543e-06,
+      "loss": 3.6945,
+      "step": 18016
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2.1482142857142856e-06,
+      "loss": 2.5047,
+      "step": 18017
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1481668773704173e-06,
+      "loss": 2.2294,
+      "step": 18018
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.148119469026549e-06,
+      "loss": 3.6509,
+      "step": 18019
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1480720606826803e-06,
+      "loss": 2.6199,
+      "step": 18020
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1480246523388116e-06,
+      "loss": 3.9701,
+      "step": 18021
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1479772439949433e-06,
+      "loss": 2.4965,
+      "step": 18022
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1479298356510745e-06,
+      "loss": 2.6407,
+      "step": 18023
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1478824273072063e-06,
+      "loss": 3.5254,
+      "step": 18024
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1478350189633375e-06,
+      "loss": 3.4714,
+      "step": 18025
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1477876106194692e-06,
+      "loss": 2.9222,
+      "step": 18026
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1477402022756005e-06,
+      "loss": 1.9488,
+      "step": 18027
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.147692793931732e-06,
+      "loss": 3.4487,
+      "step": 18028
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1476453855878635e-06,
+      "loss": 2.8671,
+      "step": 18029
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1475979772439948e-06,
+      "loss": 2.4731,
+      "step": 18030
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1475505689001265e-06,
+      "loss": 1.5952,
+      "step": 18031
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.147503160556258e-06,
+      "loss": 2.8259,
+      "step": 18032
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1474557522123895e-06,
+      "loss": 3.2694,
+      "step": 18033
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1474083438685208e-06,
+      "loss": 2.6374,
+      "step": 18034
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1473609355246525e-06,
+      "loss": 3.2191,
+      "step": 18035
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1473135271807837e-06,
+      "loss": 2.1694,
+      "step": 18036
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1472661188369154e-06,
+      "loss": 3.1639,
+      "step": 18037
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.147218710493047e-06,
+      "loss": 3.3108,
+      "step": 18038
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1471713021491784e-06,
+      "loss": 2.8921,
+      "step": 18039
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1471238938053097e-06,
+      "loss": 2.8673,
+      "step": 18040
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1470764854614414e-06,
+      "loss": 3.3247,
+      "step": 18041
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1470290771175727e-06,
+      "loss": 2.6828,
+      "step": 18042
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146981668773704e-06,
+      "loss": 2.5297,
+      "step": 18043
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1469342604298357e-06,
+      "loss": 2.5603,
+      "step": 18044
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1468868520859674e-06,
+      "loss": 2.8417,
+      "step": 18045
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1468394437420987e-06,
+      "loss": 3.0811,
+      "step": 18046
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.14679203539823e-06,
+      "loss": 1.7558,
+      "step": 18047
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1467446270543616e-06,
+      "loss": 2.0933,
+      "step": 18048
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146697218710493e-06,
+      "loss": 3.1633,
+      "step": 18049
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1466498103666246e-06,
+      "loss": 1.8113,
+      "step": 18050
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1466024020227563e-06,
+      "loss": 3.0354,
+      "step": 18051
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1465549936788876e-06,
+      "loss": 2.4027,
+      "step": 18052
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146507585335019e-06,
+      "loss": 3.631,
+      "step": 18053
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1464601769911506e-06,
+      "loss": 3.2528,
+      "step": 18054
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146412768647282e-06,
+      "loss": 3.4259,
+      "step": 18055
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146365360303413e-06,
+      "loss": 3.1285,
+      "step": 18056
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146317951959545e-06,
+      "loss": 3.5494,
+      "step": 18057
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1462705436156766e-06,
+      "loss": 2.9427,
+      "step": 18058
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146223135271808e-06,
+      "loss": 2.819,
+      "step": 18059
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1461757269279395e-06,
+      "loss": 4.0329,
+      "step": 18060
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146128318584071e-06,
+      "loss": 3.7097,
+      "step": 18061
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146080910240202e-06,
+      "loss": 2.4778,
+      "step": 18062
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.146033501896334e-06,
+      "loss": 3.3717,
+      "step": 18063
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1459860935524655e-06,
+      "loss": 2.8398,
+      "step": 18064
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145938685208597e-06,
+      "loss": 2.4048,
+      "step": 18065
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1458912768647285e-06,
+      "loss": 2.174,
+      "step": 18066
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1458438685208598e-06,
+      "loss": 3.3773,
+      "step": 18067
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145796460176991e-06,
+      "loss": 3.8105,
+      "step": 18068
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1457490518331223e-06,
+      "loss": 3.7703,
+      "step": 18069
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145701643489254e-06,
+      "loss": 2.7437,
+      "step": 18070
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1456542351453858e-06,
+      "loss": 2.9131,
+      "step": 18071
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145606826801517e-06,
+      "loss": 2.0619,
+      "step": 18072
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1455594184576487e-06,
+      "loss": 2.2057,
+      "step": 18073
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.14551201011378e-06,
+      "loss": 2.679,
+      "step": 18074
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1454646017699113e-06,
+      "loss": 1.8537,
+      "step": 18075
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145417193426043e-06,
+      "loss": 2.0847,
+      "step": 18076
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1453697850821747e-06,
+      "loss": 3.6937,
+      "step": 18077
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145322376738306e-06,
+      "loss": 3.6852,
+      "step": 18078
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.1452749683944377e-06,
+      "loss": 4.1209,
+      "step": 18079
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2.145227560050569e-06,
+      "loss": 3.2648,
+      "step": 18080
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1451801517067002e-06,
+      "loss": 2.6925,
+      "step": 18081
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.145132743362832e-06,
+      "loss": 2.7886,
+      "step": 18082
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1450853350189632e-06,
+      "loss": 3.5899,
+      "step": 18083
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.145037926675095e-06,
+      "loss": 3.0345,
+      "step": 18084
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1449905183312266e-06,
+      "loss": 3.1101,
+      "step": 18085
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144943109987358e-06,
+      "loss": 2.3545,
+      "step": 18086
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144895701643489e-06,
+      "loss": 2.7694,
+      "step": 18087
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1448482932996205e-06,
+      "loss": 3.3269,
+      "step": 18088
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144800884955752e-06,
+      "loss": 3.3357,
+      "step": 18089
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144753476611884e-06,
+      "loss": 3.9116,
+      "step": 18090
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144706068268015e-06,
+      "loss": 3.4829,
+      "step": 18091
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144658659924147e-06,
+      "loss": 3.3117,
+      "step": 18092
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144611251580278e-06,
+      "loss": 2.6289,
+      "step": 18093
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1445638432364094e-06,
+      "loss": 2.8926,
+      "step": 18094
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144516434892541e-06,
+      "loss": 2.8255,
+      "step": 18095
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1444690265486724e-06,
+      "loss": 3.3117,
+      "step": 18096
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144421618204804e-06,
+      "loss": 2.6092,
+      "step": 18097
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144374209860936e-06,
+      "loss": 2.452,
+      "step": 18098
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144326801517067e-06,
+      "loss": 2.6236,
+      "step": 18099
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1442793931731984e-06,
+      "loss": 2.2638,
+      "step": 18100
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.14423198482933e-06,
+      "loss": 4.2409,
+      "step": 18101
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1441845764854614e-06,
+      "loss": 2.4588,
+      "step": 18102
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144137168141593e-06,
+      "loss": 1.8673,
+      "step": 18103
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1440897597977248e-06,
+      "loss": 3.0227,
+      "step": 18104
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.144042351453856e-06,
+      "loss": 2.898,
+      "step": 18105
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1439949431099873e-06,
+      "loss": 2.4603,
+      "step": 18106
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.143947534766119e-06,
+      "loss": 2.7031,
+      "step": 18107
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1439001264222503e-06,
+      "loss": 2.9671,
+      "step": 18108
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1438527180783816e-06,
+      "loss": 3.1268,
+      "step": 18109
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1438053097345133e-06,
+      "loss": 3.864,
+      "step": 18110
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.143757901390645e-06,
+      "loss": 3.562,
+      "step": 18111
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1437104930467763e-06,
+      "loss": 2.8229,
+      "step": 18112
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1436630847029076e-06,
+      "loss": 4.0884,
+      "step": 18113
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1436156763590393e-06,
+      "loss": 2.3881,
+      "step": 18114
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1435682680151706e-06,
+      "loss": 3.0698,
+      "step": 18115
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1435208596713023e-06,
+      "loss": 3.3606,
+      "step": 18116
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.143473451327434e-06,
+      "loss": 2.995,
+      "step": 18117
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1434260429835652e-06,
+      "loss": 2.2067,
+      "step": 18118
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1433786346396965e-06,
+      "loss": 1.7168,
+      "step": 18119
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1433312262958282e-06,
+      "loss": 3.0533,
+      "step": 18120
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1432838179519595e-06,
+      "loss": 3.6359,
+      "step": 18121
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.143236409608091e-06,
+      "loss": 2.5084,
+      "step": 18122
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1431890012642225e-06,
+      "loss": 2.867,
+      "step": 18123
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.143141592920354e-06,
+      "loss": 2.318,
+      "step": 18124
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1430941845764855e-06,
+      "loss": 3.1471,
+      "step": 18125
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.143046776232617e-06,
+      "loss": 2.4087,
+      "step": 18126
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1429993678887485e-06,
+      "loss": 3.9237,
+      "step": 18127
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1429519595448797e-06,
+      "loss": 3.9015,
+      "step": 18128
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1429045512010114e-06,
+      "loss": 3.6275,
+      "step": 18129
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.142857142857143e-06,
+      "loss": 3.5898,
+      "step": 18130
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1428097345132744e-06,
+      "loss": 3.5507,
+      "step": 18131
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1427623261694057e-06,
+      "loss": 3.4133,
+      "step": 18132
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1427149178255374e-06,
+      "loss": 3.0371,
+      "step": 18133
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1426675094816687e-06,
+      "loss": 3.1101,
+      "step": 18134
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1426201011378e-06,
+      "loss": 3.1218,
+      "step": 18135
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1425726927939317e-06,
+      "loss": 3.0435,
+      "step": 18136
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1425252844500634e-06,
+      "loss": 3.0529,
+      "step": 18137
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1424778761061947e-06,
+      "loss": 3.3977,
+      "step": 18138
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1424304677623264e-06,
+      "loss": 2.6681,
+      "step": 18139
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1423830594184577e-06,
+      "loss": 2.7757,
+      "step": 18140
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.142335651074589e-06,
+      "loss": 2.6468,
+      "step": 18141
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1422882427307206e-06,
+      "loss": 4.0898,
+      "step": 18142
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1422408343868523e-06,
+      "loss": 3.5087,
+      "step": 18143
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2.1421934260429836e-06,
+      "loss": 2.6354,
+      "step": 18144
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1421460176991153e-06,
+      "loss": 3.348,
+      "step": 18145
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1420986093552466e-06,
+      "loss": 3.1257,
+      "step": 18146
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.142051201011378e-06,
+      "loss": 2.7258,
+      "step": 18147
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1420037926675096e-06,
+      "loss": 2.8417,
+      "step": 18148
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141956384323641e-06,
+      "loss": 2.7466,
+      "step": 18149
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1419089759797726e-06,
+      "loss": 3.0933,
+      "step": 18150
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1418615676359043e-06,
+      "loss": 2.8456,
+      "step": 18151
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1418141592920356e-06,
+      "loss": 2.9806,
+      "step": 18152
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141766750948167e-06,
+      "loss": 2.6033,
+      "step": 18153
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141719342604298e-06,
+      "loss": 3.8485,
+      "step": 18154
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.14167193426043e-06,
+      "loss": 3.6974,
+      "step": 18155
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1416245259165615e-06,
+      "loss": 2.0042,
+      "step": 18156
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141577117572693e-06,
+      "loss": 3.4628,
+      "step": 18157
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1415297092288245e-06,
+      "loss": 1.6465,
+      "step": 18158
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141482300884956e-06,
+      "loss": 3.3193,
+      "step": 18159
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141434892541087e-06,
+      "loss": 2.9553,
+      "step": 18160
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1413874841972188e-06,
+      "loss": 2.4296,
+      "step": 18161
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.14134007585335e-06,
+      "loss": 2.3735,
+      "step": 18162
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1412926675094818e-06,
+      "loss": 2.1472,
+      "step": 18163
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1412452591656135e-06,
+      "loss": 3.0951,
+      "step": 18164
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1411978508217447e-06,
+      "loss": 2.817,
+      "step": 18165
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141150442477876e-06,
+      "loss": 3.8119,
+      "step": 18166
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1411030341340077e-06,
+      "loss": 4.0038,
+      "step": 18167
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.141055625790139e-06,
+      "loss": 3.1427,
+      "step": 18168
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1410082174462707e-06,
+      "loss": 2.0288,
+      "step": 18169
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1409608091024024e-06,
+      "loss": 2.8695,
+      "step": 18170
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1409134007585337e-06,
+      "loss": 2.7944,
+      "step": 18171
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140865992414665e-06,
+      "loss": 3.34,
+      "step": 18172
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1408185840707967e-06,
+      "loss": 2.4088,
+      "step": 18173
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140771175726928e-06,
+      "loss": 3.8807,
+      "step": 18174
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1407237673830592e-06,
+      "loss": 2.9865,
+      "step": 18175
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140676359039191e-06,
+      "loss": 1.9156,
+      "step": 18176
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1406289506953227e-06,
+      "loss": 2.5925,
+      "step": 18177
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140581542351454e-06,
+      "loss": 2.9491,
+      "step": 18178
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140534134007585e-06,
+      "loss": 3.7088,
+      "step": 18179
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140486725663717e-06,
+      "loss": 3.2319,
+      "step": 18180
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140439317319848e-06,
+      "loss": 2.3425,
+      "step": 18181
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.14039190897598e-06,
+      "loss": 2.2789,
+      "step": 18182
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1403445006321116e-06,
+      "loss": 3.5788,
+      "step": 18183
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140297092288243e-06,
+      "loss": 3.1929,
+      "step": 18184
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140249683944374e-06,
+      "loss": 2.0535,
+      "step": 18185
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140202275600506e-06,
+      "loss": 2.2784,
+      "step": 18186
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140154867256637e-06,
+      "loss": 2.5966,
+      "step": 18187
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1401074589127684e-06,
+      "loss": 3.2449,
+      "step": 18188
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1400600505689e-06,
+      "loss": 3.0083,
+      "step": 18189
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.140012642225032e-06,
+      "loss": 3.5042,
+      "step": 18190
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139965233881163e-06,
+      "loss": 1.9493,
+      "step": 18191
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139917825537295e-06,
+      "loss": 3.5607,
+      "step": 18192
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139870417193426e-06,
+      "loss": 3.5302,
+      "step": 18193
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1398230088495574e-06,
+      "loss": 2.6373,
+      "step": 18194
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139775600505689e-06,
+      "loss": 1.9368,
+      "step": 18195
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139728192161821e-06,
+      "loss": 3.4376,
+      "step": 18196
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139680783817952e-06,
+      "loss": 2.8635,
+      "step": 18197
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1396333754740834e-06,
+      "loss": 2.1166,
+      "step": 18198
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139585967130215e-06,
+      "loss": 2.8028,
+      "step": 18199
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1395385587863463e-06,
+      "loss": 2.644,
+      "step": 18200
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1394911504424776e-06,
+      "loss": 3.7322,
+      "step": 18201
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1394437420986093e-06,
+      "loss": 3.0934,
+      "step": 18202
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139396333754741e-06,
+      "loss": 3.0135,
+      "step": 18203
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1393489254108723e-06,
+      "loss": 2.987,
+      "step": 18204
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.139301517067004e-06,
+      "loss": 1.7736,
+      "step": 18205
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1392541087231353e-06,
+      "loss": 3.7499,
+      "step": 18206
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2.1392067003792666e-06,
+      "loss": 2.9653,
+      "step": 18207
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1391592920353983e-06,
+      "loss": 3.2343,
+      "step": 18208
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.13911188369153e-06,
+      "loss": 2.8946,
+      "step": 18209
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1390644753476613e-06,
+      "loss": 1.9253,
+      "step": 18210
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.139017067003793e-06,
+      "loss": 3.2417,
+      "step": 18211
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1389696586599242e-06,
+      "loss": 2.5389,
+      "step": 18212
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1389222503160555e-06,
+      "loss": 2.7806,
+      "step": 18213
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1388748419721872e-06,
+      "loss": 2.2026,
+      "step": 18214
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1388274336283185e-06,
+      "loss": 3.5686,
+      "step": 18215
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.13878002528445e-06,
+      "loss": 3.7945,
+      "step": 18216
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.138732616940582e-06,
+      "loss": 3.0977,
+      "step": 18217
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.138685208596713e-06,
+      "loss": 2.4992,
+      "step": 18218
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1386378002528445e-06,
+      "loss": 2.6892,
+      "step": 18219
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1385903919089758e-06,
+      "loss": 2.7816,
+      "step": 18220
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1385429835651075e-06,
+      "loss": 2.3488,
+      "step": 18221
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.138495575221239e-06,
+      "loss": 3.1425,
+      "step": 18222
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1384481668773704e-06,
+      "loss": 3.1657,
+      "step": 18223
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.138400758533502e-06,
+      "loss": 3.0194,
+      "step": 18224
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1383533501896334e-06,
+      "loss": 2.7506,
+      "step": 18225
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1383059418457647e-06,
+      "loss": 2.5866,
+      "step": 18226
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1382585335018964e-06,
+      "loss": 3.6958,
+      "step": 18227
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1382111251580277e-06,
+      "loss": 3.0878,
+      "step": 18228
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1381637168141594e-06,
+      "loss": 1.8173,
+      "step": 18229
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.138116308470291e-06,
+      "loss": 2.7012,
+      "step": 18230
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1380689001264224e-06,
+      "loss": 3.2036,
+      "step": 18231
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1380214917825537e-06,
+      "loss": 1.9703,
+      "step": 18232
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1379740834386854e-06,
+      "loss": 2.9857,
+      "step": 18233
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1379266750948166e-06,
+      "loss": 2.0425,
+      "step": 18234
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137879266750948e-06,
+      "loss": 1.7774,
+      "step": 18235
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.13783185840708e-06,
+      "loss": 3.2563,
+      "step": 18236
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1377844500632113e-06,
+      "loss": 4.0892,
+      "step": 18237
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1377370417193426e-06,
+      "loss": 2.392,
+      "step": 18238
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1376896333754743e-06,
+      "loss": 3.6627,
+      "step": 18239
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1376422250316056e-06,
+      "loss": 3.5959,
+      "step": 18240
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137594816687737e-06,
+      "loss": 2.967,
+      "step": 18241
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1375474083438686e-06,
+      "loss": 3.5579,
+      "step": 18242
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1375000000000003e-06,
+      "loss": 3.9176,
+      "step": 18243
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1374525916561316e-06,
+      "loss": 2.6492,
+      "step": 18244
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137405183312263e-06,
+      "loss": 2.0565,
+      "step": 18245
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1373577749683946e-06,
+      "loss": 2.4847,
+      "step": 18246
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137310366624526e-06,
+      "loss": 3.0977,
+      "step": 18247
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137262958280657e-06,
+      "loss": 3.1373,
+      "step": 18248
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137215549936789e-06,
+      "loss": 1.2541,
+      "step": 18249
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1371681415929205e-06,
+      "loss": 2.6175,
+      "step": 18250
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.137120733249052e-06,
+      "loss": 3.1859,
+      "step": 18251
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1370733249051835e-06,
+      "loss": 2.6377,
+      "step": 18252
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1370259165613148e-06,
+      "loss": 2.3554,
+      "step": 18253
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.136978508217446e-06,
+      "loss": 2.8965,
+      "step": 18254
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1369310998735778e-06,
+      "loss": 2.8474,
+      "step": 18255
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1368836915297095e-06,
+      "loss": 3.4422,
+      "step": 18256
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1368362831858408e-06,
+      "loss": 3.4666,
+      "step": 18257
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1367888748419725e-06,
+      "loss": 3.2944,
+      "step": 18258
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1367414664981037e-06,
+      "loss": 2.4207,
+      "step": 18259
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.136694058154235e-06,
+      "loss": 1.9654,
+      "step": 18260
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1366466498103667e-06,
+      "loss": 2.918,
+      "step": 18261
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.136599241466498e-06,
+      "loss": 3.3893,
+      "step": 18262
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1365518331226297e-06,
+      "loss": 2.4271,
+      "step": 18263
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.136504424778761e-06,
+      "loss": 3.825,
+      "step": 18264
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1364570164348927e-06,
+      "loss": 2.182,
+      "step": 18265
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.136409608091024e-06,
+      "loss": 2.8452,
+      "step": 18266
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1363621997471553e-06,
+      "loss": 1.7107,
+      "step": 18267
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.136314791403287e-06,
+      "loss": 2.8949,
+      "step": 18268
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.1362673830594187e-06,
+      "loss": 2.3667,
+      "step": 18269
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2.13621997471555e-06,
+      "loss": 2.462,
+      "step": 18270
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1361725663716816e-06,
+      "loss": 2.4422,
+      "step": 18271
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.136125158027813e-06,
+      "loss": 1.9451,
+      "step": 18272
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.136077749683944e-06,
+      "loss": 3.1918,
+      "step": 18273
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.136030341340076e-06,
+      "loss": 3.2585,
+      "step": 18274
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135982932996207e-06,
+      "loss": 3.1701,
+      "step": 18275
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135935524652339e-06,
+      "loss": 3.06,
+      "step": 18276
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1358881163084706e-06,
+      "loss": 2.0982,
+      "step": 18277
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135840707964602e-06,
+      "loss": 1.8995,
+      "step": 18278
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135793299620733e-06,
+      "loss": 2.7902,
+      "step": 18279
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135745891276865e-06,
+      "loss": 2.9778,
+      "step": 18280
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135698482932996e-06,
+      "loss": 2.3559,
+      "step": 18281
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135651074589128e-06,
+      "loss": 3.2056,
+      "step": 18282
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1356036662452596e-06,
+      "loss": 3.2592,
+      "step": 18283
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135556257901391e-06,
+      "loss": 2.3207,
+      "step": 18284
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135508849557522e-06,
+      "loss": 3.7766,
+      "step": 18285
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1354614412136534e-06,
+      "loss": 2.7696,
+      "step": 18286
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135414032869785e-06,
+      "loss": 2.6628,
+      "step": 18287
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1353666245259164e-06,
+      "loss": 3.4486,
+      "step": 18288
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135319216182048e-06,
+      "loss": 3.2892,
+      "step": 18289
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1352718078381798e-06,
+      "loss": 2.3228,
+      "step": 18290
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135224399494311e-06,
+      "loss": 3.9645,
+      "step": 18291
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1351769911504423e-06,
+      "loss": 2.6093,
+      "step": 18292
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135129582806574e-06,
+      "loss": 2.4577,
+      "step": 18293
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1350821744627053e-06,
+      "loss": 2.306,
+      "step": 18294
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.135034766118837e-06,
+      "loss": 2.1433,
+      "step": 18295
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1349873577749687e-06,
+      "loss": 2.4785,
+      "step": 18296
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1349399494311e-06,
+      "loss": 2.6063,
+      "step": 18297
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1348925410872313e-06,
+      "loss": 2.6297,
+      "step": 18298
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134845132743363e-06,
+      "loss": 2.923,
+      "step": 18299
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1347977243994943e-06,
+      "loss": 2.3054,
+      "step": 18300
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1347503160556256e-06,
+      "loss": 2.7935,
+      "step": 18301
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1347029077117573e-06,
+      "loss": 3.2145,
+      "step": 18302
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134655499367889e-06,
+      "loss": 1.3516,
+      "step": 18303
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1346080910240203e-06,
+      "loss": 3.2204,
+      "step": 18304
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134560682680152e-06,
+      "loss": 3.3949,
+      "step": 18305
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1345132743362832e-06,
+      "loss": 2.8562,
+      "step": 18306
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1344658659924145e-06,
+      "loss": 2.5758,
+      "step": 18307
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1344184576485462e-06,
+      "loss": 2.8514,
+      "step": 18308
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134371049304678e-06,
+      "loss": 2.261,
+      "step": 18309
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134323640960809e-06,
+      "loss": 2.8989,
+      "step": 18310
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1342762326169405e-06,
+      "loss": 2.7874,
+      "step": 18311
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134228824273072e-06,
+      "loss": 2.8733,
+      "step": 18312
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1341814159292035e-06,
+      "loss": 2.8061,
+      "step": 18313
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1341340075853347e-06,
+      "loss": 2.8089,
+      "step": 18314
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1340865992414665e-06,
+      "loss": 2.651,
+      "step": 18315
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.134039190897598e-06,
+      "loss": 2.3091,
+      "step": 18316
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1339917825537294e-06,
+      "loss": 3.7967,
+      "step": 18317
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.133944374209861e-06,
+      "loss": 3.1625,
+      "step": 18318
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1338969658659924e-06,
+      "loss": 2.292,
+      "step": 18319
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1338495575221237e-06,
+      "loss": 2.7907,
+      "step": 18320
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1338021491782554e-06,
+      "loss": 3.5002,
+      "step": 18321
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.133754740834387e-06,
+      "loss": 2.5279,
+      "step": 18322
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1337073324905184e-06,
+      "loss": 2.5349,
+      "step": 18323
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.13365992414665e-06,
+      "loss": 3.0638,
+      "step": 18324
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1336125158027814e-06,
+      "loss": 4.5786,
+      "step": 18325
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1335651074589127e-06,
+      "loss": 2.6555,
+      "step": 18326
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1335176991150444e-06,
+      "loss": 2.1777,
+      "step": 18327
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1334702907711756e-06,
+      "loss": 2.5254,
+      "step": 18328
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1334228824273073e-06,
+      "loss": 3.2737,
+      "step": 18329
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1333754740834386e-06,
+      "loss": 3.0113,
+      "step": 18330
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1333280657395703e-06,
+      "loss": 3.1051,
+      "step": 18331
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1332806573957016e-06,
+      "loss": 2.0222,
+      "step": 18332
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.133233249051833e-06,
+      "loss": 3.1942,
+      "step": 18333
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2.1331858407079646e-06,
+      "loss": 2.7203,
+      "step": 18334
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1331384323640963e-06,
+      "loss": 3.543,
+      "step": 18335
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1330910240202276e-06,
+      "loss": 2.9961,
+      "step": 18336
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1330436156763593e-06,
+      "loss": 2.4609,
+      "step": 18337
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1329962073324906e-06,
+      "loss": 3.0065,
+      "step": 18338
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.132948798988622e-06,
+      "loss": 1.8496,
+      "step": 18339
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1329013906447535e-06,
+      "loss": 2.0749,
+      "step": 18340
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.132853982300885e-06,
+      "loss": 3.288,
+      "step": 18341
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1328065739570165e-06,
+      "loss": 3.359,
+      "step": 18342
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1327591656131482e-06,
+      "loss": 3.931,
+      "step": 18343
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1327117572692795e-06,
+      "loss": 2.2436,
+      "step": 18344
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.132664348925411e-06,
+      "loss": 2.229,
+      "step": 18345
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1326169405815425e-06,
+      "loss": 3.5886,
+      "step": 18346
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1325695322376738e-06,
+      "loss": 2.7992,
+      "step": 18347
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1325221238938055e-06,
+      "loss": 3.0792,
+      "step": 18348
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.132474715549937e-06,
+      "loss": 2.4986,
+      "step": 18349
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1324273072060685e-06,
+      "loss": 2.5282,
+      "step": 18350
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1323798988621997e-06,
+      "loss": 3.8295,
+      "step": 18351
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.132332490518331e-06,
+      "loss": 2.4354,
+      "step": 18352
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1322850821744627e-06,
+      "loss": 3.6734,
+      "step": 18353
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.132237673830594e-06,
+      "loss": 1.4854,
+      "step": 18354
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1321902654867257e-06,
+      "loss": 3.2347,
+      "step": 18355
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1321428571428574e-06,
+      "loss": 3.3859,
+      "step": 18356
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1320954487989887e-06,
+      "loss": 1.7585,
+      "step": 18357
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.13204804045512e-06,
+      "loss": 2.8837,
+      "step": 18358
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1320006321112517e-06,
+      "loss": 2.6144,
+      "step": 18359
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131953223767383e-06,
+      "loss": 2.8712,
+      "step": 18360
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1319058154235147e-06,
+      "loss": 2.2708,
+      "step": 18361
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1318584070796464e-06,
+      "loss": 2.7709,
+      "step": 18362
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1318109987357777e-06,
+      "loss": 3.7694,
+      "step": 18363
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131763590391909e-06,
+      "loss": 3.0679,
+      "step": 18364
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1317161820480406e-06,
+      "loss": 3.2117,
+      "step": 18365
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131668773704172e-06,
+      "loss": 3.2933,
+      "step": 18366
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131621365360303e-06,
+      "loss": 2.8948,
+      "step": 18367
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131573957016435e-06,
+      "loss": 3.2535,
+      "step": 18368
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1315265486725666e-06,
+      "loss": 2.3887,
+      "step": 18369
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131479140328698e-06,
+      "loss": 2.7006,
+      "step": 18370
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1314317319848296e-06,
+      "loss": 3.0837,
+      "step": 18371
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131384323640961e-06,
+      "loss": 3.0154,
+      "step": 18372
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131336915297092e-06,
+      "loss": 2.9937,
+      "step": 18373
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131289506953224e-06,
+      "loss": 2.6512,
+      "step": 18374
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1312420986093556e-06,
+      "loss": 4.1987,
+      "step": 18375
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131194690265487e-06,
+      "loss": 4.3136,
+      "step": 18376
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131147281921618e-06,
+      "loss": 2.7387,
+      "step": 18377
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.13109987357775e-06,
+      "loss": 3.3573,
+      "step": 18378
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.131052465233881e-06,
+      "loss": 2.2549,
+      "step": 18379
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1310050568900124e-06,
+      "loss": 2.266,
+      "step": 18380
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130957648546144e-06,
+      "loss": 2.7609,
+      "step": 18381
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130910240202276e-06,
+      "loss": 3.1395,
+      "step": 18382
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130862831858407e-06,
+      "loss": 2.3593,
+      "step": 18383
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1308154235145388e-06,
+      "loss": 2.2788,
+      "step": 18384
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.13076801517067e-06,
+      "loss": 3.2669,
+      "step": 18385
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1307206068268013e-06,
+      "loss": 3.6733,
+      "step": 18386
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130673198482933e-06,
+      "loss": 3.6947,
+      "step": 18387
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1306257901390647e-06,
+      "loss": 3.4152,
+      "step": 18388
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130578381795196e-06,
+      "loss": 3.0743,
+      "step": 18389
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1305309734513277e-06,
+      "loss": 4.2345,
+      "step": 18390
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130483565107459e-06,
+      "loss": 3.1557,
+      "step": 18391
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1304361567635903e-06,
+      "loss": 2.6825,
+      "step": 18392
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130388748419722e-06,
+      "loss": 2.52,
+      "step": 18393
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1303413400758533e-06,
+      "loss": 2.8471,
+      "step": 18394
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130293931731985e-06,
+      "loss": 2.7915,
+      "step": 18395
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.1302465233881163e-06,
+      "loss": 4.2129,
+      "step": 18396
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2.130199115044248e-06,
+      "loss": 3.0041,
+      "step": 18397
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1301517067003792e-06,
+      "loss": 2.4027,
+      "step": 18398
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1301042983565105e-06,
+      "loss": 1.6778,
+      "step": 18399
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1300568900126422e-06,
+      "loss": 2.8551,
+      "step": 18400
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.130009481668774e-06,
+      "loss": 3.189,
+      "step": 18401
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1299620733249052e-06,
+      "loss": 1.9443,
+      "step": 18402
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129914664981037e-06,
+      "loss": 2.7127,
+      "step": 18403
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129867256637168e-06,
+      "loss": 1.7934,
+      "step": 18404
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1298198482932995e-06,
+      "loss": 3.3703,
+      "step": 18405
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129772439949431e-06,
+      "loss": 3.5788,
+      "step": 18406
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1297250316055625e-06,
+      "loss": 3.6044,
+      "step": 18407
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129677623261694e-06,
+      "loss": 3.2232,
+      "step": 18408
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129630214917826e-06,
+      "loss": 3.667,
+      "step": 18409
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129582806573957e-06,
+      "loss": 4.0369,
+      "step": 18410
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1295353982300884e-06,
+      "loss": 1.4503,
+      "step": 18411
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.12948798988622e-06,
+      "loss": 2.6362,
+      "step": 18412
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1294405815423514e-06,
+      "loss": 2.8037,
+      "step": 18413
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129393173198483e-06,
+      "loss": 3.3408,
+      "step": 18414
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129345764854615e-06,
+      "loss": 2.5953,
+      "step": 18415
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129298356510746e-06,
+      "loss": 3.6325,
+      "step": 18416
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1292509481668774e-06,
+      "loss": 1.8708,
+      "step": 18417
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1292035398230087e-06,
+      "loss": 3.6372,
+      "step": 18418
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1291561314791404e-06,
+      "loss": 2.8256,
+      "step": 18419
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1291087231352716e-06,
+      "loss": 2.6731,
+      "step": 18420
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1290613147914034e-06,
+      "loss": 2.5319,
+      "step": 18421
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.129013906447535e-06,
+      "loss": 3.3984,
+      "step": 18422
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1289664981036663e-06,
+      "loss": 3.2382,
+      "step": 18423
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1289190897597976e-06,
+      "loss": 3.1736,
+      "step": 18424
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1288716814159293e-06,
+      "loss": 2.4674,
+      "step": 18425
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1288242730720606e-06,
+      "loss": 2.9356,
+      "step": 18426
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1287768647281923e-06,
+      "loss": 2.8136,
+      "step": 18427
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.128729456384324e-06,
+      "loss": 2.2587,
+      "step": 18428
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1286820480404553e-06,
+      "loss": 4.3863,
+      "step": 18429
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1286346396965866e-06,
+      "loss": 2.8509,
+      "step": 18430
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1285872313527183e-06,
+      "loss": 2.6119,
+      "step": 18431
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1285398230088496e-06,
+      "loss": 2.2223,
+      "step": 18432
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.128492414664981e-06,
+      "loss": 3.4701,
+      "step": 18433
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1284450063211125e-06,
+      "loss": 2.924,
+      "step": 18434
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1283975979772442e-06,
+      "loss": 1.5601,
+      "step": 18435
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1283501896333755e-06,
+      "loss": 3.8101,
+      "step": 18436
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1283027812895072e-06,
+      "loss": 2.7475,
+      "step": 18437
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1282553729456385e-06,
+      "loss": 2.4473,
+      "step": 18438
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.12820796460177e-06,
+      "loss": 3.6707,
+      "step": 18439
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1281605562579015e-06,
+      "loss": 3.6545,
+      "step": 18440
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.128113147914033e-06,
+      "loss": 2.4137,
+      "step": 18441
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1280657395701645e-06,
+      "loss": 2.2711,
+      "step": 18442
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1280183312262958e-06,
+      "loss": 2.3837,
+      "step": 18443
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1279709228824275e-06,
+      "loss": 3.5051,
+      "step": 18444
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1279235145385587e-06,
+      "loss": 3.6095,
+      "step": 18445
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.12787610619469e-06,
+      "loss": 3.5372,
+      "step": 18446
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1278286978508217e-06,
+      "loss": 2.7586,
+      "step": 18447
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1277812895069534e-06,
+      "loss": 2.5098,
+      "step": 18448
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1277338811630847e-06,
+      "loss": 1.544,
+      "step": 18449
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1276864728192164e-06,
+      "loss": 2.7654,
+      "step": 18450
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1276390644753477e-06,
+      "loss": 2.629,
+      "step": 18451
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.127591656131479e-06,
+      "loss": 3.3236,
+      "step": 18452
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1275442477876107e-06,
+      "loss": 3.6181,
+      "step": 18453
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1274968394437424e-06,
+      "loss": 3.8094,
+      "step": 18454
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1274494310998737e-06,
+      "loss": 2.8398,
+      "step": 18455
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1274020227560054e-06,
+      "loss": 2.8452,
+      "step": 18456
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1273546144121366e-06,
+      "loss": 3.5792,
+      "step": 18457
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.127307206068268e-06,
+      "loss": 2.9204,
+      "step": 18458
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.1272597977243996e-06,
+      "loss": 3.3215,
+      "step": 18459
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2.127212389380531e-06,
+      "loss": 3.3106,
+      "step": 18460
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1271649810366626e-06,
+      "loss": 1.7875,
+      "step": 18461
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.127117572692794e-06,
+      "loss": 2.1644,
+      "step": 18462
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1270701643489256e-06,
+      "loss": 2.3575,
+      "step": 18463
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.127022756005057e-06,
+      "loss": 3.1704,
+      "step": 18464
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126975347661188e-06,
+      "loss": 2.1475,
+      "step": 18465
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.12692793931732e-06,
+      "loss": 2.5402,
+      "step": 18466
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1268805309734516e-06,
+      "loss": 1.5659,
+      "step": 18467
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126833122629583e-06,
+      "loss": 3.5879,
+      "step": 18468
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1267857142857146e-06,
+      "loss": 2.9851,
+      "step": 18469
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126738305941846e-06,
+      "loss": 3.0897,
+      "step": 18470
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126690897597977e-06,
+      "loss": 3.3165,
+      "step": 18471
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126643489254109e-06,
+      "loss": 3.1373,
+      "step": 18472
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.12659608091024e-06,
+      "loss": 3.1305,
+      "step": 18473
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126548672566372e-06,
+      "loss": 2.5435,
+      "step": 18474
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1265012642225035e-06,
+      "loss": 3.4239,
+      "step": 18475
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1264538558786348e-06,
+      "loss": 1.9414,
+      "step": 18476
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126406447534766e-06,
+      "loss": 1.7558,
+      "step": 18477
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1263590391908978e-06,
+      "loss": 3.0941,
+      "step": 18478
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126311630847029e-06,
+      "loss": 2.2903,
+      "step": 18479
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1262642225031608e-06,
+      "loss": 3.8032,
+      "step": 18480
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1262168141592925e-06,
+      "loss": 2.0256,
+      "step": 18481
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1261694058154237e-06,
+      "loss": 3.1461,
+      "step": 18482
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126121997471555e-06,
+      "loss": 3.3362,
+      "step": 18483
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1260745891276863e-06,
+      "loss": 2.0437,
+      "step": 18484
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.126027180783818e-06,
+      "loss": 2.3543,
+      "step": 18485
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1259797724399493e-06,
+      "loss": 3.3402,
+      "step": 18486
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125932364096081e-06,
+      "loss": 3.2988,
+      "step": 18487
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1258849557522127e-06,
+      "loss": 1.7795,
+      "step": 18488
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125837547408344e-06,
+      "loss": 2.9308,
+      "step": 18489
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1257901390644753e-06,
+      "loss": 3.7591,
+      "step": 18490
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125742730720607e-06,
+      "loss": 2.8076,
+      "step": 18491
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1256953223767382e-06,
+      "loss": 3.0022,
+      "step": 18492
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.12564791403287e-06,
+      "loss": 3.4805,
+      "step": 18493
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1256005056890016e-06,
+      "loss": 2.6362,
+      "step": 18494
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125553097345133e-06,
+      "loss": 3.0557,
+      "step": 18495
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125505689001264e-06,
+      "loss": 2.785,
+      "step": 18496
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125458280657396e-06,
+      "loss": 2.6529,
+      "step": 18497
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125410872313527e-06,
+      "loss": 2.4556,
+      "step": 18498
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1253634639696585e-06,
+      "loss": 2.2357,
+      "step": 18499
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.12531605562579e-06,
+      "loss": 4.1093,
+      "step": 18500
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125268647281922e-06,
+      "loss": 2.1922,
+      "step": 18501
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125221238938053e-06,
+      "loss": 2.2743,
+      "step": 18502
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125173830594185e-06,
+      "loss": 2.9764,
+      "step": 18503
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.125126422250316e-06,
+      "loss": 2.806,
+      "step": 18504
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1250790139064474e-06,
+      "loss": 2.2875,
+      "step": 18505
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1250316055625787e-06,
+      "loss": 3.4599,
+      "step": 18506
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.124984197218711e-06,
+      "loss": 2.324,
+      "step": 18507
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.124936788874842e-06,
+      "loss": 2.3413,
+      "step": 18508
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1248893805309734e-06,
+      "loss": 3.557,
+      "step": 18509
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.124841972187105e-06,
+      "loss": 3.9697,
+      "step": 18510
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1247945638432364e-06,
+      "loss": 3.7136,
+      "step": 18511
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1247471554993677e-06,
+      "loss": 2.972,
+      "step": 18512
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1246997471554994e-06,
+      "loss": 2.6714,
+      "step": 18513
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.124652338811631e-06,
+      "loss": 2.4792,
+      "step": 18514
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1246049304677623e-06,
+      "loss": 3.215,
+      "step": 18515
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.124557522123894e-06,
+      "loss": 3.3463,
+      "step": 18516
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1245101137800253e-06,
+      "loss": 3.0513,
+      "step": 18517
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1244627054361566e-06,
+      "loss": 2.8581,
+      "step": 18518
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1244152970922883e-06,
+      "loss": 2.6673,
+      "step": 18519
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1243678887484196e-06,
+      "loss": 2.8445,
+      "step": 18520
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1243204804045513e-06,
+      "loss": 2.521,
+      "step": 18521
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.124273072060683e-06,
+      "loss": 2.4583,
+      "step": 18522
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1242256637168143e-06,
+      "loss": 3.7138,
+      "step": 18523
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2.1241782553729456e-06,
+      "loss": 3.2643,
+      "step": 18524
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.124130847029077e-06,
+      "loss": 3.0393,
+      "step": 18525
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1240834386852085e-06,
+      "loss": 2.9837,
+      "step": 18526
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1240360303413403e-06,
+      "loss": 2.9694,
+      "step": 18527
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1239886219974715e-06,
+      "loss": 2.7728,
+      "step": 18528
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1239412136536032e-06,
+      "loss": 2.586,
+      "step": 18529
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1238938053097345e-06,
+      "loss": 1.8752,
+      "step": 18530
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.123846396965866e-06,
+      "loss": 3.0689,
+      "step": 18531
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1237989886219975e-06,
+      "loss": 3.0379,
+      "step": 18532
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1237515802781288e-06,
+      "loss": 3.0556,
+      "step": 18533
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1237041719342605e-06,
+      "loss": 2.7705,
+      "step": 18534
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.123656763590392e-06,
+      "loss": 3.0087,
+      "step": 18535
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1236093552465235e-06,
+      "loss": 3.3342,
+      "step": 18536
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1235619469026548e-06,
+      "loss": 3.4634,
+      "step": 18537
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1235145385587865e-06,
+      "loss": 3.0783,
+      "step": 18538
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1234671302149177e-06,
+      "loss": 2.8402,
+      "step": 18539
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1234197218710494e-06,
+      "loss": 3.7881,
+      "step": 18540
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.123372313527181e-06,
+      "loss": 2.9755,
+      "step": 18541
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1233249051833124e-06,
+      "loss": 3.6616,
+      "step": 18542
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1232774968394437e-06,
+      "loss": 3.3265,
+      "step": 18543
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1232300884955754e-06,
+      "loss": 3.3218,
+      "step": 18544
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1231826801517067e-06,
+      "loss": 3.4148,
+      "step": 18545
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.123135271807838e-06,
+      "loss": 2.1797,
+      "step": 18546
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1230878634639697e-06,
+      "loss": 2.9803,
+      "step": 18547
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1230404551201014e-06,
+      "loss": 1.626,
+      "step": 18548
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1229930467762327e-06,
+      "loss": 3.5106,
+      "step": 18549
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122945638432364e-06,
+      "loss": 3.1249,
+      "step": 18550
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1228982300884956e-06,
+      "loss": 3.0381,
+      "step": 18551
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122850821744627e-06,
+      "loss": 3.167,
+      "step": 18552
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1228034134007586e-06,
+      "loss": 3.2369,
+      "step": 18553
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1227560050568903e-06,
+      "loss": 2.2788,
+      "step": 18554
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1227085967130216e-06,
+      "loss": 2.366,
+      "step": 18555
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122661188369153e-06,
+      "loss": 2.408,
+      "step": 18556
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1226137800252846e-06,
+      "loss": 2.9126,
+      "step": 18557
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122566371681416e-06,
+      "loss": 3.7404,
+      "step": 18558
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122518963337547e-06,
+      "loss": 2.5527,
+      "step": 18559
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122471554993679e-06,
+      "loss": 2.9516,
+      "step": 18560
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1224241466498106e-06,
+      "loss": 2.0752,
+      "step": 18561
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122376738305942e-06,
+      "loss": 2.9044,
+      "step": 18562
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1223293299620735e-06,
+      "loss": 2.5557,
+      "step": 18563
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122281921618205e-06,
+      "loss": 1.6629,
+      "step": 18564
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122234513274336e-06,
+      "loss": 2.6433,
+      "step": 18565
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122187104930468e-06,
+      "loss": 1.9188,
+      "step": 18566
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1221396965865995e-06,
+      "loss": 2.6055,
+      "step": 18567
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122092288242731e-06,
+      "loss": 3.1791,
+      "step": 18568
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.122044879898862e-06,
+      "loss": 2.4876,
+      "step": 18569
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1219974715549938e-06,
+      "loss": 2.5703,
+      "step": 18570
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121950063211125e-06,
+      "loss": 2.5529,
+      "step": 18571
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1219026548672563e-06,
+      "loss": 3.636,
+      "step": 18572
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121855246523388e-06,
+      "loss": 2.0099,
+      "step": 18573
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1218078381795198e-06,
+      "loss": 3.5203,
+      "step": 18574
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121760429835651e-06,
+      "loss": 3.1365,
+      "step": 18575
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1217130214917827e-06,
+      "loss": 2.5377,
+      "step": 18576
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121665613147914e-06,
+      "loss": 4.5156,
+      "step": 18577
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1216182048040453e-06,
+      "loss": 2.5977,
+      "step": 18578
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121570796460177e-06,
+      "loss": 2.9221,
+      "step": 18579
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1215233881163087e-06,
+      "loss": 3.2358,
+      "step": 18580
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.12147597977244e-06,
+      "loss": 2.3522,
+      "step": 18581
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1214285714285717e-06,
+      "loss": 2.3819,
+      "step": 18582
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121381163084703e-06,
+      "loss": 3.2104,
+      "step": 18583
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1213337547408342e-06,
+      "loss": 2.5377,
+      "step": 18584
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121286346396966e-06,
+      "loss": 2.5061,
+      "step": 18585
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.1212389380530972e-06,
+      "loss": 2.107,
+      "step": 18586
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2.121191529709229e-06,
+      "loss": 3.24,
+      "step": 18587
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1211441213653606e-06,
+      "loss": 2.6238,
+      "step": 18588
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.121096713021492e-06,
+      "loss": 3.126,
+      "step": 18589
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.121049304677623e-06,
+      "loss": 3.6452,
+      "step": 18590
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1210018963337545e-06,
+      "loss": 2.8482,
+      "step": 18591
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120954487989886e-06,
+      "loss": 2.8696,
+      "step": 18592
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120907079646018e-06,
+      "loss": 3.4099,
+      "step": 18593
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120859671302149e-06,
+      "loss": 2.6155,
+      "step": 18594
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120812262958281e-06,
+      "loss": 3.1732,
+      "step": 18595
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120764854614412e-06,
+      "loss": 3.1176,
+      "step": 18596
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1207174462705434e-06,
+      "loss": 2.6066,
+      "step": 18597
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120670037926675e-06,
+      "loss": 3.3505,
+      "step": 18598
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1206226295828064e-06,
+      "loss": 3.7758,
+      "step": 18599
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120575221238938e-06,
+      "loss": 2.534,
+      "step": 18600
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.12052781289507e-06,
+      "loss": 3.8232,
+      "step": 18601
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120480404551201e-06,
+      "loss": 2.5441,
+      "step": 18602
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1204329962073324e-06,
+      "loss": 4.137,
+      "step": 18603
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120385587863464e-06,
+      "loss": 2.4937,
+      "step": 18604
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1203381795195954e-06,
+      "loss": 2.7819,
+      "step": 18605
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120290771175727e-06,
+      "loss": 3.4203,
+      "step": 18606
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1202433628318588e-06,
+      "loss": 3.5754,
+      "step": 18607
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.12019595448799e-06,
+      "loss": 2.0524,
+      "step": 18608
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1201485461441213e-06,
+      "loss": 2.6515,
+      "step": 18609
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.120101137800253e-06,
+      "loss": 2.4379,
+      "step": 18610
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1200537294563843e-06,
+      "loss": 3.152,
+      "step": 18611
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1200063211125156e-06,
+      "loss": 3.3528,
+      "step": 18612
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1199589127686473e-06,
+      "loss": 2.7741,
+      "step": 18613
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.119911504424779e-06,
+      "loss": 2.7206,
+      "step": 18614
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1198640960809103e-06,
+      "loss": 2.9137,
+      "step": 18615
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1198166877370416e-06,
+      "loss": 3.0211,
+      "step": 18616
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1197692793931733e-06,
+      "loss": 2.4039,
+      "step": 18617
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1197218710493046e-06,
+      "loss": 2.3464,
+      "step": 18618
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1196744627054363e-06,
+      "loss": 2.9818,
+      "step": 18619
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.119627054361568e-06,
+      "loss": 2.551,
+      "step": 18620
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1195796460176992e-06,
+      "loss": 2.473,
+      "step": 18621
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1195322376738305e-06,
+      "loss": 2.8972,
+      "step": 18622
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1194848293299622e-06,
+      "loss": 4.0026,
+      "step": 18623
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1194374209860935e-06,
+      "loss": 3.6922,
+      "step": 18624
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.119390012642225e-06,
+      "loss": 2.4264,
+      "step": 18625
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1193426042983565e-06,
+      "loss": 2.9648,
+      "step": 18626
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.119295195954488e-06,
+      "loss": 2.6994,
+      "step": 18627
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1192477876106195e-06,
+      "loss": 3.3491,
+      "step": 18628
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.119200379266751e-06,
+      "loss": 2.1684,
+      "step": 18629
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1191529709228825e-06,
+      "loss": 3.3414,
+      "step": 18630
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1191055625790137e-06,
+      "loss": 2.78,
+      "step": 18631
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1190581542351454e-06,
+      "loss": 3.8485,
+      "step": 18632
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.119010745891277e-06,
+      "loss": 2.9403,
+      "step": 18633
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1189633375474084e-06,
+      "loss": 2.9987,
+      "step": 18634
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1189159292035397e-06,
+      "loss": 2.0137,
+      "step": 18635
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1188685208596714e-06,
+      "loss": 2.8035,
+      "step": 18636
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1188211125158027e-06,
+      "loss": 2.4307,
+      "step": 18637
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.118773704171934e-06,
+      "loss": 3.7016,
+      "step": 18638
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1187262958280657e-06,
+      "loss": 2.5344,
+      "step": 18639
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1186788874841974e-06,
+      "loss": 2.9227,
+      "step": 18640
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1186314791403287e-06,
+      "loss": 2.1269,
+      "step": 18641
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1185840707964604e-06,
+      "loss": 2.8906,
+      "step": 18642
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1185366624525917e-06,
+      "loss": 2.6628,
+      "step": 18643
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.118489254108723e-06,
+      "loss": 2.9614,
+      "step": 18644
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1184418457648546e-06,
+      "loss": 2.8563,
+      "step": 18645
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1183944374209863e-06,
+      "loss": 2.0625,
+      "step": 18646
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1183470290771176e-06,
+      "loss": 2.2752,
+      "step": 18647
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1182996207332493e-06,
+      "loss": 2.2666,
+      "step": 18648
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.1182522123893806e-06,
+      "loss": 3.7692,
+      "step": 18649
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2.118204804045512e-06,
+      "loss": 2.005,
+      "step": 18650
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1181573957016436e-06,
+      "loss": 2.2908,
+      "step": 18651
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.118109987357775e-06,
+      "loss": 2.4444,
+      "step": 18652
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1180625790139066e-06,
+      "loss": 2.3925,
+      "step": 18653
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1180151706700383e-06,
+      "loss": 2.8935,
+      "step": 18654
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1179677623261696e-06,
+      "loss": 1.7679,
+      "step": 18655
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117920353982301e-06,
+      "loss": 2.3135,
+      "step": 18656
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117872945638432e-06,
+      "loss": 3.6718,
+      "step": 18657
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117825537294564e-06,
+      "loss": 2.4774,
+      "step": 18658
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1177781289506955e-06,
+      "loss": 2.9268,
+      "step": 18659
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117730720606827e-06,
+      "loss": 3.1304,
+      "step": 18660
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1176833122629585e-06,
+      "loss": 2.5869,
+      "step": 18661
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.11763590391909e-06,
+      "loss": 2.6704,
+      "step": 18662
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117588495575221e-06,
+      "loss": 3.776,
+      "step": 18663
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1175410872313528e-06,
+      "loss": 2.9204,
+      "step": 18664
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117493678887484e-06,
+      "loss": 3.2364,
+      "step": 18665
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1174462705436158e-06,
+      "loss": 3.1016,
+      "step": 18666
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1173988621997475e-06,
+      "loss": 3.7235,
+      "step": 18667
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1173514538558787e-06,
+      "loss": 2.9717,
+      "step": 18668
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.11730404551201e-06,
+      "loss": 2.7961,
+      "step": 18669
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1172566371681417e-06,
+      "loss": 3.7812,
+      "step": 18670
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117209228824273e-06,
+      "loss": 2.9402,
+      "step": 18671
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1171618204804047e-06,
+      "loss": 3.4614,
+      "step": 18672
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1171144121365364e-06,
+      "loss": 3.6688,
+      "step": 18673
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1170670037926677e-06,
+      "loss": 3.1045,
+      "step": 18674
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.117019595448799e-06,
+      "loss": 2.5318,
+      "step": 18675
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1169721871049307e-06,
+      "loss": 3.9821,
+      "step": 18676
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116924778761062e-06,
+      "loss": 2.3336,
+      "step": 18677
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1168773704171932e-06,
+      "loss": 3.518,
+      "step": 18678
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116829962073325e-06,
+      "loss": 3.7507,
+      "step": 18679
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1167825537294567e-06,
+      "loss": 3.1821,
+      "step": 18680
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116735145385588e-06,
+      "loss": 2.9002,
+      "step": 18681
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116687737041719e-06,
+      "loss": 2.8552,
+      "step": 18682
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116640328697851e-06,
+      "loss": 2.7115,
+      "step": 18683
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116592920353982e-06,
+      "loss": 3.0657,
+      "step": 18684
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116545512010114e-06,
+      "loss": 3.0072,
+      "step": 18685
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1164981036662456e-06,
+      "loss": 3.8177,
+      "step": 18686
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116450695322377e-06,
+      "loss": 2.6028,
+      "step": 18687
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116403286978508e-06,
+      "loss": 1.7015,
+      "step": 18688
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.11635587863464e-06,
+      "loss": 2.4941,
+      "step": 18689
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116308470290771e-06,
+      "loss": 3.6448,
+      "step": 18690
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1162610619469024e-06,
+      "loss": 2.9165,
+      "step": 18691
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116213653603034e-06,
+      "loss": 3.1453,
+      "step": 18692
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116166245259166e-06,
+      "loss": 3.3141,
+      "step": 18693
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116118836915297e-06,
+      "loss": 4.5051,
+      "step": 18694
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.116071428571429e-06,
+      "loss": 2.9344,
+      "step": 18695
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.11602402022756e-06,
+      "loss": 2.3376,
+      "step": 18696
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1159766118836914e-06,
+      "loss": 1.8709,
+      "step": 18697
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115929203539823e-06,
+      "loss": 4.0686,
+      "step": 18698
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115881795195955e-06,
+      "loss": 3.8444,
+      "step": 18699
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115834386852086e-06,
+      "loss": 2.9576,
+      "step": 18700
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1157869785082174e-06,
+      "loss": 3.0579,
+      "step": 18701
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115739570164349e-06,
+      "loss": 2.622,
+      "step": 18702
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1156921618204803e-06,
+      "loss": 1.9927,
+      "step": 18703
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1156447534766116e-06,
+      "loss": 1.9201,
+      "step": 18704
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1155973451327433e-06,
+      "loss": 2.2024,
+      "step": 18705
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115549936788875e-06,
+      "loss": 2.4223,
+      "step": 18706
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1155025284450063e-06,
+      "loss": 3.1621,
+      "step": 18707
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115455120101138e-06,
+      "loss": 2.9854,
+      "step": 18708
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1154077117572693e-06,
+      "loss": 2.2123,
+      "step": 18709
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1153603034134006e-06,
+      "loss": 2.4556,
+      "step": 18710
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1153128950695323e-06,
+      "loss": 3.3432,
+      "step": 18711
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115265486725664e-06,
+      "loss": 2.9946,
+      "step": 18712
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.1152180783817953e-06,
+      "loss": 3.7004,
+      "step": 18713
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2.115170670037927e-06,
+      "loss": 3.0427,
+      "step": 18714
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1151232616940582e-06,
+      "loss": 3.3803,
+      "step": 18715
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1150758533501895e-06,
+      "loss": 3.1538,
+      "step": 18716
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1150284450063212e-06,
+      "loss": 2.6327,
+      "step": 18717
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1149810366624525e-06,
+      "loss": 2.863,
+      "step": 18718
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.114933628318584e-06,
+      "loss": 3.5513,
+      "step": 18719
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.114886219974716e-06,
+      "loss": 2.1318,
+      "step": 18720
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.114838811630847e-06,
+      "loss": 3.5736,
+      "step": 18721
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1147914032869785e-06,
+      "loss": 3.4644,
+      "step": 18722
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1147439949431098e-06,
+      "loss": 3.1336,
+      "step": 18723
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1146965865992415e-06,
+      "loss": 2.5057,
+      "step": 18724
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.114649178255373e-06,
+      "loss": 1.9543,
+      "step": 18725
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1146017699115044e-06,
+      "loss": 2.8398,
+      "step": 18726
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.114554361567636e-06,
+      "loss": 3.8704,
+      "step": 18727
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1145069532237674e-06,
+      "loss": 2.6669,
+      "step": 18728
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1144595448798987e-06,
+      "loss": 2.3142,
+      "step": 18729
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1144121365360304e-06,
+      "loss": 2.7811,
+      "step": 18730
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1143647281921617e-06,
+      "loss": 3.8703,
+      "step": 18731
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1143173198482934e-06,
+      "loss": 3.2344,
+      "step": 18732
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.114269911504425e-06,
+      "loss": 1.8157,
+      "step": 18733
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1142225031605564e-06,
+      "loss": 2.3537,
+      "step": 18734
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1141750948166877e-06,
+      "loss": 3.7633,
+      "step": 18735
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1141276864728194e-06,
+      "loss": 2.3871,
+      "step": 18736
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1140802781289506e-06,
+      "loss": 3.0514,
+      "step": 18737
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1140328697850823e-06,
+      "loss": 3.1058,
+      "step": 18738
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.113985461441214e-06,
+      "loss": 2.879,
+      "step": 18739
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1139380530973453e-06,
+      "loss": 3.4254,
+      "step": 18740
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1138906447534766e-06,
+      "loss": 1.4935,
+      "step": 18741
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1138432364096083e-06,
+      "loss": 2.5454,
+      "step": 18742
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1137958280657396e-06,
+      "loss": 1.6153,
+      "step": 18743
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.113748419721871e-06,
+      "loss": 1.8677,
+      "step": 18744
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1137010113780026e-06,
+      "loss": 3.6892,
+      "step": 18745
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1136536030341343e-06,
+      "loss": 3.2611,
+      "step": 18746
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1136061946902656e-06,
+      "loss": 3.3869,
+      "step": 18747
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.113558786346397e-06,
+      "loss": 3.2412,
+      "step": 18748
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1135113780025286e-06,
+      "loss": 2.849,
+      "step": 18749
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.11346396965866e-06,
+      "loss": 2.9537,
+      "step": 18750
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1134165613147915e-06,
+      "loss": 2.7183,
+      "step": 18751
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1133691529709232e-06,
+      "loss": 3.2469,
+      "step": 18752
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1133217446270545e-06,
+      "loss": 3.3222,
+      "step": 18753
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.113274336283186e-06,
+      "loss": 2.4525,
+      "step": 18754
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1132269279393175e-06,
+      "loss": 3.4556,
+      "step": 18755
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1131795195954488e-06,
+      "loss": 2.7699,
+      "step": 18756
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.11313211125158e-06,
+      "loss": 3.7999,
+      "step": 18757
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1130847029077118e-06,
+      "loss": 3.3329,
+      "step": 18758
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1130372945638435e-06,
+      "loss": 2.2482,
+      "step": 18759
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1129898862199748e-06,
+      "loss": 2.5624,
+      "step": 18760
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1129424778761065e-06,
+      "loss": 3.358,
+      "step": 18761
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1128950695322377e-06,
+      "loss": 2.2023,
+      "step": 18762
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112847661188369e-06,
+      "loss": 2.8143,
+      "step": 18763
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1128002528445007e-06,
+      "loss": 3.0688,
+      "step": 18764
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1127528445006324e-06,
+      "loss": 3.6854,
+      "step": 18765
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1127054361567637e-06,
+      "loss": 2.5046,
+      "step": 18766
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112658027812895e-06,
+      "loss": 3.5157,
+      "step": 18767
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1126106194690267e-06,
+      "loss": 2.2034,
+      "step": 18768
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112563211125158e-06,
+      "loss": 2.4308,
+      "step": 18769
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1125158027812893e-06,
+      "loss": 3.5998,
+      "step": 18770
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112468394437421e-06,
+      "loss": 3.6156,
+      "step": 18771
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1124209860935527e-06,
+      "loss": 3.107,
+      "step": 18772
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112373577749684e-06,
+      "loss": 1.9776,
+      "step": 18773
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.1123261694058156e-06,
+      "loss": 1.9737,
+      "step": 18774
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112278761061947e-06,
+      "loss": 3.8903,
+      "step": 18775
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.112231352718078e-06,
+      "loss": 2.0367,
+      "step": 18776
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2.11218394437421e-06,
+      "loss": 2.6502,
+      "step": 18777
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1121365360303416e-06,
+      "loss": 2.8794,
+      "step": 18778
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.112089127686473e-06,
+      "loss": 2.9156,
+      "step": 18779
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1120417193426046e-06,
+      "loss": 3.4723,
+      "step": 18780
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111994310998736e-06,
+      "loss": 3.8409,
+      "step": 18781
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111946902654867e-06,
+      "loss": 3.1839,
+      "step": 18782
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111899494310999e-06,
+      "loss": 3.4303,
+      "step": 18783
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.11185208596713e-06,
+      "loss": 1.9917,
+      "step": 18784
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111804677623262e-06,
+      "loss": 1.7578,
+      "step": 18785
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1117572692793936e-06,
+      "loss": 3.1181,
+      "step": 18786
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111709860935525e-06,
+      "loss": 2.9187,
+      "step": 18787
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111662452591656e-06,
+      "loss": 3.5834,
+      "step": 18788
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1116150442477874e-06,
+      "loss": 3.3799,
+      "step": 18789
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111567635903919e-06,
+      "loss": 2.4233,
+      "step": 18790
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1115202275600504e-06,
+      "loss": 3.5463,
+      "step": 18791
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111472819216182e-06,
+      "loss": 2.5507,
+      "step": 18792
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1114254108723138e-06,
+      "loss": 2.1712,
+      "step": 18793
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111378002528445e-06,
+      "loss": 2.7353,
+      "step": 18794
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1113305941845763e-06,
+      "loss": 3.0112,
+      "step": 18795
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111283185840708e-06,
+      "loss": 3.2117,
+      "step": 18796
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1112357774968393e-06,
+      "loss": 2.9305,
+      "step": 18797
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111188369152971e-06,
+      "loss": 2.8517,
+      "step": 18798
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1111409608091027e-06,
+      "loss": 3.0982,
+      "step": 18799
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.111093552465234e-06,
+      "loss": 3.6522,
+      "step": 18800
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1110461441213653e-06,
+      "loss": 3.1986,
+      "step": 18801
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110998735777497e-06,
+      "loss": 3.1005,
+      "step": 18802
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1109513274336283e-06,
+      "loss": 3.7658,
+      "step": 18803
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1109039190897596e-06,
+      "loss": 2.8017,
+      "step": 18804
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1108565107458913e-06,
+      "loss": 3.7769,
+      "step": 18805
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110809102402023e-06,
+      "loss": 2.6333,
+      "step": 18806
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1107616940581543e-06,
+      "loss": 2.5773,
+      "step": 18807
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110714285714286e-06,
+      "loss": 2.4637,
+      "step": 18808
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1106668773704172e-06,
+      "loss": 3.2999,
+      "step": 18809
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1106194690265485e-06,
+      "loss": 3.0615,
+      "step": 18810
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1105720606826802e-06,
+      "loss": 4.0665,
+      "step": 18811
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110524652338812e-06,
+      "loss": 3.6795,
+      "step": 18812
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110477243994943e-06,
+      "loss": 3.6066,
+      "step": 18813
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1104298356510745e-06,
+      "loss": 3.2126,
+      "step": 18814
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110382427307206e-06,
+      "loss": 3.1974,
+      "step": 18815
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1103350189633375e-06,
+      "loss": 2.5107,
+      "step": 18816
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1102876106194687e-06,
+      "loss": 2.7873,
+      "step": 18817
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1102402022756005e-06,
+      "loss": 3.429,
+      "step": 18818
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110192793931732e-06,
+      "loss": 3.0085,
+      "step": 18819
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1101453855878634e-06,
+      "loss": 2.9536,
+      "step": 18820
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.110097977243995e-06,
+      "loss": 3.0419,
+      "step": 18821
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1100505689001264e-06,
+      "loss": 2.8654,
+      "step": 18822
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1100031605562577e-06,
+      "loss": 3.8341,
+      "step": 18823
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1099557522123894e-06,
+      "loss": 3.8853,
+      "step": 18824
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.109908343868521e-06,
+      "loss": 3.8882,
+      "step": 18825
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1098609355246524e-06,
+      "loss": 3.5179,
+      "step": 18826
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.109813527180784e-06,
+      "loss": 2.0539,
+      "step": 18827
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1097661188369154e-06,
+      "loss": 3.9533,
+      "step": 18828
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1097187104930467e-06,
+      "loss": 1.3278,
+      "step": 18829
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1096713021491784e-06,
+      "loss": 2.2191,
+      "step": 18830
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1096238938053096e-06,
+      "loss": 2.4733,
+      "step": 18831
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1095764854614413e-06,
+      "loss": 4.0142,
+      "step": 18832
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1095290771175726e-06,
+      "loss": 2.8432,
+      "step": 18833
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1094816687737043e-06,
+      "loss": 2.6323,
+      "step": 18834
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1094342604298356e-06,
+      "loss": 3.3275,
+      "step": 18835
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.109386852085967e-06,
+      "loss": 3.5626,
+      "step": 18836
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1093394437420986e-06,
+      "loss": 1.9175,
+      "step": 18837
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1092920353982303e-06,
+      "loss": 1.8432,
+      "step": 18838
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1092446270543616e-06,
+      "loss": 2.9954,
+      "step": 18839
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2.1091972187104933e-06,
+      "loss": 2.6156,
+      "step": 18840
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1091498103666246e-06,
+      "loss": 3.693,
+      "step": 18841
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.109102402022756e-06,
+      "loss": 2.8907,
+      "step": 18842
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1090549936788875e-06,
+      "loss": 1.6134,
+      "step": 18843
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.109007585335019e-06,
+      "loss": 2.4012,
+      "step": 18844
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1089601769911505e-06,
+      "loss": 3.1845,
+      "step": 18845
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1089127686472822e-06,
+      "loss": 3.48,
+      "step": 18846
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1088653603034135e-06,
+      "loss": 2.6387,
+      "step": 18847
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.108817951959545e-06,
+      "loss": 3.8652,
+      "step": 18848
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1087705436156765e-06,
+      "loss": 2.9443,
+      "step": 18849
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1087231352718078e-06,
+      "loss": 2.9414,
+      "step": 18850
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1086757269279395e-06,
+      "loss": 2.0761,
+      "step": 18851
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.108628318584071e-06,
+      "loss": 2.6392,
+      "step": 18852
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1085809102402025e-06,
+      "loss": 3.5913,
+      "step": 18853
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1085335018963337e-06,
+      "loss": 2.9309,
+      "step": 18854
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.108486093552465e-06,
+      "loss": 3.4064,
+      "step": 18855
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1084386852085967e-06,
+      "loss": 3.6639,
+      "step": 18856
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.108391276864728e-06,
+      "loss": 2.9212,
+      "step": 18857
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1083438685208597e-06,
+      "loss": 2.6759,
+      "step": 18858
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1082964601769914e-06,
+      "loss": 1.5516,
+      "step": 18859
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1082490518331227e-06,
+      "loss": 1.7454,
+      "step": 18860
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.108201643489254e-06,
+      "loss": 4.1679,
+      "step": 18861
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1081542351453857e-06,
+      "loss": 3.5887,
+      "step": 18862
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.108106826801517e-06,
+      "loss": 1.6702,
+      "step": 18863
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1080594184576487e-06,
+      "loss": 3.0047,
+      "step": 18864
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1080120101137804e-06,
+      "loss": 2.3239,
+      "step": 18865
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1079646017699117e-06,
+      "loss": 2.7368,
+      "step": 18866
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107917193426043e-06,
+      "loss": 3.1598,
+      "step": 18867
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1078697850821746e-06,
+      "loss": 3.8596,
+      "step": 18868
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107822376738306e-06,
+      "loss": 2.3903,
+      "step": 18869
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107774968394437e-06,
+      "loss": 2.1002,
+      "step": 18870
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107727560050569e-06,
+      "loss": 2.8057,
+      "step": 18871
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1076801517067006e-06,
+      "loss": 1.9993,
+      "step": 18872
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107632743362832e-06,
+      "loss": 3.1024,
+      "step": 18873
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1075853350189636e-06,
+      "loss": 3.5883,
+      "step": 18874
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107537926675095e-06,
+      "loss": 3.6877,
+      "step": 18875
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107490518331226e-06,
+      "loss": 3.8098,
+      "step": 18876
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107443109987358e-06,
+      "loss": 2.3016,
+      "step": 18877
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1073957016434896e-06,
+      "loss": 3.4596,
+      "step": 18878
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107348293299621e-06,
+      "loss": 2.1194,
+      "step": 18879
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107300884955752e-06,
+      "loss": 2.5977,
+      "step": 18880
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107253476611884e-06,
+      "loss": 2.7596,
+      "step": 18881
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107206068268015e-06,
+      "loss": 3.4578,
+      "step": 18882
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1071586599241464e-06,
+      "loss": 2.4198,
+      "step": 18883
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107111251580278e-06,
+      "loss": 2.484,
+      "step": 18884
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.10706384323641e-06,
+      "loss": 3.5719,
+      "step": 18885
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.107016434892541e-06,
+      "loss": 2.778,
+      "step": 18886
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1069690265486728e-06,
+      "loss": 3.6571,
+      "step": 18887
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106921618204804e-06,
+      "loss": 3.4545,
+      "step": 18888
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1068742098609353e-06,
+      "loss": 3.2853,
+      "step": 18889
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106826801517067e-06,
+      "loss": 1.7938,
+      "step": 18890
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1067793931731987e-06,
+      "loss": 2.4917,
+      "step": 18891
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.10673198482933e-06,
+      "loss": 2.323,
+      "step": 18892
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1066845764854617e-06,
+      "loss": 2.7417,
+      "step": 18893
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106637168141593e-06,
+      "loss": 2.5152,
+      "step": 18894
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1065897597977243e-06,
+      "loss": 2.1826,
+      "step": 18895
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106542351453856e-06,
+      "loss": 1.9515,
+      "step": 18896
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1064949431099873e-06,
+      "loss": 2.2805,
+      "step": 18897
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106447534766119e-06,
+      "loss": 3.5315,
+      "step": 18898
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1064001264222503e-06,
+      "loss": 3.1106,
+      "step": 18899
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106352718078382e-06,
+      "loss": 1.2706,
+      "step": 18900
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1063053097345132e-06,
+      "loss": 3.1386,
+      "step": 18901
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1062579013906445e-06,
+      "loss": 2.3599,
+      "step": 18902
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.1062104930467762e-06,
+      "loss": 2.9629,
+      "step": 18903
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2.106163084702908e-06,
+      "loss": 3.1143,
+      "step": 18904
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1061156763590392e-06,
+      "loss": 2.0901,
+      "step": 18905
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.106068268015171e-06,
+      "loss": 1.6062,
+      "step": 18906
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.106020859671302e-06,
+      "loss": 3.6834,
+      "step": 18907
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1059734513274335e-06,
+      "loss": 1.9846,
+      "step": 18908
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105926042983565e-06,
+      "loss": 2.7662,
+      "step": 18909
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1058786346396965e-06,
+      "loss": 3.234,
+      "step": 18910
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105831226295828e-06,
+      "loss": 2.6373,
+      "step": 18911
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.10578381795196e-06,
+      "loss": 3.5108,
+      "step": 18912
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105736409608091e-06,
+      "loss": 3.7122,
+      "step": 18913
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1056890012642224e-06,
+      "loss": 3.5264,
+      "step": 18914
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105641592920354e-06,
+      "loss": 2.8019,
+      "step": 18915
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1055941845764854e-06,
+      "loss": 2.8932,
+      "step": 18916
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105546776232617e-06,
+      "loss": 3.0947,
+      "step": 18917
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105499367888749e-06,
+      "loss": 2.3778,
+      "step": 18918
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.10545195954488e-06,
+      "loss": 2.6759,
+      "step": 18919
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1054045512010114e-06,
+      "loss": 3.2181,
+      "step": 18920
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1053571428571427e-06,
+      "loss": 2.8959,
+      "step": 18921
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1053097345132744e-06,
+      "loss": 3.185,
+      "step": 18922
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1052623261694056e-06,
+      "loss": 3.1736,
+      "step": 18923
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1052149178255374e-06,
+      "loss": 2.396,
+      "step": 18924
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.105167509481669e-06,
+      "loss": 2.8465,
+      "step": 18925
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1051201011378003e-06,
+      "loss": 2.6975,
+      "step": 18926
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1050726927939316e-06,
+      "loss": 3.1135,
+      "step": 18927
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1050252844500633e-06,
+      "loss": 2.7466,
+      "step": 18928
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1049778761061946e-06,
+      "loss": 3.9971,
+      "step": 18929
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1049304677623263e-06,
+      "loss": 2.2348,
+      "step": 18930
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.104883059418458e-06,
+      "loss": 3.188,
+      "step": 18931
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1048356510745893e-06,
+      "loss": 2.8161,
+      "step": 18932
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1047882427307206e-06,
+      "loss": 3.5269,
+      "step": 18933
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1047408343868523e-06,
+      "loss": 1.9981,
+      "step": 18934
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1046934260429836e-06,
+      "loss": 3.1726,
+      "step": 18935
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.104646017699115e-06,
+      "loss": 3.5055,
+      "step": 18936
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1045986093552465e-06,
+      "loss": 1.8715,
+      "step": 18937
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1045512010113782e-06,
+      "loss": 2.2195,
+      "step": 18938
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1045037926675095e-06,
+      "loss": 3.5241,
+      "step": 18939
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1044563843236412e-06,
+      "loss": 1.8142,
+      "step": 18940
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1044089759797725e-06,
+      "loss": 3.8083,
+      "step": 18941
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.104361567635904e-06,
+      "loss": 3.2637,
+      "step": 18942
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1043141592920355e-06,
+      "loss": 2.6506,
+      "step": 18943
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.104266750948167e-06,
+      "loss": 1.9828,
+      "step": 18944
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1042193426042985e-06,
+      "loss": 2.9001,
+      "step": 18945
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1041719342604298e-06,
+      "loss": 3.5229,
+      "step": 18946
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1041245259165615e-06,
+      "loss": 2.5577,
+      "step": 18947
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1040771175726927e-06,
+      "loss": 2.5023,
+      "step": 18948
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.104029709228824e-06,
+      "loss": 2.8705,
+      "step": 18949
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1039823008849557e-06,
+      "loss": 3.3023,
+      "step": 18950
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1039348925410874e-06,
+      "loss": 2.8778,
+      "step": 18951
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1038874841972187e-06,
+      "loss": 2.5694,
+      "step": 18952
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1038400758533504e-06,
+      "loss": 2.4806,
+      "step": 18953
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1037926675094817e-06,
+      "loss": 2.7092,
+      "step": 18954
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.103745259165613e-06,
+      "loss": 3.5564,
+      "step": 18955
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1036978508217447e-06,
+      "loss": 2.8612,
+      "step": 18956
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1036504424778764e-06,
+      "loss": 2.5792,
+      "step": 18957
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1036030341340077e-06,
+      "loss": 2.5992,
+      "step": 18958
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1035556257901394e-06,
+      "loss": 2.5593,
+      "step": 18959
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1035082174462706e-06,
+      "loss": 3.3164,
+      "step": 18960
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.103460809102402e-06,
+      "loss": 3.5014,
+      "step": 18961
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.103413400758533e-06,
+      "loss": 3.8241,
+      "step": 18962
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.103365992414665e-06,
+      "loss": 2.2168,
+      "step": 18963
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1033185840707966e-06,
+      "loss": 2.7752,
+      "step": 18964
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.103271175726928e-06,
+      "loss": 2.3525,
+      "step": 18965
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.1032237673830596e-06,
+      "loss": 2.4177,
+      "step": 18966
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2.103176359039191e-06,
+      "loss": 3.2867,
+      "step": 18967
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.103128950695322e-06,
+      "loss": 2.8058,
+      "step": 18968
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.103081542351454e-06,
+      "loss": 2.9084,
+      "step": 18969
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1030341340075856e-06,
+      "loss": 3.0075,
+      "step": 18970
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102986725663717e-06,
+      "loss": 3.2696,
+      "step": 18971
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1029393173198486e-06,
+      "loss": 2.9441,
+      "step": 18972
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.10289190897598e-06,
+      "loss": 1.955,
+      "step": 18973
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102844500632111e-06,
+      "loss": 3.216,
+      "step": 18974
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102797092288243e-06,
+      "loss": 3.819,
+      "step": 18975
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102749683944374e-06,
+      "loss": 2.1915,
+      "step": 18976
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102702275600506e-06,
+      "loss": 4.1364,
+      "step": 18977
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1026548672566375e-06,
+      "loss": 2.1094,
+      "step": 18978
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1026074589127688e-06,
+      "loss": 3.1378,
+      "step": 18979
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1025600505689e-06,
+      "loss": 3.1998,
+      "step": 18980
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1025126422250318e-06,
+      "loss": 2.2935,
+      "step": 18981
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102465233881163e-06,
+      "loss": 3.836,
+      "step": 18982
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1024178255372948e-06,
+      "loss": 2.3311,
+      "step": 18983
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102370417193426e-06,
+      "loss": 2.1432,
+      "step": 18984
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1023230088495577e-06,
+      "loss": 1.7913,
+      "step": 18985
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102275600505689e-06,
+      "loss": 2.6122,
+      "step": 18986
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1022281921618203e-06,
+      "loss": 2.0532,
+      "step": 18987
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102180783817952e-06,
+      "loss": 2.5335,
+      "step": 18988
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1021333754740833e-06,
+      "loss": 2.8966,
+      "step": 18989
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.102085967130215e-06,
+      "loss": 2.696,
+      "step": 18990
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1020385587863467e-06,
+      "loss": 3.5131,
+      "step": 18991
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101991150442478e-06,
+      "loss": 1.9614,
+      "step": 18992
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1019437420986093e-06,
+      "loss": 2.4577,
+      "step": 18993
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101896333754741e-06,
+      "loss": 2.5724,
+      "step": 18994
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1018489254108722e-06,
+      "loss": 3.3322,
+      "step": 18995
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101801517067004e-06,
+      "loss": 1.9289,
+      "step": 18996
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1017541087231356e-06,
+      "loss": 2.4434,
+      "step": 18997
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101706700379267e-06,
+      "loss": 1.7253,
+      "step": 18998
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101659292035398e-06,
+      "loss": 2.5682,
+      "step": 18999
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.3321492007104796,
+      "eval_loss": 2.7323355674743652,
+      "eval_runtime": 151.6977,
+      "eval_samples_per_second": 18.557,
+      "eval_steps_per_second": 4.641,
+      "step": 18999
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.10161188369153e-06,
+      "loss": 3.1803,
+      "step": 19000
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101564475347661e-06,
+      "loss": 3.7411,
+      "step": 19001
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1015170670037925e-06,
+      "loss": 3.2719,
+      "step": 19002
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101469658659924e-06,
+      "loss": 2.9604,
+      "step": 19003
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101422250316056e-06,
+      "loss": 3.056,
+      "step": 19004
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101374841972187e-06,
+      "loss": 2.3854,
+      "step": 19005
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1013274336283184e-06,
+      "loss": 2.5128,
+      "step": 19006
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.10128002528445e-06,
+      "loss": 3.5364,
+      "step": 19007
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1012326169405814e-06,
+      "loss": 2.5926,
+      "step": 19008
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101185208596713e-06,
+      "loss": 1.9898,
+      "step": 19009
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101137800252845e-06,
+      "loss": 1.7684,
+      "step": 19010
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.101090391908976e-06,
+      "loss": 3.6138,
+      "step": 19011
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1010429835651074e-06,
+      "loss": 1.8533,
+      "step": 19012
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.100995575221239e-06,
+      "loss": 3.0296,
+      "step": 19013
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1009481668773704e-06,
+      "loss": 3.1115,
+      "step": 19014
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1009007585335017e-06,
+      "loss": 2.3917,
+      "step": 19015
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1008533501896334e-06,
+      "loss": 3.7118,
+      "step": 19016
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.100805941845765e-06,
+      "loss": 3.3495,
+      "step": 19017
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1007585335018963e-06,
+      "loss": 2.3018,
+      "step": 19018
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.100711125158028e-06,
+      "loss": 2.477,
+      "step": 19019
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1006637168141593e-06,
+      "loss": 2.4378,
+      "step": 19020
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1006163084702906e-06,
+      "loss": 2.1216,
+      "step": 19021
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1005689001264223e-06,
+      "loss": 2.8008,
+      "step": 19022
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.100521491782554e-06,
+      "loss": 3.4816,
+      "step": 19023
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1004740834386853e-06,
+      "loss": 3.5737,
+      "step": 19024
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.100426675094817e-06,
+      "loss": 2.9567,
+      "step": 19025
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1003792667509483e-06,
+      "loss": 2.4946,
+      "step": 19026
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1003318584070796e-06,
+      "loss": 2.4113,
+      "step": 19027
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.100284450063211e-06,
+      "loss": 1.7275,
+      "step": 19028
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1002370417193425e-06,
+      "loss": 3.4883,
+      "step": 19029
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 2.1001896333754743e-06,
+      "loss": 2.281,
+      "step": 19030
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.1001422250316055e-06,
+      "loss": 2.3201,
+      "step": 19031
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.1000948166877372e-06,
+      "loss": 3.1739,
+      "step": 19032
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.1000474083438685e-06,
+      "loss": 2.5542,
+      "step": 19033
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.1e-06,
+      "loss": 3.1852,
+      "step": 19034
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0999525916561315e-06,
+      "loss": 2.9599,
+      "step": 19035
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.099905183312263e-06,
+      "loss": 2.0275,
+      "step": 19036
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0998577749683945e-06,
+      "loss": 3.0621,
+      "step": 19037
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.099810366624526e-06,
+      "loss": 3.5241,
+      "step": 19038
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0997629582806575e-06,
+      "loss": 2.5585,
+      "step": 19039
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0997155499367888e-06,
+      "loss": 3.4517,
+      "step": 19040
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0996681415929205e-06,
+      "loss": 3.1233,
+      "step": 19041
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0996207332490517e-06,
+      "loss": 3.3307,
+      "step": 19042
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0995733249051834e-06,
+      "loss": 3.3404,
+      "step": 19043
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.099525916561315e-06,
+      "loss": 2.4503,
+      "step": 19044
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0994785082174464e-06,
+      "loss": 2.6513,
+      "step": 19045
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0994310998735777e-06,
+      "loss": 2.9742,
+      "step": 19046
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0993836915297094e-06,
+      "loss": 3.8986,
+      "step": 19047
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0993362831858407e-06,
+      "loss": 2.4643,
+      "step": 19048
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0992888748419724e-06,
+      "loss": 2.2759,
+      "step": 19049
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0992414664981037e-06,
+      "loss": 2.2289,
+      "step": 19050
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0991940581542354e-06,
+      "loss": 3.747,
+      "step": 19051
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0991466498103667e-06,
+      "loss": 2.4183,
+      "step": 19052
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.099099241466498e-06,
+      "loss": 2.4487,
+      "step": 19053
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0990518331226296e-06,
+      "loss": 3.0195,
+      "step": 19054
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.099004424778761e-06,
+      "loss": 2.6052,
+      "step": 19055
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0989570164348926e-06,
+      "loss": 2.5219,
+      "step": 19056
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0989096080910243e-06,
+      "loss": 2.0153,
+      "step": 19057
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0988621997471556e-06,
+      "loss": 2.5763,
+      "step": 19058
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098814791403287e-06,
+      "loss": 1.8773,
+      "step": 19059
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0987673830594186e-06,
+      "loss": 3.6889,
+      "step": 19060
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.09871997471555e-06,
+      "loss": 2.3059,
+      "step": 19061
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098672566371681e-06,
+      "loss": 2.6947,
+      "step": 19062
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0986251580278133e-06,
+      "loss": 1.9957,
+      "step": 19063
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0985777496839446e-06,
+      "loss": 2.0626,
+      "step": 19064
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098530341340076e-06,
+      "loss": 3.3522,
+      "step": 19065
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0984829329962075e-06,
+      "loss": 3.1194,
+      "step": 19066
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098435524652339e-06,
+      "loss": 3.3722,
+      "step": 19067
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.09838811630847e-06,
+      "loss": 3.4124,
+      "step": 19068
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098340707964602e-06,
+      "loss": 3.0835,
+      "step": 19069
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0982932996207335e-06,
+      "loss": 2.8786,
+      "step": 19070
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098245891276865e-06,
+      "loss": 2.6963,
+      "step": 19071
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098198482932996e-06,
+      "loss": 2.0804,
+      "step": 19072
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0981510745891278e-06,
+      "loss": 2.3905,
+      "step": 19073
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098103666245259e-06,
+      "loss": 2.1351,
+      "step": 19074
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0980562579013903e-06,
+      "loss": 4.0549,
+      "step": 19075
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.098008849557522e-06,
+      "loss": 3.8676,
+      "step": 19076
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0979614412136538e-06,
+      "loss": 3.1134,
+      "step": 19077
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097914032869785e-06,
+      "loss": 4.2161,
+      "step": 19078
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0978666245259167e-06,
+      "loss": 2.5735,
+      "step": 19079
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097819216182048e-06,
+      "loss": 2.8103,
+      "step": 19080
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0977718078381793e-06,
+      "loss": 2.8357,
+      "step": 19081
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097724399494311e-06,
+      "loss": 3.2083,
+      "step": 19082
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0976769911504427e-06,
+      "loss": 2.8887,
+      "step": 19083
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097629582806574e-06,
+      "loss": 2.069,
+      "step": 19084
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0975821744627057e-06,
+      "loss": 1.6309,
+      "step": 19085
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097534766118837e-06,
+      "loss": 3.9609,
+      "step": 19086
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0974873577749682e-06,
+      "loss": 3.0047,
+      "step": 19087
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0974399494311e-06,
+      "loss": 2.5079,
+      "step": 19088
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0973925410872312e-06,
+      "loss": 3.9776,
+      "step": 19089
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097345132743363e-06,
+      "loss": 4.141,
+      "step": 19090
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.0972977243994946e-06,
+      "loss": 2.4629,
+      "step": 19091
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097250316055626e-06,
+      "loss": 2.9601,
+      "step": 19092
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 2.097202907711757e-06,
+      "loss": 2.9508,
+      "step": 19093
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0971554993678885e-06,
+      "loss": 2.4734,
+      "step": 19094
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.09710809102402e-06,
+      "loss": 2.7325,
+      "step": 19095
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.097060682680152e-06,
+      "loss": 3.2875,
+      "step": 19096
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.097013274336283e-06,
+      "loss": 1.8693,
+      "step": 19097
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096965865992415e-06,
+      "loss": 1.7297,
+      "step": 19098
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096918457648546e-06,
+      "loss": 2.8563,
+      "step": 19099
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0968710493046774e-06,
+      "loss": 3.8905,
+      "step": 19100
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096823640960809e-06,
+      "loss": 3.164,
+      "step": 19101
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0967762326169404e-06,
+      "loss": 2.457,
+      "step": 19102
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096728824273072e-06,
+      "loss": 3.3498,
+      "step": 19103
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096681415929204e-06,
+      "loss": 2.598,
+      "step": 19104
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096634007585335e-06,
+      "loss": 3.1278,
+      "step": 19105
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0965865992414664e-06,
+      "loss": 2.993,
+      "step": 19106
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096539190897598e-06,
+      "loss": 3.4996,
+      "step": 19107
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0964917825537294e-06,
+      "loss": 3.2853,
+      "step": 19108
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096444374209861e-06,
+      "loss": 2.1304,
+      "step": 19109
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0963969658659928e-06,
+      "loss": 3.4108,
+      "step": 19110
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096349557522124e-06,
+      "loss": 3.5304,
+      "step": 19111
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0963021491782553e-06,
+      "loss": 2.9451,
+      "step": 19112
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096254740834387e-06,
+      "loss": 2.5156,
+      "step": 19113
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0962073324905183e-06,
+      "loss": 2.7029,
+      "step": 19114
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0961599241466496e-06,
+      "loss": 2.3813,
+      "step": 19115
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0961125158027813e-06,
+      "loss": 3.8629,
+      "step": 19116
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.096065107458913e-06,
+      "loss": 3.0184,
+      "step": 19117
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0960176991150443e-06,
+      "loss": 2.0938,
+      "step": 19118
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0959702907711756e-06,
+      "loss": 3.3825,
+      "step": 19119
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0959228824273073e-06,
+      "loss": 2.5188,
+      "step": 19120
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0958754740834386e-06,
+      "loss": 1.7053,
+      "step": 19121
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0958280657395703e-06,
+      "loss": 2.5794,
+      "step": 19122
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.095780657395702e-06,
+      "loss": 2.612,
+      "step": 19123
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0957332490518332e-06,
+      "loss": 2.1558,
+      "step": 19124
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0956858407079645e-06,
+      "loss": 2.8833,
+      "step": 19125
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0956384323640962e-06,
+      "loss": 4.352,
+      "step": 19126
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0955910240202275e-06,
+      "loss": 3.0045,
+      "step": 19127
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.095543615676359e-06,
+      "loss": 3.2054,
+      "step": 19128
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0954962073324905e-06,
+      "loss": 2.2707,
+      "step": 19129
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.095448798988622e-06,
+      "loss": 2.7595,
+      "step": 19130
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0954013906447535e-06,
+      "loss": 1.8325,
+      "step": 19131
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.095353982300885e-06,
+      "loss": 2.8897,
+      "step": 19132
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0953065739570165e-06,
+      "loss": 2.6285,
+      "step": 19133
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0952591656131477e-06,
+      "loss": 3.2748,
+      "step": 19134
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0952117572692795e-06,
+      "loss": 3.8213,
+      "step": 19135
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.095164348925411e-06,
+      "loss": 2.3182,
+      "step": 19136
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0951169405815424e-06,
+      "loss": 3.1131,
+      "step": 19137
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0950695322376737e-06,
+      "loss": 2.9843,
+      "step": 19138
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0950221238938054e-06,
+      "loss": 2.7858,
+      "step": 19139
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0949747155499367e-06,
+      "loss": 2.4029,
+      "step": 19140
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.094927307206068e-06,
+      "loss": 2.8336,
+      "step": 19141
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0948798988621997e-06,
+      "loss": 2.8862,
+      "step": 19142
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0948324905183314e-06,
+      "loss": 2.7182,
+      "step": 19143
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0947850821744627e-06,
+      "loss": 3.0062,
+      "step": 19144
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0947376738305944e-06,
+      "loss": 2.626,
+      "step": 19145
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0946902654867257e-06,
+      "loss": 2.3508,
+      "step": 19146
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.094642857142857e-06,
+      "loss": 2.4601,
+      "step": 19147
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0945954487989886e-06,
+      "loss": 2.5467,
+      "step": 19148
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0945480404551203e-06,
+      "loss": 3.3144,
+      "step": 19149
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0945006321112516e-06,
+      "loss": 3.0249,
+      "step": 19150
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0944532237673833e-06,
+      "loss": 1.5673,
+      "step": 19151
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0944058154235146e-06,
+      "loss": 2.5637,
+      "step": 19152
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.094358407079646e-06,
+      "loss": 3.8707,
+      "step": 19153
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0943109987357776e-06,
+      "loss": 2.678,
+      "step": 19154
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.094263590391909e-06,
+      "loss": 2.3922,
+      "step": 19155
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0942161820480406e-06,
+      "loss": 3.2359,
+      "step": 19156
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.0941687737041723e-06,
+      "loss": 3.3159,
+      "step": 19157
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0941213653603036e-06,
+      "loss": 2.9021,
+      "step": 19158
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.094073957016435e-06,
+      "loss": 3.2926,
+      "step": 19159
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.094026548672566e-06,
+      "loss": 2.0585,
+      "step": 19160
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093979140328698e-06,
+      "loss": 3.0024,
+      "step": 19161
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0939317319848295e-06,
+      "loss": 3.487,
+      "step": 19162
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093884323640961e-06,
+      "loss": 2.7534,
+      "step": 19163
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0938369152970925e-06,
+      "loss": 2.4281,
+      "step": 19164
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093789506953224e-06,
+      "loss": 3.5118,
+      "step": 19165
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093742098609355e-06,
+      "loss": 3.416,
+      "step": 19166
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0936946902654868e-06,
+      "loss": 2.7187,
+      "step": 19167
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093647281921618e-06,
+      "loss": 1.5833,
+      "step": 19168
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0935998735777498e-06,
+      "loss": 2.8966,
+      "step": 19169
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0935524652338815e-06,
+      "loss": 3.0898,
+      "step": 19170
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0935050568900127e-06,
+      "loss": 4.0277,
+      "step": 19171
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093457648546144e-06,
+      "loss": 2.1393,
+      "step": 19172
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0934102402022757e-06,
+      "loss": 2.177,
+      "step": 19173
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093362831858407e-06,
+      "loss": 3.6697,
+      "step": 19174
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0933154235145387e-06,
+      "loss": 3.6982,
+      "step": 19175
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0932680151706704e-06,
+      "loss": 2.5366,
+      "step": 19176
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0932206068268017e-06,
+      "loss": 2.3903,
+      "step": 19177
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093173198482933e-06,
+      "loss": 2.7792,
+      "step": 19178
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0931257901390647e-06,
+      "loss": 2.3242,
+      "step": 19179
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.093078381795196e-06,
+      "loss": 1.8981,
+      "step": 19180
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0930309734513272e-06,
+      "loss": 1.8433,
+      "step": 19181
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092983565107459e-06,
+      "loss": 3.467,
+      "step": 19182
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0929361567635907e-06,
+      "loss": 4.2234,
+      "step": 19183
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092888748419722e-06,
+      "loss": 3.1931,
+      "step": 19184
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0928413400758532e-06,
+      "loss": 2.2888,
+      "step": 19185
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092793931731985e-06,
+      "loss": 3.0073,
+      "step": 19186
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092746523388116e-06,
+      "loss": 2.3251,
+      "step": 19187
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092699115044248e-06,
+      "loss": 3.1853,
+      "step": 19188
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0926517067003796e-06,
+      "loss": 1.7537,
+      "step": 19189
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092604298356511e-06,
+      "loss": 3.6077,
+      "step": 19190
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092556890012642e-06,
+      "loss": 3.8489,
+      "step": 19191
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092509481668774e-06,
+      "loss": 1.8319,
+      "step": 19192
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092462073324905e-06,
+      "loss": 2.4897,
+      "step": 19193
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0924146649810364e-06,
+      "loss": 2.8005,
+      "step": 19194
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092367256637168e-06,
+      "loss": 2.5257,
+      "step": 19195
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0923198482933e-06,
+      "loss": 2.9269,
+      "step": 19196
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092272439949431e-06,
+      "loss": 3.2885,
+      "step": 19197
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092225031605563e-06,
+      "loss": 2.2925,
+      "step": 19198
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092177623261694e-06,
+      "loss": 2.9525,
+      "step": 19199
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0921302149178254e-06,
+      "loss": 3.409,
+      "step": 19200
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092082806573957e-06,
+      "loss": 3.3369,
+      "step": 19201
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.092035398230089e-06,
+      "loss": 3.0621,
+      "step": 19202
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.09198798988622e-06,
+      "loss": 2.9989,
+      "step": 19203
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0919405815423514e-06,
+      "loss": 2.2048,
+      "step": 19204
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.091893173198483e-06,
+      "loss": 2.6128,
+      "step": 19205
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0918457648546143e-06,
+      "loss": 3.118,
+      "step": 19206
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0917983565107456e-06,
+      "loss": 1.4296,
+      "step": 19207
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0917509481668773e-06,
+      "loss": 2.3346,
+      "step": 19208
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.091703539823009e-06,
+      "loss": 2.4576,
+      "step": 19209
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0916561314791403e-06,
+      "loss": 3.6552,
+      "step": 19210
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.091608723135272e-06,
+      "loss": 1.5341,
+      "step": 19211
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0915613147914033e-06,
+      "loss": 3.2072,
+      "step": 19212
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0915139064475346e-06,
+      "loss": 3.2424,
+      "step": 19213
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0914664981036663e-06,
+      "loss": 3.4064,
+      "step": 19214
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.091419089759798e-06,
+      "loss": 2.8512,
+      "step": 19215
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0913716814159293e-06,
+      "loss": 2.871,
+      "step": 19216
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.091324273072061e-06,
+      "loss": 2.955,
+      "step": 19217
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0912768647281922e-06,
+      "loss": 2.8624,
+      "step": 19218
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0912294563843235e-06,
+      "loss": 1.4707,
+      "step": 19219
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.0911820480404552e-06,
+      "loss": 1.5802,
+      "step": 19220
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0911346396965865e-06,
+      "loss": 2.9736,
+      "step": 19221
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.091087231352718e-06,
+      "loss": 3.7012,
+      "step": 19222
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.09103982300885e-06,
+      "loss": 3.2079,
+      "step": 19223
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.090992414664981e-06,
+      "loss": 2.4538,
+      "step": 19224
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0909450063211125e-06,
+      "loss": 2.7865,
+      "step": 19225
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0908975979772438e-06,
+      "loss": 3.9109,
+      "step": 19226
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0908501896333755e-06,
+      "loss": 4.3949,
+      "step": 19227
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.090802781289507e-06,
+      "loss": 3.2236,
+      "step": 19228
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0907553729456384e-06,
+      "loss": 4.0568,
+      "step": 19229
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.09070796460177e-06,
+      "loss": 3.6154,
+      "step": 19230
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0906605562579014e-06,
+      "loss": 3.7924,
+      "step": 19231
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0906131479140327e-06,
+      "loss": 1.9335,
+      "step": 19232
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0905657395701644e-06,
+      "loss": 3.1305,
+      "step": 19233
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0905183312262957e-06,
+      "loss": 2.7261,
+      "step": 19234
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0904709228824274e-06,
+      "loss": 2.7291,
+      "step": 19235
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.090423514538559e-06,
+      "loss": 2.073,
+      "step": 19236
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0903761061946904e-06,
+      "loss": 3.6913,
+      "step": 19237
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0903286978508217e-06,
+      "loss": 2.2865,
+      "step": 19238
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0902812895069534e-06,
+      "loss": 1.7833,
+      "step": 19239
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0902338811630846e-06,
+      "loss": 1.8832,
+      "step": 19240
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0901864728192164e-06,
+      "loss": 2.5413,
+      "step": 19241
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.090139064475348e-06,
+      "loss": 2.4858,
+      "step": 19242
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0900916561314793e-06,
+      "loss": 2.5913,
+      "step": 19243
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0900442477876106e-06,
+      "loss": 2.3699,
+      "step": 19244
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0899968394437423e-06,
+      "loss": 3.1449,
+      "step": 19245
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0899494310998736e-06,
+      "loss": 2.0769,
+      "step": 19246
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.089902022756005e-06,
+      "loss": 3.5022,
+      "step": 19247
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0898546144121366e-06,
+      "loss": 3.0796,
+      "step": 19248
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0898072060682683e-06,
+      "loss": 2.2185,
+      "step": 19249
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0897597977243996e-06,
+      "loss": 2.9106,
+      "step": 19250
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.089712389380531e-06,
+      "loss": 2.1541,
+      "step": 19251
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0896649810366626e-06,
+      "loss": 2.0052,
+      "step": 19252
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.089617572692794e-06,
+      "loss": 3.4648,
+      "step": 19253
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0895701643489255e-06,
+      "loss": 3.2211,
+      "step": 19254
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0895227560050572e-06,
+      "loss": 3.5993,
+      "step": 19255
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0894753476611885e-06,
+      "loss": 2.9441,
+      "step": 19256
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.08942793931732e-06,
+      "loss": 2.8352,
+      "step": 19257
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0893805309734515e-06,
+      "loss": 1.509,
+      "step": 19258
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0893331226295828e-06,
+      "loss": 3.3427,
+      "step": 19259
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.089285714285714e-06,
+      "loss": 2.9175,
+      "step": 19260
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0892383059418458e-06,
+      "loss": 2.027,
+      "step": 19261
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0891908975979775e-06,
+      "loss": 2.191,
+      "step": 19262
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0891434892541088e-06,
+      "loss": 2.2172,
+      "step": 19263
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0890960809102405e-06,
+      "loss": 1.7792,
+      "step": 19264
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0890486725663717e-06,
+      "loss": 3.4055,
+      "step": 19265
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.089001264222503e-06,
+      "loss": 2.2342,
+      "step": 19266
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0889538558786347e-06,
+      "loss": 2.1807,
+      "step": 19267
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0889064475347664e-06,
+      "loss": 2.4003,
+      "step": 19268
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0888590391908977e-06,
+      "loss": 2.2717,
+      "step": 19269
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088811630847029e-06,
+      "loss": 1.9234,
+      "step": 19270
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0887642225031607e-06,
+      "loss": 3.721,
+      "step": 19271
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088716814159292e-06,
+      "loss": 2.4427,
+      "step": 19272
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0886694058154233e-06,
+      "loss": 4.1751,
+      "step": 19273
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088621997471555e-06,
+      "loss": 3.1463,
+      "step": 19274
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0885745891276867e-06,
+      "loss": 2.7242,
+      "step": 19275
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088527180783818e-06,
+      "loss": 3.2534,
+      "step": 19276
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0884797724399496e-06,
+      "loss": 3.2191,
+      "step": 19277
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088432364096081e-06,
+      "loss": 3.3261,
+      "step": 19278
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088384955752212e-06,
+      "loss": 3.2354,
+      "step": 19279
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088337547408344e-06,
+      "loss": 3.2003,
+      "step": 19280
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0882901390644756e-06,
+      "loss": 3.0322,
+      "step": 19281
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.088242730720607e-06,
+      "loss": 3.3038,
+      "step": 19282
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.0881953223767386e-06,
+      "loss": 3.5451,
+      "step": 19283
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.08814791403287e-06,
+      "loss": 2.2614,
+      "step": 19284
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.088100505689001e-06,
+      "loss": 3.0545,
+      "step": 19285
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.088053097345133e-06,
+      "loss": 1.6218,
+      "step": 19286
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.088005689001264e-06,
+      "loss": 3.1027,
+      "step": 19287
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087958280657396e-06,
+      "loss": 2.1648,
+      "step": 19288
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0879108723135276e-06,
+      "loss": 2.568,
+      "step": 19289
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087863463969659e-06,
+      "loss": 3.4636,
+      "step": 19290
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.08781605562579e-06,
+      "loss": 1.6985,
+      "step": 19291
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0877686472819214e-06,
+      "loss": 2.5981,
+      "step": 19292
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087721238938053e-06,
+      "loss": 1.4457,
+      "step": 19293
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087673830594185e-06,
+      "loss": 3.0236,
+      "step": 19294
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087626422250316e-06,
+      "loss": 2.974,
+      "step": 19295
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0875790139064478e-06,
+      "loss": 3.7684,
+      "step": 19296
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087531605562579e-06,
+      "loss": 1.8328,
+      "step": 19297
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0874841972187103e-06,
+      "loss": 1.4592,
+      "step": 19298
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087436788874842e-06,
+      "loss": 3.8367,
+      "step": 19299
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0873893805309733e-06,
+      "loss": 2.8694,
+      "step": 19300
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087341972187105e-06,
+      "loss": 1.6975,
+      "step": 19301
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0872945638432367e-06,
+      "loss": 3.0535,
+      "step": 19302
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087247155499368e-06,
+      "loss": 3.3591,
+      "step": 19303
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0871997471554993e-06,
+      "loss": 2.5643,
+      "step": 19304
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087152338811631e-06,
+      "loss": 3.1381,
+      "step": 19305
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0871049304677623e-06,
+      "loss": 2.6763,
+      "step": 19306
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.087057522123894e-06,
+      "loss": 2.4704,
+      "step": 19307
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0870101137800257e-06,
+      "loss": 2.3556,
+      "step": 19308
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086962705436157e-06,
+      "loss": 1.8432,
+      "step": 19309
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0869152970922883e-06,
+      "loss": 3.3534,
+      "step": 19310
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.08686788874842e-06,
+      "loss": 4.2015,
+      "step": 19311
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0868204804045512e-06,
+      "loss": 1.9091,
+      "step": 19312
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0867730720606825e-06,
+      "loss": 3.1639,
+      "step": 19313
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0867256637168142e-06,
+      "loss": 3.8091,
+      "step": 19314
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086678255372946e-06,
+      "loss": 2.7194,
+      "step": 19315
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086630847029077e-06,
+      "loss": 2.6718,
+      "step": 19316
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0865834386852085e-06,
+      "loss": 3.5557,
+      "step": 19317
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.08653603034134e-06,
+      "loss": 3.473,
+      "step": 19318
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0864886219974715e-06,
+      "loss": 1.9735,
+      "step": 19319
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086441213653603e-06,
+      "loss": 1.665,
+      "step": 19320
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086393805309735e-06,
+      "loss": 2.3263,
+      "step": 19321
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086346396965866e-06,
+      "loss": 2.7441,
+      "step": 19322
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0862989886219974e-06,
+      "loss": 2.6309,
+      "step": 19323
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086251580278129e-06,
+      "loss": 2.6522,
+      "step": 19324
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0862041719342604e-06,
+      "loss": 3.1088,
+      "step": 19325
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0861567635903917e-06,
+      "loss": 2.3779,
+      "step": 19326
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0861093552465234e-06,
+      "loss": 2.6988,
+      "step": 19327
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.086061946902655e-06,
+      "loss": 2.148,
+      "step": 19328
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0860145385587864e-06,
+      "loss": 3.6488,
+      "step": 19329
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.085967130214918e-06,
+      "loss": 2.8382,
+      "step": 19330
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0859197218710494e-06,
+      "loss": 3.5374,
+      "step": 19331
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0858723135271807e-06,
+      "loss": 2.9655,
+      "step": 19332
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0858249051833124e-06,
+      "loss": 2.1527,
+      "step": 19333
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.085777496839444e-06,
+      "loss": 3.3651,
+      "step": 19334
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0857300884955753e-06,
+      "loss": 2.3793,
+      "step": 19335
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0856826801517066e-06,
+      "loss": 2.9876,
+      "step": 19336
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0856352718078383e-06,
+      "loss": 2.8145,
+      "step": 19337
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0855878634639696e-06,
+      "loss": 3.039,
+      "step": 19338
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.085540455120101e-06,
+      "loss": 2.4596,
+      "step": 19339
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0854930467762326e-06,
+      "loss": 3.332,
+      "step": 19340
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0854456384323643e-06,
+      "loss": 2.8301,
+      "step": 19341
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0853982300884956e-06,
+      "loss": 2.2218,
+      "step": 19342
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0853508217446273e-06,
+      "loss": 3.1342,
+      "step": 19343
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0853034134007586e-06,
+      "loss": 3.3385,
+      "step": 19344
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.08525600505689e-06,
+      "loss": 2.9306,
+      "step": 19345
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.0852085967130215e-06,
+      "loss": 2.7115,
+      "step": 19346
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.085161188369153e-06,
+      "loss": 2.8249,
+      "step": 19347
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0851137800252845e-06,
+      "loss": 2.1045,
+      "step": 19348
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0850663716814162e-06,
+      "loss": 2.8934,
+      "step": 19349
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0850189633375475e-06,
+      "loss": 2.8011,
+      "step": 19350
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084971554993679e-06,
+      "loss": 2.3203,
+      "step": 19351
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0849241466498105e-06,
+      "loss": 2.6947,
+      "step": 19352
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0848767383059418e-06,
+      "loss": 1.5043,
+      "step": 19353
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0848293299620735e-06,
+      "loss": 2.8377,
+      "step": 19354
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084781921618205e-06,
+      "loss": 2.2989,
+      "step": 19355
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0847345132743365e-06,
+      "loss": 1.7953,
+      "step": 19356
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0846871049304677e-06,
+      "loss": 2.3894,
+      "step": 19357
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084639696586599e-06,
+      "loss": 4.2749,
+      "step": 19358
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0845922882427307e-06,
+      "loss": 3.0402,
+      "step": 19359
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084544879898862e-06,
+      "loss": 3.1247,
+      "step": 19360
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0844974715549937e-06,
+      "loss": 2.3438,
+      "step": 19361
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0844500632111254e-06,
+      "loss": 1.7248,
+      "step": 19362
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0844026548672567e-06,
+      "loss": 3.2499,
+      "step": 19363
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084355246523388e-06,
+      "loss": 2.8597,
+      "step": 19364
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0843078381795197e-06,
+      "loss": 2.8605,
+      "step": 19365
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084260429835651e-06,
+      "loss": 3.5622,
+      "step": 19366
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0842130214917827e-06,
+      "loss": 1.4744,
+      "step": 19367
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0841656131479144e-06,
+      "loss": 1.6052,
+      "step": 19368
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0841182048040457e-06,
+      "loss": 2.5563,
+      "step": 19369
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.084070796460177e-06,
+      "loss": 3.7633,
+      "step": 19370
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0840233881163086e-06,
+      "loss": 3.3116,
+      "step": 19371
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.08397597977244e-06,
+      "loss": 2.1228,
+      "step": 19372
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083928571428571e-06,
+      "loss": 2.5457,
+      "step": 19373
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083881163084703e-06,
+      "loss": 3.0109,
+      "step": 19374
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0838337547408346e-06,
+      "loss": 1.8903,
+      "step": 19375
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083786346396966e-06,
+      "loss": 2.7055,
+      "step": 19376
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0837389380530976e-06,
+      "loss": 2.5209,
+      "step": 19377
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083691529709229e-06,
+      "loss": 3.4867,
+      "step": 19378
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.08364412136536e-06,
+      "loss": 2.9036,
+      "step": 19379
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083596713021492e-06,
+      "loss": 1.7219,
+      "step": 19380
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0835493046776236e-06,
+      "loss": 2.0787,
+      "step": 19381
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083501896333755e-06,
+      "loss": 2.3371,
+      "step": 19382
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083454487989886e-06,
+      "loss": 2.2385,
+      "step": 19383
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083407079646018e-06,
+      "loss": 2.4034,
+      "step": 19384
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083359671302149e-06,
+      "loss": 2.5423,
+      "step": 19385
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0833122629582804e-06,
+      "loss": 1.7304,
+      "step": 19386
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083264854614412e-06,
+      "loss": 2.0235,
+      "step": 19387
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083217446270544e-06,
+      "loss": 2.7726,
+      "step": 19388
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083170037926675e-06,
+      "loss": 2.4745,
+      "step": 19389
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0831226295828068e-06,
+      "loss": 3.0645,
+      "step": 19390
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.083075221238938e-06,
+      "loss": 3.258,
+      "step": 19391
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0830278128950693e-06,
+      "loss": 3.8109,
+      "step": 19392
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082980404551201e-06,
+      "loss": 2.6263,
+      "step": 19393
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0829329962073327e-06,
+      "loss": 3.848,
+      "step": 19394
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082885587863464e-06,
+      "loss": 2.3143,
+      "step": 19395
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0828381795195957e-06,
+      "loss": 3.3976,
+      "step": 19396
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082790771175727e-06,
+      "loss": 3.59,
+      "step": 19397
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0827433628318583e-06,
+      "loss": 2.3196,
+      "step": 19398
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0826959544879896e-06,
+      "loss": 3.2574,
+      "step": 19399
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0826485461441213e-06,
+      "loss": 2.6148,
+      "step": 19400
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082601137800253e-06,
+      "loss": 1.3959,
+      "step": 19401
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0825537294563843e-06,
+      "loss": 2.4359,
+      "step": 19402
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082506321112516e-06,
+      "loss": 3.0267,
+      "step": 19403
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0824589127686472e-06,
+      "loss": 2.4243,
+      "step": 19404
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0824115044247785e-06,
+      "loss": 2.0228,
+      "step": 19405
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0823640960809102e-06,
+      "loss": 3.3315,
+      "step": 19406
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082316687737042e-06,
+      "loss": 3.8819,
+      "step": 19407
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.0822692793931732e-06,
+      "loss": 2.0152,
+      "step": 19408
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082221871049305e-06,
+      "loss": 3.0497,
+      "step": 19409
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.082174462705436e-06,
+      "loss": 3.781,
+      "step": 19410
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0821270543615675e-06,
+      "loss": 3.9674,
+      "step": 19411
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.082079646017699e-06,
+      "loss": 2.7033,
+      "step": 19412
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0820322376738305e-06,
+      "loss": 2.9822,
+      "step": 19413
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081984829329962e-06,
+      "loss": 2.2353,
+      "step": 19414
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081937420986094e-06,
+      "loss": 3.1966,
+      "step": 19415
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081890012642225e-06,
+      "loss": 3.3806,
+      "step": 19416
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0818426042983564e-06,
+      "loss": 3.242,
+      "step": 19417
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081795195954488e-06,
+      "loss": 3.8361,
+      "step": 19418
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0817477876106194e-06,
+      "loss": 2.9806,
+      "step": 19419
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081700379266751e-06,
+      "loss": 2.6201,
+      "step": 19420
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0816529709228824e-06,
+      "loss": 2.29,
+      "step": 19421
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081605562579014e-06,
+      "loss": 2.8314,
+      "step": 19422
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0815581542351454e-06,
+      "loss": 3.234,
+      "step": 19423
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0815107458912767e-06,
+      "loss": 3.3742,
+      "step": 19424
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0814633375474084e-06,
+      "loss": 2.6551,
+      "step": 19425
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0814159292035397e-06,
+      "loss": 2.1228,
+      "step": 19426
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0813685208596714e-06,
+      "loss": 1.8359,
+      "step": 19427
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081321112515803e-06,
+      "loss": 3.2844,
+      "step": 19428
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0812737041719343e-06,
+      "loss": 2.8239,
+      "step": 19429
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0812262958280656e-06,
+      "loss": 1.6404,
+      "step": 19430
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0811788874841973e-06,
+      "loss": 1.945,
+      "step": 19431
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0811314791403286e-06,
+      "loss": 2.7568,
+      "step": 19432
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0810840707964603e-06,
+      "loss": 3.4806,
+      "step": 19433
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.081036662452592e-06,
+      "loss": 2.4757,
+      "step": 19434
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0809892541087233e-06,
+      "loss": 3.2545,
+      "step": 19435
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0809418457648546e-06,
+      "loss": 3.3756,
+      "step": 19436
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0808944374209863e-06,
+      "loss": 2.6903,
+      "step": 19437
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0808470290771176e-06,
+      "loss": 2.6781,
+      "step": 19438
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.080799620733249e-06,
+      "loss": 2.442,
+      "step": 19439
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0807522123893805e-06,
+      "loss": 1.9473,
+      "step": 19440
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0807048040455122e-06,
+      "loss": 2.4462,
+      "step": 19441
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0806573957016435e-06,
+      "loss": 2.3828,
+      "step": 19442
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.080609987357775e-06,
+      "loss": 2.5252,
+      "step": 19443
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0805625790139065e-06,
+      "loss": 1.4771,
+      "step": 19444
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.080515170670038e-06,
+      "loss": 3.8438,
+      "step": 19445
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0804677623261695e-06,
+      "loss": 1.9638,
+      "step": 19446
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.080420353982301e-06,
+      "loss": 3.4335,
+      "step": 19447
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0803729456384325e-06,
+      "loss": 3.1229,
+      "step": 19448
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0803255372945638e-06,
+      "loss": 2.2309,
+      "step": 19449
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0802781289506955e-06,
+      "loss": 2.9126,
+      "step": 19450
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0802307206068267e-06,
+      "loss": 2.4597,
+      "step": 19451
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.080183312262958e-06,
+      "loss": 3.5312,
+      "step": 19452
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0801359039190897e-06,
+      "loss": 1.6105,
+      "step": 19453
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0800884955752214e-06,
+      "loss": 3.5828,
+      "step": 19454
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0800410872313527e-06,
+      "loss": 3.0114,
+      "step": 19455
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0799936788874844e-06,
+      "loss": 2.4526,
+      "step": 19456
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0799462705436157e-06,
+      "loss": 3.6106,
+      "step": 19457
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079898862199747e-06,
+      "loss": 3.3903,
+      "step": 19458
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0798514538558787e-06,
+      "loss": 2.811,
+      "step": 19459
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0798040455120104e-06,
+      "loss": 2.2524,
+      "step": 19460
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0797566371681417e-06,
+      "loss": 3.4625,
+      "step": 19461
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0797092288242734e-06,
+      "loss": 2.8996,
+      "step": 19462
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0796618204804046e-06,
+      "loss": 3.3664,
+      "step": 19463
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079614412136536e-06,
+      "loss": 3.5179,
+      "step": 19464
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079567003792667e-06,
+      "loss": 3.8942,
+      "step": 19465
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079519595448799e-06,
+      "loss": 1.8073,
+      "step": 19466
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0794721871049306e-06,
+      "loss": 2.4732,
+      "step": 19467
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079424778761062e-06,
+      "loss": 2.4057,
+      "step": 19468
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0793773704171936e-06,
+      "loss": 2.1581,
+      "step": 19469
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079329962073325e-06,
+      "loss": 2.5619,
+      "step": 19470
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079282553729456e-06,
+      "loss": 2.3967,
+      "step": 19471
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.079235145385588e-06,
+      "loss": 1.9038,
+      "step": 19472
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.0791877370417196e-06,
+      "loss": 4.2677,
+      "step": 19473
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.079140328697851e-06,
+      "loss": 3.209,
+      "step": 19474
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0790929203539826e-06,
+      "loss": 2.9022,
+      "step": 19475
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.079045512010114e-06,
+      "loss": 4.0753,
+      "step": 19476
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078998103666245e-06,
+      "loss": 3.0426,
+      "step": 19477
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078950695322377e-06,
+      "loss": 2.3343,
+      "step": 19478
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078903286978508e-06,
+      "loss": 3.4097,
+      "step": 19479
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.07885587863464e-06,
+      "loss": 3.1639,
+      "step": 19480
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0788084702907715e-06,
+      "loss": 3.715,
+      "step": 19481
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078761061946903e-06,
+      "loss": 3.0378,
+      "step": 19482
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078713653603034e-06,
+      "loss": 2.7802,
+      "step": 19483
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0786662452591658e-06,
+      "loss": 2.5842,
+      "step": 19484
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078618836915297e-06,
+      "loss": 3.7892,
+      "step": 19485
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0785714285714288e-06,
+      "loss": 3.5809,
+      "step": 19486
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.07852402022756e-06,
+      "loss": 3.7046,
+      "step": 19487
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0784766118836917e-06,
+      "loss": 2.4145,
+      "step": 19488
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078429203539823e-06,
+      "loss": 3.2075,
+      "step": 19489
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0783817951959543e-06,
+      "loss": 2.5897,
+      "step": 19490
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078334386852086e-06,
+      "loss": 1.994,
+      "step": 19491
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0782869785082173e-06,
+      "loss": 4.2848,
+      "step": 19492
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078239570164349e-06,
+      "loss": 2.6557,
+      "step": 19493
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0781921618204807e-06,
+      "loss": 3.458,
+      "step": 19494
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078144753476612e-06,
+      "loss": 2.8932,
+      "step": 19495
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0780973451327433e-06,
+      "loss": 1.3318,
+      "step": 19496
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.078049936788875e-06,
+      "loss": 2.7332,
+      "step": 19497
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0780025284450062e-06,
+      "loss": 4.2342,
+      "step": 19498
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077955120101138e-06,
+      "loss": 1.855,
+      "step": 19499
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0779077117572696e-06,
+      "loss": 3.9723,
+      "step": 19500
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077860303413401e-06,
+      "loss": 2.8599,
+      "step": 19501
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077812895069532e-06,
+      "loss": 2.9652,
+      "step": 19502
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077765486725664e-06,
+      "loss": 2.9909,
+      "step": 19503
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077718078381795e-06,
+      "loss": 2.4581,
+      "step": 19504
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0776706700379265e-06,
+      "loss": 1.7789,
+      "step": 19505
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077623261694058e-06,
+      "loss": 2.5981,
+      "step": 19506
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.07757585335019e-06,
+      "loss": 2.2209,
+      "step": 19507
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077528445006321e-06,
+      "loss": 2.744,
+      "step": 19508
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0774810366624524e-06,
+      "loss": 2.8871,
+      "step": 19509
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077433628318584e-06,
+      "loss": 3.9005,
+      "step": 19510
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0773862199747154e-06,
+      "loss": 3.1131,
+      "step": 19511
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077338811630847e-06,
+      "loss": 3.3974,
+      "step": 19512
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077291403286979e-06,
+      "loss": 2.8841,
+      "step": 19513
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.07724399494311e-06,
+      "loss": 3.2623,
+      "step": 19514
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0771965865992414e-06,
+      "loss": 3.4644,
+      "step": 19515
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.077149178255373e-06,
+      "loss": 2.0689,
+      "step": 19516
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0771017699115044e-06,
+      "loss": 2.3383,
+      "step": 19517
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0770543615676357e-06,
+      "loss": 3.1746,
+      "step": 19518
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0770069532237674e-06,
+      "loss": 2.9,
+      "step": 19519
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.076959544879899e-06,
+      "loss": 3.7114,
+      "step": 19520
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0769121365360303e-06,
+      "loss": 2.8832,
+      "step": 19521
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.076864728192162e-06,
+      "loss": 2.421,
+      "step": 19522
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0768173198482933e-06,
+      "loss": 2.8766,
+      "step": 19523
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0767699115044246e-06,
+      "loss": 1.5685,
+      "step": 19524
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0767225031605563e-06,
+      "loss": 3.1432,
+      "step": 19525
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.076675094816688e-06,
+      "loss": 2.9767,
+      "step": 19526
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0766276864728193e-06,
+      "loss": 2.6023,
+      "step": 19527
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.076580278128951e-06,
+      "loss": 2.6292,
+      "step": 19528
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0765328697850823e-06,
+      "loss": 4.1312,
+      "step": 19529
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0764854614412136e-06,
+      "loss": 2.588,
+      "step": 19530
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.076438053097345e-06,
+      "loss": 1.962,
+      "step": 19531
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0763906447534766e-06,
+      "loss": 3.7398,
+      "step": 19532
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0763432364096083e-06,
+      "loss": 1.0636,
+      "step": 19533
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0762958280657395e-06,
+      "loss": 3.4088,
+      "step": 19534
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0762484197218712e-06,
+      "loss": 2.387,
+      "step": 19535
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.0762010113780025e-06,
+      "loss": 3.4276,
+      "step": 19536
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.076153603034134e-06,
+      "loss": 2.57,
+      "step": 19537
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0761061946902655e-06,
+      "loss": 2.6803,
+      "step": 19538
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.076058786346397e-06,
+      "loss": 2.5551,
+      "step": 19539
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0760113780025285e-06,
+      "loss": 1.7143,
+      "step": 19540
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.07596396965866e-06,
+      "loss": 3.2399,
+      "step": 19541
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0759165613147915e-06,
+      "loss": 3.1514,
+      "step": 19542
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0758691529709228e-06,
+      "loss": 3.0847,
+      "step": 19543
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0758217446270545e-06,
+      "loss": 2.4466,
+      "step": 19544
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0757743362831857e-06,
+      "loss": 1.6629,
+      "step": 19545
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0757269279393174e-06,
+      "loss": 2.3809,
+      "step": 19546
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.075679519595449e-06,
+      "loss": 3.0465,
+      "step": 19547
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0756321112515804e-06,
+      "loss": 3.7279,
+      "step": 19548
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0755847029077117e-06,
+      "loss": 3.0885,
+      "step": 19549
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0755372945638434e-06,
+      "loss": 3.1404,
+      "step": 19550
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0754898862199747e-06,
+      "loss": 2.493,
+      "step": 19551
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0754424778761064e-06,
+      "loss": 3.0067,
+      "step": 19552
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0753950695322377e-06,
+      "loss": 2.7119,
+      "step": 19553
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0753476611883694e-06,
+      "loss": 1.8945,
+      "step": 19554
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0753002528445007e-06,
+      "loss": 2.8431,
+      "step": 19555
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.075252844500632e-06,
+      "loss": 3.5994,
+      "step": 19556
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0752054361567636e-06,
+      "loss": 3.6278,
+      "step": 19557
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.075158027812895e-06,
+      "loss": 2.7414,
+      "step": 19558
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0751106194690266e-06,
+      "loss": 2.5318,
+      "step": 19559
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0750632111251583e-06,
+      "loss": 2.9862,
+      "step": 19560
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0750158027812896e-06,
+      "loss": 3.3661,
+      "step": 19561
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074968394437421e-06,
+      "loss": 2.2249,
+      "step": 19562
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0749209860935526e-06,
+      "loss": 2.5446,
+      "step": 19563
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074873577749684e-06,
+      "loss": 2.7462,
+      "step": 19564
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0748261694058156e-06,
+      "loss": 3.4369,
+      "step": 19565
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0747787610619473e-06,
+      "loss": 3.4064,
+      "step": 19566
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0747313527180786e-06,
+      "loss": 1.4594,
+      "step": 19567
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.07468394437421e-06,
+      "loss": 2.2769,
+      "step": 19568
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0746365360303415e-06,
+      "loss": 1.7552,
+      "step": 19569
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074589127686473e-06,
+      "loss": 2.5011,
+      "step": 19570
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074541719342604e-06,
+      "loss": 2.2357,
+      "step": 19571
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074494310998736e-06,
+      "loss": 3.0269,
+      "step": 19572
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0744469026548675e-06,
+      "loss": 2.6165,
+      "step": 19573
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074399494310999e-06,
+      "loss": 3.0755,
+      "step": 19574
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.07435208596713e-06,
+      "loss": 2.4559,
+      "step": 19575
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0743046776232618e-06,
+      "loss": 2.3832,
+      "step": 19576
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074257269279393e-06,
+      "loss": 3.0774,
+      "step": 19577
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0742098609355248e-06,
+      "loss": 1.5135,
+      "step": 19578
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0741624525916565e-06,
+      "loss": 2.9941,
+      "step": 19579
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0741150442477878e-06,
+      "loss": 3.1498,
+      "step": 19580
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.074067635903919e-06,
+      "loss": 2.449,
+      "step": 19581
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0740202275600507e-06,
+      "loss": 1.8857,
+      "step": 19582
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073972819216182e-06,
+      "loss": 2.5446,
+      "step": 19583
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0739254108723133e-06,
+      "loss": 3.4363,
+      "step": 19584
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073878002528445e-06,
+      "loss": 2.3677,
+      "step": 19585
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0738305941845767e-06,
+      "loss": 1.8238,
+      "step": 19586
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073783185840708e-06,
+      "loss": 2.1654,
+      "step": 19587
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0737357774968397e-06,
+      "loss": 2.5461,
+      "step": 19588
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073688369152971e-06,
+      "loss": 2.7702,
+      "step": 19589
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0736409608091022e-06,
+      "loss": 2.8579,
+      "step": 19590
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073593552465234e-06,
+      "loss": 3.087,
+      "step": 19591
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0735461441213657e-06,
+      "loss": 2.0682,
+      "step": 19592
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073498735777497e-06,
+      "loss": 3.8988,
+      "step": 19593
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0734513274336286e-06,
+      "loss": 3.3493,
+      "step": 19594
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.07340391908976e-06,
+      "loss": 3.5885,
+      "step": 19595
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073356510745891e-06,
+      "loss": 3.8899,
+      "step": 19596
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.0733091024020225e-06,
+      "loss": 2.9783,
+      "step": 19597
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073261694058154e-06,
+      "loss": 3.9536,
+      "step": 19598
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073214285714286e-06,
+      "loss": 2.8685,
+      "step": 19599
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.073166877370417e-06,
+      "loss": 2.8832,
+      "step": 19600
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.073119469026549e-06,
+      "loss": 3.3426,
+      "step": 19601
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.07307206068268e-06,
+      "loss": 2.9791,
+      "step": 19602
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0730246523388114e-06,
+      "loss": 3.4948,
+      "step": 19603
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072977243994943e-06,
+      "loss": 1.7961,
+      "step": 19604
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072929835651075e-06,
+      "loss": 2.0328,
+      "step": 19605
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072882427307206e-06,
+      "loss": 2.9918,
+      "step": 19606
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072835018963338e-06,
+      "loss": 3.8235,
+      "step": 19607
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072787610619469e-06,
+      "loss": 2.8491,
+      "step": 19608
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0727402022756004e-06,
+      "loss": 1.7634,
+      "step": 19609
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072692793931732e-06,
+      "loss": 2.1165,
+      "step": 19610
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0726453855878634e-06,
+      "loss": 2.1216,
+      "step": 19611
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072597977243995e-06,
+      "loss": 2.3336,
+      "step": 19612
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0725505689001268e-06,
+      "loss": 2.9511,
+      "step": 19613
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072503160556258e-06,
+      "loss": 2.6727,
+      "step": 19614
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0724557522123893e-06,
+      "loss": 2.5254,
+      "step": 19615
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072408343868521e-06,
+      "loss": 3.1702,
+      "step": 19616
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0723609355246523e-06,
+      "loss": 3.847,
+      "step": 19617
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0723135271807836e-06,
+      "loss": 3.6237,
+      "step": 19618
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0722661188369153e-06,
+      "loss": 2.8817,
+      "step": 19619
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.072218710493047e-06,
+      "loss": 3.1281,
+      "step": 19620
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0721713021491783e-06,
+      "loss": 2.8847,
+      "step": 19621
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0721238938053096e-06,
+      "loss": 3.2078,
+      "step": 19622
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0720764854614413e-06,
+      "loss": 2.1908,
+      "step": 19623
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0720290771175726e-06,
+      "loss": 2.9093,
+      "step": 19624
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0719816687737043e-06,
+      "loss": 2.6677,
+      "step": 19625
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.071934260429836e-06,
+      "loss": 2.8517,
+      "step": 19626
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0718868520859672e-06,
+      "loss": 3.1009,
+      "step": 19627
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0718394437420985e-06,
+      "loss": 3.6212,
+      "step": 19628
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0717920353982302e-06,
+      "loss": 3.113,
+      "step": 19629
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0717446270543615e-06,
+      "loss": 2.8115,
+      "step": 19630
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.071697218710493e-06,
+      "loss": 2.2629,
+      "step": 19631
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0716498103666245e-06,
+      "loss": 2.1628,
+      "step": 19632
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.071602402022756e-06,
+      "loss": 2.2237,
+      "step": 19633
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0715549936788875e-06,
+      "loss": 2.6827,
+      "step": 19634
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.071507585335019e-06,
+      "loss": 2.1942,
+      "step": 19635
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0714601769911505e-06,
+      "loss": 2.5201,
+      "step": 19636
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0714127686472817e-06,
+      "loss": 3.0087,
+      "step": 19637
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0713653603034135e-06,
+      "loss": 2.5477,
+      "step": 19638
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.071317951959545e-06,
+      "loss": 4.0481,
+      "step": 19639
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0712705436156764e-06,
+      "loss": 2.0206,
+      "step": 19640
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0712231352718077e-06,
+      "loss": 2.2467,
+      "step": 19641
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0711757269279394e-06,
+      "loss": 2.5834,
+      "step": 19642
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0711283185840707e-06,
+      "loss": 3.4847,
+      "step": 19643
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.071080910240202e-06,
+      "loss": 3.1315,
+      "step": 19644
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0710335018963337e-06,
+      "loss": 2.2568,
+      "step": 19645
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0709860935524654e-06,
+      "loss": 1.4009,
+      "step": 19646
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0709386852085967e-06,
+      "loss": 2.8136,
+      "step": 19647
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0708912768647284e-06,
+      "loss": 2.4225,
+      "step": 19648
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0708438685208597e-06,
+      "loss": 3.7071,
+      "step": 19649
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.070796460176991e-06,
+      "loss": 2.561,
+      "step": 19650
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0707490518331226e-06,
+      "loss": 2.3531,
+      "step": 19651
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0707016434892543e-06,
+      "loss": 2.2424,
+      "step": 19652
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0706542351453856e-06,
+      "loss": 3.9301,
+      "step": 19653
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0706068268015173e-06,
+      "loss": 2.0446,
+      "step": 19654
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0705594184576486e-06,
+      "loss": 2.3276,
+      "step": 19655
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.07051201011378e-06,
+      "loss": 3.8321,
+      "step": 19656
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0704646017699116e-06,
+      "loss": 1.9882,
+      "step": 19657
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.070417193426043e-06,
+      "loss": 1.9444,
+      "step": 19658
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0703697850821746e-06,
+      "loss": 2.5527,
+      "step": 19659
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0703223767383063e-06,
+      "loss": 2.5965,
+      "step": 19660
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0702749683944376e-06,
+      "loss": 3.5965,
+      "step": 19661
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.070227560050569e-06,
+      "loss": 2.4752,
+      "step": 19662
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.0701801517067e-06,
+      "loss": 3.2351,
+      "step": 19663
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.070132743362832e-06,
+      "loss": 2.744,
+      "step": 19664
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0700853350189635e-06,
+      "loss": 3.0401,
+      "step": 19665
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.070037926675095e-06,
+      "loss": 2.6121,
+      "step": 19666
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0699905183312265e-06,
+      "loss": 1.9025,
+      "step": 19667
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069943109987358e-06,
+      "loss": 2.0701,
+      "step": 19668
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069895701643489e-06,
+      "loss": 2.3037,
+      "step": 19669
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0698482932996208e-06,
+      "loss": 3.5124,
+      "step": 19670
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069800884955752e-06,
+      "loss": 3.1253,
+      "step": 19671
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0697534766118838e-06,
+      "loss": 3.4256,
+      "step": 19672
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0697060682680155e-06,
+      "loss": 2.819,
+      "step": 19673
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0696586599241467e-06,
+      "loss": 3.5214,
+      "step": 19674
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069611251580278e-06,
+      "loss": 2.6998,
+      "step": 19675
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0695638432364097e-06,
+      "loss": 1.5105,
+      "step": 19676
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069516434892541e-06,
+      "loss": 1.9288,
+      "step": 19677
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0694690265486727e-06,
+      "loss": 2.5578,
+      "step": 19678
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0694216182048044e-06,
+      "loss": 2.4508,
+      "step": 19679
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0693742098609357e-06,
+      "loss": 2.7601,
+      "step": 19680
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069326801517067e-06,
+      "loss": 3.4474,
+      "step": 19681
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0692793931731987e-06,
+      "loss": 2.7786,
+      "step": 19682
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.06923198482933e-06,
+      "loss": 3.321,
+      "step": 19683
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0691845764854612e-06,
+      "loss": 1.7829,
+      "step": 19684
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069137168141593e-06,
+      "loss": 2.2774,
+      "step": 19685
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0690897597977247e-06,
+      "loss": 2.8767,
+      "step": 19686
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.069042351453856e-06,
+      "loss": 1.7147,
+      "step": 19687
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0689949431099872e-06,
+      "loss": 2.8067,
+      "step": 19688
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068947534766119e-06,
+      "loss": 1.9172,
+      "step": 19689
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.06890012642225e-06,
+      "loss": 2.1516,
+      "step": 19690
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068852718078382e-06,
+      "loss": 1.7311,
+      "step": 19691
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0688053097345136e-06,
+      "loss": 3.508,
+      "step": 19692
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068757901390645e-06,
+      "loss": 1.7708,
+      "step": 19693
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068710493046776e-06,
+      "loss": 2.5121,
+      "step": 19694
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068663084702908e-06,
+      "loss": 3.4699,
+      "step": 19695
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068615676359039e-06,
+      "loss": 1.6723,
+      "step": 19696
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0685682680151704e-06,
+      "loss": 2.1673,
+      "step": 19697
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068520859671302e-06,
+      "loss": 3.0196,
+      "step": 19698
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068473451327434e-06,
+      "loss": 3.1371,
+      "step": 19699
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068426042983565e-06,
+      "loss": 3.8021,
+      "step": 19700
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068378634639697e-06,
+      "loss": 2.3997,
+      "step": 19701
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068331226295828e-06,
+      "loss": 2.7692,
+      "step": 19702
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0682838179519594e-06,
+      "loss": 3.3498,
+      "step": 19703
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068236409608091e-06,
+      "loss": 3.1039,
+      "step": 19704
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068189001264223e-06,
+      "loss": 3.1175,
+      "step": 19705
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068141592920354e-06,
+      "loss": 3.4313,
+      "step": 19706
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0680941845764854e-06,
+      "loss": 2.83,
+      "step": 19707
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.068046776232617e-06,
+      "loss": 2.0323,
+      "step": 19708
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0679993678887483e-06,
+      "loss": 3.6765,
+      "step": 19709
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0679519595448796e-06,
+      "loss": 4.1814,
+      "step": 19710
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0679045512010113e-06,
+      "loss": 2.0299,
+      "step": 19711
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067857142857143e-06,
+      "loss": 2.8023,
+      "step": 19712
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0678097345132743e-06,
+      "loss": 2.6884,
+      "step": 19713
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067762326169406e-06,
+      "loss": 2.076,
+      "step": 19714
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0677149178255373e-06,
+      "loss": 3.57,
+      "step": 19715
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0676675094816686e-06,
+      "loss": 2.5517,
+      "step": 19716
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0676201011378003e-06,
+      "loss": 3.8892,
+      "step": 19717
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067572692793932e-06,
+      "loss": 1.4899,
+      "step": 19718
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0675252844500633e-06,
+      "loss": 2.8628,
+      "step": 19719
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067477876106195e-06,
+      "loss": 3.063,
+      "step": 19720
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0674304677623262e-06,
+      "loss": 2.7784,
+      "step": 19721
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0673830594184575e-06,
+      "loss": 2.0257,
+      "step": 19722
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0673356510745892e-06,
+      "loss": 3.4594,
+      "step": 19723
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.0672882427307205e-06,
+      "loss": 3.0978,
+      "step": 19724
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067240834386852e-06,
+      "loss": 3.0366,
+      "step": 19725
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067193426042984e-06,
+      "loss": 2.8401,
+      "step": 19726
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.067146017699115e-06,
+      "loss": 2.73,
+      "step": 19727
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0670986093552465e-06,
+      "loss": 2.3657,
+      "step": 19728
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0670512010113778e-06,
+      "loss": 2.7302,
+      "step": 19729
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0670037926675095e-06,
+      "loss": 3.343,
+      "step": 19730
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.066956384323641e-06,
+      "loss": 3.3468,
+      "step": 19731
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0669089759797724e-06,
+      "loss": 3.4403,
+      "step": 19732
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.066861567635904e-06,
+      "loss": 4.0284,
+      "step": 19733
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0668141592920354e-06,
+      "loss": 3.2774,
+      "step": 19734
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0667667509481667e-06,
+      "loss": 2.7093,
+      "step": 19735
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0667193426042984e-06,
+      "loss": 3.2056,
+      "step": 19736
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0666719342604297e-06,
+      "loss": 2.1342,
+      "step": 19737
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0666245259165614e-06,
+      "loss": 3.563,
+      "step": 19738
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.066577117572693e-06,
+      "loss": 2.6981,
+      "step": 19739
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0665297092288244e-06,
+      "loss": 3.044,
+      "step": 19740
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0664823008849557e-06,
+      "loss": 3.1308,
+      "step": 19741
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0664348925410874e-06,
+      "loss": 2.96,
+      "step": 19742
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0663874841972186e-06,
+      "loss": 2.3658,
+      "step": 19743
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0663400758533504e-06,
+      "loss": 3.659,
+      "step": 19744
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.066292667509482e-06,
+      "loss": 2.3719,
+      "step": 19745
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0662452591656133e-06,
+      "loss": 3.0803,
+      "step": 19746
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0661978508217446e-06,
+      "loss": 3.3978,
+      "step": 19747
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0661504424778763e-06,
+      "loss": 3.1873,
+      "step": 19748
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0661030341340076e-06,
+      "loss": 2.9877,
+      "step": 19749
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.066055625790139e-06,
+      "loss": 3.1227,
+      "step": 19750
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0660082174462706e-06,
+      "loss": 1.1675,
+      "step": 19751
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0659608091024023e-06,
+      "loss": 1.8572,
+      "step": 19752
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0659134007585336e-06,
+      "loss": 2.66,
+      "step": 19753
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.065865992414665e-06,
+      "loss": 2.7296,
+      "step": 19754
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0658185840707966e-06,
+      "loss": 3.9375,
+      "step": 19755
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.065771175726928e-06,
+      "loss": 2.5957,
+      "step": 19756
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0657237673830595e-06,
+      "loss": 2.4692,
+      "step": 19757
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0656763590391912e-06,
+      "loss": 3.1344,
+      "step": 19758
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0656289506953225e-06,
+      "loss": 3.9699,
+      "step": 19759
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.065581542351454e-06,
+      "loss": 3.7561,
+      "step": 19760
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0655341340075855e-06,
+      "loss": 2.4764,
+      "step": 19761
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0654867256637168e-06,
+      "loss": 2.4626,
+      "step": 19762
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.065439317319848e-06,
+      "loss": 3.51,
+      "step": 19763
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0653919089759798e-06,
+      "loss": 3.8968,
+      "step": 19764
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0653445006321115e-06,
+      "loss": 3.6751,
+      "step": 19765
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0652970922882428e-06,
+      "loss": 2.8584,
+      "step": 19766
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0652496839443745e-06,
+      "loss": 2.8179,
+      "step": 19767
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0652022756005057e-06,
+      "loss": 2.4843,
+      "step": 19768
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.065154867256637e-06,
+      "loss": 2.9071,
+      "step": 19769
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0651074589127687e-06,
+      "loss": 3.3428,
+      "step": 19770
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0650600505689004e-06,
+      "loss": 1.8119,
+      "step": 19771
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0650126422250317e-06,
+      "loss": 2.4787,
+      "step": 19772
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064965233881163e-06,
+      "loss": 2.6215,
+      "step": 19773
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0649178255372947e-06,
+      "loss": 3.6964,
+      "step": 19774
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064870417193426e-06,
+      "loss": 2.2148,
+      "step": 19775
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0648230088495573e-06,
+      "loss": 2.0402,
+      "step": 19776
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064775600505689e-06,
+      "loss": 1.9195,
+      "step": 19777
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0647281921618207e-06,
+      "loss": 2.8036,
+      "step": 19778
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064680783817952e-06,
+      "loss": 3.2728,
+      "step": 19779
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0646333754740836e-06,
+      "loss": 1.5747,
+      "step": 19780
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064585967130215e-06,
+      "loss": 2.5896,
+      "step": 19781
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064538558786346e-06,
+      "loss": 3.5986,
+      "step": 19782
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064491150442478e-06,
+      "loss": 1.4412,
+      "step": 19783
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0644437420986096e-06,
+      "loss": 2.7581,
+      "step": 19784
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064396333754741e-06,
+      "loss": 3.4729,
+      "step": 19785
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.0643489254108726e-06,
+      "loss": 2.0938,
+      "step": 19786
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064301517067004e-06,
+      "loss": 4.0097,
+      "step": 19787
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064254108723135e-06,
+      "loss": 3.1674,
+      "step": 19788
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064206700379267e-06,
+      "loss": 3.1887,
+      "step": 19789
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.064159292035398e-06,
+      "loss": 2.5159,
+      "step": 19790
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.06411188369153e-06,
+      "loss": 3.5497,
+      "step": 19791
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0640644753476616e-06,
+      "loss": 2.5593,
+      "step": 19792
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.064017067003793e-06,
+      "loss": 2.2298,
+      "step": 19793
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063969658659924e-06,
+      "loss": 1.825,
+      "step": 19794
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0639222503160554e-06,
+      "loss": 2.3657,
+      "step": 19795
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063874841972187e-06,
+      "loss": 1.8416,
+      "step": 19796
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063827433628319e-06,
+      "loss": 2.4624,
+      "step": 19797
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.06378002528445e-06,
+      "loss": 2.8118,
+      "step": 19798
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0637326169405818e-06,
+      "loss": 2.6961,
+      "step": 19799
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063685208596713e-06,
+      "loss": 2.7777,
+      "step": 19800
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0636378002528443e-06,
+      "loss": 3.2829,
+      "step": 19801
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063590391908976e-06,
+      "loss": 3.0197,
+      "step": 19802
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0635429835651073e-06,
+      "loss": 2.0513,
+      "step": 19803
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063495575221239e-06,
+      "loss": 2.2783,
+      "step": 19804
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0634481668773707e-06,
+      "loss": 2.4444,
+      "step": 19805
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063400758533502e-06,
+      "loss": 3.7787,
+      "step": 19806
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0633533501896333e-06,
+      "loss": 2.3489,
+      "step": 19807
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063305941845765e-06,
+      "loss": 1.7399,
+      "step": 19808
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0632585335018963e-06,
+      "loss": 3.4358,
+      "step": 19809
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063211125158028e-06,
+      "loss": 3.5225,
+      "step": 19810
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0631637168141597e-06,
+      "loss": 2.2562,
+      "step": 19811
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063116308470291e-06,
+      "loss": 2.643,
+      "step": 19812
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0630689001264223e-06,
+      "loss": 3.463,
+      "step": 19813
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.063021491782554e-06,
+      "loss": 2.7688,
+      "step": 19814
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0629740834386852e-06,
+      "loss": 3.0798,
+      "step": 19815
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0629266750948165e-06,
+      "loss": 2.3198,
+      "step": 19816
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0628792667509482e-06,
+      "loss": 3.0669,
+      "step": 19817
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.06283185840708e-06,
+      "loss": 1.9188,
+      "step": 19818
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062784450063211e-06,
+      "loss": 3.5955,
+      "step": 19819
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0627370417193425e-06,
+      "loss": 3.7772,
+      "step": 19820
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062689633375474e-06,
+      "loss": 2.9845,
+      "step": 19821
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0626422250316055e-06,
+      "loss": 4.2059,
+      "step": 19822
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062594816687737e-06,
+      "loss": 2.86,
+      "step": 19823
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062547408343869e-06,
+      "loss": 3.7437,
+      "step": 19824
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0625e-06,
+      "loss": 1.9669,
+      "step": 19825
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0624525916561314e-06,
+      "loss": 2.0879,
+      "step": 19826
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062405183312263e-06,
+      "loss": 1.8244,
+      "step": 19827
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0623577749683944e-06,
+      "loss": 2.6963,
+      "step": 19828
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0623103666245257e-06,
+      "loss": 3.4237,
+      "step": 19829
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0622629582806574e-06,
+      "loss": 3.0605,
+      "step": 19830
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062215549936789e-06,
+      "loss": 4.7348,
+      "step": 19831
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0621681415929204e-06,
+      "loss": 3.1542,
+      "step": 19832
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.062120733249052e-06,
+      "loss": 2.0674,
+      "step": 19833
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0620733249051834e-06,
+      "loss": 2.798,
+      "step": 19834
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0620259165613147e-06,
+      "loss": 3.9498,
+      "step": 19835
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0619785082174464e-06,
+      "loss": 3.1052,
+      "step": 19836
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.061931099873578e-06,
+      "loss": 2.022,
+      "step": 19837
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0618836915297093e-06,
+      "loss": 1.2796,
+      "step": 19838
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0618362831858406e-06,
+      "loss": 3.8437,
+      "step": 19839
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0617888748419723e-06,
+      "loss": 2.1632,
+      "step": 19840
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0617414664981036e-06,
+      "loss": 3.4193,
+      "step": 19841
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.061694058154235e-06,
+      "loss": 3.631,
+      "step": 19842
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0616466498103666e-06,
+      "loss": 2.2549,
+      "step": 19843
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0615992414664983e-06,
+      "loss": 2.7567,
+      "step": 19844
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0615518331226296e-06,
+      "loss": 3.7452,
+      "step": 19845
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0615044247787613e-06,
+      "loss": 2.8083,
+      "step": 19846
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0614570164348926e-06,
+      "loss": 2.7462,
+      "step": 19847
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.061409608091024e-06,
+      "loss": 3.88,
+      "step": 19848
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0613621997471555e-06,
+      "loss": 2.8352,
+      "step": 19849
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0613147914032873e-06,
+      "loss": 2.5245,
+      "step": 19850
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0612673830594185e-06,
+      "loss": 1.9353,
+      "step": 19851
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0612199747155502e-06,
+      "loss": 3.6896,
+      "step": 19852
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.0611725663716815e-06,
+      "loss": 3.1851,
+      "step": 19853
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.061125158027813e-06,
+      "loss": 2.4405,
+      "step": 19854
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0610777496839445e-06,
+      "loss": 3.0403,
+      "step": 19855
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0610303413400758e-06,
+      "loss": 4.0855,
+      "step": 19856
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0609829329962075e-06,
+      "loss": 2.0271,
+      "step": 19857
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0609355246523388e-06,
+      "loss": 3.1341,
+      "step": 19858
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0608881163084705e-06,
+      "loss": 2.1085,
+      "step": 19859
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0608407079646017e-06,
+      "loss": 2.5664,
+      "step": 19860
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.060793299620733e-06,
+      "loss": 2.572,
+      "step": 19861
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0607458912768647e-06,
+      "loss": 2.8084,
+      "step": 19862
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0606984829329964e-06,
+      "loss": 3.0665,
+      "step": 19863
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0606510745891277e-06,
+      "loss": 3.6,
+      "step": 19864
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0606036662452594e-06,
+      "loss": 2.1677,
+      "step": 19865
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0605562579013907e-06,
+      "loss": 2.1694,
+      "step": 19866
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.060508849557522e-06,
+      "loss": 3.6533,
+      "step": 19867
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0604614412136537e-06,
+      "loss": 2.3334,
+      "step": 19868
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.060414032869785e-06,
+      "loss": 2.882,
+      "step": 19869
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0603666245259167e-06,
+      "loss": 2.2423,
+      "step": 19870
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0603192161820484e-06,
+      "loss": 2.9792,
+      "step": 19871
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0602718078381797e-06,
+      "loss": 3.0193,
+      "step": 19872
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.060224399494311e-06,
+      "loss": 1.6246,
+      "step": 19873
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0601769911504426e-06,
+      "loss": 4.0589,
+      "step": 19874
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.060129582806574e-06,
+      "loss": 2.278,
+      "step": 19875
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0600821744627056e-06,
+      "loss": 2.0494,
+      "step": 19876
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0600347661188373e-06,
+      "loss": 3.4653,
+      "step": 19877
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0599873577749686e-06,
+      "loss": 3.1402,
+      "step": 19878
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0599399494311e-06,
+      "loss": 2.6721,
+      "step": 19879
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059892541087231e-06,
+      "loss": 3.1875,
+      "step": 19880
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059845132743363e-06,
+      "loss": 2.1694,
+      "step": 19881
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059797724399494e-06,
+      "loss": 3.1798,
+      "step": 19882
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059750316055626e-06,
+      "loss": 2.4008,
+      "step": 19883
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0597029077117576e-06,
+      "loss": 3.1934,
+      "step": 19884
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059655499367889e-06,
+      "loss": 2.708,
+      "step": 19885
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.05960809102402e-06,
+      "loss": 2.3868,
+      "step": 19886
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059560682680152e-06,
+      "loss": 3.1446,
+      "step": 19887
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059513274336283e-06,
+      "loss": 2.4773,
+      "step": 19888
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0594658659924144e-06,
+      "loss": 2.5124,
+      "step": 19889
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0594184576485465e-06,
+      "loss": 3.5042,
+      "step": 19890
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059371049304678e-06,
+      "loss": 3.9246,
+      "step": 19891
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059323640960809e-06,
+      "loss": 2.9927,
+      "step": 19892
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0592762326169408e-06,
+      "loss": 2.9366,
+      "step": 19893
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059228824273072e-06,
+      "loss": 2.3948,
+      "step": 19894
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0591814159292033e-06,
+      "loss": 2.3486,
+      "step": 19895
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059134007585335e-06,
+      "loss": 2.8785,
+      "step": 19896
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0590865992414667e-06,
+      "loss": 2.935,
+      "step": 19897
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.059039190897598e-06,
+      "loss": 2.881,
+      "step": 19898
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0589917825537297e-06,
+      "loss": 1.4448,
+      "step": 19899
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.058944374209861e-06,
+      "loss": 3.4106,
+      "step": 19900
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0588969658659923e-06,
+      "loss": 3.0105,
+      "step": 19901
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0588495575221236e-06,
+      "loss": 2.5428,
+      "step": 19902
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0588021491782553e-06,
+      "loss": 3.4021,
+      "step": 19903
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.058754740834387e-06,
+      "loss": 4.1824,
+      "step": 19904
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0587073324905183e-06,
+      "loss": 2.9952,
+      "step": 19905
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.05865992414665e-06,
+      "loss": 3.4904,
+      "step": 19906
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0586125158027812e-06,
+      "loss": 2.262,
+      "step": 19907
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0585651074589125e-06,
+      "loss": 3.2679,
+      "step": 19908
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0585176991150442e-06,
+      "loss": 2.0432,
+      "step": 19909
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.058470290771176e-06,
+      "loss": 2.8427,
+      "step": 19910
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0584228824273072e-06,
+      "loss": 2.2832,
+      "step": 19911
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.058375474083439e-06,
+      "loss": 3.1819,
+      "step": 19912
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.05832806573957e-06,
+      "loss": 3.3399,
+      "step": 19913
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0582806573957015e-06,
+      "loss": 2.1639,
+      "step": 19914
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.058233249051833e-06,
+      "loss": 3.3905,
+      "step": 19915
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.0581858407079645e-06,
+      "loss": 1.3454,
+      "step": 19916
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.058138432364096e-06,
+      "loss": 1.774,
+      "step": 19917
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.058091024020228e-06,
+      "loss": 3.7787,
+      "step": 19918
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.058043615676359e-06,
+      "loss": 2.8521,
+      "step": 19919
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0579962073324904e-06,
+      "loss": 3.3337,
+      "step": 19920
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.057948798988622e-06,
+      "loss": 3.5122,
+      "step": 19921
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0579013906447534e-06,
+      "loss": 3.5531,
+      "step": 19922
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.057853982300885e-06,
+      "loss": 2.2776,
+      "step": 19923
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0578065739570164e-06,
+      "loss": 3.0423,
+      "step": 19924
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.057759165613148e-06,
+      "loss": 3.1434,
+      "step": 19925
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0577117572692794e-06,
+      "loss": 2.0222,
+      "step": 19926
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0576643489254107e-06,
+      "loss": 3.1493,
+      "step": 19927
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0576169405815424e-06,
+      "loss": 2.2266,
+      "step": 19928
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0575695322376737e-06,
+      "loss": 2.9997,
+      "step": 19929
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0575221238938054e-06,
+      "loss": 2.338,
+      "step": 19930
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.057474715549937e-06,
+      "loss": 3.1403,
+      "step": 19931
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0574273072060683e-06,
+      "loss": 1.1736,
+      "step": 19932
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0573798988621996e-06,
+      "loss": 2.7318,
+      "step": 19933
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0573324905183313e-06,
+      "loss": 1.9851,
+      "step": 19934
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0572850821744626e-06,
+      "loss": 2.5122,
+      "step": 19935
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0572376738305943e-06,
+      "loss": 1.919,
+      "step": 19936
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.057190265486726e-06,
+      "loss": 3.9221,
+      "step": 19937
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0571428571428573e-06,
+      "loss": 3.406,
+      "step": 19938
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0570954487989886e-06,
+      "loss": 3.3976,
+      "step": 19939
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0570480404551203e-06,
+      "loss": 3.3489,
+      "step": 19940
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0570006321112516e-06,
+      "loss": 2.939,
+      "step": 19941
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.056953223767383e-06,
+      "loss": 2.5758,
+      "step": 19942
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0569058154235145e-06,
+      "loss": 1.9226,
+      "step": 19943
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0568584070796462e-06,
+      "loss": 3.4966,
+      "step": 19944
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0568109987357775e-06,
+      "loss": 2.7021,
+      "step": 19945
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.056763590391909e-06,
+      "loss": 2.7696,
+      "step": 19946
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0567161820480405e-06,
+      "loss": 2.7717,
+      "step": 19947
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.056668773704172e-06,
+      "loss": 2.2284,
+      "step": 19948
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0566213653603035e-06,
+      "loss": 2.3104,
+      "step": 19949
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.056573957016435e-06,
+      "loss": 1.8856,
+      "step": 19950
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0565265486725665e-06,
+      "loss": 3.3269,
+      "step": 19951
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0564791403286978e-06,
+      "loss": 2.1427,
+      "step": 19952
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0564317319848295e-06,
+      "loss": 2.1538,
+      "step": 19953
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0563843236409607e-06,
+      "loss": 1.5602,
+      "step": 19954
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.056336915297092e-06,
+      "loss": 2.2089,
+      "step": 19955
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0562895069532237e-06,
+      "loss": 3.5617,
+      "step": 19956
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0562420986093554e-06,
+      "loss": 3.5268,
+      "step": 19957
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0561946902654867e-06,
+      "loss": 2.9683,
+      "step": 19958
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0561472819216184e-06,
+      "loss": 2.8079,
+      "step": 19959
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0560998735777497e-06,
+      "loss": 2.4519,
+      "step": 19960
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.056052465233881e-06,
+      "loss": 3.0729,
+      "step": 19961
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0560050568900127e-06,
+      "loss": 2.452,
+      "step": 19962
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0559576485461444e-06,
+      "loss": 3.1463,
+      "step": 19963
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0559102402022757e-06,
+      "loss": 2.5321,
+      "step": 19964
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0558628318584074e-06,
+      "loss": 3.1101,
+      "step": 19965
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0558154235145386e-06,
+      "loss": 2.4797,
+      "step": 19966
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.05576801517067e-06,
+      "loss": 3.5453,
+      "step": 19967
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055720606826801e-06,
+      "loss": 2.355,
+      "step": 19968
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055673198482933e-06,
+      "loss": 1.8042,
+      "step": 19969
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0556257901390646e-06,
+      "loss": 2.6557,
+      "step": 19970
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055578381795196e-06,
+      "loss": 2.0741,
+      "step": 19971
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0555309734513276e-06,
+      "loss": 2.7736,
+      "step": 19972
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055483565107459e-06,
+      "loss": 2.9161,
+      "step": 19973
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.05543615676359e-06,
+      "loss": 2.2981,
+      "step": 19974
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055388748419722e-06,
+      "loss": 3.8692,
+      "step": 19975
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0553413400758536e-06,
+      "loss": 1.4822,
+      "step": 19976
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055293931731985e-06,
+      "loss": 2.8901,
+      "step": 19977
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.0552465233881166e-06,
+      "loss": 2.9437,
+      "step": 19978
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055199115044248e-06,
+      "loss": 2.8475,
+      "step": 19979
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.055151706700379e-06,
+      "loss": 2.1833,
+      "step": 19980
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.055104298356511e-06,
+      "loss": 3.348,
+      "step": 19981
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.055056890012642e-06,
+      "loss": 2.4015,
+      "step": 19982
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.055009481668774e-06,
+      "loss": 2.9716,
+      "step": 19983
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0549620733249055e-06,
+      "loss": 3.3341,
+      "step": 19984
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054914664981037e-06,
+      "loss": 2.0709,
+      "step": 19985
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054867256637168e-06,
+      "loss": 3.1787,
+      "step": 19986
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0548198482932998e-06,
+      "loss": 0.9483,
+      "step": 19987
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054772439949431e-06,
+      "loss": 2.5202,
+      "step": 19988
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0547250316055628e-06,
+      "loss": 2.5897,
+      "step": 19989
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054677623261694e-06,
+      "loss": 2.1557,
+      "step": 19990
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0546302149178257e-06,
+      "loss": 2.4316,
+      "step": 19991
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054582806573957e-06,
+      "loss": 2.326,
+      "step": 19992
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0545353982300883e-06,
+      "loss": 3.2381,
+      "step": 19993
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.05448798988622e-06,
+      "loss": 2.689,
+      "step": 19994
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0544405815423513e-06,
+      "loss": 2.4875,
+      "step": 19995
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054393173198483e-06,
+      "loss": 3.593,
+      "step": 19996
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0543457648546147e-06,
+      "loss": 3.0721,
+      "step": 19997
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054298356510746e-06,
+      "loss": 3.0057,
+      "step": 19998
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0542509481668773e-06,
+      "loss": 2.9687,
+      "step": 19999
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054203539823009e-06,
+      "loss": 1.8995,
+      "step": 20000
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0541561314791402e-06,
+      "loss": 3.1095,
+      "step": 20001
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054108723135272e-06,
+      "loss": 2.5737,
+      "step": 20002
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0540613147914036e-06,
+      "loss": 2.3927,
+      "step": 20003
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.054013906447535e-06,
+      "loss": 2.7906,
+      "step": 20004
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053966498103666e-06,
+      "loss": 3.5106,
+      "step": 20005
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053919089759798e-06,
+      "loss": 1.9789,
+      "step": 20006
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053871681415929e-06,
+      "loss": 2.2679,
+      "step": 20007
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0538242730720605e-06,
+      "loss": 2.6593,
+      "step": 20008
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053776864728192e-06,
+      "loss": 4.0738,
+      "step": 20009
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053729456384324e-06,
+      "loss": 2.7979,
+      "step": 20010
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053682048040455e-06,
+      "loss": 1.9216,
+      "step": 20011
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0536346396965864e-06,
+      "loss": 2.5873,
+      "step": 20012
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053587231352718e-06,
+      "loss": 1.362,
+      "step": 20013
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0535398230088494e-06,
+      "loss": 3.9565,
+      "step": 20014
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053492414664981e-06,
+      "loss": 3.4211,
+      "step": 20015
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053445006321113e-06,
+      "loss": 1.8493,
+      "step": 20016
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053397597977244e-06,
+      "loss": 2.7889,
+      "step": 20017
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0533501896333754e-06,
+      "loss": 2.6121,
+      "step": 20018
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053302781289507e-06,
+      "loss": 4.0684,
+      "step": 20019
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0532553729456384e-06,
+      "loss": 2.7393,
+      "step": 20020
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0532079646017697e-06,
+      "loss": 2.187,
+      "step": 20021
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0531605562579014e-06,
+      "loss": 2.5591,
+      "step": 20022
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053113147914033e-06,
+      "loss": 3.7984,
+      "step": 20023
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0530657395701643e-06,
+      "loss": 2.1668,
+      "step": 20024
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.053018331226296e-06,
+      "loss": 2.0993,
+      "step": 20025
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0529709228824273e-06,
+      "loss": 2.9558,
+      "step": 20026
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0529235145385586e-06,
+      "loss": 2.8956,
+      "step": 20027
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0528761061946903e-06,
+      "loss": 3.642,
+      "step": 20028
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.052828697850822e-06,
+      "loss": 1.6107,
+      "step": 20029
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0527812895069533e-06,
+      "loss": 2.5293,
+      "step": 20030
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.052733881163085e-06,
+      "loss": 3.4653,
+      "step": 20031
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0526864728192163e-06,
+      "loss": 2.683,
+      "step": 20032
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0526390644753476e-06,
+      "loss": 3.6986,
+      "step": 20033
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.052591656131479e-06,
+      "loss": 2.502,
+      "step": 20034
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0525442477876106e-06,
+      "loss": 3.1578,
+      "step": 20035
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0524968394437423e-06,
+      "loss": 1.203,
+      "step": 20036
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0524494310998735e-06,
+      "loss": 2.685,
+      "step": 20037
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0524020227560052e-06,
+      "loss": 3.6292,
+      "step": 20038
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0523546144121365e-06,
+      "loss": 2.8709,
+      "step": 20039
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.052307206068268e-06,
+      "loss": 2.4773,
+      "step": 20040
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0522597977243995e-06,
+      "loss": 1.6768,
+      "step": 20041
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.052212389380531e-06,
+      "loss": 2.107,
+      "step": 20042
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.0521649810366625e-06,
+      "loss": 3.2216,
+      "step": 20043
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.052117572692794e-06,
+      "loss": 2.3681,
+      "step": 20044
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0520701643489255e-06,
+      "loss": 2.7736,
+      "step": 20045
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0520227560050568e-06,
+      "loss": 3.5985,
+      "step": 20046
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0519753476611885e-06,
+      "loss": 3.5474,
+      "step": 20047
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0519279393173197e-06,
+      "loss": 3.5075,
+      "step": 20048
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0518805309734514e-06,
+      "loss": 1.7458,
+      "step": 20049
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.051833122629583e-06,
+      "loss": 2.1701,
+      "step": 20050
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0517857142857144e-06,
+      "loss": 3.4951,
+      "step": 20051
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0517383059418457e-06,
+      "loss": 2.2876,
+      "step": 20052
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0516908975979774e-06,
+      "loss": 2.9565,
+      "step": 20053
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0516434892541087e-06,
+      "loss": 3.4895,
+      "step": 20054
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0515960809102404e-06,
+      "loss": 3.2045,
+      "step": 20055
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0515486725663717e-06,
+      "loss": 1.0821,
+      "step": 20056
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0515012642225034e-06,
+      "loss": 3.4384,
+      "step": 20057
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0514538558786347e-06,
+      "loss": 2.9227,
+      "step": 20058
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.051406447534766e-06,
+      "loss": 3.0279,
+      "step": 20059
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0513590391908976e-06,
+      "loss": 3.0844,
+      "step": 20060
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.051311630847029e-06,
+      "loss": 2.5669,
+      "step": 20061
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0512642225031606e-06,
+      "loss": 2.1525,
+      "step": 20062
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0512168141592923e-06,
+      "loss": 2.2333,
+      "step": 20063
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0511694058154236e-06,
+      "loss": 3.6195,
+      "step": 20064
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.051121997471555e-06,
+      "loss": 2.4651,
+      "step": 20065
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0510745891276866e-06,
+      "loss": 2.6929,
+      "step": 20066
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.051027180783818e-06,
+      "loss": 3.055,
+      "step": 20067
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0509797724399496e-06,
+      "loss": 2.9434,
+      "step": 20068
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0509323640960813e-06,
+      "loss": 2.8141,
+      "step": 20069
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0508849557522126e-06,
+      "loss": 2.9482,
+      "step": 20070
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050837547408344e-06,
+      "loss": 3.3225,
+      "step": 20071
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0507901390644755e-06,
+      "loss": 1.6995,
+      "step": 20072
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050742730720607e-06,
+      "loss": 3.9832,
+      "step": 20073
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050695322376738e-06,
+      "loss": 2.7333,
+      "step": 20074
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.05064791403287e-06,
+      "loss": 2.6228,
+      "step": 20075
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0506005056890015e-06,
+      "loss": 2.8119,
+      "step": 20076
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050553097345133e-06,
+      "loss": 3.4645,
+      "step": 20077
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050505689001264e-06,
+      "loss": 2.3533,
+      "step": 20078
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0504582806573958e-06,
+      "loss": 3.1566,
+      "step": 20079
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050410872313527e-06,
+      "loss": 3.0396,
+      "step": 20080
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0503634639696588e-06,
+      "loss": 2.3939,
+      "step": 20081
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0503160556257905e-06,
+      "loss": 2.9918,
+      "step": 20082
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0502686472819218e-06,
+      "loss": 3.3628,
+      "step": 20083
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050221238938053e-06,
+      "loss": 2.9049,
+      "step": 20084
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0501738305941847e-06,
+      "loss": 3.0051,
+      "step": 20085
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050126422250316e-06,
+      "loss": 3.5134,
+      "step": 20086
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0500790139064473e-06,
+      "loss": 3.0572,
+      "step": 20087
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.050031605562579e-06,
+      "loss": 1.5105,
+      "step": 20088
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0499841972187107e-06,
+      "loss": 3.4,
+      "step": 20089
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049936788874842e-06,
+      "loss": 3.7059,
+      "step": 20090
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0498893805309737e-06,
+      "loss": 1.345,
+      "step": 20091
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049841972187105e-06,
+      "loss": 1.6959,
+      "step": 20092
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0497945638432362e-06,
+      "loss": 2.6043,
+      "step": 20093
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049747155499368e-06,
+      "loss": 1.9274,
+      "step": 20094
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0496997471554997e-06,
+      "loss": 2.61,
+      "step": 20095
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049652338811631e-06,
+      "loss": 1.4181,
+      "step": 20096
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0496049304677626e-06,
+      "loss": 3.9911,
+      "step": 20097
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049557522123894e-06,
+      "loss": 3.179,
+      "step": 20098
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049510113780025e-06,
+      "loss": 2.3129,
+      "step": 20099
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0494627054361565e-06,
+      "loss": 1.8358,
+      "step": 20100
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049415297092288e-06,
+      "loss": 2.3811,
+      "step": 20101
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.04936788874842e-06,
+      "loss": 2.0628,
+      "step": 20102
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049320480404551e-06,
+      "loss": 3.2334,
+      "step": 20103
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049273072060683e-06,
+      "loss": 3.3321,
+      "step": 20104
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049225663716814e-06,
+      "loss": 3.4559,
+      "step": 20105
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.0491782553729454e-06,
+      "loss": 3.5411,
+      "step": 20106
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.049130847029077e-06,
+      "loss": 2.0024,
+      "step": 20107
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.049083438685209e-06,
+      "loss": 2.8085,
+      "step": 20108
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.04903603034134e-06,
+      "loss": 3.8224,
+      "step": 20109
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048988621997472e-06,
+      "loss": 2.5125,
+      "step": 20110
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048941213653603e-06,
+      "loss": 2.6683,
+      "step": 20111
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0488938053097344e-06,
+      "loss": 2.3493,
+      "step": 20112
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048846396965866e-06,
+      "loss": 3.5652,
+      "step": 20113
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0487989886219974e-06,
+      "loss": 2.5989,
+      "step": 20114
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048751580278129e-06,
+      "loss": 3.6064,
+      "step": 20115
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0487041719342608e-06,
+      "loss": 2.7569,
+      "step": 20116
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048656763590392e-06,
+      "loss": 1.2579,
+      "step": 20117
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0486093552465233e-06,
+      "loss": 3.1565,
+      "step": 20118
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048561946902655e-06,
+      "loss": 2.6631,
+      "step": 20119
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0485145385587863e-06,
+      "loss": 3.1119,
+      "step": 20120
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048467130214918e-06,
+      "loss": 2.0941,
+      "step": 20121
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0484197218710493e-06,
+      "loss": 3.2586,
+      "step": 20122
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.048372313527181e-06,
+      "loss": 2.6877,
+      "step": 20123
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0483249051833123e-06,
+      "loss": 3.3706,
+      "step": 20124
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0482774968394436e-06,
+      "loss": 1.505,
+      "step": 20125
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0482300884955753e-06,
+      "loss": 2.561,
+      "step": 20126
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0481826801517066e-06,
+      "loss": 2.4664,
+      "step": 20127
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0481352718078383e-06,
+      "loss": 2.5798,
+      "step": 20128
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.04808786346397e-06,
+      "loss": 2.95,
+      "step": 20129
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0480404551201012e-06,
+      "loss": 4.3108,
+      "step": 20130
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0479930467762325e-06,
+      "loss": 2.1072,
+      "step": 20131
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0479456384323642e-06,
+      "loss": 2.705,
+      "step": 20132
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0478982300884955e-06,
+      "loss": 3.2669,
+      "step": 20133
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0478508217446272e-06,
+      "loss": 2.7474,
+      "step": 20134
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.047803413400759e-06,
+      "loss": 3.4659,
+      "step": 20135
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.04775600505689e-06,
+      "loss": 1.8809,
+      "step": 20136
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0477085967130215e-06,
+      "loss": 2.8914,
+      "step": 20137
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.047661188369153e-06,
+      "loss": 2.8926,
+      "step": 20138
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0476137800252845e-06,
+      "loss": 1.7019,
+      "step": 20139
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0475663716814157e-06,
+      "loss": 2.2401,
+      "step": 20140
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0475189633375475e-06,
+      "loss": 3.2325,
+      "step": 20141
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.047471554993679e-06,
+      "loss": 2.5363,
+      "step": 20142
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0474241466498104e-06,
+      "loss": 2.6544,
+      "step": 20143
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0473767383059417e-06,
+      "loss": 4.1475,
+      "step": 20144
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0473293299620734e-06,
+      "loss": 2.5443,
+      "step": 20145
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0472819216182047e-06,
+      "loss": 2.8649,
+      "step": 20146
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0472345132743364e-06,
+      "loss": 3.5594,
+      "step": 20147
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.047187104930468e-06,
+      "loss": 3.0649,
+      "step": 20148
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0471396965865994e-06,
+      "loss": 2.4074,
+      "step": 20149
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0470922882427307e-06,
+      "loss": 1.6456,
+      "step": 20150
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0470448798988624e-06,
+      "loss": 3.0046,
+      "step": 20151
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0469974715549937e-06,
+      "loss": 3.7561,
+      "step": 20152
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.046950063211125e-06,
+      "loss": 1.7781,
+      "step": 20153
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0469026548672566e-06,
+      "loss": 3.7458,
+      "step": 20154
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0468552465233883e-06,
+      "loss": 2.1888,
+      "step": 20155
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0468078381795196e-06,
+      "loss": 3.2371,
+      "step": 20156
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0467604298356513e-06,
+      "loss": 3.6561,
+      "step": 20157
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0467130214917826e-06,
+      "loss": 3.839,
+      "step": 20158
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.046665613147914e-06,
+      "loss": 1.667,
+      "step": 20159
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0466182048040456e-06,
+      "loss": 2.7635,
+      "step": 20160
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0465707964601773e-06,
+      "loss": 2.1625,
+      "step": 20161
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0465233881163086e-06,
+      "loss": 3.6309,
+      "step": 20162
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0464759797724403e-06,
+      "loss": 2.5824,
+      "step": 20163
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0464285714285716e-06,
+      "loss": 2.4403,
+      "step": 20164
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.046381163084703e-06,
+      "loss": 3.5038,
+      "step": 20165
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.046333754740834e-06,
+      "loss": 3.8224,
+      "step": 20166
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.046286346396966e-06,
+      "loss": 3.536,
+      "step": 20167
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0462389380530975e-06,
+      "loss": 2.1793,
+      "step": 20168
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.046191529709229e-06,
+      "loss": 4.2003,
+      "step": 20169
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.0461441213653605e-06,
+      "loss": 2.5612,
+      "step": 20170
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.046096713021492e-06,
+      "loss": 3.3101,
+      "step": 20171
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.046049304677623e-06,
+      "loss": 2.7179,
+      "step": 20172
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0460018963337548e-06,
+      "loss": 2.0313,
+      "step": 20173
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045954487989886e-06,
+      "loss": 2.2947,
+      "step": 20174
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0459070796460178e-06,
+      "loss": 2.8949,
+      "step": 20175
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0458596713021495e-06,
+      "loss": 2.8804,
+      "step": 20176
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0458122629582807e-06,
+      "loss": 1.9765,
+      "step": 20177
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045764854614412e-06,
+      "loss": 2.6153,
+      "step": 20178
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0457174462705437e-06,
+      "loss": 3.7839,
+      "step": 20179
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045670037926675e-06,
+      "loss": 3.5013,
+      "step": 20180
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0456226295828067e-06,
+      "loss": 4.2359,
+      "step": 20181
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0455752212389384e-06,
+      "loss": 3.3778,
+      "step": 20182
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0455278128950697e-06,
+      "loss": 2.9688,
+      "step": 20183
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045480404551201e-06,
+      "loss": 3.7379,
+      "step": 20184
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0454329962073327e-06,
+      "loss": 2.5596,
+      "step": 20185
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045385587863464e-06,
+      "loss": 2.0286,
+      "step": 20186
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0453381795195952e-06,
+      "loss": 3.7231,
+      "step": 20187
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045290771175727e-06,
+      "loss": 3.2173,
+      "step": 20188
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0452433628318587e-06,
+      "loss": 2.9669,
+      "step": 20189
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.04519595448799e-06,
+      "loss": 3.8958,
+      "step": 20190
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0451485461441212e-06,
+      "loss": 2.6792,
+      "step": 20191
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045101137800253e-06,
+      "loss": 3.3376,
+      "step": 20192
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045053729456384e-06,
+      "loss": 3.2556,
+      "step": 20193
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.045006321112516e-06,
+      "loss": 2.9711,
+      "step": 20194
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0449589127686476e-06,
+      "loss": 2.808,
+      "step": 20195
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044911504424779e-06,
+      "loss": 3.3314,
+      "step": 20196
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.04486409608091e-06,
+      "loss": 3.7588,
+      "step": 20197
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044816687737042e-06,
+      "loss": 3.4737,
+      "step": 20198
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044769279393173e-06,
+      "loss": 2.945,
+      "step": 20199
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0447218710493044e-06,
+      "loss": 3.0451,
+      "step": 20200
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044674462705436e-06,
+      "loss": 2.6337,
+      "step": 20201
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044627054361568e-06,
+      "loss": 3.6061,
+      "step": 20202
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044579646017699e-06,
+      "loss": 2.296,
+      "step": 20203
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044532237673831e-06,
+      "loss": 3.0082,
+      "step": 20204
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044484829329962e-06,
+      "loss": 3.2259,
+      "step": 20205
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0444374209860934e-06,
+      "loss": 1.9965,
+      "step": 20206
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044390012642225e-06,
+      "loss": 2.8002,
+      "step": 20207
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044342604298357e-06,
+      "loss": 2.5007,
+      "step": 20208
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044295195954488e-06,
+      "loss": 1.8179,
+      "step": 20209
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0442477876106194e-06,
+      "loss": 2.6426,
+      "step": 20210
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044200379266751e-06,
+      "loss": 2.4701,
+      "step": 20211
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0441529709228823e-06,
+      "loss": 3.388,
+      "step": 20212
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0441055625790136e-06,
+      "loss": 2.9005,
+      "step": 20213
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0440581542351453e-06,
+      "loss": 3.1297,
+      "step": 20214
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.044010745891277e-06,
+      "loss": 2.4414,
+      "step": 20215
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0439633375474083e-06,
+      "loss": 2.2216,
+      "step": 20216
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.04391592920354e-06,
+      "loss": 3.1223,
+      "step": 20217
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0438685208596713e-06,
+      "loss": 2.4879,
+      "step": 20218
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0438211125158026e-06,
+      "loss": 2.5845,
+      "step": 20219
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0437737041719343e-06,
+      "loss": 1.5606,
+      "step": 20220
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.043726295828066e-06,
+      "loss": 3.4248,
+      "step": 20221
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0436788874841973e-06,
+      "loss": 3.6266,
+      "step": 20222
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.043631479140329e-06,
+      "loss": 2.7831,
+      "step": 20223
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0435840707964602e-06,
+      "loss": 2.2374,
+      "step": 20224
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0435366624525915e-06,
+      "loss": 3.8919,
+      "step": 20225
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0434892541087232e-06,
+      "loss": 3.0832,
+      "step": 20226
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0434418457648545e-06,
+      "loss": 3.7905,
+      "step": 20227
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.043394437420986e-06,
+      "loss": 3.3824,
+      "step": 20228
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.043347029077118e-06,
+      "loss": 2.4656,
+      "step": 20229
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.043299620733249e-06,
+      "loss": 3.593,
+      "step": 20230
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0432522123893805e-06,
+      "loss": 3.4538,
+      "step": 20231
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0432048040455118e-06,
+      "loss": 1.5462,
+      "step": 20232
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.0431573957016435e-06,
+      "loss": 2.7216,
+      "step": 20233
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.043109987357775e-06,
+      "loss": 2.5238,
+      "step": 20234
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0430625790139064e-06,
+      "loss": 3.6832,
+      "step": 20235
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.043015170670038e-06,
+      "loss": 3.5054,
+      "step": 20236
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0429677623261694e-06,
+      "loss": 2.2115,
+      "step": 20237
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0429203539823007e-06,
+      "loss": 2.5626,
+      "step": 20238
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0428729456384324e-06,
+      "loss": 3.4965,
+      "step": 20239
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0428255372945637e-06,
+      "loss": 1.2891,
+      "step": 20240
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0427781289506954e-06,
+      "loss": 3.5109,
+      "step": 20241
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.042730720606827e-06,
+      "loss": 3.1447,
+      "step": 20242
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0426833122629584e-06,
+      "loss": 2.8818,
+      "step": 20243
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0426359039190897e-06,
+      "loss": 1.9247,
+      "step": 20244
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0425884955752214e-06,
+      "loss": 2.3812,
+      "step": 20245
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0425410872313526e-06,
+      "loss": 3.0112,
+      "step": 20246
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0424936788874844e-06,
+      "loss": 2.8415,
+      "step": 20247
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.042446270543616e-06,
+      "loss": 2.9861,
+      "step": 20248
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0423988621997473e-06,
+      "loss": 2.1901,
+      "step": 20249
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0423514538558786e-06,
+      "loss": 2.6021,
+      "step": 20250
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0423040455120103e-06,
+      "loss": 3.2072,
+      "step": 20251
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0422566371681416e-06,
+      "loss": 2.6643,
+      "step": 20252
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.042209228824273e-06,
+      "loss": 2.5592,
+      "step": 20253
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0421618204804046e-06,
+      "loss": 3.4466,
+      "step": 20254
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0421144121365363e-06,
+      "loss": 2.6295,
+      "step": 20255
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0420670037926676e-06,
+      "loss": 1.3797,
+      "step": 20256
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.042019595448799e-06,
+      "loss": 3.0756,
+      "step": 20257
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0419721871049306e-06,
+      "loss": 2.257,
+      "step": 20258
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.041924778761062e-06,
+      "loss": 3.392,
+      "step": 20259
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0418773704171935e-06,
+      "loss": 3.2019,
+      "step": 20260
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0418299620733252e-06,
+      "loss": 2.5556,
+      "step": 20261
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0417825537294565e-06,
+      "loss": 3.5789,
+      "step": 20262
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.041735145385588e-06,
+      "loss": 2.9321,
+      "step": 20263
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0416877370417195e-06,
+      "loss": 2.1354,
+      "step": 20264
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0416403286978508e-06,
+      "loss": 3.2294,
+      "step": 20265
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.041592920353982e-06,
+      "loss": 2.6294,
+      "step": 20266
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0415455120101138e-06,
+      "loss": 2.6159,
+      "step": 20267
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0414981036662455e-06,
+      "loss": 2.3062,
+      "step": 20268
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0414506953223768e-06,
+      "loss": 3.0175,
+      "step": 20269
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0414032869785085e-06,
+      "loss": 3.2594,
+      "step": 20270
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0413558786346397e-06,
+      "loss": 2.013,
+      "step": 20271
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.041308470290771e-06,
+      "loss": 2.7536,
+      "step": 20272
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0412610619469027e-06,
+      "loss": 1.8629,
+      "step": 20273
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0412136536030344e-06,
+      "loss": 2.7424,
+      "step": 20274
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0411662452591657e-06,
+      "loss": 2.1303,
+      "step": 20275
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.041118836915297e-06,
+      "loss": 1.7398,
+      "step": 20276
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0410714285714287e-06,
+      "loss": 1.9026,
+      "step": 20277
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.04102402022756e-06,
+      "loss": 2.7927,
+      "step": 20278
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0409766118836913e-06,
+      "loss": 3.011,
+      "step": 20279
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040929203539823e-06,
+      "loss": 3.1616,
+      "step": 20280
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0408817951959547e-06,
+      "loss": 3.5,
+      "step": 20281
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040834386852086e-06,
+      "loss": 2.9476,
+      "step": 20282
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0407869785082176e-06,
+      "loss": 2.8916,
+      "step": 20283
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040739570164349e-06,
+      "loss": 3.1023,
+      "step": 20284
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.04069216182048e-06,
+      "loss": 2.8801,
+      "step": 20285
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040644753476612e-06,
+      "loss": 2.8188,
+      "step": 20286
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0405973451327436e-06,
+      "loss": 3.1569,
+      "step": 20287
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040549936788875e-06,
+      "loss": 3.6428,
+      "step": 20288
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.0405025284450066e-06,
+      "loss": 3.0008,
+      "step": 20289
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040455120101138e-06,
+      "loss": 3.0267,
+      "step": 20290
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040407711757269e-06,
+      "loss": 2.7771,
+      "step": 20291
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040360303413401e-06,
+      "loss": 2.6164,
+      "step": 20292
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040312895069532e-06,
+      "loss": 3.6065,
+      "step": 20293
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040265486725664e-06,
+      "loss": 2.4723,
+      "step": 20294
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040218078381795e-06,
+      "loss": 2.8599,
+      "step": 20295
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040170670037927e-06,
+      "loss": 3.0311,
+      "step": 20296
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.040123261694058e-06,
+      "loss": 3.6274,
+      "step": 20297
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0400758533501894e-06,
+      "loss": 2.402,
+      "step": 20298
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.040028445006321e-06,
+      "loss": 2.8885,
+      "step": 20299
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039981036662453e-06,
+      "loss": 3.0289,
+      "step": 20300
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039933628318584e-06,
+      "loss": 2.7243,
+      "step": 20301
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0398862199747158e-06,
+      "loss": 2.8572,
+      "step": 20302
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039838811630847e-06,
+      "loss": 2.851,
+      "step": 20303
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0397914032869783e-06,
+      "loss": 3.3959,
+      "step": 20304
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.03974399494311e-06,
+      "loss": 1.5049,
+      "step": 20305
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0396965865992413e-06,
+      "loss": 3.5743,
+      "step": 20306
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039649178255373e-06,
+      "loss": 2.5329,
+      "step": 20307
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0396017699115047e-06,
+      "loss": 2.2946,
+      "step": 20308
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039554361567636e-06,
+      "loss": 3.109,
+      "step": 20309
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0395069532237673e-06,
+      "loss": 2.8004,
+      "step": 20310
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039459544879899e-06,
+      "loss": 2.9761,
+      "step": 20311
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0394121365360303e-06,
+      "loss": 2.7232,
+      "step": 20312
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039364728192162e-06,
+      "loss": 2.5732,
+      "step": 20313
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0393173198482937e-06,
+      "loss": 2.9838,
+      "step": 20314
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.039269911504425e-06,
+      "loss": 2.6203,
+      "step": 20315
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0392225031605563e-06,
+      "loss": 3.5285,
+      "step": 20316
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0391750948166875e-06,
+      "loss": 2.8193,
+      "step": 20317
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0391276864728192e-06,
+      "loss": 3.4972,
+      "step": 20318
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0390802781289505e-06,
+      "loss": 3.4495,
+      "step": 20319
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0390328697850822e-06,
+      "loss": 2.8321,
+      "step": 20320
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038985461441214e-06,
+      "loss": 3.7302,
+      "step": 20321
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038938053097345e-06,
+      "loss": 2.5134,
+      "step": 20322
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0388906447534765e-06,
+      "loss": 3.2405,
+      "step": 20323
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038843236409608e-06,
+      "loss": 2.5525,
+      "step": 20324
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0387958280657395e-06,
+      "loss": 2.6231,
+      "step": 20325
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038748419721871e-06,
+      "loss": 2.3378,
+      "step": 20326
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038701011378003e-06,
+      "loss": 1.4772,
+      "step": 20327
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038653603034134e-06,
+      "loss": 2.5887,
+      "step": 20328
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0386061946902654e-06,
+      "loss": 2.0966,
+      "step": 20329
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038558786346397e-06,
+      "loss": 3.4028,
+      "step": 20330
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0385113780025284e-06,
+      "loss": 3.5921,
+      "step": 20331
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0384639696586597e-06,
+      "loss": 2.797,
+      "step": 20332
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0384165613147914e-06,
+      "loss": 3.0701,
+      "step": 20333
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038369152970923e-06,
+      "loss": 2.9313,
+      "step": 20334
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0383217446270544e-06,
+      "loss": 3.4702,
+      "step": 20335
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038274336283186e-06,
+      "loss": 2.2272,
+      "step": 20336
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0382269279393174e-06,
+      "loss": 1.5442,
+      "step": 20337
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0381795195954487e-06,
+      "loss": 3.3734,
+      "step": 20338
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0381321112515804e-06,
+      "loss": 2.9634,
+      "step": 20339
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.038084702907712e-06,
+      "loss": 2.1334,
+      "step": 20340
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0380372945638433e-06,
+      "loss": 3.1832,
+      "step": 20341
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0379898862199746e-06,
+      "loss": 2.8369,
+      "step": 20342
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0379424778761063e-06,
+      "loss": 2.9316,
+      "step": 20343
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0378950695322376e-06,
+      "loss": 1.6094,
+      "step": 20344
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.037847661188369e-06,
+      "loss": 3.0559,
+      "step": 20345
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0378002528445006e-06,
+      "loss": 2.2778,
+      "step": 20346
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0377528445006323e-06,
+      "loss": 2.5855,
+      "step": 20347
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0377054361567636e-06,
+      "loss": 2.7154,
+      "step": 20348
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0376580278128953e-06,
+      "loss": 3.3194,
+      "step": 20349
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0376106194690266e-06,
+      "loss": 2.8982,
+      "step": 20350
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.037563211125158e-06,
+      "loss": 1.2206,
+      "step": 20351
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0375158027812895e-06,
+      "loss": 3.2961,
+      "step": 20352
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0374683944374213e-06,
+      "loss": 3.2714,
+      "step": 20353
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0374209860935525e-06,
+      "loss": 3.1368,
+      "step": 20354
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0373735777496842e-06,
+      "loss": 2.4506,
+      "step": 20355
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0373261694058155e-06,
+      "loss": 5.0029,
+      "step": 20356
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.037278761061947e-06,
+      "loss": 1.9841,
+      "step": 20357
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0372313527180785e-06,
+      "loss": 2.6073,
+      "step": 20358
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0371839443742098e-06,
+      "loss": 2.8594,
+      "step": 20359
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.0371365360303415e-06,
+      "loss": 2.2334,
+      "step": 20360
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0370891276864728e-06,
+      "loss": 2.7352,
+      "step": 20361
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0370417193426045e-06,
+      "loss": 3.9199,
+      "step": 20362
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0369943109987357e-06,
+      "loss": 2.0445,
+      "step": 20363
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036946902654867e-06,
+      "loss": 2.8474,
+      "step": 20364
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0368994943109987e-06,
+      "loss": 1.6935,
+      "step": 20365
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0368520859671304e-06,
+      "loss": 2.6466,
+      "step": 20366
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0368046776232617e-06,
+      "loss": 3.7692,
+      "step": 20367
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0367572692793934e-06,
+      "loss": 2.1904,
+      "step": 20368
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0367098609355247e-06,
+      "loss": 3.1668,
+      "step": 20369
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036662452591656e-06,
+      "loss": 2.5851,
+      "step": 20370
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0366150442477877e-06,
+      "loss": 3.1921,
+      "step": 20371
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036567635903919e-06,
+      "loss": 3.457,
+      "step": 20372
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0365202275600507e-06,
+      "loss": 2.8595,
+      "step": 20373
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0364728192161824e-06,
+      "loss": 4.1244,
+      "step": 20374
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0364254108723137e-06,
+      "loss": 2.4504,
+      "step": 20375
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036378002528445e-06,
+      "loss": 2.0601,
+      "step": 20376
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0363305941845766e-06,
+      "loss": 3.1149,
+      "step": 20377
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036283185840708e-06,
+      "loss": 4.7139,
+      "step": 20378
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0362357774968396e-06,
+      "loss": 3.3024,
+      "step": 20379
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0361883691529713e-06,
+      "loss": 3.4122,
+      "step": 20380
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0361409608091026e-06,
+      "loss": 2.7382,
+      "step": 20381
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036093552465234e-06,
+      "loss": 3.159,
+      "step": 20382
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.036046144121365e-06,
+      "loss": 2.0491,
+      "step": 20383
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035998735777497e-06,
+      "loss": 2.5959,
+      "step": 20384
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035951327433628e-06,
+      "loss": 3.7202,
+      "step": 20385
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.03590391908976e-06,
+      "loss": 2.5771,
+      "step": 20386
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0358565107458916e-06,
+      "loss": 3.0724,
+      "step": 20387
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035809102402023e-06,
+      "loss": 2.7475,
+      "step": 20388
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035761694058154e-06,
+      "loss": 3.5163,
+      "step": 20389
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035714285714286e-06,
+      "loss": 2.5418,
+      "step": 20390
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035666877370417e-06,
+      "loss": 1.6623,
+      "step": 20391
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035619469026549e-06,
+      "loss": 2.3047,
+      "step": 20392
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0355720606826805e-06,
+      "loss": 1.9657,
+      "step": 20393
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035524652338812e-06,
+      "loss": 2.6965,
+      "step": 20394
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035477243994943e-06,
+      "loss": 2.2128,
+      "step": 20395
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0354298356510748e-06,
+      "loss": 3.479,
+      "step": 20396
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035382427307206e-06,
+      "loss": 3.39,
+      "step": 20397
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0353350189633373e-06,
+      "loss": 2.6757,
+      "step": 20398
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035287610619469e-06,
+      "loss": 3.1145,
+      "step": 20399
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0352402022756007e-06,
+      "loss": 2.9094,
+      "step": 20400
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035192793931732e-06,
+      "loss": 1.7629,
+      "step": 20401
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0351453855878637e-06,
+      "loss": 2.2136,
+      "step": 20402
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035097977243995e-06,
+      "loss": 2.8526,
+      "step": 20403
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0350505689001263e-06,
+      "loss": 3.3827,
+      "step": 20404
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.035003160556258e-06,
+      "loss": 2.8592,
+      "step": 20405
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0349557522123897e-06,
+      "loss": 3.5109,
+      "step": 20406
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034908343868521e-06,
+      "loss": 2.5257,
+      "step": 20407
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0348609355246523e-06,
+      "loss": 2.388,
+      "step": 20408
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034813527180784e-06,
+      "loss": 2.1763,
+      "step": 20409
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0347661188369152e-06,
+      "loss": 3.5189,
+      "step": 20410
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0347187104930465e-06,
+      "loss": 1.1848,
+      "step": 20411
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0346713021491782e-06,
+      "loss": 1.2864,
+      "step": 20412
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.03462389380531e-06,
+      "loss": 2.8869,
+      "step": 20413
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0345764854614412e-06,
+      "loss": 1.8105,
+      "step": 20414
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034529077117573e-06,
+      "loss": 3.1402,
+      "step": 20415
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034481668773704e-06,
+      "loss": 2.5997,
+      "step": 20416
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0344342604298355e-06,
+      "loss": 3.5437,
+      "step": 20417
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034386852085967e-06,
+      "loss": 1.8803,
+      "step": 20418
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034339443742099e-06,
+      "loss": 3.5453,
+      "step": 20419
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.03429203539823e-06,
+      "loss": 2.1948,
+      "step": 20420
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034244627054362e-06,
+      "loss": 1.6105,
+      "step": 20421
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.034197218710493e-06,
+      "loss": 3.6261,
+      "step": 20422
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.0341498103666244e-06,
+      "loss": 2.5246,
+      "step": 20423
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.034102402022756e-06,
+      "loss": 2.5107,
+      "step": 20424
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0340549936788874e-06,
+      "loss": 1.7517,
+      "step": 20425
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.034007585335019e-06,
+      "loss": 2.4117,
+      "step": 20426
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0339601769911504e-06,
+      "loss": 1.4668,
+      "step": 20427
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.033912768647282e-06,
+      "loss": 3.2018,
+      "step": 20428
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0338653603034134e-06,
+      "loss": 2.9483,
+      "step": 20429
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0338179519595447e-06,
+      "loss": 4.6279,
+      "step": 20430
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0337705436156764e-06,
+      "loss": 3.4159,
+      "step": 20431
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.033723135271808e-06,
+      "loss": 2.9327,
+      "step": 20432
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0336757269279394e-06,
+      "loss": 2.8446,
+      "step": 20433
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.033628318584071e-06,
+      "loss": 2.8966,
+      "step": 20434
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0335809102402023e-06,
+      "loss": 3.2094,
+      "step": 20435
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0335335018963336e-06,
+      "loss": 3.7155,
+      "step": 20436
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0334860935524653e-06,
+      "loss": 3.7264,
+      "step": 20437
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0334386852085966e-06,
+      "loss": 3.0126,
+      "step": 20438
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0333912768647283e-06,
+      "loss": 2.6721,
+      "step": 20439
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.03334386852086e-06,
+      "loss": 2.0728,
+      "step": 20440
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0332964601769913e-06,
+      "loss": 2.6457,
+      "step": 20441
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0332490518331226e-06,
+      "loss": 2.2738,
+      "step": 20442
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0332016434892543e-06,
+      "loss": 3.1239,
+      "step": 20443
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0331542351453856e-06,
+      "loss": 2.8623,
+      "step": 20444
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.033106826801517e-06,
+      "loss": 2.4729,
+      "step": 20445
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.033059418457649e-06,
+      "loss": 1.0112,
+      "step": 20446
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0330120101137802e-06,
+      "loss": 2.0386,
+      "step": 20447
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0329646017699115e-06,
+      "loss": 3.2854,
+      "step": 20448
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.032917193426043e-06,
+      "loss": 3.2718,
+      "step": 20449
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0328697850821745e-06,
+      "loss": 2.4346,
+      "step": 20450
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.032822376738306e-06,
+      "loss": 2.5195,
+      "step": 20451
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0327749683944375e-06,
+      "loss": 3.3201,
+      "step": 20452
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.032727560050569e-06,
+      "loss": 2.1639,
+      "step": 20453
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0326801517067005e-06,
+      "loss": 2.5435,
+      "step": 20454
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0326327433628318e-06,
+      "loss": 3.2121,
+      "step": 20455
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0325853350189635e-06,
+      "loss": 2.5779,
+      "step": 20456
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0325379266750947e-06,
+      "loss": 1.9063,
+      "step": 20457
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.032490518331226e-06,
+      "loss": 4.0821,
+      "step": 20458
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0324431099873577e-06,
+      "loss": 2.5445,
+      "step": 20459
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0323957016434894e-06,
+      "loss": 3.7264,
+      "step": 20460
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0323482932996207e-06,
+      "loss": 4.0134,
+      "step": 20461
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0323008849557524e-06,
+      "loss": 2.7155,
+      "step": 20462
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0322534766118837e-06,
+      "loss": 1.704,
+      "step": 20463
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.032206068268015e-06,
+      "loss": 3.6453,
+      "step": 20464
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0321586599241467e-06,
+      "loss": 3.5404,
+      "step": 20465
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0321112515802784e-06,
+      "loss": 3.5288,
+      "step": 20466
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0320638432364097e-06,
+      "loss": 2.661,
+      "step": 20467
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0320164348925414e-06,
+      "loss": 2.1148,
+      "step": 20468
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0319690265486726e-06,
+      "loss": 3.9761,
+      "step": 20469
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031921618204804e-06,
+      "loss": 3.4842,
+      "step": 20470
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031874209860935e-06,
+      "loss": 2.5244,
+      "step": 20471
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031826801517067e-06,
+      "loss": 2.4573,
+      "step": 20472
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0317793931731986e-06,
+      "loss": 2.9863,
+      "step": 20473
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.03173198482933e-06,
+      "loss": 1.9492,
+      "step": 20474
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0316845764854616e-06,
+      "loss": 3.0169,
+      "step": 20475
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031637168141593e-06,
+      "loss": 3.3888,
+      "step": 20476
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031589759797724e-06,
+      "loss": 3.1209,
+      "step": 20477
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031542351453856e-06,
+      "loss": 3.339,
+      "step": 20478
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0314949431099876e-06,
+      "loss": 3.3494,
+      "step": 20479
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031447534766119e-06,
+      "loss": 1.8672,
+      "step": 20480
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0314001264222506e-06,
+      "loss": 2.2076,
+      "step": 20481
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031352718078382e-06,
+      "loss": 3.0441,
+      "step": 20482
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031305309734513e-06,
+      "loss": 2.563,
+      "step": 20483
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031257901390645e-06,
+      "loss": 1.9674,
+      "step": 20484
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031210493046776e-06,
+      "loss": 2.6413,
+      "step": 20485
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.031163084702908e-06,
+      "loss": 1.9104,
+      "step": 20486
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.0311156763590395e-06,
+      "loss": 2.6643,
+      "step": 20487
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.031068268015171e-06,
+      "loss": 3.0304,
+      "step": 20488
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.031020859671302e-06,
+      "loss": 1.5736,
+      "step": 20489
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0309734513274338e-06,
+      "loss": 3.3438,
+      "step": 20490
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030926042983565e-06,
+      "loss": 2.6122,
+      "step": 20491
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0308786346396968e-06,
+      "loss": 2.2876,
+      "step": 20492
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030831226295828e-06,
+      "loss": 3.0866,
+      "step": 20493
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0307838179519597e-06,
+      "loss": 3.0539,
+      "step": 20494
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030736409608091e-06,
+      "loss": 1.7531,
+      "step": 20495
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0306890012642223e-06,
+      "loss": 3.042,
+      "step": 20496
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030641592920354e-06,
+      "loss": 1.8109,
+      "step": 20497
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0305941845764853e-06,
+      "loss": 3.7743,
+      "step": 20498
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030546776232617e-06,
+      "loss": 2.9308,
+      "step": 20499
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0304993678887487e-06,
+      "loss": 1.6077,
+      "step": 20500
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.03045195954488e-06,
+      "loss": 3.6016,
+      "step": 20501
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0304045512010113e-06,
+      "loss": 3.6809,
+      "step": 20502
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030357142857143e-06,
+      "loss": 3.3093,
+      "step": 20503
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0303097345132742e-06,
+      "loss": 2.8579,
+      "step": 20504
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030262326169406e-06,
+      "loss": 2.5467,
+      "step": 20505
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0302149178255376e-06,
+      "loss": 2.5157,
+      "step": 20506
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030167509481669e-06,
+      "loss": 3.6952,
+      "step": 20507
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0301201011378e-06,
+      "loss": 2.1999,
+      "step": 20508
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030072692793932e-06,
+      "loss": 2.3387,
+      "step": 20509
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.030025284450063e-06,
+      "loss": 2.7634,
+      "step": 20510
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0299778761061945e-06,
+      "loss": 2.7863,
+      "step": 20511
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029930467762326e-06,
+      "loss": 2.4567,
+      "step": 20512
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029883059418458e-06,
+      "loss": 2.89,
+      "step": 20513
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029835651074589e-06,
+      "loss": 2.5576,
+      "step": 20514
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0297882427307204e-06,
+      "loss": 2.8943,
+      "step": 20515
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029740834386852e-06,
+      "loss": 2.6101,
+      "step": 20516
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0296934260429834e-06,
+      "loss": 3.6875,
+      "step": 20517
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029646017699115e-06,
+      "loss": 2.4754,
+      "step": 20518
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029598609355247e-06,
+      "loss": 2.2299,
+      "step": 20519
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029551201011378e-06,
+      "loss": 2.9947,
+      "step": 20520
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0295037926675094e-06,
+      "loss": 2.2446,
+      "step": 20521
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029456384323641e-06,
+      "loss": 2.4485,
+      "step": 20522
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0294089759797724e-06,
+      "loss": 2.1728,
+      "step": 20523
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0293615676359037e-06,
+      "loss": 2.7809,
+      "step": 20524
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0293141592920354e-06,
+      "loss": 1.3744,
+      "step": 20525
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.029266750948167e-06,
+      "loss": 2.549,
+      "step": 20526
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0292193426042983e-06,
+      "loss": 2.4711,
+      "step": 20527
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.02917193426043e-06,
+      "loss": 3.1062,
+      "step": 20528
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0291245259165613e-06,
+      "loss": 1.5654,
+      "step": 20529
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0290771175726926e-06,
+      "loss": 2.9268,
+      "step": 20530
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0290297092288243e-06,
+      "loss": 1.8084,
+      "step": 20531
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.028982300884956e-06,
+      "loss": 3.1591,
+      "step": 20532
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0289348925410873e-06,
+      "loss": 1.607,
+      "step": 20533
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.028887484197219e-06,
+      "loss": 3.9899,
+      "step": 20534
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0288400758533503e-06,
+      "loss": 1.3749,
+      "step": 20535
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0287926675094816e-06,
+      "loss": 2.5269,
+      "step": 20536
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.028745259165613e-06,
+      "loss": 2.3729,
+      "step": 20537
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0286978508217446e-06,
+      "loss": 3.9406,
+      "step": 20538
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0286504424778763e-06,
+      "loss": 1.9687,
+      "step": 20539
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0286030341340075e-06,
+      "loss": 3.0112,
+      "step": 20540
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0285556257901392e-06,
+      "loss": 3.1516,
+      "step": 20541
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0285082174462705e-06,
+      "loss": 1.8859,
+      "step": 20542
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.028460809102402e-06,
+      "loss": 3.6568,
+      "step": 20543
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0284134007585335e-06,
+      "loss": 3.1499,
+      "step": 20544
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.028365992414665e-06,
+      "loss": 2.9332,
+      "step": 20545
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0283185840707965e-06,
+      "loss": 2.5441,
+      "step": 20546
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.028271175726928e-06,
+      "loss": 3.4024,
+      "step": 20547
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0282237673830595e-06,
+      "loss": 2.0084,
+      "step": 20548
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0281763590391908e-06,
+      "loss": 2.8964,
+      "step": 20549
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.0281289506953225e-06,
+      "loss": 1.8327,
+      "step": 20550
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0280815423514537e-06,
+      "loss": 2.6195,
+      "step": 20551
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0280341340075854e-06,
+      "loss": 3.5635,
+      "step": 20552
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.027986725663717e-06,
+      "loss": 2.1985,
+      "step": 20553
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0279393173198484e-06,
+      "loss": 2.982,
+      "step": 20554
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0278919089759797e-06,
+      "loss": 1.7327,
+      "step": 20555
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0278445006321114e-06,
+      "loss": 3.9151,
+      "step": 20556
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0277970922882427e-06,
+      "loss": 2.3981,
+      "step": 20557
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0277496839443744e-06,
+      "loss": 2.7804,
+      "step": 20558
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0277022756005057e-06,
+      "loss": 3.8517,
+      "step": 20559
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0276548672566374e-06,
+      "loss": 3.3576,
+      "step": 20560
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0276074589127687e-06,
+      "loss": 3.729,
+      "step": 20561
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0275600505689e-06,
+      "loss": 2.1321,
+      "step": 20562
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0275126422250316e-06,
+      "loss": 2.5236,
+      "step": 20563
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.027465233881163e-06,
+      "loss": 2.8052,
+      "step": 20564
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0274178255372946e-06,
+      "loss": 3.2588,
+      "step": 20565
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0273704171934263e-06,
+      "loss": 2.7813,
+      "step": 20566
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0273230088495576e-06,
+      "loss": 2.4751,
+      "step": 20567
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.027275600505689e-06,
+      "loss": 3.3101,
+      "step": 20568
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0272281921618206e-06,
+      "loss": 2.3113,
+      "step": 20569
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.027180783817952e-06,
+      "loss": 1.7609,
+      "step": 20570
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0271333754740836e-06,
+      "loss": 2.6408,
+      "step": 20571
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0270859671302153e-06,
+      "loss": 3.5087,
+      "step": 20572
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0270385587863466e-06,
+      "loss": 3.6054,
+      "step": 20573
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026991150442478e-06,
+      "loss": 3.2107,
+      "step": 20574
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0269437420986095e-06,
+      "loss": 2.578,
+      "step": 20575
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026896333754741e-06,
+      "loss": 2.265,
+      "step": 20576
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026848925410872e-06,
+      "loss": 1.8059,
+      "step": 20577
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026801517067004e-06,
+      "loss": 3.4858,
+      "step": 20578
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0267541087231355e-06,
+      "loss": 2.3674,
+      "step": 20579
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026706700379267e-06,
+      "loss": 3.1875,
+      "step": 20580
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026659292035398e-06,
+      "loss": 2.9299,
+      "step": 20581
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0266118836915298e-06,
+      "loss": 1.9456,
+      "step": 20582
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026564475347661e-06,
+      "loss": 3.058,
+      "step": 20583
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0265170670037928e-06,
+      "loss": 1.9403,
+      "step": 20584
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0264696586599245e-06,
+      "loss": 2.4169,
+      "step": 20585
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0264222503160558e-06,
+      "loss": 2.2705,
+      "step": 20586
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026374841972187e-06,
+      "loss": 3.3481,
+      "step": 20587
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0263274336283187e-06,
+      "loss": 2.2642,
+      "step": 20588
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.02628002528445e-06,
+      "loss": 2.4725,
+      "step": 20589
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0262326169405813e-06,
+      "loss": 2.4329,
+      "step": 20590
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026185208596713e-06,
+      "loss": 3.0295,
+      "step": 20591
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0261378002528447e-06,
+      "loss": 3.7227,
+      "step": 20592
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.026090391908976e-06,
+      "loss": 2.164,
+      "step": 20593
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0260429835651077e-06,
+      "loss": 2.6272,
+      "step": 20594
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025995575221239e-06,
+      "loss": 2.5372,
+      "step": 20595
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0259481668773703e-06,
+      "loss": 2.9575,
+      "step": 20596
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025900758533502e-06,
+      "loss": 1.8481,
+      "step": 20597
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0258533501896337e-06,
+      "loss": 3.3212,
+      "step": 20598
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025805941845765e-06,
+      "loss": 2.9551,
+      "step": 20599
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0257585335018966e-06,
+      "loss": 2.8777,
+      "step": 20600
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025711125158028e-06,
+      "loss": 1.9167,
+      "step": 20601
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025663716814159e-06,
+      "loss": 2.8496,
+      "step": 20602
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0256163084702905e-06,
+      "loss": 2.405,
+      "step": 20603
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025568900126422e-06,
+      "loss": 2.7656,
+      "step": 20604
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025521491782554e-06,
+      "loss": 1.8316,
+      "step": 20605
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025474083438685e-06,
+      "loss": 2.1327,
+      "step": 20606
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025426675094817e-06,
+      "loss": 3.0755,
+      "step": 20607
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025379266750948e-06,
+      "loss": 3.0015,
+      "step": 20608
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.0253318584070794e-06,
+      "loss": 1.694,
+      "step": 20609
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025284450063211e-06,
+      "loss": 2.5473,
+      "step": 20610
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025237041719343e-06,
+      "loss": 1.9757,
+      "step": 20611
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025189633375474e-06,
+      "loss": 2.1026,
+      "step": 20612
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.025142225031606e-06,
+      "loss": 2.74,
+      "step": 20613
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.025094816687737e-06,
+      "loss": 3.0339,
+      "step": 20614
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0250474083438684e-06,
+      "loss": 2.2054,
+      "step": 20615
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.025e-06,
+      "loss": 3.2441,
+      "step": 20616
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0249525916561314e-06,
+      "loss": 1.2183,
+      "step": 20617
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.024905183312263e-06,
+      "loss": 1.8915,
+      "step": 20618
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0248577749683948e-06,
+      "loss": 2.892,
+      "step": 20619
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.024810366624526e-06,
+      "loss": 3.154,
+      "step": 20620
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0247629582806573e-06,
+      "loss": 2.2655,
+      "step": 20621
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.024715549936789e-06,
+      "loss": 3.3026,
+      "step": 20622
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0246681415929203e-06,
+      "loss": 3.2949,
+      "step": 20623
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.024620733249052e-06,
+      "loss": 2.1404,
+      "step": 20624
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0245733249051833e-06,
+      "loss": 2.1722,
+      "step": 20625
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.024525916561315e-06,
+      "loss": 2.7085,
+      "step": 20626
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0244785082174463e-06,
+      "loss": 2.4593,
+      "step": 20627
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0244310998735776e-06,
+      "loss": 2.6149,
+      "step": 20628
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0243836915297093e-06,
+      "loss": 2.4628,
+      "step": 20629
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0243362831858406e-06,
+      "loss": 2.8336,
+      "step": 20630
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0242888748419723e-06,
+      "loss": 2.1307,
+      "step": 20631
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.024241466498104e-06,
+      "loss": 3.8589,
+      "step": 20632
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0241940581542352e-06,
+      "loss": 1.9004,
+      "step": 20633
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0241466498103665e-06,
+      "loss": 2.1263,
+      "step": 20634
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0240992414664982e-06,
+      "loss": 2.7776,
+      "step": 20635
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0240518331226295e-06,
+      "loss": 2.5764,
+      "step": 20636
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0240044247787612e-06,
+      "loss": 2.9661,
+      "step": 20637
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.023957016434893e-06,
+      "loss": 1.7453,
+      "step": 20638
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.023909608091024e-06,
+      "loss": 2.6324,
+      "step": 20639
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0238621997471555e-06,
+      "loss": 3.5675,
+      "step": 20640
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.023814791403287e-06,
+      "loss": 5.0641,
+      "step": 20641
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0237673830594185e-06,
+      "loss": 2.9237,
+      "step": 20642
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0237199747155497e-06,
+      "loss": 3.2092,
+      "step": 20643
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0236725663716815e-06,
+      "loss": 2.6058,
+      "step": 20644
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.023625158027813e-06,
+      "loss": 2.5026,
+      "step": 20645
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0235777496839444e-06,
+      "loss": 2.2884,
+      "step": 20646
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0235303413400757e-06,
+      "loss": 2.4759,
+      "step": 20647
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0234829329962074e-06,
+      "loss": 3.2116,
+      "step": 20648
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0234355246523387e-06,
+      "loss": 1.3289,
+      "step": 20649
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0233881163084704e-06,
+      "loss": 3.4263,
+      "step": 20650
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.023340707964602e-06,
+      "loss": 3.3444,
+      "step": 20651
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0232932996207334e-06,
+      "loss": 2.9338,
+      "step": 20652
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0232458912768647e-06,
+      "loss": 2.784,
+      "step": 20653
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0231984829329964e-06,
+      "loss": 2.7961,
+      "step": 20654
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0231510745891277e-06,
+      "loss": 1.8647,
+      "step": 20655
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.023103666245259e-06,
+      "loss": 2.8056,
+      "step": 20656
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0230562579013906e-06,
+      "loss": 2.7035,
+      "step": 20657
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0230088495575223e-06,
+      "loss": 2.8207,
+      "step": 20658
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0229614412136536e-06,
+      "loss": 2.2952,
+      "step": 20659
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0229140328697853e-06,
+      "loss": 3.7356,
+      "step": 20660
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0228666245259166e-06,
+      "loss": 3.1464,
+      "step": 20661
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.022819216182048e-06,
+      "loss": 1.8535,
+      "step": 20662
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0227718078381796e-06,
+      "loss": 2.212,
+      "step": 20663
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0227243994943113e-06,
+      "loss": 2.0879,
+      "step": 20664
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0226769911504426e-06,
+      "loss": 3.5706,
+      "step": 20665
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0226295828065743e-06,
+      "loss": 2.1691,
+      "step": 20666
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0225821744627056e-06,
+      "loss": 2.197,
+      "step": 20667
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.022534766118837e-06,
+      "loss": 2.7624,
+      "step": 20668
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.022487357774968e-06,
+      "loss": 2.9913,
+      "step": 20669
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0224399494311e-06,
+      "loss": 2.7485,
+      "step": 20670
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0223925410872315e-06,
+      "loss": 2.1358,
+      "step": 20671
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.022345132743363e-06,
+      "loss": 2.3424,
+      "step": 20672
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0222977243994945e-06,
+      "loss": 1.8031,
+      "step": 20673
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.022250316055626e-06,
+      "loss": 1.3537,
+      "step": 20674
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.022202907711757e-06,
+      "loss": 2.8845,
+      "step": 20675
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0221554993678888e-06,
+      "loss": 3.3236,
+      "step": 20676
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.0221080910240205e-06,
+      "loss": 2.6526,
+      "step": 20677
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0220606826801518e-06,
+      "loss": 2.7194,
+      "step": 20678
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0220132743362835e-06,
+      "loss": 2.7456,
+      "step": 20679
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0219658659924147e-06,
+      "loss": 1.947,
+      "step": 20680
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021918457648546e-06,
+      "loss": 1.4732,
+      "step": 20681
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0218710493046777e-06,
+      "loss": 2.4086,
+      "step": 20682
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021823640960809e-06,
+      "loss": 3.1378,
+      "step": 20683
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0217762326169407e-06,
+      "loss": 2.5689,
+      "step": 20684
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0217288242730724e-06,
+      "loss": 1.1385,
+      "step": 20685
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0216814159292037e-06,
+      "loss": 2.815,
+      "step": 20686
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021634007585335e-06,
+      "loss": 2.7354,
+      "step": 20687
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0215865992414667e-06,
+      "loss": 2.6288,
+      "step": 20688
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021539190897598e-06,
+      "loss": 3.0449,
+      "step": 20689
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0214917825537297e-06,
+      "loss": 2.3291,
+      "step": 20690
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021444374209861e-06,
+      "loss": 3.1383,
+      "step": 20691
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0213969658659927e-06,
+      "loss": 1.9934,
+      "step": 20692
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021349557522124e-06,
+      "loss": 1.6535,
+      "step": 20693
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0213021491782552e-06,
+      "loss": 2.6481,
+      "step": 20694
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021254740834387e-06,
+      "loss": 2.3475,
+      "step": 20695
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021207332490518e-06,
+      "loss": 1.4531,
+      "step": 20696
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.02115992414665e-06,
+      "loss": 2.1529,
+      "step": 20697
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0211125158027816e-06,
+      "loss": 1.209,
+      "step": 20698
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021065107458913e-06,
+      "loss": 2.869,
+      "step": 20699
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.021017699115044e-06,
+      "loss": 2.2857,
+      "step": 20700
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020970290771176e-06,
+      "loss": 2.6859,
+      "step": 20701
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020922882427307e-06,
+      "loss": 3.3492,
+      "step": 20702
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020875474083439e-06,
+      "loss": 3.3062,
+      "step": 20703
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0208280657395706e-06,
+      "loss": 2.2638,
+      "step": 20704
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020780657395702e-06,
+      "loss": 3.8089,
+      "step": 20705
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020733249051833e-06,
+      "loss": 3.396,
+      "step": 20706
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020685840707965e-06,
+      "loss": 3.0391,
+      "step": 20707
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020638432364096e-06,
+      "loss": 1.5116,
+      "step": 20708
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0205910240202274e-06,
+      "loss": 2.6645,
+      "step": 20709
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020543615676359e-06,
+      "loss": 2.7223,
+      "step": 20710
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020496207332491e-06,
+      "loss": 1.4241,
+      "step": 20711
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020448798988622e-06,
+      "loss": 2.0169,
+      "step": 20712
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0204013906447534e-06,
+      "loss": 2.6571,
+      "step": 20713
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020353982300885e-06,
+      "loss": 2.0437,
+      "step": 20714
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0203065739570163e-06,
+      "loss": 4.0422,
+      "step": 20715
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020259165613148e-06,
+      "loss": 3.4466,
+      "step": 20716
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0202117572692797e-06,
+      "loss": 3.4762,
+      "step": 20717
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020164348925411e-06,
+      "loss": 2.3515,
+      "step": 20718
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0201169405815423e-06,
+      "loss": 3.9573,
+      "step": 20719
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.020069532237674e-06,
+      "loss": 3.3609,
+      "step": 20720
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0200221238938053e-06,
+      "loss": 2.1983,
+      "step": 20721
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0199747155499366e-06,
+      "loss": 2.9037,
+      "step": 20722
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0199273072060683e-06,
+      "loss": 2.2011,
+      "step": 20723
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0198798988622e-06,
+      "loss": 2.1432,
+      "step": 20724
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0198324905183313e-06,
+      "loss": 2.2167,
+      "step": 20725
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.019785082174463e-06,
+      "loss": 2.2506,
+      "step": 20726
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0197376738305942e-06,
+      "loss": 3.1401,
+      "step": 20727
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0196902654867255e-06,
+      "loss": 3.4952,
+      "step": 20728
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0196428571428572e-06,
+      "loss": 2.6464,
+      "step": 20729
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.019595448798989e-06,
+      "loss": 2.3272,
+      "step": 20730
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.01954804045512e-06,
+      "loss": 2.5855,
+      "step": 20731
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0195006321112515e-06,
+      "loss": 1.4305,
+      "step": 20732
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.019453223767383e-06,
+      "loss": 3.9764,
+      "step": 20733
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0194058154235145e-06,
+      "loss": 2.5659,
+      "step": 20734
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0193584070796458e-06,
+      "loss": 2.6501,
+      "step": 20735
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0193109987357775e-06,
+      "loss": 2.8314,
+      "step": 20736
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.019263590391909e-06,
+      "loss": 2.5581,
+      "step": 20737
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0192161820480404e-06,
+      "loss": 2.9415,
+      "step": 20738
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.019168773704172e-06,
+      "loss": 0.9392,
+      "step": 20739
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.0191213653603034e-06,
+      "loss": 2.9081,
+      "step": 20740
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0190739570164347e-06,
+      "loss": 3.1254,
+      "step": 20741
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0190265486725664e-06,
+      "loss": 2.1809,
+      "step": 20742
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0189791403286977e-06,
+      "loss": 2.5714,
+      "step": 20743
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0189317319848294e-06,
+      "loss": 2.7172,
+      "step": 20744
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.018884323640961e-06,
+      "loss": 2.523,
+      "step": 20745
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0188369152970924e-06,
+      "loss": 2.9597,
+      "step": 20746
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0187895069532237e-06,
+      "loss": 1.5557,
+      "step": 20747
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0187420986093554e-06,
+      "loss": 2.9093,
+      "step": 20748
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0186946902654866e-06,
+      "loss": 2.4717,
+      "step": 20749
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0186472819216184e-06,
+      "loss": 3.3005,
+      "step": 20750
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.01859987357775e-06,
+      "loss": 3.3987,
+      "step": 20751
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0185524652338813e-06,
+      "loss": 3.6436,
+      "step": 20752
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0185050568900126e-06,
+      "loss": 3.5238,
+      "step": 20753
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.018457648546144e-06,
+      "loss": 2.5394,
+      "step": 20754
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0184102402022756e-06,
+      "loss": 2.8145,
+      "step": 20755
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.018362831858407e-06,
+      "loss": 3.9926,
+      "step": 20756
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0183154235145386e-06,
+      "loss": 2.8532,
+      "step": 20757
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0182680151706703e-06,
+      "loss": 3.1413,
+      "step": 20758
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0182206068268016e-06,
+      "loss": 2.9342,
+      "step": 20759
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.018173198482933e-06,
+      "loss": 3.7601,
+      "step": 20760
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0181257901390646e-06,
+      "loss": 1.8555,
+      "step": 20761
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.018078381795196e-06,
+      "loss": 2.7772,
+      "step": 20762
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0180309734513275e-06,
+      "loss": 3.2249,
+      "step": 20763
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0179835651074592e-06,
+      "loss": 3.7707,
+      "step": 20764
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0179361567635905e-06,
+      "loss": 2.5705,
+      "step": 20765
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.017888748419722e-06,
+      "loss": 3.0744,
+      "step": 20766
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0178413400758535e-06,
+      "loss": 3.0065,
+      "step": 20767
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0177939317319848e-06,
+      "loss": 2.5791,
+      "step": 20768
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.017746523388116e-06,
+      "loss": 3.0288,
+      "step": 20769
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0176991150442478e-06,
+      "loss": 2.3729,
+      "step": 20770
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0176517067003795e-06,
+      "loss": 2.6872,
+      "step": 20771
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0176042983565108e-06,
+      "loss": 2.6948,
+      "step": 20772
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0175568900126425e-06,
+      "loss": 2.1996,
+      "step": 20773
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0175094816687737e-06,
+      "loss": 2.0766,
+      "step": 20774
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.017462073324905e-06,
+      "loss": 2.7929,
+      "step": 20775
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0174146649810367e-06,
+      "loss": 1.953,
+      "step": 20776
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0173672566371684e-06,
+      "loss": 1.8223,
+      "step": 20777
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0173198482932997e-06,
+      "loss": 1.5064,
+      "step": 20778
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.017272439949431e-06,
+      "loss": 2.2791,
+      "step": 20779
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0172250316055627e-06,
+      "loss": 3.9011,
+      "step": 20780
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.017177623261694e-06,
+      "loss": 2.4995,
+      "step": 20781
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0171302149178253e-06,
+      "loss": 2.9705,
+      "step": 20782
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.017082806573957e-06,
+      "loss": 3.1985,
+      "step": 20783
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0170353982300887e-06,
+      "loss": 4.629,
+      "step": 20784
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.01698798988622e-06,
+      "loss": 1.31,
+      "step": 20785
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0169405815423516e-06,
+      "loss": 2.9317,
+      "step": 20786
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016893173198483e-06,
+      "loss": 3.1859,
+      "step": 20787
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016845764854614e-06,
+      "loss": 1.3727,
+      "step": 20788
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016798356510746e-06,
+      "loss": 2.0556,
+      "step": 20789
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0167509481668776e-06,
+      "loss": 1.9072,
+      "step": 20790
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016703539823009e-06,
+      "loss": 2.8983,
+      "step": 20791
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0166561314791406e-06,
+      "loss": 1.7965,
+      "step": 20792
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016608723135272e-06,
+      "loss": 2.5979,
+      "step": 20793
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016561314791403e-06,
+      "loss": 2.086,
+      "step": 20794
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016513906447535e-06,
+      "loss": 3.8503,
+      "step": 20795
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016466498103666e-06,
+      "loss": 4.485,
+      "step": 20796
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016419089759798e-06,
+      "loss": 3.1246,
+      "step": 20797
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016371681415929e-06,
+      "loss": 2.3611,
+      "step": 20798
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016324273072061e-06,
+      "loss": 1.8678,
+      "step": 20799
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016276864728192e-06,
+      "loss": 2.917,
+      "step": 20800
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.0162294563843234e-06,
+      "loss": 2.4364,
+      "step": 20801
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016182048040455e-06,
+      "loss": 3.0564,
+      "step": 20802
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.016134639696587e-06,
+      "loss": 2.6805,
+      "step": 20803
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.016087231352718e-06,
+      "loss": 3.2788,
+      "step": 20804
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0160398230088498e-06,
+      "loss": 2.9942,
+      "step": 20805
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015992414664981e-06,
+      "loss": 2.0285,
+      "step": 20806
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0159450063211123e-06,
+      "loss": 2.7368,
+      "step": 20807
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015897597977244e-06,
+      "loss": 2.4612,
+      "step": 20808
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0158501896333753e-06,
+      "loss": 2.6953,
+      "step": 20809
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015802781289507e-06,
+      "loss": 2.5881,
+      "step": 20810
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0157553729456387e-06,
+      "loss": 3.4104,
+      "step": 20811
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.01570796460177e-06,
+      "loss": 2.451,
+      "step": 20812
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0156605562579013e-06,
+      "loss": 2.8325,
+      "step": 20813
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015613147914033e-06,
+      "loss": 2.9412,
+      "step": 20814
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0155657395701643e-06,
+      "loss": 1.3445,
+      "step": 20815
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015518331226296e-06,
+      "loss": 3.4672,
+      "step": 20816
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0154709228824277e-06,
+      "loss": 2.6144,
+      "step": 20817
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015423514538559e-06,
+      "loss": 2.8651,
+      "step": 20818
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0153761061946903e-06,
+      "loss": 3.5847,
+      "step": 20819
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0153286978508215e-06,
+      "loss": 3.025,
+      "step": 20820
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0152812895069532e-06,
+      "loss": 2.413,
+      "step": 20821
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0152338811630845e-06,
+      "loss": 3.4556,
+      "step": 20822
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0151864728192162e-06,
+      "loss": 2.4675,
+      "step": 20823
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015139064475348e-06,
+      "loss": 2.4711,
+      "step": 20824
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.015091656131479e-06,
+      "loss": 2.7881,
+      "step": 20825
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0150442477876105e-06,
+      "loss": 2.6128,
+      "step": 20826
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014996839443742e-06,
+      "loss": 2.7258,
+      "step": 20827
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0149494310998735e-06,
+      "loss": 2.0873,
+      "step": 20828
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014902022756005e-06,
+      "loss": 2.7674,
+      "step": 20829
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014854614412137e-06,
+      "loss": 2.5141,
+      "step": 20830
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014807206068268e-06,
+      "loss": 3.0149,
+      "step": 20831
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0147597977243994e-06,
+      "loss": 2.645,
+      "step": 20832
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014712389380531e-06,
+      "loss": 2.6426,
+      "step": 20833
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0146649810366624e-06,
+      "loss": 3.2922,
+      "step": 20834
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0146175726927937e-06,
+      "loss": 3.3021,
+      "step": 20835
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0145701643489254e-06,
+      "loss": 3.6726,
+      "step": 20836
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014522756005057e-06,
+      "loss": 2.5545,
+      "step": 20837
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0144753476611884e-06,
+      "loss": 3.8315,
+      "step": 20838
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.01442793931732e-06,
+      "loss": 2.5353,
+      "step": 20839
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0143805309734514e-06,
+      "loss": 1.4818,
+      "step": 20840
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0143331226295827e-06,
+      "loss": 2.5341,
+      "step": 20841
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0142857142857144e-06,
+      "loss": 2.6543,
+      "step": 20842
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014238305941846e-06,
+      "loss": 2.4542,
+      "step": 20843
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0141908975979773e-06,
+      "loss": 2.0885,
+      "step": 20844
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0141434892541086e-06,
+      "loss": 4.0697,
+      "step": 20845
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0140960809102403e-06,
+      "loss": 3.1219,
+      "step": 20846
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0140486725663716e-06,
+      "loss": 2.6642,
+      "step": 20847
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.014001264222503e-06,
+      "loss": 3.3688,
+      "step": 20848
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0139538558786346e-06,
+      "loss": 2.8336,
+      "step": 20849
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0139064475347663e-06,
+      "loss": 2.1183,
+      "step": 20850
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0138590391908976e-06,
+      "loss": 1.8495,
+      "step": 20851
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0138116308470293e-06,
+      "loss": 3.3468,
+      "step": 20852
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0137642225031606e-06,
+      "loss": 2.2436,
+      "step": 20853
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.013716814159292e-06,
+      "loss": 3.1528,
+      "step": 20854
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0136694058154235e-06,
+      "loss": 3.3403,
+      "step": 20855
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0136219974715553e-06,
+      "loss": 2.8695,
+      "step": 20856
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0135745891276865e-06,
+      "loss": 1.7127,
+      "step": 20857
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0135271807838182e-06,
+      "loss": 2.8388,
+      "step": 20858
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0134797724399495e-06,
+      "loss": 3.3382,
+      "step": 20859
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.013432364096081e-06,
+      "loss": 3.0274,
+      "step": 20860
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0133849557522125e-06,
+      "loss": 1.4594,
+      "step": 20861
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0133375474083438e-06,
+      "loss": 3.2293,
+      "step": 20862
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0132901390644755e-06,
+      "loss": 2.4158,
+      "step": 20863
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0132427307206068e-06,
+      "loss": 3.3183,
+      "step": 20864
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0131953223767385e-06,
+      "loss": 2.9736,
+      "step": 20865
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.0131479140328697e-06,
+      "loss": 2.5364,
+      "step": 20866
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.013100505689001e-06,
+      "loss": 1.9752,
+      "step": 20867
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0130530973451327e-06,
+      "loss": 2.6632,
+      "step": 20868
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0130056890012644e-06,
+      "loss": 2.8252,
+      "step": 20869
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0129582806573957e-06,
+      "loss": 2.7019,
+      "step": 20870
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0129108723135274e-06,
+      "loss": 2.8044,
+      "step": 20871
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0128634639696587e-06,
+      "loss": 2.3272,
+      "step": 20872
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.01281605562579e-06,
+      "loss": 1.435,
+      "step": 20873
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0127686472819217e-06,
+      "loss": 2.7215,
+      "step": 20874
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012721238938053e-06,
+      "loss": 2.8295,
+      "step": 20875
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0126738305941847e-06,
+      "loss": 3.2557,
+      "step": 20876
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0126264222503164e-06,
+      "loss": 2.4334,
+      "step": 20877
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0125790139064477e-06,
+      "loss": 3.6671,
+      "step": 20878
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012531605562579e-06,
+      "loss": 2.2368,
+      "step": 20879
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0124841972187106e-06,
+      "loss": 2.0232,
+      "step": 20880
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012436788874842e-06,
+      "loss": 4.3874,
+      "step": 20881
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0123893805309736e-06,
+      "loss": 2.6324,
+      "step": 20882
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0123419721871053e-06,
+      "loss": 2.6571,
+      "step": 20883
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0122945638432366e-06,
+      "loss": 1.8618,
+      "step": 20884
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012247155499368e-06,
+      "loss": 2.7761,
+      "step": 20885
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012199747155499e-06,
+      "loss": 3.1357,
+      "step": 20886
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012152338811631e-06,
+      "loss": 1.5149,
+      "step": 20887
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012104930467762e-06,
+      "loss": 3.4768,
+      "step": 20888
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.012057522123894e-06,
+      "loss": 2.9812,
+      "step": 20889
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0120101137800256e-06,
+      "loss": 4.0582,
+      "step": 20890
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011962705436157e-06,
+      "loss": 2.7618,
+      "step": 20891
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011915297092288e-06,
+      "loss": 2.2913,
+      "step": 20892
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.01186788874842e-06,
+      "loss": 2.6059,
+      "step": 20893
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011820480404551e-06,
+      "loss": 2.8432,
+      "step": 20894
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011773072060683e-06,
+      "loss": 3.6702,
+      "step": 20895
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0117256637168145e-06,
+      "loss": 1.6837,
+      "step": 20896
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011678255372946e-06,
+      "loss": 2.4688,
+      "step": 20897
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011630847029077e-06,
+      "loss": 1.8092,
+      "step": 20898
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0115834386852088e-06,
+      "loss": 2.0493,
+      "step": 20899
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.01153603034134e-06,
+      "loss": 2.3149,
+      "step": 20900
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0114886219974713e-06,
+      "loss": 2.8803,
+      "step": 20901
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011441213653603e-06,
+      "loss": 2.6507,
+      "step": 20902
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0113938053097347e-06,
+      "loss": 3.6063,
+      "step": 20903
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011346396965866e-06,
+      "loss": 2.8594,
+      "step": 20904
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0112989886219977e-06,
+      "loss": 2.6113,
+      "step": 20905
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011251580278129e-06,
+      "loss": 2.3046,
+      "step": 20906
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0112041719342603e-06,
+      "loss": 2.0145,
+      "step": 20907
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011156763590392e-06,
+      "loss": 3.6724,
+      "step": 20908
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0111093552465237e-06,
+      "loss": 2.4683,
+      "step": 20909
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.011061946902655e-06,
+      "loss": 3.1254,
+      "step": 20910
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0110145385587863e-06,
+      "loss": 3.6405,
+      "step": 20911
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010967130214918e-06,
+      "loss": 3.2608,
+      "step": 20912
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0109197218710492e-06,
+      "loss": 2.5936,
+      "step": 20913
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0108723135271805e-06,
+      "loss": 3.1336,
+      "step": 20914
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0108249051833122e-06,
+      "loss": 3.0952,
+      "step": 20915
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010777496839444e-06,
+      "loss": 2.7057,
+      "step": 20916
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0107300884955752e-06,
+      "loss": 1.8812,
+      "step": 20917
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010682680151707e-06,
+      "loss": 1.5853,
+      "step": 20918
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010635271807838e-06,
+      "loss": 3.7289,
+      "step": 20919
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0105878634639695e-06,
+      "loss": 1.5821,
+      "step": 20920
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010540455120101e-06,
+      "loss": 3.0001,
+      "step": 20921
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010493046776233e-06,
+      "loss": 2.8223,
+      "step": 20922
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010445638432364e-06,
+      "loss": 2.0778,
+      "step": 20923
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010398230088496e-06,
+      "loss": 3.1736,
+      "step": 20924
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010350821744627e-06,
+      "loss": 3.5558,
+      "step": 20925
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0103034134007584e-06,
+      "loss": 3.0904,
+      "step": 20926
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.01025600505689e-06,
+      "loss": 3.3108,
+      "step": 20927
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0102085967130214e-06,
+      "loss": 1.7715,
+      "step": 20928
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.010161188369153e-06,
+      "loss": 1.7485,
+      "step": 20929
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.0101137800252844e-06,
+      "loss": 1.2709,
+      "step": 20930
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.010066371681416e-06,
+      "loss": 1.5539,
+      "step": 20931
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0100189633375474e-06,
+      "loss": 2.5581,
+      "step": 20932
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0099715549936787e-06,
+      "loss": 2.7537,
+      "step": 20933
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0099241466498104e-06,
+      "loss": 2.9771,
+      "step": 20934
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.009876738305942e-06,
+      "loss": 3.8797,
+      "step": 20935
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0098293299620734e-06,
+      "loss": 1.9128,
+      "step": 20936
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.009781921618205e-06,
+      "loss": 2.1044,
+      "step": 20937
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0097345132743363e-06,
+      "loss": 1.63,
+      "step": 20938
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0096871049304676e-06,
+      "loss": 2.7276,
+      "step": 20939
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0096396965865993e-06,
+      "loss": 2.4163,
+      "step": 20940
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0095922882427306e-06,
+      "loss": 3.0089,
+      "step": 20941
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0095448798988623e-06,
+      "loss": 2.484,
+      "step": 20942
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.009497471554994e-06,
+      "loss": 2.1397,
+      "step": 20943
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0094500632111253e-06,
+      "loss": 2.5147,
+      "step": 20944
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0094026548672566e-06,
+      "loss": 2.6346,
+      "step": 20945
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0093552465233883e-06,
+      "loss": 1.643,
+      "step": 20946
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0093078381795196e-06,
+      "loss": 2.2727,
+      "step": 20947
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0092604298356513e-06,
+      "loss": 2.4167,
+      "step": 20948
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.009213021491783e-06,
+      "loss": 2.5692,
+      "step": 20949
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0091656131479142e-06,
+      "loss": 2.3636,
+      "step": 20950
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0091182048040455e-06,
+      "loss": 2.8556,
+      "step": 20951
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.009070796460177e-06,
+      "loss": 2.2606,
+      "step": 20952
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0090233881163085e-06,
+      "loss": 4.0337,
+      "step": 20953
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.00897597977244e-06,
+      "loss": 2.4157,
+      "step": 20954
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0089285714285715e-06,
+      "loss": 2.8314,
+      "step": 20955
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.008881163084703e-06,
+      "loss": 2.9084,
+      "step": 20956
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0088337547408345e-06,
+      "loss": 3.2949,
+      "step": 20957
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0087863463969658e-06,
+      "loss": 3.0075,
+      "step": 20958
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0087389380530975e-06,
+      "loss": 2.4017,
+      "step": 20959
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0086915297092287e-06,
+      "loss": 2.9779,
+      "step": 20960
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0086441213653604e-06,
+      "loss": 2.4431,
+      "step": 20961
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.008596713021492e-06,
+      "loss": 2.8115,
+      "step": 20962
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0085493046776234e-06,
+      "loss": 2.6686,
+      "step": 20963
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0085018963337547e-06,
+      "loss": 3.0577,
+      "step": 20964
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0084544879898864e-06,
+      "loss": 3.473,
+      "step": 20965
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0084070796460177e-06,
+      "loss": 3.2408,
+      "step": 20966
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.008359671302149e-06,
+      "loss": 1.1039,
+      "step": 20967
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0083122629582807e-06,
+      "loss": 3.344,
+      "step": 20968
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0082648546144124e-06,
+      "loss": 2.0266,
+      "step": 20969
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0082174462705437e-06,
+      "loss": 1.5793,
+      "step": 20970
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0081700379266754e-06,
+      "loss": 2.7928,
+      "step": 20971
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0081226295828066e-06,
+      "loss": 3.761,
+      "step": 20972
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.008075221238938e-06,
+      "loss": 3.9085,
+      "step": 20973
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0080278128950696e-06,
+      "loss": 2.7806,
+      "step": 20974
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0079804045512013e-06,
+      "loss": 2.9892,
+      "step": 20975
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0079329962073326e-06,
+      "loss": 2.4538,
+      "step": 20976
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007885587863464e-06,
+      "loss": 3.3452,
+      "step": 20977
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0078381795195956e-06,
+      "loss": 2.0875,
+      "step": 20978
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007790771175727e-06,
+      "loss": 2.8436,
+      "step": 20979
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007743362831858e-06,
+      "loss": 3.7664,
+      "step": 20980
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.00769595448799e-06,
+      "loss": 2.5881,
+      "step": 20981
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0076485461441216e-06,
+      "loss": 3.8779,
+      "step": 20982
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007601137800253e-06,
+      "loss": 2.1127,
+      "step": 20983
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0075537294563846e-06,
+      "loss": 1.6196,
+      "step": 20984
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007506321112516e-06,
+      "loss": 2.1026,
+      "step": 20985
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007458912768647e-06,
+      "loss": 2.6806,
+      "step": 20986
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007411504424779e-06,
+      "loss": 2.5503,
+      "step": 20987
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0073640960809105e-06,
+      "loss": 2.1545,
+      "step": 20988
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007316687737042e-06,
+      "loss": 3.8955,
+      "step": 20989
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0072692793931735e-06,
+      "loss": 2.8757,
+      "step": 20990
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007221871049305e-06,
+      "loss": 2.4212,
+      "step": 20991
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.007174462705436e-06,
+      "loss": 2.4125,
+      "step": 20992
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.0071270543615678e-06,
+      "loss": 3.3663,
+      "step": 20993
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.007079646017699e-06,
+      "loss": 2.8657,
+      "step": 20994
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0070322376738308e-06,
+      "loss": 2.1048,
+      "step": 20995
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006984829329962e-06,
+      "loss": 2.7057,
+      "step": 20996
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0069374209860937e-06,
+      "loss": 3.0618,
+      "step": 20997
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006890012642225e-06,
+      "loss": 2.5486,
+      "step": 20998
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0068426042983563e-06,
+      "loss": 2.8404,
+      "step": 20999
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006795195954488e-06,
+      "loss": 1.9702,
+      "step": 21000
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0067477876106197e-06,
+      "loss": 2.8701,
+      "step": 21001
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006700379266751e-06,
+      "loss": 2.6387,
+      "step": 21002
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0066529709228827e-06,
+      "loss": 2.1758,
+      "step": 21003
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006605562579014e-06,
+      "loss": 2.8003,
+      "step": 21004
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0065581542351453e-06,
+      "loss": 1.831,
+      "step": 21005
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006510745891277e-06,
+      "loss": 3.7567,
+      "step": 21006
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0064633375474082e-06,
+      "loss": 1.8733,
+      "step": 21007
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.00641592920354e-06,
+      "loss": 1.8169,
+      "step": 21008
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0063685208596716e-06,
+      "loss": 2.3578,
+      "step": 21009
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006321112515803e-06,
+      "loss": 1.9218,
+      "step": 21010
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006273704171934e-06,
+      "loss": 1.4871,
+      "step": 21011
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006226295828066e-06,
+      "loss": 1.9366,
+      "step": 21012
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006178887484197e-06,
+      "loss": 2.7904,
+      "step": 21013
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0061314791403285e-06,
+      "loss": 2.904,
+      "step": 21014
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0060840707964606e-06,
+      "loss": 2.8576,
+      "step": 21015
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.006036662452592e-06,
+      "loss": 2.4951,
+      "step": 21016
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005989254108723e-06,
+      "loss": 2.742,
+      "step": 21017
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0059418457648544e-06,
+      "loss": 2.3418,
+      "step": 21018
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005894437420986e-06,
+      "loss": 2.8681,
+      "step": 21019
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0058470290771174e-06,
+      "loss": 3.4005,
+      "step": 21020
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005799620733249e-06,
+      "loss": 2.5147,
+      "step": 21021
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005752212389381e-06,
+      "loss": 2.8905,
+      "step": 21022
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005704804045512e-06,
+      "loss": 3.5387,
+      "step": 21023
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0056573957016434e-06,
+      "loss": 1.6692,
+      "step": 21024
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005609987357775e-06,
+      "loss": 2.8917,
+      "step": 21025
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0055625790139064e-06,
+      "loss": 2.946,
+      "step": 21026
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0055151706700377e-06,
+      "loss": 3.0918,
+      "step": 21027
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0054677623261694e-06,
+      "loss": 2.3637,
+      "step": 21028
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005420353982301e-06,
+      "loss": 3.4128,
+      "step": 21029
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0053729456384323e-06,
+      "loss": 3.0887,
+      "step": 21030
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005325537294564e-06,
+      "loss": 2.2616,
+      "step": 21031
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0052781289506953e-06,
+      "loss": 2.7528,
+      "step": 21032
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0052307206068266e-06,
+      "loss": 2.9212,
+      "step": 21033
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0051833122629583e-06,
+      "loss": 2.2712,
+      "step": 21034
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.00513590391909e-06,
+      "loss": 4.4057,
+      "step": 21035
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0050884955752213e-06,
+      "loss": 3.7934,
+      "step": 21036
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.005041087231353e-06,
+      "loss": 2.8459,
+      "step": 21037
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0049936788874843e-06,
+      "loss": 2.3757,
+      "step": 21038
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0049462705436156e-06,
+      "loss": 3.3445,
+      "step": 21039
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.004898862199747e-06,
+      "loss": 2.8493,
+      "step": 21040
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0048514538558786e-06,
+      "loss": 2.4984,
+      "step": 21041
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0048040455120103e-06,
+      "loss": 2.4453,
+      "step": 21042
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0047566371681415e-06,
+      "loss": 2.6985,
+      "step": 21043
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0047092288242732e-06,
+      "loss": 2.3885,
+      "step": 21044
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0046618204804045e-06,
+      "loss": 2.1466,
+      "step": 21045
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.004614412136536e-06,
+      "loss": 2.5897,
+      "step": 21046
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0045670037926675e-06,
+      "loss": 4.0826,
+      "step": 21047
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.004519595448799e-06,
+      "loss": 2.235,
+      "step": 21048
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0044721871049305e-06,
+      "loss": 2.3276,
+      "step": 21049
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.004424778761062e-06,
+      "loss": 2.4145,
+      "step": 21050
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0043773704171935e-06,
+      "loss": 2.7941,
+      "step": 21051
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0043299620733248e-06,
+      "loss": 2.7881,
+      "step": 21052
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0042825537294565e-06,
+      "loss": 2.2141,
+      "step": 21053
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0042351453855877e-06,
+      "loss": 2.9957,
+      "step": 21054
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0041877370417194e-06,
+      "loss": 2.3519,
+      "step": 21055
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.004140328697851e-06,
+      "loss": 2.26,
+      "step": 21056
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.0040929203539824e-06,
+      "loss": 2.2757,
+      "step": 21057
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0040455120101137e-06,
+      "loss": 2.7887,
+      "step": 21058
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0039981036662454e-06,
+      "loss": 2.2216,
+      "step": 21059
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0039506953223767e-06,
+      "loss": 1.9847,
+      "step": 21060
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0039032869785084e-06,
+      "loss": 2.262,
+      "step": 21061
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0038558786346397e-06,
+      "loss": 2.8785,
+      "step": 21062
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0038084702907714e-06,
+      "loss": 3.094,
+      "step": 21063
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0037610619469027e-06,
+      "loss": 2.5047,
+      "step": 21064
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003713653603034e-06,
+      "loss": 2.8863,
+      "step": 21065
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0036662452591656e-06,
+      "loss": 2.9584,
+      "step": 21066
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003618836915297e-06,
+      "loss": 3.5908,
+      "step": 21067
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0035714285714286e-06,
+      "loss": 2.4509,
+      "step": 21068
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0035240202275603e-06,
+      "loss": 3.496,
+      "step": 21069
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0034766118836916e-06,
+      "loss": 2.1274,
+      "step": 21070
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003429203539823e-06,
+      "loss": 2.4875,
+      "step": 21071
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0033817951959546e-06,
+      "loss": 2.2608,
+      "step": 21072
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003334386852086e-06,
+      "loss": 1.3178,
+      "step": 21073
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0032869785082176e-06,
+      "loss": 4.0051,
+      "step": 21074
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0032395701643493e-06,
+      "loss": 2.8778,
+      "step": 21075
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0031921618204806e-06,
+      "loss": 3.0937,
+      "step": 21076
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003144753476612e-06,
+      "loss": 2.1322,
+      "step": 21077
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0030973451327436e-06,
+      "loss": 2.7679,
+      "step": 21078
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003049936788875e-06,
+      "loss": 2.798,
+      "step": 21079
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.003002528445006e-06,
+      "loss": 3.2127,
+      "step": 21080
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002955120101138e-06,
+      "loss": 1.9888,
+      "step": 21081
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0029077117572695e-06,
+      "loss": 3.0192,
+      "step": 21082
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002860303413401e-06,
+      "loss": 3.3606,
+      "step": 21083
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002812895069532e-06,
+      "loss": 3.2082,
+      "step": 21084
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0027654867256638e-06,
+      "loss": 3.1874,
+      "step": 21085
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002718078381795e-06,
+      "loss": 3.5905,
+      "step": 21086
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0026706700379268e-06,
+      "loss": 2.1777,
+      "step": 21087
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0026232616940585e-06,
+      "loss": 2.2329,
+      "step": 21088
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0025758533501898e-06,
+      "loss": 2.4108,
+      "step": 21089
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002528445006321e-06,
+      "loss": 3.8544,
+      "step": 21090
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0024810366624527e-06,
+      "loss": 2.4678,
+      "step": 21091
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002433628318584e-06,
+      "loss": 3.5688,
+      "step": 21092
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0023862199747153e-06,
+      "loss": 2.3963,
+      "step": 21093
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002338811630847e-06,
+      "loss": 3.3225,
+      "step": 21094
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0022914032869787e-06,
+      "loss": 2.5555,
+      "step": 21095
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.00224399494311e-06,
+      "loss": 3.1099,
+      "step": 21096
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0021965865992417e-06,
+      "loss": 3.8729,
+      "step": 21097
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002149178255373e-06,
+      "loss": 3.3361,
+      "step": 21098
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0021017699115043e-06,
+      "loss": 1.6857,
+      "step": 21099
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.002054361567636e-06,
+      "loss": 1.6111,
+      "step": 21100
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0020069532237677e-06,
+      "loss": 1.7594,
+      "step": 21101
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001959544879899e-06,
+      "loss": 3.2398,
+      "step": 21102
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0019121365360306e-06,
+      "loss": 3.399,
+      "step": 21103
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001864728192162e-06,
+      "loss": 2.9797,
+      "step": 21104
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001817319848293e-06,
+      "loss": 3.1115,
+      "step": 21105
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0017699115044245e-06,
+      "loss": 1.6387,
+      "step": 21106
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001722503160556e-06,
+      "loss": 1.9219,
+      "step": 21107
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001675094816688e-06,
+      "loss": 3.1244,
+      "step": 21108
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001627686472819e-06,
+      "loss": 2.3949,
+      "step": 21109
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001580278128951e-06,
+      "loss": 2.7675,
+      "step": 21110
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001532869785082e-06,
+      "loss": 2.5274,
+      "step": 21111
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0014854614412134e-06,
+      "loss": 3.4847,
+      "step": 21112
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001438053097345e-06,
+      "loss": 3.3054,
+      "step": 21113
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001390644753477e-06,
+      "loss": 2.4954,
+      "step": 21114
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001343236409608e-06,
+      "loss": 3.207,
+      "step": 21115
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.00129582806574e-06,
+      "loss": 3.4316,
+      "step": 21116
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001248419721871e-06,
+      "loss": 4.066,
+      "step": 21117
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0012010113780024e-06,
+      "loss": 2.2089,
+      "step": 21118
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.001153603034134e-06,
+      "loss": 3.3332,
+      "step": 21119
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.0011061946902654e-06,
+      "loss": 2.7414,
+      "step": 21120
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.001058786346397e-06,
+      "loss": 2.7366,
+      "step": 21121
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0010113780025288e-06,
+      "loss": 2.7959,
+      "step": 21122
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.00096396965866e-06,
+      "loss": 2.002,
+      "step": 21123
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0009165613147913e-06,
+      "loss": 3.1097,
+      "step": 21124
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.000869152970923e-06,
+      "loss": 2.8144,
+      "step": 21125
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0008217446270543e-06,
+      "loss": 1.9259,
+      "step": 21126
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.000774336283186e-06,
+      "loss": 2.9059,
+      "step": 21127
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0007269279393173e-06,
+      "loss": 2.1762,
+      "step": 21128
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.000679519595449e-06,
+      "loss": 1.5871,
+      "step": 21129
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0006321112515803e-06,
+      "loss": 1.5696,
+      "step": 21130
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0005847029077116e-06,
+      "loss": 1.7903,
+      "step": 21131
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0005372945638433e-06,
+      "loss": 1.8397,
+      "step": 21132
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0004898862199746e-06,
+      "loss": 2.7818,
+      "step": 21133
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0004424778761063e-06,
+      "loss": 2.9749,
+      "step": 21134
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.000395069532238e-06,
+      "loss": 3.2533,
+      "step": 21135
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0003476611883692e-06,
+      "loss": 2.6991,
+      "step": 21136
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0003002528445005e-06,
+      "loss": 3.1271,
+      "step": 21137
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0002528445006322e-06,
+      "loss": 2.3126,
+      "step": 21138
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0002054361567635e-06,
+      "loss": 4.0788,
+      "step": 21139
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0001580278128952e-06,
+      "loss": 2.4258,
+      "step": 21140
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.000110619469027e-06,
+      "loss": 3.2714,
+      "step": 21141
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.000063211125158e-06,
+      "loss": 2.4719,
+      "step": 21142
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.0000158027812895e-06,
+      "loss": 1.717,
+      "step": 21143
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.999968394437421e-06,
+      "loss": 3.2007,
+      "step": 21144
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9999209860935525e-06,
+      "loss": 3.435,
+      "step": 21145
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9998735777496837e-06,
+      "loss": 1.5653,
+      "step": 21146
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9998261694058155e-06,
+      "loss": 2.9969,
+      "step": 21147
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.999778761061947e-06,
+      "loss": 2.1723,
+      "step": 21148
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9997313527180784e-06,
+      "loss": 2.9419,
+      "step": 21149
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9996839443742097e-06,
+      "loss": 1.7368,
+      "step": 21150
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9996365360303414e-06,
+      "loss": 2.1993,
+      "step": 21151
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9995891276864727e-06,
+      "loss": 2.6148,
+      "step": 21152
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9995417193426044e-06,
+      "loss": 2.4412,
+      "step": 21153
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.999494310998736e-06,
+      "loss": 2.5407,
+      "step": 21154
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9994469026548674e-06,
+      "loss": 2.1612,
+      "step": 21155
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9993994943109987e-06,
+      "loss": 1.919,
+      "step": 21156
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9993520859671304e-06,
+      "loss": 3.0854,
+      "step": 21157
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9993046776232617e-06,
+      "loss": 4.0672,
+      "step": 21158
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.999257269279393e-06,
+      "loss": 2.2461,
+      "step": 21159
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9992098609355246e-06,
+      "loss": 2.1552,
+      "step": 21160
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9991624525916563e-06,
+      "loss": 3.2919,
+      "step": 21161
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9991150442477876e-06,
+      "loss": 3.4522,
+      "step": 21162
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9990676359039193e-06,
+      "loss": 3.5682,
+      "step": 21163
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9990202275600506e-06,
+      "loss": 3.7513,
+      "step": 21164
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998972819216182e-06,
+      "loss": 2.6043,
+      "step": 21165
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9989254108723136e-06,
+      "loss": 3.8081,
+      "step": 21166
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9988780025284453e-06,
+      "loss": 1.7683,
+      "step": 21167
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9988305941845766e-06,
+      "loss": 2.0135,
+      "step": 21168
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998783185840708e-06,
+      "loss": 3.4811,
+      "step": 21169
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9987357774968396e-06,
+      "loss": 3.5227,
+      "step": 21170
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998688369152971e-06,
+      "loss": 2.2777,
+      "step": 21171
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998640960809102e-06,
+      "loss": 2.4192,
+      "step": 21172
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998593552465234e-06,
+      "loss": 2.2199,
+      "step": 21173
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9985461441213655e-06,
+      "loss": 2.8676,
+      "step": 21174
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998498735777497e-06,
+      "loss": 2.457,
+      "step": 21175
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9984513274336285e-06,
+      "loss": 3.112,
+      "step": 21176
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.99840391908976e-06,
+      "loss": 3.0763,
+      "step": 21177
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.998356510745891e-06,
+      "loss": 3.1352,
+      "step": 21178
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9983091024020228e-06,
+      "loss": 2.3063,
+      "step": 21179
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9982616940581545e-06,
+      "loss": 1.7347,
+      "step": 21180
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9982142857142858e-06,
+      "loss": 3.5839,
+      "step": 21181
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9981668773704175e-06,
+      "loss": 1.8155,
+      "step": 21182
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.9981194690265487e-06,
+      "loss": 2.4505,
+      "step": 21183
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.99807206068268e-06,
+      "loss": 2.4725,
+      "step": 21184
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9980246523388117e-06,
+      "loss": 3.8563,
+      "step": 21185
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997977243994943e-06,
+      "loss": 2.5721,
+      "step": 21186
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9979298356510747e-06,
+      "loss": 3.5916,
+      "step": 21187
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9978824273072064e-06,
+      "loss": 3.4061,
+      "step": 21188
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9978350189633377e-06,
+      "loss": 2.3839,
+      "step": 21189
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997787610619469e-06,
+      "loss": 2.815,
+      "step": 21190
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9977402022756003e-06,
+      "loss": 2.4919,
+      "step": 21191
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997692793931732e-06,
+      "loss": 2.9823,
+      "step": 21192
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9976453855878637e-06,
+      "loss": 2.6212,
+      "step": 21193
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997597977243995e-06,
+      "loss": 3.772,
+      "step": 21194
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9975505689001267e-06,
+      "loss": 2.0141,
+      "step": 21195
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997503160556258e-06,
+      "loss": 3.9731,
+      "step": 21196
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9974557522123892e-06,
+      "loss": 2.3255,
+      "step": 21197
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997408343868521e-06,
+      "loss": 3.0424,
+      "step": 21198
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997360935524652e-06,
+      "loss": 3.8425,
+      "step": 21199
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997313527180784e-06,
+      "loss": 2.8418,
+      "step": 21200
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9972661188369156e-06,
+      "loss": 3.375,
+      "step": 21201
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997218710493047e-06,
+      "loss": 3.9919,
+      "step": 21202
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997171302149178e-06,
+      "loss": 2.9589,
+      "step": 21203
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.99712389380531e-06,
+      "loss": 2.2469,
+      "step": 21204
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997076485461441e-06,
+      "loss": 2.5956,
+      "step": 21205
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.997029077117573e-06,
+      "loss": 2.9068,
+      "step": 21206
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9969816687737046e-06,
+      "loss": 4.3278,
+      "step": 21207
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996934260429836e-06,
+      "loss": 2.5879,
+      "step": 21208
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996886852085967e-06,
+      "loss": 1.7926,
+      "step": 21209
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996839443742099e-06,
+      "loss": 2.8963,
+      "step": 21210
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.99679203539823e-06,
+      "loss": 2.5177,
+      "step": 21211
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9967446270543614e-06,
+      "loss": 2.5721,
+      "step": 21212
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996697218710493e-06,
+      "loss": 3.2847,
+      "step": 21213
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996649810366625e-06,
+      "loss": 1.8707,
+      "step": 21214
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996602402022756e-06,
+      "loss": 1.9093,
+      "step": 21215
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9965549936788874e-06,
+      "loss": 3.1273,
+      "step": 21216
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996507585335019e-06,
+      "loss": 2.9411,
+      "step": 21217
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9964601769911503e-06,
+      "loss": 3.7664,
+      "step": 21218
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996412768647282e-06,
+      "loss": 2.5309,
+      "step": 21219
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9963653603034137e-06,
+      "loss": 4.4985,
+      "step": 21220
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996317951959545e-06,
+      "loss": 3.1758,
+      "step": 21221
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9962705436156763e-06,
+      "loss": 2.3779,
+      "step": 21222
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996223135271808e-06,
+      "loss": 2.4207,
+      "step": 21223
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9961757269279393e-06,
+      "loss": 3.9064,
+      "step": 21224
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9961283185840706e-06,
+      "loss": 2.8128,
+      "step": 21225
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9960809102402023e-06,
+      "loss": 3.1814,
+      "step": 21226
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.996033501896334e-06,
+      "loss": 0.9381,
+      "step": 21227
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9959860935524653e-06,
+      "loss": 3.8214,
+      "step": 21228
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.995938685208597e-06,
+      "loss": 3.0444,
+      "step": 21229
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9958912768647282e-06,
+      "loss": 3.5015,
+      "step": 21230
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9958438685208595e-06,
+      "loss": 3.3685,
+      "step": 21231
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9957964601769912e-06,
+      "loss": 3.4283,
+      "step": 21232
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.995749051833123e-06,
+      "loss": 2.9084,
+      "step": 21233
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9957016434892542e-06,
+      "loss": 2.5653,
+      "step": 21234
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9956542351453855e-06,
+      "loss": 3.6052,
+      "step": 21235
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.995606826801517e-06,
+      "loss": 3.4891,
+      "step": 21236
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9955594184576485e-06,
+      "loss": 3.2346,
+      "step": 21237
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9955120101137798e-06,
+      "loss": 2.5173,
+      "step": 21238
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9954646017699115e-06,
+      "loss": 2.9857,
+      "step": 21239
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.995417193426043e-06,
+      "loss": 3.4745,
+      "step": 21240
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9953697850821744e-06,
+      "loss": 2.7679,
+      "step": 21241
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.995322376738306e-06,
+      "loss": 2.5047,
+      "step": 21242
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9952749683944374e-06,
+      "loss": 2.6545,
+      "step": 21243
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9952275600505687e-06,
+      "loss": 1.9375,
+      "step": 21244
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9951801517067004e-06,
+      "loss": 2.0081,
+      "step": 21245
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.995132743362832e-06,
+      "loss": 2.4321,
+      "step": 21246
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.9950853350189634e-06,
+      "loss": 3.2028,
+      "step": 21247
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.995037926675095e-06,
+      "loss": 2.037,
+      "step": 21248
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9949905183312264e-06,
+      "loss": 2.051,
+      "step": 21249
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9949431099873577e-06,
+      "loss": 2.0313,
+      "step": 21250
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9948957016434894e-06,
+      "loss": 1.6546,
+      "step": 21251
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9948482932996206e-06,
+      "loss": 2.649,
+      "step": 21252
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9948008849557524e-06,
+      "loss": 2.6427,
+      "step": 21253
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.994753476611884e-06,
+      "loss": 2.3455,
+      "step": 21254
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9947060682680153e-06,
+      "loss": 3.7516,
+      "step": 21255
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9946586599241466e-06,
+      "loss": 2.872,
+      "step": 21256
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.994611251580278e-06,
+      "loss": 3.2004,
+      "step": 21257
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9945638432364096e-06,
+      "loss": 2.0851,
+      "step": 21258
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9945164348925413e-06,
+      "loss": 3.5586,
+      "step": 21259
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9944690265486726e-06,
+      "loss": 2.8476,
+      "step": 21260
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9944216182048043e-06,
+      "loss": 2.2798,
+      "step": 21261
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9943742098609356e-06,
+      "loss": 2.7978,
+      "step": 21262
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.994326801517067e-06,
+      "loss": 2.4596,
+      "step": 21263
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9942793931731986e-06,
+      "loss": 3.1748,
+      "step": 21264
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.99423198482933e-06,
+      "loss": 2.6122,
+      "step": 21265
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9941845764854615e-06,
+      "loss": 2.8643,
+      "step": 21266
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9941371681415932e-06,
+      "loss": 3.4691,
+      "step": 21267
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9940897597977245e-06,
+      "loss": 2.8086,
+      "step": 21268
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.994042351453856e-06,
+      "loss": 2.734,
+      "step": 21269
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9939949431099875e-06,
+      "loss": 2.717,
+      "step": 21270
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9939475347661188e-06,
+      "loss": 1.9825,
+      "step": 21271
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9939001264222505e-06,
+      "loss": 2.308,
+      "step": 21272
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.993852718078382e-06,
+      "loss": 1.6419,
+      "step": 21273
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9938053097345135e-06,
+      "loss": 2.3372,
+      "step": 21274
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9937579013906448e-06,
+      "loss": 2.9649,
+      "step": 21275
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9937104930467765e-06,
+      "loss": 3.435,
+      "step": 21276
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9936630847029077e-06,
+      "loss": 2.433,
+      "step": 21277
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.993615676359039e-06,
+      "loss": 2.0291,
+      "step": 21278
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9935682680151707e-06,
+      "loss": 1.9483,
+      "step": 21279
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9935208596713024e-06,
+      "loss": 2.9681,
+      "step": 21280
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9934734513274337e-06,
+      "loss": 2.7858,
+      "step": 21281
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.993426042983565e-06,
+      "loss": 2.4886,
+      "step": 21282
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9933786346396967e-06,
+      "loss": 3.0531,
+      "step": 21283
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.993331226295828e-06,
+      "loss": 4.2183,
+      "step": 21284
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9932838179519593e-06,
+      "loss": 1.6152,
+      "step": 21285
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9932364096080914e-06,
+      "loss": 3.5102,
+      "step": 21286
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9931890012642227e-06,
+      "loss": 2.8418,
+      "step": 21287
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.993141592920354e-06,
+      "loss": 2.9153,
+      "step": 21288
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9930941845764856e-06,
+      "loss": 3.1572,
+      "step": 21289
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.993046776232617e-06,
+      "loss": 3.1454,
+      "step": 21290
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992999367888748e-06,
+      "loss": 2.149,
+      "step": 21291
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.99295195954488e-06,
+      "loss": 2.2428,
+      "step": 21292
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9929045512010116e-06,
+      "loss": 1.8742,
+      "step": 21293
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992857142857143e-06,
+      "loss": 2.2739,
+      "step": 21294
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9928097345132746e-06,
+      "loss": 2.4268,
+      "step": 21295
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992762326169406e-06,
+      "loss": 1.3016,
+      "step": 21296
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992714917825537e-06,
+      "loss": 2.107,
+      "step": 21297
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992667509481669e-06,
+      "loss": 3.4545,
+      "step": 21298
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9926201011378e-06,
+      "loss": 4.675,
+      "step": 21299
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992572692793932e-06,
+      "loss": 1.8417,
+      "step": 21300
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992525284450063e-06,
+      "loss": 2.729,
+      "step": 21301
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992477876106195e-06,
+      "loss": 3.0547,
+      "step": 21302
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992430467762326e-06,
+      "loss": 2.0513,
+      "step": 21303
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9923830594184574e-06,
+      "loss": 3.2159,
+      "step": 21304
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992335651074589e-06,
+      "loss": 3.9186,
+      "step": 21305
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992288242730721e-06,
+      "loss": 2.5515,
+      "step": 21306
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992240834386852e-06,
+      "loss": 2.6646,
+      "step": 21307
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9921934260429838e-06,
+      "loss": 2.2847,
+      "step": 21308
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.992146017699115e-06,
+      "loss": 2.6834,
+      "step": 21309
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.9920986093552463e-06,
+      "loss": 2.7652,
+      "step": 21310
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.992051201011378e-06,
+      "loss": 2.2745,
+      "step": 21311
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9920037926675093e-06,
+      "loss": 2.9278,
+      "step": 21312
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991956384323641e-06,
+      "loss": 3.3123,
+      "step": 21313
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9919089759797727e-06,
+      "loss": 1.7836,
+      "step": 21314
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991861567635904e-06,
+      "loss": 3.716,
+      "step": 21315
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9918141592920353e-06,
+      "loss": 2.7591,
+      "step": 21316
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991766750948167e-06,
+      "loss": 3.0116,
+      "step": 21317
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9917193426042983e-06,
+      "loss": 3.1044,
+      "step": 21318
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.99167193426043e-06,
+      "loss": 3.1797,
+      "step": 21319
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9916245259165617e-06,
+      "loss": 3.4243,
+      "step": 21320
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991577117572693e-06,
+      "loss": 2.7315,
+      "step": 21321
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9915297092288243e-06,
+      "loss": 2.4146,
+      "step": 21322
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9914823008849555e-06,
+      "loss": 1.9521,
+      "step": 21323
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9914348925410872e-06,
+      "loss": 2.5739,
+      "step": 21324
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9913874841972185e-06,
+      "loss": 2.3643,
+      "step": 21325
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9913400758533502e-06,
+      "loss": 3.5941,
+      "step": 21326
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991292667509482e-06,
+      "loss": 2.9364,
+      "step": 21327
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991245259165613e-06,
+      "loss": 3.5456,
+      "step": 21328
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9911978508217445e-06,
+      "loss": 2.85,
+      "step": 21329
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991150442477876e-06,
+      "loss": 3.0642,
+      "step": 21330
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9911030341340075e-06,
+      "loss": 3.1509,
+      "step": 21331
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991055625790139e-06,
+      "loss": 3.327,
+      "step": 21332
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.991008217446271e-06,
+      "loss": 1.8333,
+      "step": 21333
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.990960809102402e-06,
+      "loss": 1.3871,
+      "step": 21334
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9909134007585334e-06,
+      "loss": 2.6853,
+      "step": 21335
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.990865992414665e-06,
+      "loss": 2.4329,
+      "step": 21336
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9908185840707964e-06,
+      "loss": 2.5751,
+      "step": 21337
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9907711757269277e-06,
+      "loss": 3.4024,
+      "step": 21338
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9907237673830594e-06,
+      "loss": 1.487,
+      "step": 21339
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.990676359039191e-06,
+      "loss": 3.6028,
+      "step": 21340
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9906289506953224e-06,
+      "loss": 2.0647,
+      "step": 21341
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.990581542351454e-06,
+      "loss": 4.0623,
+      "step": 21342
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9905341340075854e-06,
+      "loss": 1.9695,
+      "step": 21343
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9904867256637167e-06,
+      "loss": 1.9966,
+      "step": 21344
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9904393173198484e-06,
+      "loss": 3.3463,
+      "step": 21345
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.99039190897598e-06,
+      "loss": 2.4861,
+      "step": 21346
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9903445006321113e-06,
+      "loss": 2.5413,
+      "step": 21347
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9902970922882426e-06,
+      "loss": 2.911,
+      "step": 21348
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9902496839443743e-06,
+      "loss": 2.6982,
+      "step": 21349
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9902022756005056e-06,
+      "loss": 3.2437,
+      "step": 21350
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.990154867256637e-06,
+      "loss": 3.0052,
+      "step": 21351
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9901074589127686e-06,
+      "loss": 3.1992,
+      "step": 21352
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9900600505689003e-06,
+      "loss": 2.225,
+      "step": 21353
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9900126422250316e-06,
+      "loss": 2.4081,
+      "step": 21354
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9899652338811633e-06,
+      "loss": 2.4574,
+      "step": 21355
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9899178255372946e-06,
+      "loss": 2.8277,
+      "step": 21356
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.989870417193426e-06,
+      "loss": 1.719,
+      "step": 21357
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9898230088495575e-06,
+      "loss": 3.2496,
+      "step": 21358
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9897756005056893e-06,
+      "loss": 2.2306,
+      "step": 21359
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9897281921618205e-06,
+      "loss": 1.1975,
+      "step": 21360
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9896807838179522e-06,
+      "loss": 2.6038,
+      "step": 21361
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9896333754740835e-06,
+      "loss": 1.7321,
+      "step": 21362
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.989585967130215e-06,
+      "loss": 2.2445,
+      "step": 21363
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9895385587863465e-06,
+      "loss": 3.9992,
+      "step": 21364
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9894911504424778e-06,
+      "loss": 4.0285,
+      "step": 21365
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9894437420986095e-06,
+      "loss": 2.5636,
+      "step": 21366
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9893963337547408e-06,
+      "loss": 2.2019,
+      "step": 21367
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9893489254108725e-06,
+      "loss": 2.8948,
+      "step": 21368
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9893015170670037e-06,
+      "loss": 3.3524,
+      "step": 21369
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.989254108723135e-06,
+      "loss": 2.7127,
+      "step": 21370
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9892067003792667e-06,
+      "loss": 2.0303,
+      "step": 21371
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9891592920353984e-06,
+      "loss": 2.4535,
+      "step": 21372
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.9891118836915297e-06,
+      "loss": 1.8795,
+      "step": 21373
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9890644753476614e-06,
+      "loss": 2.6518,
+      "step": 21374
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9890170670037927e-06,
+      "loss": 2.9077,
+      "step": 21375
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988969658659924e-06,
+      "loss": 3.4157,
+      "step": 21376
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9889222503160557e-06,
+      "loss": 3.2612,
+      "step": 21377
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988874841972187e-06,
+      "loss": 2.4652,
+      "step": 21378
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9888274336283187e-06,
+      "loss": 2.4424,
+      "step": 21379
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9887800252844504e-06,
+      "loss": 1.9564,
+      "step": 21380
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9887326169405817e-06,
+      "loss": 2.7162,
+      "step": 21381
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988685208596713e-06,
+      "loss": 3.3876,
+      "step": 21382
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9886378002528446e-06,
+      "loss": 2.1701,
+      "step": 21383
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988590391908976e-06,
+      "loss": 1.6014,
+      "step": 21384
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9885429835651076e-06,
+      "loss": 3.7242,
+      "step": 21385
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9884955752212393e-06,
+      "loss": 3.6961,
+      "step": 21386
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9884481668773706e-06,
+      "loss": 3.017,
+      "step": 21387
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988400758533502e-06,
+      "loss": 2.0899,
+      "step": 21388
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988353350189633e-06,
+      "loss": 0.9606,
+      "step": 21389
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988305941845765e-06,
+      "loss": 3.1106,
+      "step": 21390
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988258533501896e-06,
+      "loss": 2.8809,
+      "step": 21391
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988211125158028e-06,
+      "loss": 2.4494,
+      "step": 21392
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9881637168141596e-06,
+      "loss": 2.8783,
+      "step": 21393
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988116308470291e-06,
+      "loss": 2.9222,
+      "step": 21394
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988068900126422e-06,
+      "loss": 3.5835,
+      "step": 21395
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.988021491782554e-06,
+      "loss": 2.4797,
+      "step": 21396
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987974083438685e-06,
+      "loss": 3.1559,
+      "step": 21397
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987926675094817e-06,
+      "loss": 2.6215,
+      "step": 21398
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9878792667509485e-06,
+      "loss": 2.6676,
+      "step": 21399
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.98783185840708e-06,
+      "loss": 3.5544,
+      "step": 21400
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987784450063211e-06,
+      "loss": 3.3834,
+      "step": 21401
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9877370417193428e-06,
+      "loss": 2.4709,
+      "step": 21402
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987689633375474e-06,
+      "loss": 2.6577,
+      "step": 21403
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9876422250316053e-06,
+      "loss": 2.1684,
+      "step": 21404
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987594816687737e-06,
+      "loss": 1.2855,
+      "step": 21405
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9875474083438687e-06,
+      "loss": 1.6056,
+      "step": 21406
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9875e-06,
+      "loss": 2.9668,
+      "step": 21407
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9874525916561317e-06,
+      "loss": 3.1117,
+      "step": 21408
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987405183312263e-06,
+      "loss": 2.1932,
+      "step": 21409
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9873577749683943e-06,
+      "loss": 2.4271,
+      "step": 21410
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987310366624526e-06,
+      "loss": 1.7939,
+      "step": 21411
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9872629582806577e-06,
+      "loss": 2.2832,
+      "step": 21412
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987215549936789e-06,
+      "loss": 2.2784,
+      "step": 21413
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9871681415929203e-06,
+      "loss": 2.4047,
+      "step": 21414
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.987120733249052e-06,
+      "loss": 3.2943,
+      "step": 21415
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9870733249051832e-06,
+      "loss": 3.3541,
+      "step": 21416
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9870259165613145e-06,
+      "loss": 2.3333,
+      "step": 21417
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9869785082174462e-06,
+      "loss": 2.1201,
+      "step": 21418
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986931099873578e-06,
+      "loss": 2.4172,
+      "step": 21419
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9868836915297092e-06,
+      "loss": 1.2134,
+      "step": 21420
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986836283185841e-06,
+      "loss": 3.2305,
+      "step": 21421
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986788874841972e-06,
+      "loss": 4.0837,
+      "step": 21422
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9867414664981035e-06,
+      "loss": 2.4137,
+      "step": 21423
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986694058154235e-06,
+      "loss": 2.927,
+      "step": 21424
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986646649810367e-06,
+      "loss": 3.4534,
+      "step": 21425
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986599241466498e-06,
+      "loss": 2.9533,
+      "step": 21426
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.98655183312263e-06,
+      "loss": 3.4895,
+      "step": 21427
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986504424778761e-06,
+      "loss": 2.5889,
+      "step": 21428
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9864570164348924e-06,
+      "loss": 3.8471,
+      "step": 21429
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986409608091024e-06,
+      "loss": 2.0494,
+      "step": 21430
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9863621997471554e-06,
+      "loss": 2.1446,
+      "step": 21431
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.986314791403287e-06,
+      "loss": 3.165,
+      "step": 21432
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9862673830594184e-06,
+      "loss": 3.411,
+      "step": 21433
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.98621997471555e-06,
+      "loss": 2.6341,
+      "step": 21434
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9861725663716814e-06,
+      "loss": 1.8487,
+      "step": 21435
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9861251580278127e-06,
+      "loss": 2.4708,
+      "step": 21436
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.9860777496839444e-06,
+      "loss": 2.6618,
+      "step": 21437
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.986030341340076e-06,
+      "loss": 1.384,
+      "step": 21438
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9859829329962074e-06,
+      "loss": 2.015,
+      "step": 21439
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.985935524652339e-06,
+      "loss": 2.4974,
+      "step": 21440
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9858881163084703e-06,
+      "loss": 3.9598,
+      "step": 21441
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9858407079646016e-06,
+      "loss": 2.6175,
+      "step": 21442
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9857932996207333e-06,
+      "loss": 2.1625,
+      "step": 21443
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9857458912768646e-06,
+      "loss": 1.7478,
+      "step": 21444
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9856984829329963e-06,
+      "loss": 1.6803,
+      "step": 21445
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.985651074589128e-06,
+      "loss": 2.0347,
+      "step": 21446
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9856036662452593e-06,
+      "loss": 2.0594,
+      "step": 21447
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9855562579013906e-06,
+      "loss": 2.6157,
+      "step": 21448
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9855088495575223e-06,
+      "loss": 3.4467,
+      "step": 21449
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9854614412136536e-06,
+      "loss": 3.0196,
+      "step": 21450
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9854140328697853e-06,
+      "loss": 4.0021,
+      "step": 21451
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.985366624525917e-06,
+      "loss": 2.3599,
+      "step": 21452
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9853192161820482e-06,
+      "loss": 3.2753,
+      "step": 21453
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9852718078381795e-06,
+      "loss": 1.4667,
+      "step": 21454
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.985224399494311e-06,
+      "loss": 3.0877,
+      "step": 21455
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9851769911504425e-06,
+      "loss": 1.9274,
+      "step": 21456
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.985129582806574e-06,
+      "loss": 1.9608,
+      "step": 21457
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9850821744627055e-06,
+      "loss": 3.6255,
+      "step": 21458
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.985034766118837e-06,
+      "loss": 3.1471,
+      "step": 21459
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9849873577749685e-06,
+      "loss": 2.3183,
+      "step": 21460
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9849399494310998e-06,
+      "loss": 2.2109,
+      "step": 21461
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9848925410872315e-06,
+      "loss": 1.5274,
+      "step": 21462
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9848451327433627e-06,
+      "loss": 3.2352,
+      "step": 21463
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9847977243994944e-06,
+      "loss": 2.5401,
+      "step": 21464
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.984750316055626e-06,
+      "loss": 2.2576,
+      "step": 21465
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9847029077117574e-06,
+      "loss": 1.7458,
+      "step": 21466
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9846554993678887e-06,
+      "loss": 1.1814,
+      "step": 21467
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9846080910240204e-06,
+      "loss": 3.4395,
+      "step": 21468
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9845606826801517e-06,
+      "loss": 3.8854,
+      "step": 21469
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.984513274336283e-06,
+      "loss": 2.7779,
+      "step": 21470
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9844658659924147e-06,
+      "loss": 3.3399,
+      "step": 21471
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9844184576485464e-06,
+      "loss": 2.8923,
+      "step": 21472
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9843710493046777e-06,
+      "loss": 2.0767,
+      "step": 21473
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9843236409608094e-06,
+      "loss": 1.3012,
+      "step": 21474
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9842762326169407e-06,
+      "loss": 3.0152,
+      "step": 21475
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.984228824273072e-06,
+      "loss": 2.3637,
+      "step": 21476
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9841814159292036e-06,
+      "loss": 3.7086,
+      "step": 21477
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9841340075853353e-06,
+      "loss": 2.7601,
+      "step": 21478
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9840865992414666e-06,
+      "loss": 1.6872,
+      "step": 21479
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.984039190897598e-06,
+      "loss": 3.9682,
+      "step": 21480
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9839917825537296e-06,
+      "loss": 2.259,
+      "step": 21481
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983944374209861e-06,
+      "loss": 3.7456,
+      "step": 21482
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983896965865992e-06,
+      "loss": 3.3492,
+      "step": 21483
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983849557522124e-06,
+      "loss": 2.493,
+      "step": 21484
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9838021491782556e-06,
+      "loss": 3.4994,
+      "step": 21485
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983754740834387e-06,
+      "loss": 2.3475,
+      "step": 21486
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9837073324905186e-06,
+      "loss": 3.1703,
+      "step": 21487
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.98365992414665e-06,
+      "loss": 2.6464,
+      "step": 21488
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983612515802781e-06,
+      "loss": 2.4506,
+      "step": 21489
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983565107458913e-06,
+      "loss": 2.7929,
+      "step": 21490
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9835176991150445e-06,
+      "loss": 2.6661,
+      "step": 21491
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983470290771176e-06,
+      "loss": 3.0423,
+      "step": 21492
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9834228824273075e-06,
+      "loss": 1.8115,
+      "step": 21493
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983375474083439e-06,
+      "loss": 2.0643,
+      "step": 21494
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.98332806573957e-06,
+      "loss": 2.932,
+      "step": 21495
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9832806573957018e-06,
+      "loss": 3.1272,
+      "step": 21496
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983233249051833e-06,
+      "loss": 2.7348,
+      "step": 21497
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9831858407079648e-06,
+      "loss": 2.6805,
+      "step": 21498
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.983138432364096e-06,
+      "loss": 2.277,
+      "step": 21499
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.9830910240202277e-06,
+      "loss": 3.5833,
+      "step": 21500
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.983043615676359e-06,
+      "loss": 3.2701,
+      "step": 21501
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9829962073324903e-06,
+      "loss": 2.835,
+      "step": 21502
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982948798988622e-06,
+      "loss": 2.053,
+      "step": 21503
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9829013906447537e-06,
+      "loss": 3.4612,
+      "step": 21504
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982853982300885e-06,
+      "loss": 2.2531,
+      "step": 21505
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9828065739570167e-06,
+      "loss": 2.3512,
+      "step": 21506
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982759165613148e-06,
+      "loss": 0.6326,
+      "step": 21507
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9827117572692793e-06,
+      "loss": 3.113,
+      "step": 21508
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982664348925411e-06,
+      "loss": 1.8259,
+      "step": 21509
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9826169405815422e-06,
+      "loss": 2.6647,
+      "step": 21510
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982569532237674e-06,
+      "loss": 2.0799,
+      "step": 21511
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9825221238938056e-06,
+      "loss": 3.1531,
+      "step": 21512
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982474715549937e-06,
+      "loss": 3.6711,
+      "step": 21513
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982427307206068e-06,
+      "loss": 2.3227,
+      "step": 21514
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9823798988622e-06,
+      "loss": 2.5734,
+      "step": 21515
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982332490518331e-06,
+      "loss": 2.5394,
+      "step": 21516
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982285082174463e-06,
+      "loss": 1.6693,
+      "step": 21517
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9822376738305946e-06,
+      "loss": 2.8909,
+      "step": 21518
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982190265486726e-06,
+      "loss": 3.0093,
+      "step": 21519
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.982142857142857e-06,
+      "loss": 2.2292,
+      "step": 21520
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9820954487989884e-06,
+      "loss": 1.5217,
+      "step": 21521
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.98204804045512e-06,
+      "loss": 2.8303,
+      "step": 21522
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9820006321112514e-06,
+      "loss": 2.8198,
+      "step": 21523
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981953223767383e-06,
+      "loss": 3.8839,
+      "step": 21524
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981905815423515e-06,
+      "loss": 3.1562,
+      "step": 21525
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981858407079646e-06,
+      "loss": 2.0791,
+      "step": 21526
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9818109987357774e-06,
+      "loss": 2.1736,
+      "step": 21527
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981763590391909e-06,
+      "loss": 3.0503,
+      "step": 21528
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9817161820480404e-06,
+      "loss": 3.1861,
+      "step": 21529
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981668773704172e-06,
+      "loss": 3.0821,
+      "step": 21530
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981621365360304e-06,
+      "loss": 2.2914,
+      "step": 21531
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981573957016435e-06,
+      "loss": 3.0879,
+      "step": 21532
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9815265486725663e-06,
+      "loss": 2.4364,
+      "step": 21533
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981479140328698e-06,
+      "loss": 2.5676,
+      "step": 21534
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9814317319848293e-06,
+      "loss": 1.4299,
+      "step": 21535
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9813843236409606e-06,
+      "loss": 2.1642,
+      "step": 21536
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9813369152970923e-06,
+      "loss": 2.8205,
+      "step": 21537
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981289506953224e-06,
+      "loss": 1.7408,
+      "step": 21538
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9812420986093553e-06,
+      "loss": 3.1863,
+      "step": 21539
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981194690265487e-06,
+      "loss": 3.4038,
+      "step": 21540
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9811472819216183e-06,
+      "loss": 3.0595,
+      "step": 21541
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9810998735777496e-06,
+      "loss": 1.8633,
+      "step": 21542
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9810524652338813e-06,
+      "loss": 2.3161,
+      "step": 21543
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.981005056890013e-06,
+      "loss": 2.9494,
+      "step": 21544
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9809576485461443e-06,
+      "loss": 3.1599,
+      "step": 21545
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9809102402022755e-06,
+      "loss": 3.0998,
+      "step": 21546
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9808628318584072e-06,
+      "loss": 2.204,
+      "step": 21547
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9808154235145385e-06,
+      "loss": 2.5517,
+      "step": 21548
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.98076801517067e-06,
+      "loss": 2.0652,
+      "step": 21549
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9807206068268015e-06,
+      "loss": 2.4115,
+      "step": 21550
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.980673198482933e-06,
+      "loss": 2.9925,
+      "step": 21551
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9806257901390645e-06,
+      "loss": 3.6397,
+      "step": 21552
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.980578381795196e-06,
+      "loss": 2.4078,
+      "step": 21553
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9805309734513275e-06,
+      "loss": 1.5432,
+      "step": 21554
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9804835651074588e-06,
+      "loss": 2.5648,
+      "step": 21555
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9804361567635905e-06,
+      "loss": 3.9618,
+      "step": 21556
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.980388748419722e-06,
+      "loss": 1.6888,
+      "step": 21557
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9803413400758534e-06,
+      "loss": 1.6211,
+      "step": 21558
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.980293931731985e-06,
+      "loss": 2.8391,
+      "step": 21559
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9802465233881164e-06,
+      "loss": 2.2584,
+      "step": 21560
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9801991150442477e-06,
+      "loss": 2.6167,
+      "step": 21561
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9801517067003794e-06,
+      "loss": 2.9327,
+      "step": 21562
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.9801042983565107e-06,
+      "loss": 1.9783,
+      "step": 21563
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9800568900126424e-06,
+      "loss": 1.3223,
+      "step": 21564
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9800094816687737e-06,
+      "loss": 3.7632,
+      "step": 21565
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9799620733249054e-06,
+      "loss": 2.9138,
+      "step": 21566
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9799146649810367e-06,
+      "loss": 3.0159,
+      "step": 21567
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979867256637168e-06,
+      "loss": 2.6255,
+      "step": 21568
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9798198482932996e-06,
+      "loss": 2.3653,
+      "step": 21569
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979772439949431e-06,
+      "loss": 1.9483,
+      "step": 21570
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9797250316055626e-06,
+      "loss": 3.0163,
+      "step": 21571
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9796776232616943e-06,
+      "loss": 3.055,
+      "step": 21572
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9796302149178256e-06,
+      "loss": 3.8013,
+      "step": 21573
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979582806573957e-06,
+      "loss": 1.8577,
+      "step": 21574
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9795353982300886e-06,
+      "loss": 3.4835,
+      "step": 21575
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.97948798988622e-06,
+      "loss": 2.6093,
+      "step": 21576
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9794405815423516e-06,
+      "loss": 2.8838,
+      "step": 21577
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9793931731984833e-06,
+      "loss": 2.547,
+      "step": 21578
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9793457648546146e-06,
+      "loss": 2.0766,
+      "step": 21579
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979298356510746e-06,
+      "loss": 3.0921,
+      "step": 21580
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9792509481668776e-06,
+      "loss": 2.2089,
+      "step": 21581
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979203539823009e-06,
+      "loss": 3.0129,
+      "step": 21582
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.97915613147914e-06,
+      "loss": 1.8913,
+      "step": 21583
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979108723135272e-06,
+      "loss": 3.1235,
+      "step": 21584
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9790613147914035e-06,
+      "loss": 2.6759,
+      "step": 21585
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.979013906447535e-06,
+      "loss": 1.4571,
+      "step": 21586
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978966498103666e-06,
+      "loss": 2.9962,
+      "step": 21587
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9789190897597978e-06,
+      "loss": 2.3291,
+      "step": 21588
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978871681415929e-06,
+      "loss": 2.1087,
+      "step": 21589
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9788242730720608e-06,
+      "loss": 2.0666,
+      "step": 21590
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9787768647281925e-06,
+      "loss": 2.839,
+      "step": 21591
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9787294563843238e-06,
+      "loss": 2.9328,
+      "step": 21592
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978682048040455e-06,
+      "loss": 3.8986,
+      "step": 21593
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9786346396965867e-06,
+      "loss": 3.0524,
+      "step": 21594
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978587231352718e-06,
+      "loss": 1.7891,
+      "step": 21595
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9785398230088493e-06,
+      "loss": 2.9485,
+      "step": 21596
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978492414664981e-06,
+      "loss": 2.1304,
+      "step": 21597
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9784450063211127e-06,
+      "loss": 3.4684,
+      "step": 21598
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978397597977244e-06,
+      "loss": 2.0889,
+      "step": 21599
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9783501896333757e-06,
+      "loss": 3.6082,
+      "step": 21600
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978302781289507e-06,
+      "loss": 3.3368,
+      "step": 21601
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9782553729456383e-06,
+      "loss": 2.8595,
+      "step": 21602
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.97820796460177e-06,
+      "loss": 2.8931,
+      "step": 21603
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9781605562579017e-06,
+      "loss": 3.3184,
+      "step": 21604
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978113147914033e-06,
+      "loss": 2.4445,
+      "step": 21605
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9780657395701642e-06,
+      "loss": 2.3493,
+      "step": 21606
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.978018331226296e-06,
+      "loss": 2.4385,
+      "step": 21607
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977970922882427e-06,
+      "loss": 2.7815,
+      "step": 21608
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9779235145385585e-06,
+      "loss": 3.236,
+      "step": 21609
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.97787610619469e-06,
+      "loss": 2.6228,
+      "step": 21610
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977828697850822e-06,
+      "loss": 3.6199,
+      "step": 21611
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977781289506953e-06,
+      "loss": 3.1833,
+      "step": 21612
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977733881163085e-06,
+      "loss": 2.977,
+      "step": 21613
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977686472819216e-06,
+      "loss": 1.7048,
+      "step": 21614
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9776390644753474e-06,
+      "loss": 2.735,
+      "step": 21615
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977591656131479e-06,
+      "loss": 2.5986,
+      "step": 21616
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977544247787611e-06,
+      "loss": 3.366,
+      "step": 21617
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977496839443742e-06,
+      "loss": 2.3143,
+      "step": 21618
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977449431099874e-06,
+      "loss": 2.0088,
+      "step": 21619
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977402022756005e-06,
+      "loss": 2.4742,
+      "step": 21620
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9773546144121364e-06,
+      "loss": 1.7372,
+      "step": 21621
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977307206068268e-06,
+      "loss": 3.0634,
+      "step": 21622
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9772597977243994e-06,
+      "loss": 3.0585,
+      "step": 21623
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977212389380531e-06,
+      "loss": 2.6634,
+      "step": 21624
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9771649810366628e-06,
+      "loss": 2.6087,
+      "step": 21625
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.977117572692794e-06,
+      "loss": 3.1299,
+      "step": 21626
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.9770701643489253e-06,
+      "loss": 3.1894,
+      "step": 21627
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9770227560050566e-06,
+      "loss": 2.2278,
+      "step": 21628
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9769753476611883e-06,
+      "loss": 2.5128,
+      "step": 21629
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.97692793931732e-06,
+      "loss": 2.6111,
+      "step": 21630
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9768805309734513e-06,
+      "loss": 2.763,
+      "step": 21631
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.976833122629583e-06,
+      "loss": 2.091,
+      "step": 21632
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9767857142857143e-06,
+      "loss": 3.2431,
+      "step": 21633
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9767383059418456e-06,
+      "loss": 2.8968,
+      "step": 21634
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9766908975979773e-06,
+      "loss": 2.6467,
+      "step": 21635
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9766434892541086e-06,
+      "loss": 1.3365,
+      "step": 21636
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9765960809102403e-06,
+      "loss": 2.3493,
+      "step": 21637
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.976548672566372e-06,
+      "loss": 2.6908,
+      "step": 21638
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9765012642225032e-06,
+      "loss": 2.1043,
+      "step": 21639
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9764538558786345e-06,
+      "loss": 1.5924,
+      "step": 21640
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9764064475347662e-06,
+      "loss": 3.4097,
+      "step": 21641
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9763590391908975e-06,
+      "loss": 2.5742,
+      "step": 21642
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9763116308470292e-06,
+      "loss": 2.2506,
+      "step": 21643
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.976264222503161e-06,
+      "loss": 3.8083,
+      "step": 21644
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.976216814159292e-06,
+      "loss": 1.6644,
+      "step": 21645
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9761694058154235e-06,
+      "loss": 2.9692,
+      "step": 21646
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.976121997471555e-06,
+      "loss": 1.3808,
+      "step": 21647
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9760745891276865e-06,
+      "loss": 2.4976,
+      "step": 21648
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9760271807838177e-06,
+      "loss": 3.9814,
+      "step": 21649
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9759797724399495e-06,
+      "loss": 2.7542,
+      "step": 21650
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.975932364096081e-06,
+      "loss": 1.6543,
+      "step": 21651
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9758849557522124e-06,
+      "loss": 2.5705,
+      "step": 21652
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9758375474083437e-06,
+      "loss": 2.6153,
+      "step": 21653
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9757901390644754e-06,
+      "loss": 2.8789,
+      "step": 21654
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9757427307206067e-06,
+      "loss": 3.4321,
+      "step": 21655
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9756953223767384e-06,
+      "loss": 3.1564,
+      "step": 21656
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.97564791403287e-06,
+      "loss": 1.7154,
+      "step": 21657
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9756005056890014e-06,
+      "loss": 2.9203,
+      "step": 21658
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9755530973451327e-06,
+      "loss": 2.1946,
+      "step": 21659
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9755056890012644e-06,
+      "loss": 3.0376,
+      "step": 21660
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9754582806573957e-06,
+      "loss": 3.4589,
+      "step": 21661
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.975410872313527e-06,
+      "loss": 1.5216,
+      "step": 21662
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9753634639696586e-06,
+      "loss": 3.2195,
+      "step": 21663
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9753160556257903e-06,
+      "loss": 2.5753,
+      "step": 21664
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9752686472819216e-06,
+      "loss": 4.0412,
+      "step": 21665
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9752212389380533e-06,
+      "loss": 3.182,
+      "step": 21666
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9751738305941846e-06,
+      "loss": 2.9843,
+      "step": 21667
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.975126422250316e-06,
+      "loss": 2.2535,
+      "step": 21668
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9750790139064476e-06,
+      "loss": 3.2862,
+      "step": 21669
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9750316055625793e-06,
+      "loss": 1.5417,
+      "step": 21670
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9749841972187106e-06,
+      "loss": 1.6375,
+      "step": 21671
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974936788874842e-06,
+      "loss": 2.3013,
+      "step": 21672
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9748893805309736e-06,
+      "loss": 2.9205,
+      "step": 21673
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974841972187105e-06,
+      "loss": 1.6529,
+      "step": 21674
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974794563843236e-06,
+      "loss": 3.8733,
+      "step": 21675
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974747155499368e-06,
+      "loss": 3.6895,
+      "step": 21676
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9746997471554995e-06,
+      "loss": 3.3532,
+      "step": 21677
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974652338811631e-06,
+      "loss": 3.1759,
+      "step": 21678
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9746049304677625e-06,
+      "loss": 2.0789,
+      "step": 21679
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974557522123894e-06,
+      "loss": 1.8697,
+      "step": 21680
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974510113780025e-06,
+      "loss": 2.0648,
+      "step": 21681
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9744627054361568e-06,
+      "loss": 2.2483,
+      "step": 21682
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9744152970922885e-06,
+      "loss": 2.6491,
+      "step": 21683
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9743678887484198e-06,
+      "loss": 2.4203,
+      "step": 21684
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9743204804045515e-06,
+      "loss": 2.6012,
+      "step": 21685
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9742730720606827e-06,
+      "loss": 1.9354,
+      "step": 21686
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974225663716814e-06,
+      "loss": 2.6098,
+      "step": 21687
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9741782553729457e-06,
+      "loss": 3.5066,
+      "step": 21688
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.974130847029077e-06,
+      "loss": 3.1606,
+      "step": 21689
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.9740834386852087e-06,
+      "loss": 2.89,
+      "step": 21690
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9740360303413404e-06,
+      "loss": 2.1223,
+      "step": 21691
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9739886219974717e-06,
+      "loss": 3.1559,
+      "step": 21692
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973941213653603e-06,
+      "loss": 3.3358,
+      "step": 21693
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9738938053097343e-06,
+      "loss": 3.5059,
+      "step": 21694
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973846396965866e-06,
+      "loss": 2.8755,
+      "step": 21695
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9737989886219977e-06,
+      "loss": 2.4462,
+      "step": 21696
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973751580278129e-06,
+      "loss": 3.3929,
+      "step": 21697
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9737041719342607e-06,
+      "loss": 1.5548,
+      "step": 21698
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973656763590392e-06,
+      "loss": 2.3193,
+      "step": 21699
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9736093552465232e-06,
+      "loss": 2.3583,
+      "step": 21700
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973561946902655e-06,
+      "loss": 2.0786,
+      "step": 21701
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973514538558786e-06,
+      "loss": 2.8068,
+      "step": 21702
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973467130214918e-06,
+      "loss": 2.9147,
+      "step": 21703
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9734197218710496e-06,
+      "loss": 3.2331,
+      "step": 21704
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973372313527181e-06,
+      "loss": 1.7798,
+      "step": 21705
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973324905183312e-06,
+      "loss": 2.0146,
+      "step": 21706
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973277496839444e-06,
+      "loss": 3.2676,
+      "step": 21707
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973230088495575e-06,
+      "loss": 3.4411,
+      "step": 21708
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973182680151707e-06,
+      "loss": 2.5757,
+      "step": 21709
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9731352718078386e-06,
+      "loss": 3.0756,
+      "step": 21710
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.97308786346397e-06,
+      "loss": 3.1634,
+      "step": 21711
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.973040455120101e-06,
+      "loss": 2.9046,
+      "step": 21712
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972993046776233e-06,
+      "loss": 2.7665,
+      "step": 21713
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972945638432364e-06,
+      "loss": 2.6516,
+      "step": 21714
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9728982300884954e-06,
+      "loss": 2.702,
+      "step": 21715
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972850821744627e-06,
+      "loss": 2.6937,
+      "step": 21716
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972803413400759e-06,
+      "loss": 1.8971,
+      "step": 21717
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.97275600505689e-06,
+      "loss": 2.5694,
+      "step": 21718
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9727085967130214e-06,
+      "loss": 3.4285,
+      "step": 21719
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972661188369153e-06,
+      "loss": 3.0401,
+      "step": 21720
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9726137800252843e-06,
+      "loss": 2.7532,
+      "step": 21721
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972566371681416e-06,
+      "loss": 3.2458,
+      "step": 21722
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9725189633375477e-06,
+      "loss": 2.6366,
+      "step": 21723
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972471554993679e-06,
+      "loss": 2.7345,
+      "step": 21724
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9724241466498103e-06,
+      "loss": 2.748,
+      "step": 21725
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972376738305942e-06,
+      "loss": 2.1341,
+      "step": 21726
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9723293299620733e-06,
+      "loss": 2.3557,
+      "step": 21727
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9722819216182046e-06,
+      "loss": 3.9255,
+      "step": 21728
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9722345132743363e-06,
+      "loss": 2.2686,
+      "step": 21729
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972187104930468e-06,
+      "loss": 1.9109,
+      "step": 21730
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9721396965865993e-06,
+      "loss": 3.51,
+      "step": 21731
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.972092288242731e-06,
+      "loss": 2.4711,
+      "step": 21732
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9720448798988622e-06,
+      "loss": 3.2994,
+      "step": 21733
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9719974715549935e-06,
+      "loss": 2.8218,
+      "step": 21734
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9719500632111252e-06,
+      "loss": 2.4759,
+      "step": 21735
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.971902654867257e-06,
+      "loss": 2.25,
+      "step": 21736
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9718552465233882e-06,
+      "loss": 3.6624,
+      "step": 21737
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9718078381795195e-06,
+      "loss": 3.0217,
+      "step": 21738
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.971760429835651e-06,
+      "loss": 2.6349,
+      "step": 21739
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9717130214917825e-06,
+      "loss": 2.77,
+      "step": 21740
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9716656131479138e-06,
+      "loss": 1.9395,
+      "step": 21741
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9716182048040455e-06,
+      "loss": 1.4577,
+      "step": 21742
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.971570796460177e-06,
+      "loss": 4.0711,
+      "step": 21743
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9715233881163084e-06,
+      "loss": 1.7481,
+      "step": 21744
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.97147597977244e-06,
+      "loss": 1.681,
+      "step": 21745
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9714285714285714e-06,
+      "loss": 3.0794,
+      "step": 21746
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9713811630847027e-06,
+      "loss": 2.8417,
+      "step": 21747
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9713337547408344e-06,
+      "loss": 1.8833,
+      "step": 21748
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.971286346396966e-06,
+      "loss": 3.0323,
+      "step": 21749
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9712389380530974e-06,
+      "loss": 2.3401,
+      "step": 21750
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.971191529709229e-06,
+      "loss": 2.0256,
+      "step": 21751
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9711441213653604e-06,
+      "loss": 2.8115,
+      "step": 21752
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.9710967130214917e-06,
+      "loss": 2.2828,
+      "step": 21753
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9710493046776234e-06,
+      "loss": 3.2018,
+      "step": 21754
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9710018963337546e-06,
+      "loss": 2.7237,
+      "step": 21755
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9709544879898864e-06,
+      "loss": 2.34,
+      "step": 21756
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.970907079646018e-06,
+      "loss": 3.1446,
+      "step": 21757
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9708596713021493e-06,
+      "loss": 2.6003,
+      "step": 21758
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9708122629582806e-06,
+      "loss": 2.5204,
+      "step": 21759
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.970764854614412e-06,
+      "loss": 2.4765,
+      "step": 21760
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9707174462705436e-06,
+      "loss": 2.2055,
+      "step": 21761
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9706700379266753e-06,
+      "loss": 2.5558,
+      "step": 21762
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9706226295828066e-06,
+      "loss": 3.256,
+      "step": 21763
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9705752212389383e-06,
+      "loss": 2.306,
+      "step": 21764
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9705278128950696e-06,
+      "loss": 3.6617,
+      "step": 21765
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.970480404551201e-06,
+      "loss": 3.0172,
+      "step": 21766
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9704329962073326e-06,
+      "loss": 4.0181,
+      "step": 21767
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.970385587863464e-06,
+      "loss": 3.0121,
+      "step": 21768
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9703381795195955e-06,
+      "loss": 2.0946,
+      "step": 21769
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9702907711757272e-06,
+      "loss": 2.1892,
+      "step": 21770
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9702433628318585e-06,
+      "loss": 3.2884,
+      "step": 21771
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.97019595448799e-06,
+      "loss": 2.9931,
+      "step": 21772
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9701485461441215e-06,
+      "loss": 2.8839,
+      "step": 21773
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9701011378002528e-06,
+      "loss": 3.0367,
+      "step": 21774
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9700537294563845e-06,
+      "loss": 2.2142,
+      "step": 21775
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.970006321112516e-06,
+      "loss": 2.9849,
+      "step": 21776
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9699589127686475e-06,
+      "loss": 2.1307,
+      "step": 21777
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9699115044247788e-06,
+      "loss": 2.8868,
+      "step": 21778
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9698640960809105e-06,
+      "loss": 3.0555,
+      "step": 21779
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9698166877370417e-06,
+      "loss": 1.9037,
+      "step": 21780
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969769279393173e-06,
+      "loss": 2.7952,
+      "step": 21781
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9697218710493047e-06,
+      "loss": 2.0473,
+      "step": 21782
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9696744627054364e-06,
+      "loss": 1.5907,
+      "step": 21783
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9696270543615677e-06,
+      "loss": 2.5503,
+      "step": 21784
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969579646017699e-06,
+      "loss": 2.9149,
+      "step": 21785
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9695322376738307e-06,
+      "loss": 3.0246,
+      "step": 21786
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969484829329962e-06,
+      "loss": 2.9466,
+      "step": 21787
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9694374209860937e-06,
+      "loss": 2.6012,
+      "step": 21788
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9693900126422254e-06,
+      "loss": 2.2412,
+      "step": 21789
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9693426042983567e-06,
+      "loss": 2.3721,
+      "step": 21790
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969295195954488e-06,
+      "loss": 2.4736,
+      "step": 21791
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9692477876106196e-06,
+      "loss": 3.3742,
+      "step": 21792
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969200379266751e-06,
+      "loss": 3.0587,
+      "step": 21793
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969152970922882e-06,
+      "loss": 1.3222,
+      "step": 21794
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969105562579014e-06,
+      "loss": 3.3563,
+      "step": 21795
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9690581542351456e-06,
+      "loss": 2.2801,
+      "step": 21796
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.969010745891277e-06,
+      "loss": 3.5081,
+      "step": 21797
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9689633375474086e-06,
+      "loss": 3.1147,
+      "step": 21798
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.96891592920354e-06,
+      "loss": 3.756,
+      "step": 21799
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968868520859671e-06,
+      "loss": 2.0521,
+      "step": 21800
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968821112515803e-06,
+      "loss": 2.7629,
+      "step": 21801
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9687737041719346e-06,
+      "loss": 3.5333,
+      "step": 21802
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968726295828066e-06,
+      "loss": 2.9043,
+      "step": 21803
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968678887484197e-06,
+      "loss": 2.196,
+      "step": 21804
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968631479140329e-06,
+      "loss": 2.3043,
+      "step": 21805
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.96858407079646e-06,
+      "loss": 1.4489,
+      "step": 21806
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9685366624525914e-06,
+      "loss": 2.3991,
+      "step": 21807
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968489254108723e-06,
+      "loss": 3.0745,
+      "step": 21808
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968441845764855e-06,
+      "loss": 3.6925,
+      "step": 21809
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968394437420986e-06,
+      "loss": 2.4646,
+      "step": 21810
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9683470290771178e-06,
+      "loss": 3.3933,
+      "step": 21811
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968299620733249e-06,
+      "loss": 2.8427,
+      "step": 21812
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9682522123893803e-06,
+      "loss": 2.3006,
+      "step": 21813
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968204804045512e-06,
+      "loss": 3.3891,
+      "step": 21814
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9681573957016438e-06,
+      "loss": 3.4004,
+      "step": 21815
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.968109987357775e-06,
+      "loss": 3.1197,
+      "step": 21816
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.9680625790139067e-06,
+      "loss": 3.6341,
+      "step": 21817
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.968015170670038e-06,
+      "loss": 3.4607,
+      "step": 21818
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9679677623261693e-06,
+      "loss": 2.6327,
+      "step": 21819
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967920353982301e-06,
+      "loss": 3.9451,
+      "step": 21820
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9678729456384323e-06,
+      "loss": 2.8881,
+      "step": 21821
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967825537294564e-06,
+      "loss": 2.1323,
+      "step": 21822
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9677781289506957e-06,
+      "loss": 1.8663,
+      "step": 21823
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967730720606827e-06,
+      "loss": 2.2715,
+      "step": 21824
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9676833122629583e-06,
+      "loss": 2.7738,
+      "step": 21825
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9676359039190895e-06,
+      "loss": 3.0418,
+      "step": 21826
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9675884955752212e-06,
+      "loss": 2.2841,
+      "step": 21827
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967541087231353e-06,
+      "loss": 2.5847,
+      "step": 21828
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9674936788874842e-06,
+      "loss": 2.1175,
+      "step": 21829
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967446270543616e-06,
+      "loss": 2.2941,
+      "step": 21830
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967398862199747e-06,
+      "loss": 2.0173,
+      "step": 21831
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9673514538558785e-06,
+      "loss": 2.8571,
+      "step": 21832
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.96730404551201e-06,
+      "loss": 3.669,
+      "step": 21833
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9672566371681415e-06,
+      "loss": 1.5491,
+      "step": 21834
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967209228824273e-06,
+      "loss": 2.415,
+      "step": 21835
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967161820480405e-06,
+      "loss": 3.2961,
+      "step": 21836
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967114412136536e-06,
+      "loss": 2.8857,
+      "step": 21837
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9670670037926674e-06,
+      "loss": 2.6949,
+      "step": 21838
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.967019595448799e-06,
+      "loss": 4.1615,
+      "step": 21839
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9669721871049304e-06,
+      "loss": 2.2553,
+      "step": 21840
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9669247787610617e-06,
+      "loss": 2.3714,
+      "step": 21841
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.966877370417194e-06,
+      "loss": 1.6766,
+      "step": 21842
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.966829962073325e-06,
+      "loss": 2.9729,
+      "step": 21843
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9667825537294564e-06,
+      "loss": 2.1976,
+      "step": 21844
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.966735145385588e-06,
+      "loss": 1.982,
+      "step": 21845
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9666877370417194e-06,
+      "loss": 3.2242,
+      "step": 21846
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9666403286978507e-06,
+      "loss": 1.6034,
+      "step": 21847
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9665929203539824e-06,
+      "loss": 2.6874,
+      "step": 21848
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.966545512010114e-06,
+      "loss": 2.2966,
+      "step": 21849
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9664981036662453e-06,
+      "loss": 2.357,
+      "step": 21850
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9664506953223766e-06,
+      "loss": 2.7178,
+      "step": 21851
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9664032869785083e-06,
+      "loss": 1.9791,
+      "step": 21852
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9663558786346396e-06,
+      "loss": 3.1743,
+      "step": 21853
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.966308470290771e-06,
+      "loss": 2.5728,
+      "step": 21854
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9662610619469026e-06,
+      "loss": 3.2662,
+      "step": 21855
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9662136536030343e-06,
+      "loss": 2.6457,
+      "step": 21856
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9661662452591656e-06,
+      "loss": 2.9976,
+      "step": 21857
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9661188369152973e-06,
+      "loss": 3.2015,
+      "step": 21858
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9660714285714286e-06,
+      "loss": 3.2997,
+      "step": 21859
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.96602402022756e-06,
+      "loss": 2.7852,
+      "step": 21860
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9659766118836915e-06,
+      "loss": 2.7858,
+      "step": 21861
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9659292035398233e-06,
+      "loss": 2.8835,
+      "step": 21862
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9658817951959545e-06,
+      "loss": 3.4887,
+      "step": 21863
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9658343868520862e-06,
+      "loss": 2.5159,
+      "step": 21864
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9657869785082175e-06,
+      "loss": 3.2727,
+      "step": 21865
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.965739570164349e-06,
+      "loss": 1.8469,
+      "step": 21866
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9656921618204805e-06,
+      "loss": 2.5294,
+      "step": 21867
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9656447534766118e-06,
+      "loss": 2.8768,
+      "step": 21868
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9655973451327435e-06,
+      "loss": 2.6091,
+      "step": 21869
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9655499367888748e-06,
+      "loss": 1.305,
+      "step": 21870
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9655025284450065e-06,
+      "loss": 1.8446,
+      "step": 21871
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9654551201011378e-06,
+      "loss": 2.6985,
+      "step": 21872
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.965407711757269e-06,
+      "loss": 2.8169,
+      "step": 21873
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9653603034134007e-06,
+      "loss": 2.641,
+      "step": 21874
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9653128950695324e-06,
+      "loss": 2.6989,
+      "step": 21875
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9652654867256637e-06,
+      "loss": 2.2918,
+      "step": 21876
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9652180783817954e-06,
+      "loss": 2.9058,
+      "step": 21877
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9651706700379267e-06,
+      "loss": 2.3274,
+      "step": 21878
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.965123261694058e-06,
+      "loss": 4.1745,
+      "step": 21879
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.9650758533501897e-06,
+      "loss": 2.5996,
+      "step": 21880
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.965028445006321e-06,
+      "loss": 3.597,
+      "step": 21881
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9649810366624527e-06,
+      "loss": 3.4325,
+      "step": 21882
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9649336283185844e-06,
+      "loss": 2.972,
+      "step": 21883
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9648862199747157e-06,
+      "loss": 3.6037,
+      "step": 21884
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964838811630847e-06,
+      "loss": 0.8378,
+      "step": 21885
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9647914032869786e-06,
+      "loss": 2.9116,
+      "step": 21886
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.96474399494311e-06,
+      "loss": 2.0268,
+      "step": 21887
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9646965865992416e-06,
+      "loss": 1.2859,
+      "step": 21888
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9646491782553733e-06,
+      "loss": 3.3878,
+      "step": 21889
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9646017699115046e-06,
+      "loss": 2.6573,
+      "step": 21890
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964554361567636e-06,
+      "loss": 3.0995,
+      "step": 21891
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964506953223767e-06,
+      "loss": 2.5123,
+      "step": 21892
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964459544879899e-06,
+      "loss": 2.4877,
+      "step": 21893
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.96441213653603e-06,
+      "loss": 1.8846,
+      "step": 21894
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964364728192162e-06,
+      "loss": 2.6104,
+      "step": 21895
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9643173198482936e-06,
+      "loss": 2.4033,
+      "step": 21896
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964269911504425e-06,
+      "loss": 2.4245,
+      "step": 21897
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964222503160556e-06,
+      "loss": 2.3652,
+      "step": 21898
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964175094816688e-06,
+      "loss": 3.973,
+      "step": 21899
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964127686472819e-06,
+      "loss": 2.1432,
+      "step": 21900
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.964080278128951e-06,
+      "loss": 4.3332,
+      "step": 21901
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9640328697850825e-06,
+      "loss": 2.8069,
+      "step": 21902
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963985461441214e-06,
+      "loss": 2.5926,
+      "step": 21903
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963938053097345e-06,
+      "loss": 0.8135,
+      "step": 21904
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9638906447534768e-06,
+      "loss": 2.5534,
+      "step": 21905
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963843236409608e-06,
+      "loss": 2.9977,
+      "step": 21906
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9637958280657393e-06,
+      "loss": 2.4794,
+      "step": 21907
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963748419721871e-06,
+      "loss": 1.6003,
+      "step": 21908
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9637010113780027e-06,
+      "loss": 3.5355,
+      "step": 21909
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963653603034134e-06,
+      "loss": 1.9066,
+      "step": 21910
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9636061946902657e-06,
+      "loss": 2.6668,
+      "step": 21911
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963558786346397e-06,
+      "loss": 3.1819,
+      "step": 21912
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9635113780025283e-06,
+      "loss": 2.1766,
+      "step": 21913
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.96346396965866e-06,
+      "loss": 2.0158,
+      "step": 21914
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9634165613147917e-06,
+      "loss": 2.9815,
+      "step": 21915
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963369152970923e-06,
+      "loss": 2.7935,
+      "step": 21916
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9633217446270543e-06,
+      "loss": 2.8903,
+      "step": 21917
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963274336283186e-06,
+      "loss": 2.8413,
+      "step": 21918
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9632269279393172e-06,
+      "loss": 2.0906,
+      "step": 21919
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9631795195954485e-06,
+      "loss": 2.1497,
+      "step": 21920
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9631321112515802e-06,
+      "loss": 3.368,
+      "step": 21921
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.963084702907712e-06,
+      "loss": 2.5888,
+      "step": 21922
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9630372945638432e-06,
+      "loss": 2.6848,
+      "step": 21923
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962989886219975e-06,
+      "loss": 2.8053,
+      "step": 21924
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962942477876106e-06,
+      "loss": 2.7547,
+      "step": 21925
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9628950695322375e-06,
+      "loss": 2.6694,
+      "step": 21926
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962847661188369e-06,
+      "loss": 2.541,
+      "step": 21927
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962800252844501e-06,
+      "loss": 1.9512,
+      "step": 21928
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962752844500632e-06,
+      "loss": 3.0391,
+      "step": 21929
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962705436156764e-06,
+      "loss": 2.6701,
+      "step": 21930
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962658027812895e-06,
+      "loss": 3.1195,
+      "step": 21931
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9626106194690264e-06,
+      "loss": 2.3671,
+      "step": 21932
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962563211125158e-06,
+      "loss": 1.8663,
+      "step": 21933
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9625158027812894e-06,
+      "loss": 1.7455,
+      "step": 21934
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962468394437421e-06,
+      "loss": 1.9917,
+      "step": 21935
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9624209860935524e-06,
+      "loss": 3.8448,
+      "step": 21936
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962373577749684e-06,
+      "loss": 4.1464,
+      "step": 21937
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9623261694058154e-06,
+      "loss": 1.4096,
+      "step": 21938
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9622787610619467e-06,
+      "loss": 2.9234,
+      "step": 21939
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9622313527180784e-06,
+      "loss": 1.9143,
+      "step": 21940
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.96218394437421e-06,
+      "loss": 2.6677,
+      "step": 21941
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.9621365360303414e-06,
+      "loss": 1.3565,
+      "step": 21942
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.962089127686473e-06,
+      "loss": 2.8145,
+      "step": 21943
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9620417193426043e-06,
+      "loss": 2.264,
+      "step": 21944
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9619943109987356e-06,
+      "loss": 2.451,
+      "step": 21945
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9619469026548673e-06,
+      "loss": 2.1485,
+      "step": 21946
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9618994943109986e-06,
+      "loss": 3.2716,
+      "step": 21947
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9618520859671303e-06,
+      "loss": 1.8461,
+      "step": 21948
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.961804677623262e-06,
+      "loss": 2.953,
+      "step": 21949
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9617572692793933e-06,
+      "loss": 3.3034,
+      "step": 21950
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9617098609355246e-06,
+      "loss": 1.7129,
+      "step": 21951
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9616624525916563e-06,
+      "loss": 2.7574,
+      "step": 21952
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9616150442477876e-06,
+      "loss": 2.4417,
+      "step": 21953
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9615676359039193e-06,
+      "loss": 2.6299,
+      "step": 21954
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.961520227560051e-06,
+      "loss": 2.5687,
+      "step": 21955
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9614728192161822e-06,
+      "loss": 2.2791,
+      "step": 21956
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9614254108723135e-06,
+      "loss": 2.5631,
+      "step": 21957
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.961378002528445e-06,
+      "loss": 3.3916,
+      "step": 21958
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9613305941845765e-06,
+      "loss": 1.8377,
+      "step": 21959
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.961283185840708e-06,
+      "loss": 3.1289,
+      "step": 21960
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9612357774968395e-06,
+      "loss": 2.2148,
+      "step": 21961
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.961188369152971e-06,
+      "loss": 2.4267,
+      "step": 21962
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9611409608091025e-06,
+      "loss": 3.338,
+      "step": 21963
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9610935524652338e-06,
+      "loss": 2.4173,
+      "step": 21964
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9610461441213655e-06,
+      "loss": 1.783,
+      "step": 21965
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9609987357774967e-06,
+      "loss": 2.5784,
+      "step": 21966
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9609513274336284e-06,
+      "loss": 1.8855,
+      "step": 21967
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.96090391908976e-06,
+      "loss": 3.0866,
+      "step": 21968
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9608565107458914e-06,
+      "loss": 3.1117,
+      "step": 21969
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9608091024020227e-06,
+      "loss": 2.3409,
+      "step": 21970
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9607616940581544e-06,
+      "loss": 2.3303,
+      "step": 21971
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9607142857142857e-06,
+      "loss": 2.2302,
+      "step": 21972
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.960666877370417e-06,
+      "loss": 2.4483,
+      "step": 21973
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9606194690265487e-06,
+      "loss": 2.8191,
+      "step": 21974
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9605720606826804e-06,
+      "loss": 2.5812,
+      "step": 21975
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9605246523388117e-06,
+      "loss": 2.1644,
+      "step": 21976
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9604772439949434e-06,
+      "loss": 4.1414,
+      "step": 21977
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9604298356510747e-06,
+      "loss": 2.7423,
+      "step": 21978
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.960382427307206e-06,
+      "loss": 3.102,
+      "step": 21979
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9603350189633376e-06,
+      "loss": 2.6227,
+      "step": 21980
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9602876106194693e-06,
+      "loss": 3.5542,
+      "step": 21981
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9602402022756006e-06,
+      "loss": 2.1217,
+      "step": 21982
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.960192793931732e-06,
+      "loss": 1.2565,
+      "step": 21983
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9601453855878636e-06,
+      "loss": 2.594,
+      "step": 21984
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.960097977243995e-06,
+      "loss": 3.374,
+      "step": 21985
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.960050568900126e-06,
+      "loss": 3.3302,
+      "step": 21986
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.960003160556258e-06,
+      "loss": 1.9702,
+      "step": 21987
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9599557522123896e-06,
+      "loss": 3.1718,
+      "step": 21988
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959908343868521e-06,
+      "loss": 2.9741,
+      "step": 21989
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9598609355246526e-06,
+      "loss": 2.7423,
+      "step": 21990
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959813527180784e-06,
+      "loss": 2.5354,
+      "step": 21991
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959766118836915e-06,
+      "loss": 2.4308,
+      "step": 21992
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959718710493047e-06,
+      "loss": 2.4426,
+      "step": 21993
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9596713021491785e-06,
+      "loss": 2.1658,
+      "step": 21994
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.95962389380531e-06,
+      "loss": 3.6567,
+      "step": 21995
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9595764854614415e-06,
+      "loss": 3.0164,
+      "step": 21996
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959529077117573e-06,
+      "loss": 3.0504,
+      "step": 21997
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959481668773704e-06,
+      "loss": 2.9977,
+      "step": 21998
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9594342604298358e-06,
+      "loss": 2.4921,
+      "step": 21999
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959386852085967e-06,
+      "loss": 2.6513,
+      "step": 22000
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9593394437420988e-06,
+      "loss": 3.3801,
+      "step": 22001
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.95929203539823e-06,
+      "loss": 2.6082,
+      "step": 22002
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9592446270543617e-06,
+      "loss": 2.4345,
+      "step": 22003
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959197218710493e-06,
+      "loss": 3.1754,
+      "step": 22004
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9591498103666243e-06,
+      "loss": 2.893,
+      "step": 22005
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.959102402022756e-06,
+      "loss": 3.4913,
+      "step": 22006
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.9590549936788877e-06,
+      "loss": 2.1427,
+      "step": 22007
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.959007585335019e-06,
+      "loss": 2.1837,
+      "step": 22008
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9589601769911507e-06,
+      "loss": 3.968,
+      "step": 22009
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958912768647282e-06,
+      "loss": 2.9869,
+      "step": 22010
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9588653603034133e-06,
+      "loss": 3.0513,
+      "step": 22011
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958817951959545e-06,
+      "loss": 2.276,
+      "step": 22012
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9587705436156762e-06,
+      "loss": 3.3033,
+      "step": 22013
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958723135271808e-06,
+      "loss": 3.272,
+      "step": 22014
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9586757269279396e-06,
+      "loss": 2.9822,
+      "step": 22015
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958628318584071e-06,
+      "loss": 3.3179,
+      "step": 22016
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958580910240202e-06,
+      "loss": 2.543,
+      "step": 22017
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958533501896334e-06,
+      "loss": 2.4292,
+      "step": 22018
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958486093552465e-06,
+      "loss": 2.8127,
+      "step": 22019
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958438685208597e-06,
+      "loss": 2.1966,
+      "step": 22020
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9583912768647286e-06,
+      "loss": 2.617,
+      "step": 22021
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.95834386852086e-06,
+      "loss": 3.5384,
+      "step": 22022
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958296460176991e-06,
+      "loss": 3.1,
+      "step": 22023
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9582490518331224e-06,
+      "loss": 3.1704,
+      "step": 22024
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958201643489254e-06,
+      "loss": 2.9504,
+      "step": 22025
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9581542351453854e-06,
+      "loss": 3.2286,
+      "step": 22026
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958106826801517e-06,
+      "loss": 2.0055,
+      "step": 22027
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.958059418457649e-06,
+      "loss": 2.8359,
+      "step": 22028
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.95801201011378e-06,
+      "loss": 2.1493,
+      "step": 22029
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9579646017699114e-06,
+      "loss": 1.7361,
+      "step": 22030
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957917193426043e-06,
+      "loss": 3.091,
+      "step": 22031
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9578697850821744e-06,
+      "loss": 2.5819,
+      "step": 22032
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957822376738306e-06,
+      "loss": 2.0868,
+      "step": 22033
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957774968394438e-06,
+      "loss": 1.9264,
+      "step": 22034
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957727560050569e-06,
+      "loss": 3.5024,
+      "step": 22035
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9576801517067003e-06,
+      "loss": 3.4253,
+      "step": 22036
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957632743362832e-06,
+      "loss": 3.1928,
+      "step": 22037
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9575853350189633e-06,
+      "loss": 2.3047,
+      "step": 22038
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9575379266750946e-06,
+      "loss": 2.634,
+      "step": 22039
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9574905183312263e-06,
+      "loss": 2.0558,
+      "step": 22040
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957443109987358e-06,
+      "loss": 3.0027,
+      "step": 22041
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9573957016434893e-06,
+      "loss": 2.1904,
+      "step": 22042
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9573482932996206e-06,
+      "loss": 2.4408,
+      "step": 22043
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9573008849557523e-06,
+      "loss": 2.003,
+      "step": 22044
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9572534766118836e-06,
+      "loss": 3.0156,
+      "step": 22045
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9572060682680153e-06,
+      "loss": 2.1401,
+      "step": 22046
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.957158659924147e-06,
+      "loss": 3.3501,
+      "step": 22047
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9571112515802783e-06,
+      "loss": 3.569,
+      "step": 22048
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9570638432364095e-06,
+      "loss": 1.9312,
+      "step": 22049
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9570164348925412e-06,
+      "loss": 3.0948,
+      "step": 22050
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9569690265486725e-06,
+      "loss": 3.9821,
+      "step": 22051
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.956921618204804e-06,
+      "loss": 3.2148,
+      "step": 22052
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9568742098609355e-06,
+      "loss": 3.1645,
+      "step": 22053
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.956826801517067e-06,
+      "loss": 1.8456,
+      "step": 22054
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9567793931731985e-06,
+      "loss": 2.5355,
+      "step": 22055
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.95673198482933e-06,
+      "loss": 2.815,
+      "step": 22056
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9566845764854615e-06,
+      "loss": 3.2165,
+      "step": 22057
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9566371681415928e-06,
+      "loss": 3.859,
+      "step": 22058
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9565897597977245e-06,
+      "loss": 1.7762,
+      "step": 22059
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.956542351453856e-06,
+      "loss": 3.2597,
+      "step": 22060
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9564949431099874e-06,
+      "loss": 3.2892,
+      "step": 22061
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.956447534766119e-06,
+      "loss": 2.5236,
+      "step": 22062
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9564001264222504e-06,
+      "loss": 2.5415,
+      "step": 22063
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9563527180783817e-06,
+      "loss": 1.3255,
+      "step": 22064
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.956305309734513e-06,
+      "loss": 2.5111,
+      "step": 22065
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9562579013906447e-06,
+      "loss": 2.5352,
+      "step": 22066
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9562104930467764e-06,
+      "loss": 3.3761,
+      "step": 22067
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9561630847029077e-06,
+      "loss": 2.7068,
+      "step": 22068
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9561156763590394e-06,
+      "loss": 2.6814,
+      "step": 22069
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.9560682680151707e-06,
+      "loss": 2.1352,
+      "step": 22070
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.956020859671302e-06,
+      "loss": 2.9995,
+      "step": 22071
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9559734513274336e-06,
+      "loss": 2.7721,
+      "step": 22072
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9559260429835653e-06,
+      "loss": 2.2657,
+      "step": 22073
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9558786346396966e-06,
+      "loss": 2.9139,
+      "step": 22074
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9558312262958283e-06,
+      "loss": 3.2614,
+      "step": 22075
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9557838179519596e-06,
+      "loss": 2.5029,
+      "step": 22076
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.955736409608091e-06,
+      "loss": 3.7249,
+      "step": 22077
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9556890012642226e-06,
+      "loss": 3.3106,
+      "step": 22078
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.955641592920354e-06,
+      "loss": 2.6253,
+      "step": 22079
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9555941845764856e-06,
+      "loss": 2.6917,
+      "step": 22080
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9555467762326173e-06,
+      "loss": 2.4138,
+      "step": 22081
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9554993678887486e-06,
+      "loss": 1.3472,
+      "step": 22082
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.95545195954488e-06,
+      "loss": 2.8933,
+      "step": 22083
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9554045512010116e-06,
+      "loss": 3.6202,
+      "step": 22084
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.955357142857143e-06,
+      "loss": 3.6304,
+      "step": 22085
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9553097345132745e-06,
+      "loss": 2.3455,
+      "step": 22086
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.955262326169406e-06,
+      "loss": 2.7453,
+      "step": 22087
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9552149178255375e-06,
+      "loss": 3.0219,
+      "step": 22088
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.955167509481669e-06,
+      "loss": 2.3927,
+      "step": 22089
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9551201011378e-06,
+      "loss": 3.9647,
+      "step": 22090
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9550726927939318e-06,
+      "loss": 2.1669,
+      "step": 22091
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.955025284450063e-06,
+      "loss": 2.3915,
+      "step": 22092
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9549778761061948e-06,
+      "loss": 2.4424,
+      "step": 22093
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9549304677623265e-06,
+      "loss": 1.6068,
+      "step": 22094
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9548830594184578e-06,
+      "loss": 3.0767,
+      "step": 22095
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954835651074589e-06,
+      "loss": 2.8778,
+      "step": 22096
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9547882427307207e-06,
+      "loss": 1.6679,
+      "step": 22097
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954740834386852e-06,
+      "loss": 3.0353,
+      "step": 22098
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9546934260429837e-06,
+      "loss": 3.1457,
+      "step": 22099
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9546460176991154e-06,
+      "loss": 2.975,
+      "step": 22100
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9545986093552467e-06,
+      "loss": 3.5288,
+      "step": 22101
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954551201011378e-06,
+      "loss": 3.1738,
+      "step": 22102
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9545037926675097e-06,
+      "loss": 2.373,
+      "step": 22103
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954456384323641e-06,
+      "loss": 3.63,
+      "step": 22104
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9544089759797723e-06,
+      "loss": 2.7157,
+      "step": 22105
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954361567635904e-06,
+      "loss": 2.5379,
+      "step": 22106
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9543141592920357e-06,
+      "loss": 2.5689,
+      "step": 22107
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954266750948167e-06,
+      "loss": 3.8098,
+      "step": 22108
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9542193426042982e-06,
+      "loss": 2.4498,
+      "step": 22109
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.95417193426043e-06,
+      "loss": 3.5383,
+      "step": 22110
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.954124525916561e-06,
+      "loss": 2.6074,
+      "step": 22111
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9540771175726925e-06,
+      "loss": 2.2506,
+      "step": 22112
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9540297092288246e-06,
+      "loss": 2.3493,
+      "step": 22113
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953982300884956e-06,
+      "loss": 3.1806,
+      "step": 22114
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953934892541087e-06,
+      "loss": 2.5533,
+      "step": 22115
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953887484197219e-06,
+      "loss": 3.4567,
+      "step": 22116
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.95384007585335e-06,
+      "loss": 2.5444,
+      "step": 22117
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9537926675094814e-06,
+      "loss": 2.9557,
+      "step": 22118
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953745259165613e-06,
+      "loss": 3.7789,
+      "step": 22119
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953697850821745e-06,
+      "loss": 3.1143,
+      "step": 22120
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953650442477876e-06,
+      "loss": 2.0774,
+      "step": 22121
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953603034134008e-06,
+      "loss": 2.0406,
+      "step": 22122
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953555625790139e-06,
+      "loss": 1.5538,
+      "step": 22123
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9535082174462704e-06,
+      "loss": 3.0605,
+      "step": 22124
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953460809102402e-06,
+      "loss": 2.994,
+      "step": 22125
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9534134007585334e-06,
+      "loss": 3.3273,
+      "step": 22126
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953365992414665e-06,
+      "loss": 1.8968,
+      "step": 22127
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9533185840707968e-06,
+      "loss": 2.7297,
+      "step": 22128
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953271175726928e-06,
+      "loss": 2.8603,
+      "step": 22129
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9532237673830593e-06,
+      "loss": 2.2117,
+      "step": 22130
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9531763590391906e-06,
+      "loss": 2.4847,
+      "step": 22131
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.9531289506953223e-06,
+      "loss": 3.4325,
+      "step": 22132
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.953081542351454e-06,
+      "loss": 2.5385,
+      "step": 22133
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9530341340075853e-06,
+      "loss": 3.7779,
+      "step": 22134
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.952986725663717e-06,
+      "loss": 4.885,
+      "step": 22135
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9529393173198483e-06,
+      "loss": 2.4808,
+      "step": 22136
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9528919089759796e-06,
+      "loss": 2.4336,
+      "step": 22137
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9528445006321113e-06,
+      "loss": 1.9377,
+      "step": 22138
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9527970922882426e-06,
+      "loss": 3.2127,
+      "step": 22139
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9527496839443743e-06,
+      "loss": 3.1158,
+      "step": 22140
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.952702275600506e-06,
+      "loss": 3.963,
+      "step": 22141
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9526548672566372e-06,
+      "loss": 3.2285,
+      "step": 22142
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9526074589127685e-06,
+      "loss": 2.6301,
+      "step": 22143
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9525600505689002e-06,
+      "loss": 1.4188,
+      "step": 22144
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9525126422250315e-06,
+      "loss": 2.435,
+      "step": 22145
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9524652338811632e-06,
+      "loss": 3.3469,
+      "step": 22146
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.952417825537295e-06,
+      "loss": 3.466,
+      "step": 22147
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.952370417193426e-06,
+      "loss": 2.215,
+      "step": 22148
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9523230088495575e-06,
+      "loss": 2.9577,
+      "step": 22149
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.952275600505689e-06,
+      "loss": 3.4969,
+      "step": 22150
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9522281921618205e-06,
+      "loss": 3.1126,
+      "step": 22151
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9521807838179517e-06,
+      "loss": 2.4714,
+      "step": 22152
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9521333754740835e-06,
+      "loss": 2.8188,
+      "step": 22153
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.952085967130215e-06,
+      "loss": 1.9097,
+      "step": 22154
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9520385587863464e-06,
+      "loss": 3.6167,
+      "step": 22155
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9519911504424777e-06,
+      "loss": 2.2177,
+      "step": 22156
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9519437420986094e-06,
+      "loss": 2.0105,
+      "step": 22157
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9518963337547407e-06,
+      "loss": 3.2688,
+      "step": 22158
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9518489254108724e-06,
+      "loss": 2.3857,
+      "step": 22159
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.951801517067004e-06,
+      "loss": 1.9502,
+      "step": 22160
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9517541087231354e-06,
+      "loss": 3.4751,
+      "step": 22161
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9517067003792667e-06,
+      "loss": 3.0886,
+      "step": 22162
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9516592920353984e-06,
+      "loss": 2.6429,
+      "step": 22163
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9516118836915297e-06,
+      "loss": 3.4592,
+      "step": 22164
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.951564475347661e-06,
+      "loss": 2.9964,
+      "step": 22165
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9515170670037926e-06,
+      "loss": 1.9595,
+      "step": 22166
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9514696586599243e-06,
+      "loss": 2.7421,
+      "step": 22167
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9514222503160556e-06,
+      "loss": 3.6632,
+      "step": 22168
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9513748419721873e-06,
+      "loss": 2.2967,
+      "step": 22169
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9513274336283186e-06,
+      "loss": 2.6262,
+      "step": 22170
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.95128002528445e-06,
+      "loss": 2.4799,
+      "step": 22171
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9512326169405816e-06,
+      "loss": 2.4815,
+      "step": 22172
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9511852085967133e-06,
+      "loss": 2.3987,
+      "step": 22173
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9511378002528446e-06,
+      "loss": 1.6517,
+      "step": 22174
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.951090391908976e-06,
+      "loss": 2.423,
+      "step": 22175
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9510429835651076e-06,
+      "loss": 1.8475,
+      "step": 22176
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950995575221239e-06,
+      "loss": 2.0614,
+      "step": 22177
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.95094816687737e-06,
+      "loss": 1.5447,
+      "step": 22178
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950900758533502e-06,
+      "loss": 3.2572,
+      "step": 22179
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9508533501896335e-06,
+      "loss": 3.5805,
+      "step": 22180
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950805941845765e-06,
+      "loss": 3.4342,
+      "step": 22181
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9507585335018965e-06,
+      "loss": 3.5692,
+      "step": 22182
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950711125158028e-06,
+      "loss": 2.8369,
+      "step": 22183
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950663716814159e-06,
+      "loss": 3.9197,
+      "step": 22184
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9506163084702908e-06,
+      "loss": 2.1446,
+      "step": 22185
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9505689001264225e-06,
+      "loss": 3.7741,
+      "step": 22186
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9505214917825538e-06,
+      "loss": 3.0922,
+      "step": 22187
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9504740834386855e-06,
+      "loss": 1.3712,
+      "step": 22188
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9504266750948167e-06,
+      "loss": 2.4446,
+      "step": 22189
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950379266750948e-06,
+      "loss": 2.5985,
+      "step": 22190
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9503318584070797e-06,
+      "loss": 2.5402,
+      "step": 22191
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950284450063211e-06,
+      "loss": 2.1902,
+      "step": 22192
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9502370417193427e-06,
+      "loss": 2.7993,
+      "step": 22193
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9501896333754744e-06,
+      "loss": 3.3151,
+      "step": 22194
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9501422250316057e-06,
+      "loss": 3.0574,
+      "step": 22195
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.950094816687737e-06,
+      "loss": 3.3473,
+      "step": 22196
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.9500474083438683e-06,
+      "loss": 4.2164,
+      "step": 22197
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.95e-06,
+      "loss": 2.2996,
+      "step": 22198
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9499525916561317e-06,
+      "loss": 3.471,
+      "step": 22199
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949905183312263e-06,
+      "loss": 1.4167,
+      "step": 22200
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9498577749683947e-06,
+      "loss": 2.6674,
+      "step": 22201
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949810366624526e-06,
+      "loss": 1.7623,
+      "step": 22202
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9497629582806572e-06,
+      "loss": 3.7792,
+      "step": 22203
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949715549936789e-06,
+      "loss": 3.0274,
+      "step": 22204
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.94966814159292e-06,
+      "loss": 1.9728,
+      "step": 22205
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949620733249052e-06,
+      "loss": 2.0111,
+      "step": 22206
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9495733249051836e-06,
+      "loss": 2.6854,
+      "step": 22207
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949525916561315e-06,
+      "loss": 2.0086,
+      "step": 22208
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949478508217446e-06,
+      "loss": 3.8915,
+      "step": 22209
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949431099873578e-06,
+      "loss": 2.6618,
+      "step": 22210
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949383691529709e-06,
+      "loss": 3.0821,
+      "step": 22211
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949336283185841e-06,
+      "loss": 4.1887,
+      "step": 22212
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9492888748419726e-06,
+      "loss": 2.7528,
+      "step": 22213
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949241466498104e-06,
+      "loss": 2.0511,
+      "step": 22214
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949194058154235e-06,
+      "loss": 1.8517,
+      "step": 22215
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949146649810367e-06,
+      "loss": 2.7895,
+      "step": 22216
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949099241466498e-06,
+      "loss": 2.9284,
+      "step": 22217
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9490518331226294e-06,
+      "loss": 2.2209,
+      "step": 22218
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.949004424778761e-06,
+      "loss": 2.1601,
+      "step": 22219
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948957016434893e-06,
+      "loss": 2.647,
+      "step": 22220
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948909608091024e-06,
+      "loss": 2.2849,
+      "step": 22221
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9488621997471554e-06,
+      "loss": 2.0706,
+      "step": 22222
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948814791403287e-06,
+      "loss": 2.3245,
+      "step": 22223
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9487673830594183e-06,
+      "loss": 2.601,
+      "step": 22224
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.94871997471555e-06,
+      "loss": 2.6875,
+      "step": 22225
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9486725663716817e-06,
+      "loss": 1.6896,
+      "step": 22226
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948625158027813e-06,
+      "loss": 3.0615,
+      "step": 22227
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9485777496839443e-06,
+      "loss": 2.3805,
+      "step": 22228
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948530341340076e-06,
+      "loss": 2.1324,
+      "step": 22229
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9484829329962073e-06,
+      "loss": 2.4856,
+      "step": 22230
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9484355246523386e-06,
+      "loss": 2.4049,
+      "step": 22231
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9483881163084703e-06,
+      "loss": 2.3443,
+      "step": 22232
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948340707964602e-06,
+      "loss": 2.5881,
+      "step": 22233
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9482932996207333e-06,
+      "loss": 3.4131,
+      "step": 22234
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948245891276865e-06,
+      "loss": 2.2441,
+      "step": 22235
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9481984829329962e-06,
+      "loss": 4.033,
+      "step": 22236
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9481510745891275e-06,
+      "loss": 4.5375,
+      "step": 22237
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9481036662452592e-06,
+      "loss": 4.254,
+      "step": 22238
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.948056257901391e-06,
+      "loss": 1.8346,
+      "step": 22239
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9480088495575222e-06,
+      "loss": 1.4629,
+      "step": 22240
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9479614412136535e-06,
+      "loss": 3.4662,
+      "step": 22241
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.947914032869785e-06,
+      "loss": 3.1853,
+      "step": 22242
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9478666245259165e-06,
+      "loss": 1.8655,
+      "step": 22243
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9478192161820478e-06,
+      "loss": 2.0922,
+      "step": 22244
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9477718078381795e-06,
+      "loss": 1.2528,
+      "step": 22245
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.947724399494311e-06,
+      "loss": 2.2834,
+      "step": 22246
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9476769911504424e-06,
+      "loss": 2.9803,
+      "step": 22247
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.947629582806574e-06,
+      "loss": 2.2508,
+      "step": 22248
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9475821744627054e-06,
+      "loss": 3.5515,
+      "step": 22249
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9475347661188367e-06,
+      "loss": 3.0301,
+      "step": 22250
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9474873577749684e-06,
+      "loss": 2.2695,
+      "step": 22251
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9474399494311e-06,
+      "loss": 2.749,
+      "step": 22252
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9473925410872314e-06,
+      "loss": 3.1177,
+      "step": 22253
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.947345132743363e-06,
+      "loss": 3.2484,
+      "step": 22254
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9472977243994944e-06,
+      "loss": 1.3237,
+      "step": 22255
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9472503160556257e-06,
+      "loss": 1.8084,
+      "step": 22256
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9472029077117574e-06,
+      "loss": 2.9768,
+      "step": 22257
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9471554993678886e-06,
+      "loss": 2.1424,
+      "step": 22258
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.9471080910240204e-06,
+      "loss": 3.3842,
+      "step": 22259
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.947060682680152e-06,
+      "loss": 1.5969,
+      "step": 22260
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9470132743362833e-06,
+      "loss": 1.8922,
+      "step": 22261
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9469658659924146e-06,
+      "loss": 2.3454,
+      "step": 22262
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.946918457648546e-06,
+      "loss": 1.8483,
+      "step": 22263
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9468710493046776e-06,
+      "loss": 2.4747,
+      "step": 22264
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9468236409608093e-06,
+      "loss": 2.1626,
+      "step": 22265
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9467762326169406e-06,
+      "loss": 3.4956,
+      "step": 22266
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9467288242730723e-06,
+      "loss": 3.2392,
+      "step": 22267
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9466814159292036e-06,
+      "loss": 1.6097,
+      "step": 22268
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.946634007585335e-06,
+      "loss": 2.5543,
+      "step": 22269
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9465865992414666e-06,
+      "loss": 3.0292,
+      "step": 22270
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.946539190897598e-06,
+      "loss": 3.4398,
+      "step": 22271
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9464917825537295e-06,
+      "loss": 3.3373,
+      "step": 22272
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9464443742098612e-06,
+      "loss": 3.598,
+      "step": 22273
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9463969658659925e-06,
+      "loss": 3.2893,
+      "step": 22274
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.946349557522124e-06,
+      "loss": 2.8653,
+      "step": 22275
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9463021491782555e-06,
+      "loss": 3.3686,
+      "step": 22276
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9462547408343868e-06,
+      "loss": 2.2647,
+      "step": 22277
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9462073324905185e-06,
+      "loss": 3.5644,
+      "step": 22278
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.94615992414665e-06,
+      "loss": 1.7749,
+      "step": 22279
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9461125158027815e-06,
+      "loss": 2.4285,
+      "step": 22280
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9460651074589128e-06,
+      "loss": 2.4304,
+      "step": 22281
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9460176991150445e-06,
+      "loss": 2.2869,
+      "step": 22282
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9459702907711757e-06,
+      "loss": 1.3916,
+      "step": 22283
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945922882427307e-06,
+      "loss": 3.4878,
+      "step": 22284
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9458754740834387e-06,
+      "loss": 2.0254,
+      "step": 22285
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9458280657395704e-06,
+      "loss": 2.2597,
+      "step": 22286
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9457806573957017e-06,
+      "loss": 2.1581,
+      "step": 22287
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945733249051833e-06,
+      "loss": 3.7858,
+      "step": 22288
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9456858407079647e-06,
+      "loss": 2.1489,
+      "step": 22289
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945638432364096e-06,
+      "loss": 3.2949,
+      "step": 22290
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9455910240202277e-06,
+      "loss": 2.2841,
+      "step": 22291
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9455436156763594e-06,
+      "loss": 2.0186,
+      "step": 22292
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9454962073324907e-06,
+      "loss": 3.4756,
+      "step": 22293
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945448798988622e-06,
+      "loss": 2.5264,
+      "step": 22294
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9454013906447536e-06,
+      "loss": 1.3522,
+      "step": 22295
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945353982300885e-06,
+      "loss": 1.0214,
+      "step": 22296
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945306573957016e-06,
+      "loss": 0.8836,
+      "step": 22297
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945259165613148e-06,
+      "loss": 2.5292,
+      "step": 22298
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9452117572692796e-06,
+      "loss": 2.6589,
+      "step": 22299
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945164348925411e-06,
+      "loss": 2.4951,
+      "step": 22300
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9451169405815426e-06,
+      "loss": 1.9258,
+      "step": 22301
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945069532237674e-06,
+      "loss": 3.2106,
+      "step": 22302
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.945022123893805e-06,
+      "loss": 2.2033,
+      "step": 22303
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944974715549937e-06,
+      "loss": 1.5138,
+      "step": 22304
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9449273072060686e-06,
+      "loss": 2.4744,
+      "step": 22305
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9448798988622e-06,
+      "loss": 3.3144,
+      "step": 22306
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944832490518331e-06,
+      "loss": 1.9049,
+      "step": 22307
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944785082174463e-06,
+      "loss": 1.779,
+      "step": 22308
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944737673830594e-06,
+      "loss": 2.4814,
+      "step": 22309
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9446902654867254e-06,
+      "loss": 3.0282,
+      "step": 22310
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944642857142857e-06,
+      "loss": 3.1649,
+      "step": 22311
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944595448798989e-06,
+      "loss": 2.2983,
+      "step": 22312
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.94454804045512e-06,
+      "loss": 3.8037,
+      "step": 22313
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9445006321112518e-06,
+      "loss": 2.8092,
+      "step": 22314
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944453223767383e-06,
+      "loss": 2.9223,
+      "step": 22315
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9444058154235143e-06,
+      "loss": 1.9827,
+      "step": 22316
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944358407079646e-06,
+      "loss": 3.1725,
+      "step": 22317
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9443109987357778e-06,
+      "loss": 2.6076,
+      "step": 22318
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944263590391909e-06,
+      "loss": 2.7064,
+      "step": 22319
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9442161820480407e-06,
+      "loss": 2.8444,
+      "step": 22320
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944168773704172e-06,
+      "loss": 3.0091,
+      "step": 22321
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.9441213653603033e-06,
+      "loss": 3.7374,
+      "step": 22322
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.944073957016435e-06,
+      "loss": 2.5647,
+      "step": 22323
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9440265486725663e-06,
+      "loss": 3.2069,
+      "step": 22324
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943979140328698e-06,
+      "loss": 1.5903,
+      "step": 22325
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9439317319848297e-06,
+      "loss": 3.3698,
+      "step": 22326
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943884323640961e-06,
+      "loss": 2.1807,
+      "step": 22327
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9438369152970923e-06,
+      "loss": 1.8123,
+      "step": 22328
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9437895069532235e-06,
+      "loss": 2.4415,
+      "step": 22329
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9437420986093552e-06,
+      "loss": 2.6822,
+      "step": 22330
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943694690265487e-06,
+      "loss": 3.0058,
+      "step": 22331
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9436472819216182e-06,
+      "loss": 3.2812,
+      "step": 22332
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.94359987357775e-06,
+      "loss": 3.321,
+      "step": 22333
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943552465233881e-06,
+      "loss": 3.0407,
+      "step": 22334
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9435050568900125e-06,
+      "loss": 3.5231,
+      "step": 22335
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943457648546144e-06,
+      "loss": 0.6707,
+      "step": 22336
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9434102402022755e-06,
+      "loss": 2.7724,
+      "step": 22337
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943362831858407e-06,
+      "loss": 3.2292,
+      "step": 22338
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943315423514539e-06,
+      "loss": 2.2984,
+      "step": 22339
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.94326801517067e-06,
+      "loss": 3.0098,
+      "step": 22340
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9432206068268014e-06,
+      "loss": 3.0428,
+      "step": 22341
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943173198482933e-06,
+      "loss": 1.7467,
+      "step": 22342
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9431257901390644e-06,
+      "loss": 2.3305,
+      "step": 22343
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943078381795196e-06,
+      "loss": 2.9346,
+      "step": 22344
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.943030973451328e-06,
+      "loss": 2.5587,
+      "step": 22345
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.942983565107459e-06,
+      "loss": 2.7449,
+      "step": 22346
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9429361567635904e-06,
+      "loss": 2.8396,
+      "step": 22347
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.942888748419722e-06,
+      "loss": 2.0507,
+      "step": 22348
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9428413400758534e-06,
+      "loss": 3.0166,
+      "step": 22349
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9427939317319847e-06,
+      "loss": 3.2362,
+      "step": 22350
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9427465233881164e-06,
+      "loss": 2.7648,
+      "step": 22351
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.942699115044248e-06,
+      "loss": 2.1604,
+      "step": 22352
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9426517067003793e-06,
+      "loss": 3.9339,
+      "step": 22353
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9426042983565106e-06,
+      "loss": 2.0898,
+      "step": 22354
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9425568900126423e-06,
+      "loss": 1.6089,
+      "step": 22355
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9425094816687736e-06,
+      "loss": 3.3561,
+      "step": 22356
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9424620733249053e-06,
+      "loss": 3.8856,
+      "step": 22357
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.942414664981037e-06,
+      "loss": 2.9802,
+      "step": 22358
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9423672566371683e-06,
+      "loss": 2.5082,
+      "step": 22359
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9423198482932996e-06,
+      "loss": 2.8097,
+      "step": 22360
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9422724399494313e-06,
+      "loss": 3.3768,
+      "step": 22361
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9422250316055626e-06,
+      "loss": 3.6981,
+      "step": 22362
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.942177623261694e-06,
+      "loss": 1.5495,
+      "step": 22363
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9421302149178255e-06,
+      "loss": 3.1631,
+      "step": 22364
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9420828065739573e-06,
+      "loss": 3.3603,
+      "step": 22365
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9420353982300885e-06,
+      "loss": 1.8936,
+      "step": 22366
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9419879898862202e-06,
+      "loss": 2.2446,
+      "step": 22367
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9419405815423515e-06,
+      "loss": 2.4016,
+      "step": 22368
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.941893173198483e-06,
+      "loss": 2.8279,
+      "step": 22369
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9418457648546145e-06,
+      "loss": 2.6117,
+      "step": 22370
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.941798356510746e-06,
+      "loss": 1.6565,
+      "step": 22371
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9417509481668775e-06,
+      "loss": 2.7886,
+      "step": 22372
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9417035398230088e-06,
+      "loss": 2.5215,
+      "step": 22373
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9416561314791405e-06,
+      "loss": 2.8964,
+      "step": 22374
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9416087231352718e-06,
+      "loss": 2.6078,
+      "step": 22375
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.941561314791403e-06,
+      "loss": 1.6921,
+      "step": 22376
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9415139064475347e-06,
+      "loss": 3.2775,
+      "step": 22377
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9414664981036664e-06,
+      "loss": 2.9479,
+      "step": 22378
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9414190897597977e-06,
+      "loss": 3.608,
+      "step": 22379
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9413716814159294e-06,
+      "loss": 2.5242,
+      "step": 22380
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9413242730720607e-06,
+      "loss": 3.1802,
+      "step": 22381
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.941276864728192e-06,
+      "loss": 1.4449,
+      "step": 22382
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9412294563843237e-06,
+      "loss": 3.0375,
+      "step": 22383
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9411820480404554e-06,
+      "loss": 2.1991,
+      "step": 22384
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9411346396965867e-06,
+      "loss": 2.5007,
+      "step": 22385
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9410872313527184e-06,
+      "loss": 2.1092,
+      "step": 22386
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9410398230088497e-06,
+      "loss": 3.392,
+      "step": 22387
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940992414664981e-06,
+      "loss": 3.3926,
+      "step": 22388
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9409450063211126e-06,
+      "loss": 2.6817,
+      "step": 22389
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940897597977244e-06,
+      "loss": 2.2563,
+      "step": 22390
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9408501896333756e-06,
+      "loss": 3.305,
+      "step": 22391
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9408027812895073e-06,
+      "loss": 2.6706,
+      "step": 22392
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9407553729456386e-06,
+      "loss": 3.6899,
+      "step": 22393
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.94070796460177e-06,
+      "loss": 3.1617,
+      "step": 22394
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940660556257901e-06,
+      "loss": 2.6285,
+      "step": 22395
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940613147914033e-06,
+      "loss": 1.4311,
+      "step": 22396
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940565739570164e-06,
+      "loss": 3.0434,
+      "step": 22397
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940518331226296e-06,
+      "loss": 2.1394,
+      "step": 22398
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9404709228824276e-06,
+      "loss": 3.0891,
+      "step": 22399
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940423514538559e-06,
+      "loss": 2.7726,
+      "step": 22400
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.94037610619469e-06,
+      "loss": 3.2741,
+      "step": 22401
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940328697850822e-06,
+      "loss": 2.5322,
+      "step": 22402
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940281289506953e-06,
+      "loss": 2.673,
+      "step": 22403
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940233881163085e-06,
+      "loss": 3.1228,
+      "step": 22404
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9401864728192165e-06,
+      "loss": 2.3413,
+      "step": 22405
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940139064475348e-06,
+      "loss": 2.422,
+      "step": 22406
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.940091656131479e-06,
+      "loss": 3.6015,
+      "step": 22407
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9400442477876108e-06,
+      "loss": 3.0206,
+      "step": 22408
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939996839443742e-06,
+      "loss": 2.2962,
+      "step": 22409
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9399494310998733e-06,
+      "loss": 2.3323,
+      "step": 22410
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939902022756005e-06,
+      "loss": 2.496,
+      "step": 22411
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9398546144121367e-06,
+      "loss": 3.2781,
+      "step": 22412
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939807206068268e-06,
+      "loss": 2.9349,
+      "step": 22413
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9397597977243997e-06,
+      "loss": 2.187,
+      "step": 22414
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939712389380531e-06,
+      "loss": 3.5805,
+      "step": 22415
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9396649810366623e-06,
+      "loss": 2.4757,
+      "step": 22416
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939617572692794e-06,
+      "loss": 2.5409,
+      "step": 22417
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9395701643489257e-06,
+      "loss": 1.7634,
+      "step": 22418
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939522756005057e-06,
+      "loss": 2.7845,
+      "step": 22419
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9394753476611883e-06,
+      "loss": 2.8593,
+      "step": 22420
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.93942793931732e-06,
+      "loss": 2.458,
+      "step": 22421
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9393805309734512e-06,
+      "loss": 2.3253,
+      "step": 22422
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9393331226295825e-06,
+      "loss": 2.5609,
+      "step": 22423
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9392857142857142e-06,
+      "loss": 2.6156,
+      "step": 22424
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939238305941846e-06,
+      "loss": 2.3044,
+      "step": 22425
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9391908975979772e-06,
+      "loss": 3.0633,
+      "step": 22426
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939143489254109e-06,
+      "loss": 3.7834,
+      "step": 22427
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.93909608091024e-06,
+      "loss": 3.4185,
+      "step": 22428
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9390486725663715e-06,
+      "loss": 2.2929,
+      "step": 22429
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.939001264222503e-06,
+      "loss": 2.3987,
+      "step": 22430
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938953855878635e-06,
+      "loss": 2.9327,
+      "step": 22431
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938906447534766e-06,
+      "loss": 1.8143,
+      "step": 22432
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938859039190898e-06,
+      "loss": 2.52,
+      "step": 22433
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938811630847029e-06,
+      "loss": 1.8956,
+      "step": 22434
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9387642225031604e-06,
+      "loss": 1.9822,
+      "step": 22435
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938716814159292e-06,
+      "loss": 3.2731,
+      "step": 22436
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9386694058154234e-06,
+      "loss": 2.0532,
+      "step": 22437
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938621997471555e-06,
+      "loss": 2.7686,
+      "step": 22438
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9385745891276864e-06,
+      "loss": 2.8212,
+      "step": 22439
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938527180783818e-06,
+      "loss": 2.309,
+      "step": 22440
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9384797724399494e-06,
+      "loss": 3.001,
+      "step": 22441
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9384323640960807e-06,
+      "loss": 2.0901,
+      "step": 22442
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9383849557522124e-06,
+      "loss": 2.7574,
+      "step": 22443
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938337547408344e-06,
+      "loss": 2.1515,
+      "step": 22444
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9382901390644754e-06,
+      "loss": 3.4828,
+      "step": 22445
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.938242730720607e-06,
+      "loss": 2.6248,
+      "step": 22446
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9381953223767383e-06,
+      "loss": 2.8139,
+      "step": 22447
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9381479140328696e-06,
+      "loss": 2.7569,
+      "step": 22448
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9381005056890013e-06,
+      "loss": 3.1271,
+      "step": 22449
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9380530973451326e-06,
+      "loss": 1.6119,
+      "step": 22450
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9380056890012643e-06,
+      "loss": 2.777,
+      "step": 22451
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937958280657396e-06,
+      "loss": 2.4767,
+      "step": 22452
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9379108723135273e-06,
+      "loss": 2.6166,
+      "step": 22453
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9378634639696586e-06,
+      "loss": 3.2372,
+      "step": 22454
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9378160556257903e-06,
+      "loss": 2.9606,
+      "step": 22455
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9377686472819216e-06,
+      "loss": 3.3255,
+      "step": 22456
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9377212389380533e-06,
+      "loss": 3.6076,
+      "step": 22457
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937673830594185e-06,
+      "loss": 3.1808,
+      "step": 22458
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9376264222503162e-06,
+      "loss": 1.628,
+      "step": 22459
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9375790139064475e-06,
+      "loss": 2.5797,
+      "step": 22460
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937531605562579e-06,
+      "loss": 3.3799,
+      "step": 22461
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9374841972187105e-06,
+      "loss": 2.5688,
+      "step": 22462
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937436788874842e-06,
+      "loss": 2.5968,
+      "step": 22463
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9373893805309735e-06,
+      "loss": 2.2029,
+      "step": 22464
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937341972187105e-06,
+      "loss": 1.4962,
+      "step": 22465
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9372945638432365e-06,
+      "loss": 2.1356,
+      "step": 22466
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9372471554993678e-06,
+      "loss": 1.0775,
+      "step": 22467
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9371997471554995e-06,
+      "loss": 3.2613,
+      "step": 22468
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9371523388116307e-06,
+      "loss": 3.3884,
+      "step": 22469
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9371049304677624e-06,
+      "loss": 3.2958,
+      "step": 22470
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937057522123894e-06,
+      "loss": 3.8321,
+      "step": 22471
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9370101137800254e-06,
+      "loss": 2.9463,
+      "step": 22472
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9369627054361567e-06,
+      "loss": 2.1084,
+      "step": 22473
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9369152970922884e-06,
+      "loss": 3.1335,
+      "step": 22474
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9368678887484197e-06,
+      "loss": 3.2669,
+      "step": 22475
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.936820480404551e-06,
+      "loss": 1.2391,
+      "step": 22476
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9367730720606827e-06,
+      "loss": 2.2232,
+      "step": 22477
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9367256637168144e-06,
+      "loss": 1.938,
+      "step": 22478
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9366782553729457e-06,
+      "loss": 2.4315,
+      "step": 22479
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.936630847029077e-06,
+      "loss": 2.6376,
+      "step": 22480
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9365834386852087e-06,
+      "loss": 2.7722,
+      "step": 22481
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.93653603034134e-06,
+      "loss": 1.8719,
+      "step": 22482
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9364886219974716e-06,
+      "loss": 3.1149,
+      "step": 22483
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9364412136536033e-06,
+      "loss": 2.1748,
+      "step": 22484
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9363938053097346e-06,
+      "loss": 2.204,
+      "step": 22485
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.936346396965866e-06,
+      "loss": 2.3921,
+      "step": 22486
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9362989886219976e-06,
+      "loss": 2.4421,
+      "step": 22487
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.936251580278129e-06,
+      "loss": 4.0368,
+      "step": 22488
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.93620417193426e-06,
+      "loss": 2.1277,
+      "step": 22489
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.936156763590392e-06,
+      "loss": 4.0141,
+      "step": 22490
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9361093552465236e-06,
+      "loss": 3.6676,
+      "step": 22491
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.936061946902655e-06,
+      "loss": 3.3969,
+      "step": 22492
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9360145385587866e-06,
+      "loss": 2.9227,
+      "step": 22493
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935967130214918e-06,
+      "loss": 2.2052,
+      "step": 22494
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935919721871049e-06,
+      "loss": 2.6855,
+      "step": 22495
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935872313527181e-06,
+      "loss": 3.1731,
+      "step": 22496
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9358249051833125e-06,
+      "loss": 2.095,
+      "step": 22497
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935777496839444e-06,
+      "loss": 2.5686,
+      "step": 22498
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9357300884955755e-06,
+      "loss": 2.427,
+      "step": 22499
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935682680151707e-06,
+      "loss": 2.8632,
+      "step": 22500
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935635271807838e-06,
+      "loss": 1.8206,
+      "step": 22501
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9355878634639694e-06,
+      "loss": 2.6533,
+      "step": 22502
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935540455120101e-06,
+      "loss": 3.7619,
+      "step": 22503
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9354930467762328e-06,
+      "loss": 2.9781,
+      "step": 22504
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935445638432364e-06,
+      "loss": 2.215,
+      "step": 22505
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9353982300884957e-06,
+      "loss": 2.6646,
+      "step": 22506
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935350821744627e-06,
+      "loss": 3.3471,
+      "step": 22507
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9353034134007583e-06,
+      "loss": 3.0244,
+      "step": 22508
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.93525600505689e-06,
+      "loss": 2.5912,
+      "step": 22509
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9352085967130217e-06,
+      "loss": 2.7818,
+      "step": 22510
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935161188369153e-06,
+      "loss": 2.9137,
+      "step": 22511
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.9351137800252847e-06,
+      "loss": 3.0229,
+      "step": 22512
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.935066371681416e-06,
+      "loss": 2.4275,
+      "step": 22513
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9350189633375473e-06,
+      "loss": 1.9293,
+      "step": 22514
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934971554993679e-06,
+      "loss": 2.3195,
+      "step": 22515
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9349241466498102e-06,
+      "loss": 3.285,
+      "step": 22516
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934876738305942e-06,
+      "loss": 2.9891,
+      "step": 22517
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9348293299620736e-06,
+      "loss": 2.0281,
+      "step": 22518
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934781921618205e-06,
+      "loss": 1.5161,
+      "step": 22519
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934734513274336e-06,
+      "loss": 2.7846,
+      "step": 22520
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934687104930468e-06,
+      "loss": 2.4518,
+      "step": 22521
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934639696586599e-06,
+      "loss": 2.4058,
+      "step": 22522
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934592288242731e-06,
+      "loss": 2.3306,
+      "step": 22523
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934544879898862e-06,
+      "loss": 2.1483,
+      "step": 22524
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934497471554994e-06,
+      "loss": 2.9137,
+      "step": 22525
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934450063211125e-06,
+      "loss": 2.874,
+      "step": 22526
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9344026548672564e-06,
+      "loss": 2.5228,
+      "step": 22527
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934355246523388e-06,
+      "loss": 2.455,
+      "step": 22528
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9343078381795194e-06,
+      "loss": 2.6315,
+      "step": 22529
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934260429835651e-06,
+      "loss": 2.5126,
+      "step": 22530
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934213021491783e-06,
+      "loss": 2.925,
+      "step": 22531
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934165613147914e-06,
+      "loss": 2.9345,
+      "step": 22532
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9341182048040454e-06,
+      "loss": 2.2566,
+      "step": 22533
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.934070796460177e-06,
+      "loss": 2.6183,
+      "step": 22534
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9340233881163084e-06,
+      "loss": 2.2806,
+      "step": 22535
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.93397597977244e-06,
+      "loss": 2.7944,
+      "step": 22536
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.933928571428572e-06,
+      "loss": 2.1413,
+      "step": 22537
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.933881163084703e-06,
+      "loss": 2.9313,
+      "step": 22538
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9338337547408343e-06,
+      "loss": 2.4584,
+      "step": 22539
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.933786346396966e-06,
+      "loss": 2.1563,
+      "step": 22540
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9337389380530973e-06,
+      "loss": 2.7634,
+      "step": 22541
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9336915297092286e-06,
+      "loss": 2.2371,
+      "step": 22542
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9336441213653603e-06,
+      "loss": 2.6952,
+      "step": 22543
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.933596713021492e-06,
+      "loss": 2.6539,
+      "step": 22544
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9335493046776233e-06,
+      "loss": 2.4035,
+      "step": 22545
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9335018963337546e-06,
+      "loss": 2.9821,
+      "step": 22546
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9334544879898863e-06,
+      "loss": 3.3473,
+      "step": 22547
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9334070796460176e-06,
+      "loss": 3.4009,
+      "step": 22548
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9333596713021493e-06,
+      "loss": 2.8447,
+      "step": 22549
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.933312262958281e-06,
+      "loss": 2.3441,
+      "step": 22550
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9332648546144123e-06,
+      "loss": 2.9113,
+      "step": 22551
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9332174462705435e-06,
+      "loss": 1.8047,
+      "step": 22552
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9331700379266752e-06,
+      "loss": 1.5008,
+      "step": 22553
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9331226295828065e-06,
+      "loss": 2.5807,
+      "step": 22554
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.933075221238938e-06,
+      "loss": 3.4409,
+      "step": 22555
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9330278128950695e-06,
+      "loss": 3.8402,
+      "step": 22556
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.932980404551201e-06,
+      "loss": 3.5973,
+      "step": 22557
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9329329962073325e-06,
+      "loss": 2.7131,
+      "step": 22558
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.932885587863464e-06,
+      "loss": 2.4358,
+      "step": 22559
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9328381795195955e-06,
+      "loss": 3.0489,
+      "step": 22560
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9327907711757268e-06,
+      "loss": 1.7793,
+      "step": 22561
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9327433628318585e-06,
+      "loss": 3.5429,
+      "step": 22562
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.93269595448799e-06,
+      "loss": 1.4589,
+      "step": 22563
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9326485461441214e-06,
+      "loss": 2.7342,
+      "step": 22564
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.932601137800253e-06,
+      "loss": 2.3387,
+      "step": 22565
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9325537294563844e-06,
+      "loss": 2.3746,
+      "step": 22566
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9325063211125157e-06,
+      "loss": 2.343,
+      "step": 22567
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.932458912768647e-06,
+      "loss": 2.4617,
+      "step": 22568
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9324115044247787e-06,
+      "loss": 2.583,
+      "step": 22569
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9323640960809104e-06,
+      "loss": 2.0597,
+      "step": 22570
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9323166877370417e-06,
+      "loss": 2.7686,
+      "step": 22571
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9322692793931734e-06,
+      "loss": 2.8412,
+      "step": 22572
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9322218710493047e-06,
+      "loss": 2.5607,
+      "step": 22573
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.932174462705436e-06,
+      "loss": 2.7133,
+      "step": 22574
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9321270543615676e-06,
+      "loss": 2.6091,
+      "step": 22575
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9320796460176993e-06,
+      "loss": 2.212,
+      "step": 22576
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9320322376738306e-06,
+      "loss": 2.3652,
+      "step": 22577
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9319848293299623e-06,
+      "loss": 2.0652,
+      "step": 22578
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9319374209860936e-06,
+      "loss": 2.1316,
+      "step": 22579
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931890012642225e-06,
+      "loss": 2.2068,
+      "step": 22580
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9318426042983566e-06,
+      "loss": 2.7089,
+      "step": 22581
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931795195954488e-06,
+      "loss": 1.408,
+      "step": 22582
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9317477876106196e-06,
+      "loss": 1.403,
+      "step": 22583
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9317003792667513e-06,
+      "loss": 4.0287,
+      "step": 22584
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9316529709228826e-06,
+      "loss": 2.1875,
+      "step": 22585
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931605562579014e-06,
+      "loss": 3.1116,
+      "step": 22586
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9315581542351456e-06,
+      "loss": 3.0468,
+      "step": 22587
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931510745891277e-06,
+      "loss": 2.7155,
+      "step": 22588
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9314633375474085e-06,
+      "loss": 2.5566,
+      "step": 22589
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.93141592920354e-06,
+      "loss": 3.3199,
+      "step": 22590
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9313685208596715e-06,
+      "loss": 2.6653,
+      "step": 22591
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931321112515803e-06,
+      "loss": 1.5879,
+      "step": 22592
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931273704171934e-06,
+      "loss": 2.4633,
+      "step": 22593
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9312262958280658e-06,
+      "loss": 2.2789,
+      "step": 22594
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.931178887484197e-06,
+      "loss": 2.4186,
+      "step": 22595
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9311314791403288e-06,
+      "loss": 3.1894,
+      "step": 22596
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9310840707964605e-06,
+      "loss": 2.3573,
+      "step": 22597
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9310366624525918e-06,
+      "loss": 0.9011,
+      "step": 22598
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930989254108723e-06,
+      "loss": 3.5829,
+      "step": 22599
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9309418457648547e-06,
+      "loss": 3.0438,
+      "step": 22600
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930894437420986e-06,
+      "loss": 2.7389,
+      "step": 22601
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9308470290771177e-06,
+      "loss": 2.4908,
+      "step": 22602
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9307996207332494e-06,
+      "loss": 4.0341,
+      "step": 22603
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9307522123893807e-06,
+      "loss": 2.7022,
+      "step": 22604
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930704804045512e-06,
+      "loss": 2.8738,
+      "step": 22605
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9306573957016437e-06,
+      "loss": 1.3289,
+      "step": 22606
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930609987357775e-06,
+      "loss": 2.1026,
+      "step": 22607
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9305625790139063e-06,
+      "loss": 2.2919,
+      "step": 22608
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930515170670038e-06,
+      "loss": 3.4978,
+      "step": 22609
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9304677623261697e-06,
+      "loss": 3.1758,
+      "step": 22610
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930420353982301e-06,
+      "loss": 2.1549,
+      "step": 22611
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9303729456384322e-06,
+      "loss": 2.116,
+      "step": 22612
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930325537294564e-06,
+      "loss": 2.7917,
+      "step": 22613
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930278128950695e-06,
+      "loss": 3.1236,
+      "step": 22614
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930230720606827e-06,
+      "loss": 1.6788,
+      "step": 22615
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9301833122629586e-06,
+      "loss": 2.5026,
+      "step": 22616
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.93013590391909e-06,
+      "loss": 3.2589,
+      "step": 22617
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930088495575221e-06,
+      "loss": 2.6247,
+      "step": 22618
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.930041087231353e-06,
+      "loss": 2.3194,
+      "step": 22619
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929993678887484e-06,
+      "loss": 2.8077,
+      "step": 22620
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9299462705436154e-06,
+      "loss": 2.2078,
+      "step": 22621
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929898862199747e-06,
+      "loss": 2.2595,
+      "step": 22622
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929851453855879e-06,
+      "loss": 1.9,
+      "step": 22623
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.92980404551201e-06,
+      "loss": 2.7902,
+      "step": 22624
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929756637168142e-06,
+      "loss": 2.3243,
+      "step": 22625
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929709228824273e-06,
+      "loss": 2.7142,
+      "step": 22626
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9296618204804044e-06,
+      "loss": 1.8392,
+      "step": 22627
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929614412136536e-06,
+      "loss": 1.6422,
+      "step": 22628
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929567003792668e-06,
+      "loss": 2.5709,
+      "step": 22629
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929519595448799e-06,
+      "loss": 2.1506,
+      "step": 22630
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9294721871049308e-06,
+      "loss": 4.1866,
+      "step": 22631
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929424778761062e-06,
+      "loss": 3.5743,
+      "step": 22632
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9293773704171933e-06,
+      "loss": 2.9739,
+      "step": 22633
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9293299620733246e-06,
+      "loss": 1.8759,
+      "step": 22634
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9292825537294563e-06,
+      "loss": 3.2884,
+      "step": 22635
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929235145385588e-06,
+      "loss": 2.6756,
+      "step": 22636
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9291877370417193e-06,
+      "loss": 2.6739,
+      "step": 22637
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.929140328697851e-06,
+      "loss": 3.6409,
+      "step": 22638
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9290929203539823e-06,
+      "loss": 3.0589,
+      "step": 22639
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.9290455120101136e-06,
+      "loss": 2.2416,
+      "step": 22640
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9289981036662453e-06,
+      "loss": 1.9907,
+      "step": 22641
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.928950695322377e-06,
+      "loss": 2.4139,
+      "step": 22642
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9289032869785083e-06,
+      "loss": 2.5295,
+      "step": 22643
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.92885587863464e-06,
+      "loss": 3.8617,
+      "step": 22644
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9288084702907713e-06,
+      "loss": 3.2912,
+      "step": 22645
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9287610619469025e-06,
+      "loss": 3.1989,
+      "step": 22646
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9287136536030342e-06,
+      "loss": 2.0366,
+      "step": 22647
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9286662452591655e-06,
+      "loss": 3.6404,
+      "step": 22648
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9286188369152972e-06,
+      "loss": 3.466,
+      "step": 22649
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.928571428571429e-06,
+      "loss": 2.5788,
+      "step": 22650
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.92852402022756e-06,
+      "loss": 2.0662,
+      "step": 22651
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9284766118836915e-06,
+      "loss": 3.9468,
+      "step": 22652
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.928429203539823e-06,
+      "loss": 2.056,
+      "step": 22653
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9283817951959545e-06,
+      "loss": 3.4375,
+      "step": 22654
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.928334386852086e-06,
+      "loss": 2.3826,
+      "step": 22655
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9282869785082175e-06,
+      "loss": 3.0206,
+      "step": 22656
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.928239570164349e-06,
+      "loss": 2.8909,
+      "step": 22657
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9281921618204804e-06,
+      "loss": 3.4445,
+      "step": 22658
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9281447534766117e-06,
+      "loss": 3.0095,
+      "step": 22659
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9280973451327434e-06,
+      "loss": 1.93,
+      "step": 22660
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9280499367888747e-06,
+      "loss": 3.1811,
+      "step": 22661
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9280025284450064e-06,
+      "loss": 2.7802,
+      "step": 22662
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927955120101138e-06,
+      "loss": 3.2595,
+      "step": 22663
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9279077117572694e-06,
+      "loss": 1.694,
+      "step": 22664
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9278603034134007e-06,
+      "loss": 2.7778,
+      "step": 22665
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9278128950695324e-06,
+      "loss": 1.7198,
+      "step": 22666
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9277654867256637e-06,
+      "loss": 2.5706,
+      "step": 22667
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927718078381795e-06,
+      "loss": 2.7972,
+      "step": 22668
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927670670037927e-06,
+      "loss": 2.9445,
+      "step": 22669
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9276232616940583e-06,
+      "loss": 2.3004,
+      "step": 22670
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9275758533501896e-06,
+      "loss": 3.168,
+      "step": 22671
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9275284450063213e-06,
+      "loss": 2.0731,
+      "step": 22672
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9274810366624526e-06,
+      "loss": 2.195,
+      "step": 22673
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927433628318584e-06,
+      "loss": 3.3814,
+      "step": 22674
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9273862199747156e-06,
+      "loss": 1.4725,
+      "step": 22675
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9273388116308473e-06,
+      "loss": 3.0058,
+      "step": 22676
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9272914032869786e-06,
+      "loss": 2.2401,
+      "step": 22677
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.92724399494311e-06,
+      "loss": 1.9082,
+      "step": 22678
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9271965865992416e-06,
+      "loss": 2.8982,
+      "step": 22679
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927149178255373e-06,
+      "loss": 2.8662,
+      "step": 22680
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927101769911504e-06,
+      "loss": 2.9291,
+      "step": 22681
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.927054361567636e-06,
+      "loss": 3.0237,
+      "step": 22682
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9270069532237675e-06,
+      "loss": 3.4809,
+      "step": 22683
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926959544879899e-06,
+      "loss": 2.2103,
+      "step": 22684
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9269121365360305e-06,
+      "loss": 2.9916,
+      "step": 22685
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926864728192162e-06,
+      "loss": 2.5916,
+      "step": 22686
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926817319848293e-06,
+      "loss": 2.5329,
+      "step": 22687
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9267699115044248e-06,
+      "loss": 2.4215,
+      "step": 22688
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9267225031605565e-06,
+      "loss": 3.2033,
+      "step": 22689
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9266750948166878e-06,
+      "loss": 2.691,
+      "step": 22690
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9266276864728195e-06,
+      "loss": 2.6581,
+      "step": 22691
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9265802781289507e-06,
+      "loss": 3.5981,
+      "step": 22692
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926532869785082e-06,
+      "loss": 3.7364,
+      "step": 22693
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9264854614412137e-06,
+      "loss": 1.9369,
+      "step": 22694
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926438053097345e-06,
+      "loss": 1.607,
+      "step": 22695
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9263906447534767e-06,
+      "loss": 2.7093,
+      "step": 22696
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9263432364096084e-06,
+      "loss": 3.1144,
+      "step": 22697
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9262958280657397e-06,
+      "loss": 2.4667,
+      "step": 22698
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926248419721871e-06,
+      "loss": 2.9893,
+      "step": 22699
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9262010113780023e-06,
+      "loss": 3.1813,
+      "step": 22700
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926153603034134e-06,
+      "loss": 3.3344,
+      "step": 22701
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.9261061946902657e-06,
+      "loss": 2.6988,
+      "step": 22702
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.926058786346397e-06,
+      "loss": 2.9378,
+      "step": 22703
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9260113780025287e-06,
+      "loss": 1.5653,
+      "step": 22704
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.92596396965866e-06,
+      "loss": 2.4482,
+      "step": 22705
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9259165613147912e-06,
+      "loss": 1.864,
+      "step": 22706
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925869152970923e-06,
+      "loss": 2.0259,
+      "step": 22707
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925821744627054e-06,
+      "loss": 2.5625,
+      "step": 22708
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925774336283186e-06,
+      "loss": 1.6972,
+      "step": 22709
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9257269279393176e-06,
+      "loss": 2.522,
+      "step": 22710
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925679519595449e-06,
+      "loss": 1.6296,
+      "step": 22711
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.92563211125158e-06,
+      "loss": 2.2051,
+      "step": 22712
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925584702907712e-06,
+      "loss": 2.5852,
+      "step": 22713
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925537294563843e-06,
+      "loss": 3.4278,
+      "step": 22714
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925489886219975e-06,
+      "loss": 3.4907,
+      "step": 22715
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9254424778761066e-06,
+      "loss": 1.852,
+      "step": 22716
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925395069532238e-06,
+      "loss": 1.8074,
+      "step": 22717
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925347661188369e-06,
+      "loss": 2.098,
+      "step": 22718
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925300252844501e-06,
+      "loss": 2.4782,
+      "step": 22719
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925252844500632e-06,
+      "loss": 2.0732,
+      "step": 22720
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9252054361567634e-06,
+      "loss": 2.3778,
+      "step": 22721
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925158027812895e-06,
+      "loss": 3.4818,
+      "step": 22722
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925110619469027e-06,
+      "loss": 1.6409,
+      "step": 22723
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.925063211125158e-06,
+      "loss": 2.8719,
+      "step": 22724
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9250158027812894e-06,
+      "loss": 1.8287,
+      "step": 22725
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924968394437421e-06,
+      "loss": 2.9273,
+      "step": 22726
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9249209860935523e-06,
+      "loss": 2.2558,
+      "step": 22727
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924873577749684e-06,
+      "loss": 3.2248,
+      "step": 22728
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9248261694058157e-06,
+      "loss": 2.3326,
+      "step": 22729
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924778761061947e-06,
+      "loss": 2.1974,
+      "step": 22730
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9247313527180783e-06,
+      "loss": 2.0821,
+      "step": 22731
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.92468394437421e-06,
+      "loss": 2.2726,
+      "step": 22732
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9246365360303413e-06,
+      "loss": 3.9766,
+      "step": 22733
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9245891276864726e-06,
+      "loss": 2.8207,
+      "step": 22734
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9245417193426043e-06,
+      "loss": 2.5308,
+      "step": 22735
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924494310998736e-06,
+      "loss": 3.4163,
+      "step": 22736
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9244469026548673e-06,
+      "loss": 1.8858,
+      "step": 22737
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924399494310999e-06,
+      "loss": 1.3844,
+      "step": 22738
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9243520859671302e-06,
+      "loss": 2.4561,
+      "step": 22739
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9243046776232615e-06,
+      "loss": 2.4472,
+      "step": 22740
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9242572692793932e-06,
+      "loss": 2.4882,
+      "step": 22741
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924209860935525e-06,
+      "loss": 3.8663,
+      "step": 22742
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9241624525916562e-06,
+      "loss": 2.3884,
+      "step": 22743
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9241150442477875e-06,
+      "loss": 2.7797,
+      "step": 22744
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.924067635903919e-06,
+      "loss": 3.6594,
+      "step": 22745
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9240202275600505e-06,
+      "loss": 2.3496,
+      "step": 22746
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9239728192161818e-06,
+      "loss": 4.215,
+      "step": 22747
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9239254108723135e-06,
+      "loss": 3.3994,
+      "step": 22748
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.923878002528445e-06,
+      "loss": 1.689,
+      "step": 22749
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9238305941845764e-06,
+      "loss": 3.0231,
+      "step": 22750
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.923783185840708e-06,
+      "loss": 2.4669,
+      "step": 22751
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9237357774968394e-06,
+      "loss": 2.9424,
+      "step": 22752
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9236883691529707e-06,
+      "loss": 1.8057,
+      "step": 22753
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9236409608091024e-06,
+      "loss": 2.3845,
+      "step": 22754
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.923593552465234e-06,
+      "loss": 3.1057,
+      "step": 22755
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9235461441213654e-06,
+      "loss": 1.6379,
+      "step": 22756
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.923498735777497e-06,
+      "loss": 3.495,
+      "step": 22757
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9234513274336284e-06,
+      "loss": 3.395,
+      "step": 22758
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9234039190897597e-06,
+      "loss": 1.4066,
+      "step": 22759
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9233565107458914e-06,
+      "loss": 2.491,
+      "step": 22760
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9233091024020226e-06,
+      "loss": 1.61,
+      "step": 22761
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9232616940581544e-06,
+      "loss": 3.6283,
+      "step": 22762
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.923214285714286e-06,
+      "loss": 2.4145,
+      "step": 22763
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9231668773704173e-06,
+      "loss": 2.4905,
+      "step": 22764
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9231194690265486e-06,
+      "loss": 2.2309,
+      "step": 22765
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.92307206068268e-06,
+      "loss": 3.0047,
+      "step": 22766
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.9230246523388116e-06,
+      "loss": 2.5809,
+      "step": 22767
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9229772439949433e-06,
+      "loss": 3.2606,
+      "step": 22768
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9229298356510746e-06,
+      "loss": 3.7075,
+      "step": 22769
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9228824273072063e-06,
+      "loss": 3.1859,
+      "step": 22770
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9228350189633376e-06,
+      "loss": 3.0371,
+      "step": 22771
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.922787610619469e-06,
+      "loss": 2.191,
+      "step": 22772
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9227402022756006e-06,
+      "loss": 2.6631,
+      "step": 22773
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.922692793931732e-06,
+      "loss": 2.9977,
+      "step": 22774
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9226453855878635e-06,
+      "loss": 1.876,
+      "step": 22775
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9225979772439952e-06,
+      "loss": 2.1607,
+      "step": 22776
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9225505689001265e-06,
+      "loss": 2.9546,
+      "step": 22777
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.922503160556258e-06,
+      "loss": 2.9483,
+      "step": 22778
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9224557522123895e-06,
+      "loss": 3.4107,
+      "step": 22779
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9224083438685208e-06,
+      "loss": 2.6716,
+      "step": 22780
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9223609355246525e-06,
+      "loss": 2.5146,
+      "step": 22781
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.922313527180784e-06,
+      "loss": 3.1202,
+      "step": 22782
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9222661188369155e-06,
+      "loss": 3.2049,
+      "step": 22783
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9222187104930468e-06,
+      "loss": 2.8332,
+      "step": 22784
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9221713021491785e-06,
+      "loss": 2.229,
+      "step": 22785
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9221238938053097e-06,
+      "loss": 3.2724,
+      "step": 22786
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.922076485461441e-06,
+      "loss": 2.3636,
+      "step": 22787
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9220290771175727e-06,
+      "loss": 2.6186,
+      "step": 22788
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9219816687737044e-06,
+      "loss": 2.2575,
+      "step": 22789
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9219342604298357e-06,
+      "loss": 2.9368,
+      "step": 22790
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921886852085967e-06,
+      "loss": 2.6664,
+      "step": 22791
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9218394437420987e-06,
+      "loss": 2.0921,
+      "step": 22792
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.92179203539823e-06,
+      "loss": 3.1396,
+      "step": 22793
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9217446270543617e-06,
+      "loss": 1.9742,
+      "step": 22794
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9216972187104934e-06,
+      "loss": 1.6475,
+      "step": 22795
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9216498103666247e-06,
+      "loss": 2.5341,
+      "step": 22796
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921602402022756e-06,
+      "loss": 2.2857,
+      "step": 22797
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9215549936788876e-06,
+      "loss": 3.0441,
+      "step": 22798
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921507585335019e-06,
+      "loss": 4.4683,
+      "step": 22799
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.92146017699115e-06,
+      "loss": 2.5472,
+      "step": 22800
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921412768647282e-06,
+      "loss": 4.0267,
+      "step": 22801
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9213653603034136e-06,
+      "loss": 3.6203,
+      "step": 22802
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921317951959545e-06,
+      "loss": 2.2962,
+      "step": 22803
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9212705436156766e-06,
+      "loss": 1.7871,
+      "step": 22804
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921223135271808e-06,
+      "loss": 2.8182,
+      "step": 22805
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921175726927939e-06,
+      "loss": 1.4557,
+      "step": 22806
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921128318584071e-06,
+      "loss": 1.8112,
+      "step": 22807
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9210809102402026e-06,
+      "loss": 2.7881,
+      "step": 22808
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.921033501896334e-06,
+      "loss": 2.3241,
+      "step": 22809
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920986093552465e-06,
+      "loss": 2.8146,
+      "step": 22810
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920938685208597e-06,
+      "loss": 2.4081,
+      "step": 22811
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920891276864728e-06,
+      "loss": 3.1489,
+      "step": 22812
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9208438685208594e-06,
+      "loss": 2.4759,
+      "step": 22813
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920796460176991e-06,
+      "loss": 2.6147,
+      "step": 22814
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920749051833123e-06,
+      "loss": 2.2182,
+      "step": 22815
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920701643489254e-06,
+      "loss": 3.6821,
+      "step": 22816
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9206542351453858e-06,
+      "loss": 2.8374,
+      "step": 22817
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920606826801517e-06,
+      "loss": 3.9203,
+      "step": 22818
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9205594184576483e-06,
+      "loss": 3.2469,
+      "step": 22819
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.92051201011378e-06,
+      "loss": 1.8618,
+      "step": 22820
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9204646017699118e-06,
+      "loss": 1.4356,
+      "step": 22821
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920417193426043e-06,
+      "loss": 2.8931,
+      "step": 22822
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9203697850821747e-06,
+      "loss": 2.6208,
+      "step": 22823
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920322376738306e-06,
+      "loss": 2.4379,
+      "step": 22824
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9202749683944373e-06,
+      "loss": 3.1661,
+      "step": 22825
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920227560050569e-06,
+      "loss": 2.5793,
+      "step": 22826
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9201801517067003e-06,
+      "loss": 1.2235,
+      "step": 22827
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920132743362832e-06,
+      "loss": 2.6404,
+      "step": 22828
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.9200853350189637e-06,
+      "loss": 2.9293,
+      "step": 22829
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.920037926675095e-06,
+      "loss": 1.9819,
+      "step": 22830
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9199905183312263e-06,
+      "loss": 1.3647,
+      "step": 22831
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9199431099873575e-06,
+      "loss": 1.5427,
+      "step": 22832
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9198957016434892e-06,
+      "loss": 3.0873,
+      "step": 22833
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919848293299621e-06,
+      "loss": 2.6842,
+      "step": 22834
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9198008849557522e-06,
+      "loss": 2.4857,
+      "step": 22835
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919753476611884e-06,
+      "loss": 2.2603,
+      "step": 22836
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919706068268015e-06,
+      "loss": 1.0462,
+      "step": 22837
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9196586599241465e-06,
+      "loss": 1.4476,
+      "step": 22838
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919611251580278e-06,
+      "loss": 2.4465,
+      "step": 22839
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9195638432364095e-06,
+      "loss": 3.6067,
+      "step": 22840
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919516434892541e-06,
+      "loss": 2.0279,
+      "step": 22841
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919469026548673e-06,
+      "loss": 3.1359,
+      "step": 22842
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919421618204804e-06,
+      "loss": 3.3023,
+      "step": 22843
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9193742098609354e-06,
+      "loss": 2.0329,
+      "step": 22844
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919326801517067e-06,
+      "loss": 3.5732,
+      "step": 22845
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9192793931731984e-06,
+      "loss": 3.3438,
+      "step": 22846
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.91923198482933e-06,
+      "loss": 2.9987,
+      "step": 22847
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919184576485462e-06,
+      "loss": 3.1582,
+      "step": 22848
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919137168141593e-06,
+      "loss": 2.2182,
+      "step": 22849
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9190897597977244e-06,
+      "loss": 3.066,
+      "step": 22850
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.919042351453856e-06,
+      "loss": 1.3856,
+      "step": 22851
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9189949431099874e-06,
+      "loss": 2.8608,
+      "step": 22852
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9189475347661187e-06,
+      "loss": 3.1177,
+      "step": 22853
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9189001264222504e-06,
+      "loss": 2.03,
+      "step": 22854
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.918852718078382e-06,
+      "loss": 2.6386,
+      "step": 22855
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9188053097345133e-06,
+      "loss": 2.1287,
+      "step": 22856
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9187579013906446e-06,
+      "loss": 1.5552,
+      "step": 22857
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9187104930467763e-06,
+      "loss": 3.2779,
+      "step": 22858
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9186630847029076e-06,
+      "loss": 2.0483,
+      "step": 22859
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9186156763590393e-06,
+      "loss": 2.6076,
+      "step": 22860
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.918568268015171e-06,
+      "loss": 2.8178,
+      "step": 22861
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9185208596713023e-06,
+      "loss": 2.641,
+      "step": 22862
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9184734513274336e-06,
+      "loss": 1.9207,
+      "step": 22863
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9184260429835653e-06,
+      "loss": 3.2808,
+      "step": 22864
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9183786346396966e-06,
+      "loss": 2.6602,
+      "step": 22865
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.918331226295828e-06,
+      "loss": 3.1164,
+      "step": 22866
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9182838179519595e-06,
+      "loss": 2.237,
+      "step": 22867
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9182364096080913e-06,
+      "loss": 3.3858,
+      "step": 22868
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9181890012642225e-06,
+      "loss": 3.8593,
+      "step": 22869
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9181415929203542e-06,
+      "loss": 1.9631,
+      "step": 22870
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9180941845764855e-06,
+      "loss": 2.9431,
+      "step": 22871
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.918046776232617e-06,
+      "loss": 3.1459,
+      "step": 22872
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9179993678887485e-06,
+      "loss": 2.1581,
+      "step": 22873
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.91795195954488e-06,
+      "loss": 3.2467,
+      "step": 22874
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9179045512010115e-06,
+      "loss": 3.0423,
+      "step": 22875
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9178571428571428e-06,
+      "loss": 4.1795,
+      "step": 22876
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9178097345132745e-06,
+      "loss": 1.0389,
+      "step": 22877
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9177623261694058e-06,
+      "loss": 3.238,
+      "step": 22878
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.917714917825537e-06,
+      "loss": 2.2406,
+      "step": 22879
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9176675094816687e-06,
+      "loss": 2.4932,
+      "step": 22880
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9176201011378004e-06,
+      "loss": 1.5618,
+      "step": 22881
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9175726927939317e-06,
+      "loss": 2.6643,
+      "step": 22882
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9175252844500634e-06,
+      "loss": 2.9218,
+      "step": 22883
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9174778761061947e-06,
+      "loss": 2.9632,
+      "step": 22884
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.917430467762326e-06,
+      "loss": 2.6647,
+      "step": 22885
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9173830594184577e-06,
+      "loss": 2.2497,
+      "step": 22886
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9173356510745894e-06,
+      "loss": 2.0701,
+      "step": 22887
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9172882427307207e-06,
+      "loss": 4.3774,
+      "step": 22888
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9172408343868524e-06,
+      "loss": 2.3505,
+      "step": 22889
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9171934260429837e-06,
+      "loss": 3.5008,
+      "step": 22890
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.917146017699115e-06,
+      "loss": 2.892,
+      "step": 22891
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.9170986093552466e-06,
+      "loss": 3.0108,
+      "step": 22892
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.917051201011378e-06,
+      "loss": 2.503,
+      "step": 22893
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9170037926675096e-06,
+      "loss": 3.8831,
+      "step": 22894
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9169563843236413e-06,
+      "loss": 3.0792,
+      "step": 22895
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9169089759797726e-06,
+      "loss": 3.4937,
+      "step": 22896
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916861567635904e-06,
+      "loss": 2.1242,
+      "step": 22897
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916814159292035e-06,
+      "loss": 2.0249,
+      "step": 22898
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916766750948167e-06,
+      "loss": 1.4168,
+      "step": 22899
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9167193426042986e-06,
+      "loss": 3.2708,
+      "step": 22900
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.91667193426043e-06,
+      "loss": 1.6981,
+      "step": 22901
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9166245259165616e-06,
+      "loss": 2.0824,
+      "step": 22902
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916577117572693e-06,
+      "loss": 2.8575,
+      "step": 22903
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916529709228824e-06,
+      "loss": 2.5573,
+      "step": 22904
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916482300884956e-06,
+      "loss": 3.8969,
+      "step": 22905
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916434892541087e-06,
+      "loss": 2.9004,
+      "step": 22906
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916387484197219e-06,
+      "loss": 2.4013,
+      "step": 22907
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9163400758533505e-06,
+      "loss": 3.3756,
+      "step": 22908
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916292667509482e-06,
+      "loss": 2.7162,
+      "step": 22909
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916245259165613e-06,
+      "loss": 3.4968,
+      "step": 22910
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9161978508217448e-06,
+      "loss": 3.3695,
+      "step": 22911
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.916150442477876e-06,
+      "loss": 2.2781,
+      "step": 22912
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9161030341340078e-06,
+      "loss": 2.6526,
+      "step": 22913
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9160556257901395e-06,
+      "loss": 3.6201,
+      "step": 22914
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9160082174462707e-06,
+      "loss": 2.7937,
+      "step": 22915
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915960809102402e-06,
+      "loss": 2.6254,
+      "step": 22916
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9159134007585333e-06,
+      "loss": 3.0618,
+      "step": 22917
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915865992414665e-06,
+      "loss": 1.337,
+      "step": 22918
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9158185840707963e-06,
+      "loss": 0.7672,
+      "step": 22919
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915771175726928e-06,
+      "loss": 1.5688,
+      "step": 22920
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9157237673830597e-06,
+      "loss": 3.586,
+      "step": 22921
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915676359039191e-06,
+      "loss": 2.0989,
+      "step": 22922
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9156289506953223e-06,
+      "loss": 1.923,
+      "step": 22923
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915581542351454e-06,
+      "loss": 3.6557,
+      "step": 22924
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9155341340075852e-06,
+      "loss": 3.1997,
+      "step": 22925
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915486725663717e-06,
+      "loss": 2.7541,
+      "step": 22926
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9154393173198487e-06,
+      "loss": 1.6239,
+      "step": 22927
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.91539190897598e-06,
+      "loss": 2.2376,
+      "step": 22928
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9153445006321112e-06,
+      "loss": 2.3222,
+      "step": 22929
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915297092288243e-06,
+      "loss": 2.9403,
+      "step": 22930
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915249683944374e-06,
+      "loss": 2.9137,
+      "step": 22931
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9152022756005055e-06,
+      "loss": 3.3749,
+      "step": 22932
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915154867256637e-06,
+      "loss": 4.7299,
+      "step": 22933
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915107458912769e-06,
+      "loss": 3.4882,
+      "step": 22934
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9150600505689e-06,
+      "loss": 3.0399,
+      "step": 22935
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.915012642225032e-06,
+      "loss": 2.2141,
+      "step": 22936
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.914965233881163e-06,
+      "loss": 2.967,
+      "step": 22937
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9149178255372944e-06,
+      "loss": 1.5715,
+      "step": 22938
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9148704171934257e-06,
+      "loss": 2.5098,
+      "step": 22939
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.914823008849558e-06,
+      "loss": 2.5834,
+      "step": 22940
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.914775600505689e-06,
+      "loss": 2.9303,
+      "step": 22941
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9147281921618204e-06,
+      "loss": 1.8294,
+      "step": 22942
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.914680783817952e-06,
+      "loss": 2.9379,
+      "step": 22943
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9146333754740834e-06,
+      "loss": 2.9769,
+      "step": 22944
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9145859671302147e-06,
+      "loss": 2.9133,
+      "step": 22945
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9145385587863464e-06,
+      "loss": 2.768,
+      "step": 22946
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.914491150442478e-06,
+      "loss": 2.7637,
+      "step": 22947
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9144437420986094e-06,
+      "loss": 2.2449,
+      "step": 22948
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.914396333754741e-06,
+      "loss": 2.8279,
+      "step": 22949
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9143489254108723e-06,
+      "loss": 1.9114,
+      "step": 22950
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9143015170670036e-06,
+      "loss": 1.8335,
+      "step": 22951
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9142541087231353e-06,
+      "loss": 2.2041,
+      "step": 22952
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9142067003792666e-06,
+      "loss": 2.4353,
+      "step": 22953
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9141592920353983e-06,
+      "loss": 1.7269,
+      "step": 22954
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.91411188369153e-06,
+      "loss": 1.9101,
+      "step": 22955
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9140644753476613e-06,
+      "loss": 2.3982,
+      "step": 22956
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.9140170670037926e-06,
+      "loss": 2.4744,
+      "step": 22957
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9139696586599243e-06,
+      "loss": 1.624,
+      "step": 22958
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9139222503160556e-06,
+      "loss": 3.0547,
+      "step": 22959
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9138748419721873e-06,
+      "loss": 3.6631,
+      "step": 22960
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9138274336283185e-06,
+      "loss": 3.0881,
+      "step": 22961
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9137800252844502e-06,
+      "loss": 2.9183,
+      "step": 22962
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9137326169405815e-06,
+      "loss": 2.3898,
+      "step": 22963
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.913685208596713e-06,
+      "loss": 2.7415,
+      "step": 22964
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9136378002528445e-06,
+      "loss": 2.6043,
+      "step": 22965
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.913590391908976e-06,
+      "loss": 2.4244,
+      "step": 22966
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9135429835651075e-06,
+      "loss": 3.6375,
+      "step": 22967
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.913495575221239e-06,
+      "loss": 1.4654,
+      "step": 22968
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9134481668773705e-06,
+      "loss": 2.3922,
+      "step": 22969
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9134007585335018e-06,
+      "loss": 1.9833,
+      "step": 22970
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9133533501896335e-06,
+      "loss": 2.913,
+      "step": 22971
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9133059418457647e-06,
+      "loss": 1.778,
+      "step": 22972
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9132585335018964e-06,
+      "loss": 3.1877,
+      "step": 22973
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.913211125158028e-06,
+      "loss": 2.7564,
+      "step": 22974
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9131637168141594e-06,
+      "loss": 3.4729,
+      "step": 22975
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9131163084702907e-06,
+      "loss": 2.236,
+      "step": 22976
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9130689001264224e-06,
+      "loss": 2.4184,
+      "step": 22977
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9130214917825537e-06,
+      "loss": 2.0572,
+      "step": 22978
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912974083438685e-06,
+      "loss": 2.722,
+      "step": 22979
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9129266750948167e-06,
+      "loss": 1.1937,
+      "step": 22980
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9128792667509484e-06,
+      "loss": 3.7368,
+      "step": 22981
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9128318584070797e-06,
+      "loss": 3.314,
+      "step": 22982
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912784450063211e-06,
+      "loss": 2.9928,
+      "step": 22983
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9127370417193427e-06,
+      "loss": 3.4442,
+      "step": 22984
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912689633375474e-06,
+      "loss": 2.06,
+      "step": 22985
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9126422250316056e-06,
+      "loss": 3.1816,
+      "step": 22986
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9125948166877373e-06,
+      "loss": 2.7594,
+      "step": 22987
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9125474083438686e-06,
+      "loss": 2.487,
+      "step": 22988
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9125e-06,
+      "loss": 1.6288,
+      "step": 22989
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9124525916561316e-06,
+      "loss": 3.1665,
+      "step": 22990
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912405183312263e-06,
+      "loss": 3.3411,
+      "step": 22991
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912357774968394e-06,
+      "loss": 2.9799,
+      "step": 22992
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912310366624526e-06,
+      "loss": 2.0334,
+      "step": 22993
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9122629582806576e-06,
+      "loss": 2.6917,
+      "step": 22994
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912215549936789e-06,
+      "loss": 3.7038,
+      "step": 22995
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9121681415929206e-06,
+      "loss": 2.1634,
+      "step": 22996
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912120733249052e-06,
+      "loss": 3.6159,
+      "step": 22997
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912073324905183e-06,
+      "loss": 3.1377,
+      "step": 22998
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.912025916561315e-06,
+      "loss": 2.5328,
+      "step": 22999
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9119785082174465e-06,
+      "loss": 2.7645,
+      "step": 23000
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911931099873578e-06,
+      "loss": 3.9297,
+      "step": 23001
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9118836915297095e-06,
+      "loss": 2.0035,
+      "step": 23002
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911836283185841e-06,
+      "loss": 2.2747,
+      "step": 23003
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911788874841972e-06,
+      "loss": 3.2656,
+      "step": 23004
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9117414664981034e-06,
+      "loss": 3.7558,
+      "step": 23005
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911694058154235e-06,
+      "loss": 2.4579,
+      "step": 23006
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9116466498103668e-06,
+      "loss": 2.2627,
+      "step": 23007
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911599241466498e-06,
+      "loss": 2.5122,
+      "step": 23008
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9115518331226297e-06,
+      "loss": 2.5878,
+      "step": 23009
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911504424778761e-06,
+      "loss": 3.7078,
+      "step": 23010
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9114570164348923e-06,
+      "loss": 2.5286,
+      "step": 23011
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911409608091024e-06,
+      "loss": 4.2403,
+      "step": 23012
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9113621997471557e-06,
+      "loss": 3.3977,
+      "step": 23013
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911314791403287e-06,
+      "loss": 2.5989,
+      "step": 23014
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9112673830594187e-06,
+      "loss": 1.1844,
+      "step": 23015
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.91121997471555e-06,
+      "loss": 3.5082,
+      "step": 23016
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9111725663716813e-06,
+      "loss": 1.4785,
+      "step": 23017
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911125158027813e-06,
+      "loss": 2.1922,
+      "step": 23018
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.9110777496839442e-06,
+      "loss": 2.886,
+      "step": 23019
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.911030341340076e-06,
+      "loss": 1.5063,
+      "step": 23020
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9109829329962076e-06,
+      "loss": 2.4568,
+      "step": 23021
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910935524652339e-06,
+      "loss": 1.6548,
+      "step": 23022
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.91088811630847e-06,
+      "loss": 3.1633,
+      "step": 23023
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910840707964602e-06,
+      "loss": 2.6287,
+      "step": 23024
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910793299620733e-06,
+      "loss": 2.691,
+      "step": 23025
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910745891276865e-06,
+      "loss": 2.7912,
+      "step": 23026
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910698482932996e-06,
+      "loss": 2.1982,
+      "step": 23027
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910651074589128e-06,
+      "loss": 2.1265,
+      "step": 23028
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910603666245259e-06,
+      "loss": 3.5432,
+      "step": 23029
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9105562579013904e-06,
+      "loss": 1.6052,
+      "step": 23030
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910508849557522e-06,
+      "loss": 2.4065,
+      "step": 23031
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9104614412136534e-06,
+      "loss": 4.3428,
+      "step": 23032
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910414032869785e-06,
+      "loss": 2.5843,
+      "step": 23033
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910366624525917e-06,
+      "loss": 1.5994,
+      "step": 23034
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910319216182048e-06,
+      "loss": 1.8726,
+      "step": 23035
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9102718078381794e-06,
+      "loss": 3.4001,
+      "step": 23036
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910224399494311e-06,
+      "loss": 2.8901,
+      "step": 23037
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9101769911504424e-06,
+      "loss": 4.2318,
+      "step": 23038
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910129582806574e-06,
+      "loss": 3.0299,
+      "step": 23039
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910082174462706e-06,
+      "loss": 2.1725,
+      "step": 23040
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.910034766118837e-06,
+      "loss": 2.5095,
+      "step": 23041
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9099873577749684e-06,
+      "loss": 2.4445,
+      "step": 23042
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9099399494311e-06,
+      "loss": 1.984,
+      "step": 23043
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9098925410872313e-06,
+      "loss": 3.1072,
+      "step": 23044
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9098451327433626e-06,
+      "loss": 2.6207,
+      "step": 23045
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9097977243994943e-06,
+      "loss": 1.7793,
+      "step": 23046
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.909750316055626e-06,
+      "loss": 1.4965,
+      "step": 23047
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9097029077117573e-06,
+      "loss": 2.5828,
+      "step": 23048
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9096554993678886e-06,
+      "loss": 2.4497,
+      "step": 23049
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9096080910240203e-06,
+      "loss": 1.8886,
+      "step": 23050
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9095606826801516e-06,
+      "loss": 1.6377,
+      "step": 23051
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9095132743362833e-06,
+      "loss": 1.0844,
+      "step": 23052
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.909465865992415e-06,
+      "loss": 1.1849,
+      "step": 23053
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9094184576485463e-06,
+      "loss": 1.9748,
+      "step": 23054
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9093710493046775e-06,
+      "loss": 4.0424,
+      "step": 23055
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9093236409608092e-06,
+      "loss": 3.1373,
+      "step": 23056
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9092762326169405e-06,
+      "loss": 3.6367,
+      "step": 23057
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.909228824273072e-06,
+      "loss": 2.2193,
+      "step": 23058
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9091814159292035e-06,
+      "loss": 1.8513,
+      "step": 23059
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.909134007585335e-06,
+      "loss": 1.7804,
+      "step": 23060
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9090865992414665e-06,
+      "loss": 3.5988,
+      "step": 23061
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.909039190897598e-06,
+      "loss": 1.911,
+      "step": 23062
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9089917825537295e-06,
+      "loss": 1.7186,
+      "step": 23063
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9089443742098608e-06,
+      "loss": 4.371,
+      "step": 23064
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9088969658659925e-06,
+      "loss": 2.1957,
+      "step": 23065
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.908849557522124e-06,
+      "loss": 1.8917,
+      "step": 23066
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9088021491782554e-06,
+      "loss": 2.0559,
+      "step": 23067
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.908754740834387e-06,
+      "loss": 2.3548,
+      "step": 23068
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9087073324905184e-06,
+      "loss": 3.3091,
+      "step": 23069
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9086599241466497e-06,
+      "loss": 2.8843,
+      "step": 23070
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.908612515802781e-06,
+      "loss": 3.1659,
+      "step": 23071
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9085651074589127e-06,
+      "loss": 1.4094,
+      "step": 23072
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9085176991150444e-06,
+      "loss": 2.651,
+      "step": 23073
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9084702907711757e-06,
+      "loss": 3.7125,
+      "step": 23074
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9084228824273074e-06,
+      "loss": 3.6801,
+      "step": 23075
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9083754740834387e-06,
+      "loss": 2.2639,
+      "step": 23076
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.90832806573957e-06,
+      "loss": 1.3748,
+      "step": 23077
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9082806573957016e-06,
+      "loss": 2.2348,
+      "step": 23078
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9082332490518333e-06,
+      "loss": 3.0446,
+      "step": 23079
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9081858407079646e-06,
+      "loss": 2.5816,
+      "step": 23080
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9081384323640963e-06,
+      "loss": 2.8648,
+      "step": 23081
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.9080910240202276e-06,
+      "loss": 3.2402,
+      "step": 23082
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.908043615676359e-06,
+      "loss": 2.1159,
+      "step": 23083
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9079962073324906e-06,
+      "loss": 3.0887,
+      "step": 23084
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907948798988622e-06,
+      "loss": 1.8635,
+      "step": 23085
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9079013906447536e-06,
+      "loss": 2.4712,
+      "step": 23086
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9078539823008853e-06,
+      "loss": 2.426,
+      "step": 23087
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9078065739570166e-06,
+      "loss": 3.5958,
+      "step": 23088
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907759165613148e-06,
+      "loss": 2.0,
+      "step": 23089
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9077117572692796e-06,
+      "loss": 3.1514,
+      "step": 23090
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907664348925411e-06,
+      "loss": 2.2731,
+      "step": 23091
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9076169405815425e-06,
+      "loss": 1.9233,
+      "step": 23092
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907569532237674e-06,
+      "loss": 2.8512,
+      "step": 23093
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9075221238938055e-06,
+      "loss": 1.8857,
+      "step": 23094
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907474715549937e-06,
+      "loss": 2.6135,
+      "step": 23095
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907427307206068e-06,
+      "loss": 2.5933,
+      "step": 23096
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9073798988621998e-06,
+      "loss": 2.8396,
+      "step": 23097
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9073324905183313e-06,
+      "loss": 2.3161,
+      "step": 23098
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9072850821744626e-06,
+      "loss": 3.0489,
+      "step": 23099
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9072376738305943e-06,
+      "loss": 2.1822,
+      "step": 23100
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9071902654867258e-06,
+      "loss": 3.0017,
+      "step": 23101
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.907142857142857e-06,
+      "loss": 1.8917,
+      "step": 23102
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9070954487989887e-06,
+      "loss": 2.4349,
+      "step": 23103
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9070480404551202e-06,
+      "loss": 3.1047,
+      "step": 23104
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9070006321112515e-06,
+      "loss": 2.183,
+      "step": 23105
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9069532237673832e-06,
+      "loss": 2.0415,
+      "step": 23106
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9069058154235147e-06,
+      "loss": 1.801,
+      "step": 23107
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.906858407079646e-06,
+      "loss": 2.4625,
+      "step": 23108
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9068109987357777e-06,
+      "loss": 2.7622,
+      "step": 23109
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.906763590391909e-06,
+      "loss": 3.4268,
+      "step": 23110
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9067161820480405e-06,
+      "loss": 2.5058,
+      "step": 23111
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9066687737041722e-06,
+      "loss": 2.1373,
+      "step": 23112
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9066213653603034e-06,
+      "loss": 2.1801,
+      "step": 23113
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.906573957016435e-06,
+      "loss": 2.4207,
+      "step": 23114
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9065265486725662e-06,
+      "loss": 2.9021,
+      "step": 23115
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.906479140328698e-06,
+      "loss": 2.7925,
+      "step": 23116
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9064317319848294e-06,
+      "loss": 2.9012,
+      "step": 23117
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9063843236409607e-06,
+      "loss": 3.1862,
+      "step": 23118
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9063369152970924e-06,
+      "loss": 2.8892,
+      "step": 23119
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9062895069532239e-06,
+      "loss": 2.6141,
+      "step": 23120
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9062420986093552e-06,
+      "loss": 3.1059,
+      "step": 23121
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9061946902654869e-06,
+      "loss": 2.7485,
+      "step": 23122
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9061472819216182e-06,
+      "loss": 3.4341,
+      "step": 23123
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9060998735777496e-06,
+      "loss": 2.338,
+      "step": 23124
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9060524652338814e-06,
+      "loss": 2.2034,
+      "step": 23125
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9060050568900126e-06,
+      "loss": 2.4392,
+      "step": 23126
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9059576485461441e-06,
+      "loss": 3.2884,
+      "step": 23127
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9059102402022758e-06,
+      "loss": 2.5035,
+      "step": 23128
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9058628318584071e-06,
+      "loss": 2.7716,
+      "step": 23129
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9058154235145386e-06,
+      "loss": 1.9251,
+      "step": 23130
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9057680151706703e-06,
+      "loss": 2.7797,
+      "step": 23131
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9057206068268016e-06,
+      "loss": 3.2224,
+      "step": 23132
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.905673198482933e-06,
+      "loss": 2.0921,
+      "step": 23133
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9056257901390648e-06,
+      "loss": 3.2876,
+      "step": 23134
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.905578381795196e-06,
+      "loss": 2.6475,
+      "step": 23135
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9055309734513273e-06,
+      "loss": 1.0277,
+      "step": 23136
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9054835651074588e-06,
+      "loss": 1.7157,
+      "step": 23137
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9054361567635905e-06,
+      "loss": 3.3672,
+      "step": 23138
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9053887484197218e-06,
+      "loss": 1.2994,
+      "step": 23139
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9053413400758533e-06,
+      "loss": 3.4561,
+      "step": 23140
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.905293931731985e-06,
+      "loss": 1.5983,
+      "step": 23141
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9052465233881163e-06,
+      "loss": 2.3557,
+      "step": 23142
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9051991150442478e-06,
+      "loss": 2.9551,
+      "step": 23143
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9051517067003795e-06,
+      "loss": 3.8347,
+      "step": 23144
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9051042983565108e-06,
+      "loss": 1.9552,
+      "step": 23145
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.9050568900126423e-06,
+      "loss": 2.1003,
+      "step": 23146
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.905009481668774e-06,
+      "loss": 2.5795,
+      "step": 23147
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9049620733249053e-06,
+      "loss": 2.8707,
+      "step": 23148
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9049146649810365e-06,
+      "loss": 2.2917,
+      "step": 23149
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9048672566371682e-06,
+      "loss": 3.1965,
+      "step": 23150
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9048198482932997e-06,
+      "loss": 1.9886,
+      "step": 23151
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.904772439949431e-06,
+      "loss": 2.0799,
+      "step": 23152
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9047250316055627e-06,
+      "loss": 4.4188,
+      "step": 23153
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9046776232616942e-06,
+      "loss": 2.3123,
+      "step": 23154
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9046302149178255e-06,
+      "loss": 3.0151,
+      "step": 23155
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9045828065739572e-06,
+      "loss": 1.1325,
+      "step": 23156
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9045353982300887e-06,
+      "loss": 1.8083,
+      "step": 23157
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.90448798988622e-06,
+      "loss": 3.0152,
+      "step": 23158
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9044405815423512e-06,
+      "loss": 3.71,
+      "step": 23159
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9043931731984832e-06,
+      "loss": 3.6187,
+      "step": 23160
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9043457648546144e-06,
+      "loss": 2.6338,
+      "step": 23161
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9042983565107457e-06,
+      "loss": 1.9329,
+      "step": 23162
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9042509481668774e-06,
+      "loss": 2.6672,
+      "step": 23163
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.904203539823009e-06,
+      "loss": 2.2661,
+      "step": 23164
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9041561314791402e-06,
+      "loss": 2.4756,
+      "step": 23165
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.904108723135272e-06,
+      "loss": 2.9626,
+      "step": 23166
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9040613147914034e-06,
+      "loss": 1.9929,
+      "step": 23167
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9040139064475347e-06,
+      "loss": 1.6671,
+      "step": 23168
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9039664981036664e-06,
+      "loss": 3.0365,
+      "step": 23169
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9039190897597979e-06,
+      "loss": 1.8173,
+      "step": 23170
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9038716814159291e-06,
+      "loss": 2.6917,
+      "step": 23171
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9038242730720609e-06,
+      "loss": 2.4416,
+      "step": 23172
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9037768647281921e-06,
+      "loss": 1.4013,
+      "step": 23173
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9037294563843236e-06,
+      "loss": 2.78,
+      "step": 23174
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9036820480404553e-06,
+      "loss": 1.7168,
+      "step": 23175
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9036346396965866e-06,
+      "loss": 2.2791,
+      "step": 23176
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.903587231352718e-06,
+      "loss": 3.2528,
+      "step": 23177
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9035398230088498e-06,
+      "loss": 3.0677,
+      "step": 23178
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.903492414664981e-06,
+      "loss": 2.4672,
+      "step": 23179
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9034450063211126e-06,
+      "loss": 1.6392,
+      "step": 23180
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9033975979772439e-06,
+      "loss": 3.2144,
+      "step": 23181
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9033501896333756e-06,
+      "loss": 3.3364,
+      "step": 23182
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.903302781289507e-06,
+      "loss": 3.1875,
+      "step": 23183
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9032553729456383e-06,
+      "loss": 2.415,
+      "step": 23184
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.90320796460177e-06,
+      "loss": 2.436,
+      "step": 23185
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9031605562579013e-06,
+      "loss": 3.0289,
+      "step": 23186
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9031131479140328e-06,
+      "loss": 2.0946,
+      "step": 23187
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9030657395701645e-06,
+      "loss": 2.5433,
+      "step": 23188
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9030183312262958e-06,
+      "loss": 2.0569,
+      "step": 23189
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9029709228824273e-06,
+      "loss": 2.185,
+      "step": 23190
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.902923514538559e-06,
+      "loss": 2.9606,
+      "step": 23191
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9028761061946903e-06,
+      "loss": 2.006,
+      "step": 23192
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9028286978508218e-06,
+      "loss": 2.7363,
+      "step": 23193
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9027812895069535e-06,
+      "loss": 3.7894,
+      "step": 23194
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9027338811630847e-06,
+      "loss": 3.0848,
+      "step": 23195
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9026864728192162e-06,
+      "loss": 2.5965,
+      "step": 23196
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.902639064475348e-06,
+      "loss": 2.7385,
+      "step": 23197
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9025916561314792e-06,
+      "loss": 2.4353,
+      "step": 23198
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9025442477876105e-06,
+      "loss": 1.7967,
+      "step": 23199
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9024968394437422e-06,
+      "loss": 2.2992,
+      "step": 23200
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9024494310998737e-06,
+      "loss": 3.5535,
+      "step": 23201
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.902402022756005e-06,
+      "loss": 2.97,
+      "step": 23202
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9023546144121365e-06,
+      "loss": 2.2235,
+      "step": 23203
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9023072060682682e-06,
+      "loss": 2.2575,
+      "step": 23204
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9022597977243995e-06,
+      "loss": 2.9581,
+      "step": 23205
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.902212389380531e-06,
+      "loss": 3.7478,
+      "step": 23206
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9021649810366627e-06,
+      "loss": 3.5425,
+      "step": 23207
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.902117572692794e-06,
+      "loss": 2.8489,
+      "step": 23208
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9020701643489254e-06,
+      "loss": 2.5494,
+      "step": 23209
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.9020227560050571e-06,
+      "loss": 2.0011,
+      "step": 23210
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9019753476611884e-06,
+      "loss": 2.9048,
+      "step": 23211
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9019279393173197e-06,
+      "loss": 2.8531,
+      "step": 23212
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9018805309734514e-06,
+      "loss": 3.5876,
+      "step": 23213
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9018331226295829e-06,
+      "loss": 2.7773,
+      "step": 23214
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9017857142857142e-06,
+      "loss": 2.5265,
+      "step": 23215
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9017383059418459e-06,
+      "loss": 2.1016,
+      "step": 23216
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9016908975979774e-06,
+      "loss": 3.8754,
+      "step": 23217
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9016434892541086e-06,
+      "loss": 2.032,
+      "step": 23218
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9015960809102403e-06,
+      "loss": 3.164,
+      "step": 23219
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9015486725663718e-06,
+      "loss": 2.4672,
+      "step": 23220
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9015012642225031e-06,
+      "loss": 2.7208,
+      "step": 23221
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9014538558786348e-06,
+      "loss": 2.7576,
+      "step": 23222
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9014064475347663e-06,
+      "loss": 2.5111,
+      "step": 23223
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9013590391908976e-06,
+      "loss": 2.519,
+      "step": 23224
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9013116308470289e-06,
+      "loss": 3.7068,
+      "step": 23225
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9012642225031606e-06,
+      "loss": 3.3492,
+      "step": 23226
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.901216814159292e-06,
+      "loss": 3.9293,
+      "step": 23227
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9011694058154234e-06,
+      "loss": 2.2912,
+      "step": 23228
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.901121997471555e-06,
+      "loss": 3.0687,
+      "step": 23229
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9010745891276866e-06,
+      "loss": 2.318,
+      "step": 23230
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9010271807838178e-06,
+      "loss": 4.1924,
+      "step": 23231
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9009797724399495e-06,
+      "loss": 2.413,
+      "step": 23232
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.900932364096081e-06,
+      "loss": 3.3325,
+      "step": 23233
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9008849557522123e-06,
+      "loss": 1.2059,
+      "step": 23234
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.900837547408344e-06,
+      "loss": 2.5403,
+      "step": 23235
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9007901390644755e-06,
+      "loss": 2.2887,
+      "step": 23236
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9007427307206068e-06,
+      "loss": 3.3465,
+      "step": 23237
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9006953223767385e-06,
+      "loss": 3.5555,
+      "step": 23238
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9006479140328698e-06,
+      "loss": 2.7151,
+      "step": 23239
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9006005056890013e-06,
+      "loss": 2.949,
+      "step": 23240
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.900553097345133e-06,
+      "loss": 2.8574,
+      "step": 23241
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9005056890012642e-06,
+      "loss": 2.5342,
+      "step": 23242
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9004582806573957e-06,
+      "loss": 1.7979,
+      "step": 23243
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9004108723135274e-06,
+      "loss": 3.5975,
+      "step": 23244
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9003634639696587e-06,
+      "loss": 3.1298,
+      "step": 23245
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9003160556257902e-06,
+      "loss": 2.5611,
+      "step": 23246
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9002686472819215e-06,
+      "loss": 2.7521,
+      "step": 23247
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9002212389380532e-06,
+      "loss": 3.0932,
+      "step": 23248
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9001738305941847e-06,
+      "loss": 1.973,
+      "step": 23249
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.900126422250316e-06,
+      "loss": 2.3506,
+      "step": 23250
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.9000790139064477e-06,
+      "loss": 3.7998,
+      "step": 23251
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.900031605562579e-06,
+      "loss": 1.6676,
+      "step": 23252
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8999841972187104e-06,
+      "loss": 2.2392,
+      "step": 23253
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8999367888748422e-06,
+      "loss": 2.9654,
+      "step": 23254
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8998893805309734e-06,
+      "loss": 2.0611,
+      "step": 23255
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.899841972187105e-06,
+      "loss": 2.3493,
+      "step": 23256
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8997945638432366e-06,
+      "loss": 3.9695,
+      "step": 23257
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.899747155499368e-06,
+      "loss": 2.7553,
+      "step": 23258
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8996997471554994e-06,
+      "loss": 2.9004,
+      "step": 23259
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.899652338811631e-06,
+      "loss": 2.6154,
+      "step": 23260
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8996049304677624e-06,
+      "loss": 2.2238,
+      "step": 23261
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8995575221238939e-06,
+      "loss": 2.5768,
+      "step": 23262
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8995101137800256e-06,
+      "loss": 2.7829,
+      "step": 23263
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8994627054361569e-06,
+      "loss": 1.7644,
+      "step": 23264
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8994152970922881e-06,
+      "loss": 3.8448,
+      "step": 23265
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8993678887484198e-06,
+      "loss": 2.5765,
+      "step": 23266
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8993204804045513e-06,
+      "loss": 2.989,
+      "step": 23267
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8992730720606826e-06,
+      "loss": 2.4325,
+      "step": 23268
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8992256637168141e-06,
+      "loss": 2.4408,
+      "step": 23269
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8991782553729458e-06,
+      "loss": 3.1162,
+      "step": 23270
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.899130847029077e-06,
+      "loss": 2.0316,
+      "step": 23271
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8990834386852086e-06,
+      "loss": 2.6556,
+      "step": 23272
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.8990360303413403e-06,
+      "loss": 2.1049,
+      "step": 23273
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8989886219974716e-06,
+      "loss": 1.9143,
+      "step": 23274
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898941213653603e-06,
+      "loss": 2.145,
+      "step": 23275
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8988938053097348e-06,
+      "loss": 2.8926,
+      "step": 23276
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898846396965866e-06,
+      "loss": 2.2781,
+      "step": 23277
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8987989886219973e-06,
+      "loss": 3.4493,
+      "step": 23278
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898751580278129e-06,
+      "loss": 3.9508,
+      "step": 23279
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8987041719342605e-06,
+      "loss": 1.9351,
+      "step": 23280
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8986567635903918e-06,
+      "loss": 3.4852,
+      "step": 23281
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8986093552465235e-06,
+      "loss": 2.6874,
+      "step": 23282
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898561946902655e-06,
+      "loss": 1.7775,
+      "step": 23283
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8985145385587863e-06,
+      "loss": 2.4251,
+      "step": 23284
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898467130214918e-06,
+      "loss": 2.8052,
+      "step": 23285
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8984197218710495e-06,
+      "loss": 1.9989,
+      "step": 23286
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8983723135271808e-06,
+      "loss": 3.5523,
+      "step": 23287
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8983249051833125e-06,
+      "loss": 2.3057,
+      "step": 23288
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898277496839444e-06,
+      "loss": 2.5259,
+      "step": 23289
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8982300884955752e-06,
+      "loss": 2.4789,
+      "step": 23290
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8981826801517065e-06,
+      "loss": 1.2726,
+      "step": 23291
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8981352718078382e-06,
+      "loss": 3.6577,
+      "step": 23292
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8980878634639697e-06,
+      "loss": 2.1479,
+      "step": 23293
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.898040455120101e-06,
+      "loss": 2.6357,
+      "step": 23294
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8979930467762327e-06,
+      "loss": 3.2053,
+      "step": 23295
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8979456384323642e-06,
+      "loss": 1.4957,
+      "step": 23296
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8978982300884955e-06,
+      "loss": 2.5085,
+      "step": 23297
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8978508217446272e-06,
+      "loss": 2.3121,
+      "step": 23298
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8978034134007587e-06,
+      "loss": 1.5021,
+      "step": 23299
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.89775600505689e-06,
+      "loss": 2.1296,
+      "step": 23300
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8977085967130216e-06,
+      "loss": 2.1092,
+      "step": 23301
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.897661188369153e-06,
+      "loss": 3.7888,
+      "step": 23302
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8976137800252844e-06,
+      "loss": 2.1591,
+      "step": 23303
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8975663716814161e-06,
+      "loss": 2.7342,
+      "step": 23304
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8975189633375474e-06,
+      "loss": 1.661,
+      "step": 23305
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.897471554993679e-06,
+      "loss": 2.9055,
+      "step": 23306
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8974241466498106e-06,
+      "loss": 3.1713,
+      "step": 23307
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8973767383059419e-06,
+      "loss": 1.6807,
+      "step": 23308
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8973293299620734e-06,
+      "loss": 2.2609,
+      "step": 23309
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.897281921618205e-06,
+      "loss": 3.0995,
+      "step": 23310
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8972345132743364e-06,
+      "loss": 1.7348,
+      "step": 23311
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8971871049304678e-06,
+      "loss": 1.919,
+      "step": 23312
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8971396965865991e-06,
+      "loss": 2.3344,
+      "step": 23313
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8970922882427308e-06,
+      "loss": 2.7431,
+      "step": 23314
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8970448798988621e-06,
+      "loss": 1.6721,
+      "step": 23315
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8969974715549936e-06,
+      "loss": 2.5082,
+      "step": 23316
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8969500632111253e-06,
+      "loss": 3.1236,
+      "step": 23317
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8969026548672566e-06,
+      "loss": 2.1819,
+      "step": 23318
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.896855246523388e-06,
+      "loss": 3.0839,
+      "step": 23319
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8968078381795198e-06,
+      "loss": 2.9043,
+      "step": 23320
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.896760429835651e-06,
+      "loss": 3.5475,
+      "step": 23321
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8967130214917826e-06,
+      "loss": 3.4775,
+      "step": 23322
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8966656131479143e-06,
+      "loss": 2.6756,
+      "step": 23323
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8966182048040455e-06,
+      "loss": 2.5143,
+      "step": 23324
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.896570796460177e-06,
+      "loss": 2.73,
+      "step": 23325
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8965233881163087e-06,
+      "loss": 2.5254,
+      "step": 23326
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.89647597977244e-06,
+      "loss": 2.0772,
+      "step": 23327
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8964285714285713e-06,
+      "loss": 2.2182,
+      "step": 23328
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.896381163084703e-06,
+      "loss": 2.2352,
+      "step": 23329
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8963337547408345e-06,
+      "loss": 2.626,
+      "step": 23330
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8962863463969658e-06,
+      "loss": 1.97,
+      "step": 23331
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8962389380530975e-06,
+      "loss": 2.8314,
+      "step": 23332
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.896191529709229e-06,
+      "loss": 2.1853,
+      "step": 23333
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8961441213653603e-06,
+      "loss": 3.1547,
+      "step": 23334
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8960967130214917e-06,
+      "loss": 3.8499,
+      "step": 23335
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8960493046776235e-06,
+      "loss": 2.8194,
+      "step": 23336
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.8960018963337547e-06,
+      "loss": 2.8615,
+      "step": 23337
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8959544879898862e-06,
+      "loss": 2.2997,
+      "step": 23338
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.895907079646018e-06,
+      "loss": 3.3897,
+      "step": 23339
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8958596713021492e-06,
+      "loss": 3.7584,
+      "step": 23340
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8958122629582805e-06,
+      "loss": 2.2693,
+      "step": 23341
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8957648546144122e-06,
+      "loss": 1.3698,
+      "step": 23342
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8957174462705437e-06,
+      "loss": 3.9645,
+      "step": 23343
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.895670037926675e-06,
+      "loss": 0.6189,
+      "step": 23344
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8956226295828067e-06,
+      "loss": 3.0573,
+      "step": 23345
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8955752212389382e-06,
+      "loss": 3.187,
+      "step": 23346
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8955278128950694e-06,
+      "loss": 1.4029,
+      "step": 23347
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8954804045512011e-06,
+      "loss": 3.1369,
+      "step": 23348
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8954329962073326e-06,
+      "loss": 2.8877,
+      "step": 23349
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.895385587863464e-06,
+      "loss": 1.548,
+      "step": 23350
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8953381795195956e-06,
+      "loss": 3.953,
+      "step": 23351
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8952907711757271e-06,
+      "loss": 2.2088,
+      "step": 23352
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8952433628318584e-06,
+      "loss": 3.2184,
+      "step": 23353
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8951959544879897e-06,
+      "loss": 1.7813,
+      "step": 23354
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8951485461441214e-06,
+      "loss": 2.2684,
+      "step": 23355
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8951011378002529e-06,
+      "loss": 2.6717,
+      "step": 23356
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8950537294563842e-06,
+      "loss": 3.1342,
+      "step": 23357
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8950063211125159e-06,
+      "loss": 2.2797,
+      "step": 23358
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8949589127686473e-06,
+      "loss": 3.165,
+      "step": 23359
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8949115044247786e-06,
+      "loss": 1.9304,
+      "step": 23360
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8948640960809103e-06,
+      "loss": 2.6894,
+      "step": 23361
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8948166877370418e-06,
+      "loss": 2.7413,
+      "step": 23362
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.894769279393173e-06,
+      "loss": 2.8695,
+      "step": 23363
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8947218710493048e-06,
+      "loss": 3.014,
+      "step": 23364
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8946744627054363e-06,
+      "loss": 2.9622,
+      "step": 23365
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8946270543615676e-06,
+      "loss": 2.7749,
+      "step": 23366
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8945796460176993e-06,
+      "loss": 2.8331,
+      "step": 23367
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8945322376738306e-06,
+      "loss": 2.9341,
+      "step": 23368
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.894484829329962e-06,
+      "loss": 2.5077,
+      "step": 23369
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8944374209860938e-06,
+      "loss": 3.6536,
+      "step": 23370
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.894390012642225e-06,
+      "loss": 1.7767,
+      "step": 23371
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8943426042983565e-06,
+      "loss": 3.2974,
+      "step": 23372
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8942951959544882e-06,
+      "loss": 1.3945,
+      "step": 23373
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8942477876106195e-06,
+      "loss": 2.3551,
+      "step": 23374
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.894200379266751e-06,
+      "loss": 3.426,
+      "step": 23375
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8941529709228823e-06,
+      "loss": 4.0106,
+      "step": 23376
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.894105562579014e-06,
+      "loss": 2.8676,
+      "step": 23377
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8940581542351455e-06,
+      "loss": 3.2516,
+      "step": 23378
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8940107458912768e-06,
+      "loss": 3.6802,
+      "step": 23379
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8939633375474085e-06,
+      "loss": 1.3179,
+      "step": 23380
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8939159292035398e-06,
+      "loss": 2.4157,
+      "step": 23381
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8938685208596712e-06,
+      "loss": 1.8662,
+      "step": 23382
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.893821112515803e-06,
+      "loss": 2.533,
+      "step": 23383
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8937737041719342e-06,
+      "loss": 3.0716,
+      "step": 23384
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8937262958280657e-06,
+      "loss": 2.1885,
+      "step": 23385
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8936788874841974e-06,
+      "loss": 3.8724,
+      "step": 23386
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8936314791403287e-06,
+      "loss": 2.2241,
+      "step": 23387
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8935840707964602e-06,
+      "loss": 2.5999,
+      "step": 23388
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.893536662452592e-06,
+      "loss": 3.6221,
+      "step": 23389
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8934892541087232e-06,
+      "loss": 3.3093,
+      "step": 23390
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8934418457648547e-06,
+      "loss": 3.9102,
+      "step": 23391
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8933944374209864e-06,
+      "loss": 2.9793,
+      "step": 23392
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8933470290771177e-06,
+      "loss": 2.5607,
+      "step": 23393
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.893299620733249e-06,
+      "loss": 2.3724,
+      "step": 23394
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8932522123893806e-06,
+      "loss": 2.7663,
+      "step": 23395
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8932048040455121e-06,
+      "loss": 2.1308,
+      "step": 23396
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8931573957016434e-06,
+      "loss": 3.0742,
+      "step": 23397
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.893109987357775e-06,
+      "loss": 3.5522,
+      "step": 23398
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8930625790139066e-06,
+      "loss": 2.5153,
+      "step": 23399
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.8930151706700379e-06,
+      "loss": 3.428,
+      "step": 23400
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8929677623261694e-06,
+      "loss": 2.6441,
+      "step": 23401
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.892920353982301e-06,
+      "loss": 1.5799,
+      "step": 23402
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8928729456384324e-06,
+      "loss": 1.1529,
+      "step": 23403
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8928255372945639e-06,
+      "loss": 2.1121,
+      "step": 23404
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8927781289506956e-06,
+      "loss": 2.8981,
+      "step": 23405
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8927307206068268e-06,
+      "loss": 2.6177,
+      "step": 23406
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8926833122629581e-06,
+      "loss": 3.2211,
+      "step": 23407
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8926359039190898e-06,
+      "loss": 2.9725,
+      "step": 23408
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8925884955752213e-06,
+      "loss": 2.3562,
+      "step": 23409
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8925410872313526e-06,
+      "loss": 2.4702,
+      "step": 23410
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8924936788874843e-06,
+      "loss": 2.6848,
+      "step": 23411
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8924462705436158e-06,
+      "loss": 3.0132,
+      "step": 23412
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.892398862199747e-06,
+      "loss": 1.8416,
+      "step": 23413
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8923514538558788e-06,
+      "loss": 3.0577,
+      "step": 23414
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8923040455120103e-06,
+      "loss": 3.2466,
+      "step": 23415
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8922566371681416e-06,
+      "loss": 2.4101,
+      "step": 23416
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8922092288242733e-06,
+      "loss": 3.3087,
+      "step": 23417
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8921618204804048e-06,
+      "loss": 3.5998,
+      "step": 23418
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.892114412136536e-06,
+      "loss": 3.8822,
+      "step": 23419
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8920670037926673e-06,
+      "loss": 2.8644,
+      "step": 23420
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.892019595448799e-06,
+      "loss": 3.2953,
+      "step": 23421
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8919721871049305e-06,
+      "loss": 1.5837,
+      "step": 23422
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8919247787610618e-06,
+      "loss": 4.0747,
+      "step": 23423
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8918773704171935e-06,
+      "loss": 2.4475,
+      "step": 23424
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.891829962073325e-06,
+      "loss": 2.5973,
+      "step": 23425
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8917825537294563e-06,
+      "loss": 2.5888,
+      "step": 23426
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.891735145385588e-06,
+      "loss": 2.8046,
+      "step": 23427
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8916877370417195e-06,
+      "loss": 2.5257,
+      "step": 23428
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8916403286978507e-06,
+      "loss": 2.147,
+      "step": 23429
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8915929203539824e-06,
+      "loss": 2.2434,
+      "step": 23430
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.891545512010114e-06,
+      "loss": 2.7851,
+      "step": 23431
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8914981036662452e-06,
+      "loss": 2.0813,
+      "step": 23432
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.891450695322377e-06,
+      "loss": 2.223,
+      "step": 23433
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8914032869785082e-06,
+      "loss": 3.0682,
+      "step": 23434
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8913558786346397e-06,
+      "loss": 3.475,
+      "step": 23435
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8913084702907714e-06,
+      "loss": 2.2178,
+      "step": 23436
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8912610619469027e-06,
+      "loss": 3.5841,
+      "step": 23437
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8912136536030342e-06,
+      "loss": 1.7159,
+      "step": 23438
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8911662452591659e-06,
+      "loss": 2.7928,
+      "step": 23439
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8911188369152972e-06,
+      "loss": 2.3202,
+      "step": 23440
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8910714285714286e-06,
+      "loss": 3.0943,
+      "step": 23441
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.89102402022756e-06,
+      "loss": 2.7838,
+      "step": 23442
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8909766118836916e-06,
+      "loss": 2.2493,
+      "step": 23443
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.890929203539823e-06,
+      "loss": 2.9036,
+      "step": 23444
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8908817951959544e-06,
+      "loss": 2.5259,
+      "step": 23445
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8908343868520861e-06,
+      "loss": 3.1237,
+      "step": 23446
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8907869785082174e-06,
+      "loss": 3.9183,
+      "step": 23447
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8907395701643489e-06,
+      "loss": 3.7003,
+      "step": 23448
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8906921618204806e-06,
+      "loss": 2.3176,
+      "step": 23449
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8906447534766119e-06,
+      "loss": 1.6035,
+      "step": 23450
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8905973451327434e-06,
+      "loss": 3.2421,
+      "step": 23451
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.890549936788875e-06,
+      "loss": 0.9963,
+      "step": 23452
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8905025284450063e-06,
+      "loss": 2.1619,
+      "step": 23453
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8904551201011378e-06,
+      "loss": 2.4491,
+      "step": 23454
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8904077117572695e-06,
+      "loss": 1.727,
+      "step": 23455
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8903603034134008e-06,
+      "loss": 2.3077,
+      "step": 23456
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.890312895069532e-06,
+      "loss": 2.1269,
+      "step": 23457
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8902654867256638e-06,
+      "loss": 2.4683,
+      "step": 23458
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8902180783817953e-06,
+      "loss": 3.0142,
+      "step": 23459
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8901706700379266e-06,
+      "loss": 4.0209,
+      "step": 23460
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8901232616940583e-06,
+      "loss": 3.0862,
+      "step": 23461
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.8900758533501898e-06,
+      "loss": 2.2583,
+      "step": 23462
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.890028445006321e-06,
+      "loss": 2.5166,
+      "step": 23463
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8899810366624525e-06,
+      "loss": 2.4052,
+      "step": 23464
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8899336283185842e-06,
+      "loss": 2.5763,
+      "step": 23465
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8898862199747155e-06,
+      "loss": 2.9805,
+      "step": 23466
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.889838811630847e-06,
+      "loss": 2.1134,
+      "step": 23467
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8897914032869787e-06,
+      "loss": 1.8721,
+      "step": 23468
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.88974399494311e-06,
+      "loss": 4.1972,
+      "step": 23469
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8896965865992413e-06,
+      "loss": 3.4526,
+      "step": 23470
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.889649178255373e-06,
+      "loss": 2.3129,
+      "step": 23471
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8896017699115045e-06,
+      "loss": 2.4182,
+      "step": 23472
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8895543615676358e-06,
+      "loss": 2.0387,
+      "step": 23473
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8895069532237675e-06,
+      "loss": 2.6976,
+      "step": 23474
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.889459544879899e-06,
+      "loss": 3.4377,
+      "step": 23475
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8894121365360302e-06,
+      "loss": 3.3969,
+      "step": 23476
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.889364728192162e-06,
+      "loss": 2.9348,
+      "step": 23477
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8893173198482934e-06,
+      "loss": 2.4246,
+      "step": 23478
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8892699115044247e-06,
+      "loss": 2.255,
+      "step": 23479
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8892225031605564e-06,
+      "loss": 1.9571,
+      "step": 23480
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.889175094816688e-06,
+      "loss": 2.9574,
+      "step": 23481
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8891276864728192e-06,
+      "loss": 2.5769,
+      "step": 23482
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.889080278128951e-06,
+      "loss": 1.7746,
+      "step": 23483
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8890328697850822e-06,
+      "loss": 4.1446,
+      "step": 23484
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8889854614412137e-06,
+      "loss": 2.238,
+      "step": 23485
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.888938053097345e-06,
+      "loss": 2.3068,
+      "step": 23486
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8888906447534767e-06,
+      "loss": 2.167,
+      "step": 23487
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8888432364096081e-06,
+      "loss": 2.7451,
+      "step": 23488
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8887958280657394e-06,
+      "loss": 3.47,
+      "step": 23489
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8887484197218711e-06,
+      "loss": 3.0854,
+      "step": 23490
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8887010113780026e-06,
+      "loss": 1.7584,
+      "step": 23491
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.888653603034134e-06,
+      "loss": 2.7696,
+      "step": 23492
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8886061946902656e-06,
+      "loss": 2.6628,
+      "step": 23493
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.888558786346397e-06,
+      "loss": 3.7322,
+      "step": 23494
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8885113780025284e-06,
+      "loss": 2.8594,
+      "step": 23495
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.88846396965866e-06,
+      "loss": 2.5335,
+      "step": 23496
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8884165613147914e-06,
+      "loss": 3.27,
+      "step": 23497
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8883691529709229e-06,
+      "loss": 1.6656,
+      "step": 23498
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8883217446270546e-06,
+      "loss": 2.7923,
+      "step": 23499
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8882743362831858e-06,
+      "loss": 3.5372,
+      "step": 23500
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8882269279393173e-06,
+      "loss": 3.6871,
+      "step": 23501
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.888179519595449e-06,
+      "loss": 3.5741,
+      "step": 23502
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8881321112515803e-06,
+      "loss": 2.923,
+      "step": 23503
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8880847029077118e-06,
+      "loss": 2.4625,
+      "step": 23504
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8880372945638435e-06,
+      "loss": 2.8592,
+      "step": 23505
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8879898862199748e-06,
+      "loss": 2.0533,
+      "step": 23506
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8879424778761063e-06,
+      "loss": 1.7898,
+      "step": 23507
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8878950695322376e-06,
+      "loss": 2.576,
+      "step": 23508
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8878476611883693e-06,
+      "loss": 3.3669,
+      "step": 23509
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8878002528445005e-06,
+      "loss": 2.823,
+      "step": 23510
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.887752844500632e-06,
+      "loss": 2.868,
+      "step": 23511
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8877054361567637e-06,
+      "loss": 1.6509,
+      "step": 23512
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.887658027812895e-06,
+      "loss": 2.0623,
+      "step": 23513
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8876106194690265e-06,
+      "loss": 1.6615,
+      "step": 23514
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8875632111251582e-06,
+      "loss": 2.6077,
+      "step": 23515
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8875158027812895e-06,
+      "loss": 4.5593,
+      "step": 23516
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.887468394437421e-06,
+      "loss": 2.3444,
+      "step": 23517
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8874209860935527e-06,
+      "loss": 3.0572,
+      "step": 23518
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.887373577749684e-06,
+      "loss": 2.6044,
+      "step": 23519
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8873261694058155e-06,
+      "loss": 3.7714,
+      "step": 23520
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8872787610619472e-06,
+      "loss": 2.0237,
+      "step": 23521
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8872313527180785e-06,
+      "loss": 3.4998,
+      "step": 23522
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8871839443742097e-06,
+      "loss": 3.2103,
+      "step": 23523
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8871365360303414e-06,
+      "loss": 2.6768,
+      "step": 23524
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.887089127686473e-06,
+      "loss": 1.4519,
+      "step": 23525
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.8870417193426042e-06,
+      "loss": 2.3638,
+      "step": 23526
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.886994310998736e-06,
+      "loss": 2.6733,
+      "step": 23527
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8869469026548674e-06,
+      "loss": 1.2657,
+      "step": 23528
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8868994943109987e-06,
+      "loss": 1.6089,
+      "step": 23529
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8868520859671302e-06,
+      "loss": 1.7468,
+      "step": 23530
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8868046776232619e-06,
+      "loss": 1.521,
+      "step": 23531
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8867572692793932e-06,
+      "loss": 3.345,
+      "step": 23532
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8867098609355247e-06,
+      "loss": 2.5419,
+      "step": 23533
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8866624525916564e-06,
+      "loss": 2.3053,
+      "step": 23534
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8866150442477876e-06,
+      "loss": 3.0219,
+      "step": 23535
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.886567635903919e-06,
+      "loss": 3.4494,
+      "step": 23536
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8865202275600506e-06,
+      "loss": 1.9577,
+      "step": 23537
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8864728192161821e-06,
+      "loss": 1.8929,
+      "step": 23538
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8864254108723134e-06,
+      "loss": 4.026,
+      "step": 23539
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.886378002528445e-06,
+      "loss": 2.8453,
+      "step": 23540
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8863305941845766e-06,
+      "loss": 2.2829,
+      "step": 23541
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8862831858407079e-06,
+      "loss": 2.4892,
+      "step": 23542
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8862357774968396e-06,
+      "loss": 3.6712,
+      "step": 23543
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.886188369152971e-06,
+      "loss": 3.215,
+      "step": 23544
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8861409608091024e-06,
+      "loss": 3.3641,
+      "step": 23545
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.886093552465234e-06,
+      "loss": 3.3403,
+      "step": 23546
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8860461441213655e-06,
+      "loss": 2.0879,
+      "step": 23547
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8859987357774968e-06,
+      "loss": 3.2476,
+      "step": 23548
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8859513274336285e-06,
+      "loss": 2.4332,
+      "step": 23549
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8859039190897598e-06,
+      "loss": 2.6289,
+      "step": 23550
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8858565107458913e-06,
+      "loss": 2.211,
+      "step": 23551
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8858091024020226e-06,
+      "loss": 2.6764,
+      "step": 23552
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8857616940581543e-06,
+      "loss": 3.5004,
+      "step": 23553
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8857142857142858e-06,
+      "loss": 2.9089,
+      "step": 23554
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.885666877370417e-06,
+      "loss": 3.0724,
+      "step": 23555
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8856194690265488e-06,
+      "loss": 2.1291,
+      "step": 23556
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8855720606826803e-06,
+      "loss": 2.9224,
+      "step": 23557
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8855246523388115e-06,
+      "loss": 2.536,
+      "step": 23558
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8854772439949432e-06,
+      "loss": 0.7728,
+      "step": 23559
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8854298356510747e-06,
+      "loss": 3.6117,
+      "step": 23560
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.885382427307206e-06,
+      "loss": 2.5225,
+      "step": 23561
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8853350189633377e-06,
+      "loss": 0.9492,
+      "step": 23562
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.885287610619469e-06,
+      "loss": 2.6712,
+      "step": 23563
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8852402022756005e-06,
+      "loss": 1.5674,
+      "step": 23564
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8851927939317322e-06,
+      "loss": 2.5539,
+      "step": 23565
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8851453855878635e-06,
+      "loss": 2.4334,
+      "step": 23566
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.885097977243995e-06,
+      "loss": 1.156,
+      "step": 23567
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8850505689001267e-06,
+      "loss": 2.1745,
+      "step": 23568
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.885003160556258e-06,
+      "loss": 1.7337,
+      "step": 23569
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8849557522123894e-06,
+      "loss": 2.9076,
+      "step": 23570
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8849083438685211e-06,
+      "loss": 3.0459,
+      "step": 23571
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8848609355246524e-06,
+      "loss": 2.1344,
+      "step": 23572
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8848135271807837e-06,
+      "loss": 3.566,
+      "step": 23573
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8847661188369152e-06,
+      "loss": 2.9502,
+      "step": 23574
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.884718710493047e-06,
+      "loss": 3.2072,
+      "step": 23575
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8846713021491782e-06,
+      "loss": 3.1656,
+      "step": 23576
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8846238938053097e-06,
+      "loss": 2.641,
+      "step": 23577
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8845764854614414e-06,
+      "loss": 3.368,
+      "step": 23578
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8845290771175727e-06,
+      "loss": 1.7924,
+      "step": 23579
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8844816687737042e-06,
+      "loss": 3.2717,
+      "step": 23580
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8844342604298359e-06,
+      "loss": 3.6747,
+      "step": 23581
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8843868520859671e-06,
+      "loss": 2.0078,
+      "step": 23582
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8843394437420986e-06,
+      "loss": 2.6467,
+      "step": 23583
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8842920353982303e-06,
+      "loss": 1.7964,
+      "step": 23584
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8842446270543616e-06,
+      "loss": 2.5772,
+      "step": 23585
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.884197218710493e-06,
+      "loss": 2.8458,
+      "step": 23586
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8841498103666246e-06,
+      "loss": 3.445,
+      "step": 23587
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.884102402022756e-06,
+      "loss": 3.264,
+      "step": 23588
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.8840549936788874e-06,
+      "loss": 2.8405,
+      "step": 23589
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.884007585335019e-06,
+      "loss": 1.141,
+      "step": 23590
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8839601769911506e-06,
+      "loss": 3.5134,
+      "step": 23591
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8839127686472818e-06,
+      "loss": 2.7797,
+      "step": 23592
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8838653603034136e-06,
+      "loss": 3.6321,
+      "step": 23593
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.883817951959545e-06,
+      "loss": 2.3618,
+      "step": 23594
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8837705436156763e-06,
+      "loss": 2.4816,
+      "step": 23595
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8837231352718078e-06,
+      "loss": 2.3723,
+      "step": 23596
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8836757269279395e-06,
+      "loss": 4.1792,
+      "step": 23597
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8836283185840708e-06,
+      "loss": 2.5374,
+      "step": 23598
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.883580910240202e-06,
+      "loss": 3.4563,
+      "step": 23599
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8835335018963338e-06,
+      "loss": 2.6985,
+      "step": 23600
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8834860935524653e-06,
+      "loss": 3.0824,
+      "step": 23601
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8834386852085966e-06,
+      "loss": 1.9389,
+      "step": 23602
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8833912768647283e-06,
+      "loss": 2.9679,
+      "step": 23603
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8833438685208598e-06,
+      "loss": 2.3473,
+      "step": 23604
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.883296460176991e-06,
+      "loss": 1.6931,
+      "step": 23605
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8832490518331227e-06,
+      "loss": 2.8245,
+      "step": 23606
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8832016434892542e-06,
+      "loss": 2.3482,
+      "step": 23607
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8831542351453855e-06,
+      "loss": 1.0046,
+      "step": 23608
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8831068268015172e-06,
+      "loss": 2.5805,
+      "step": 23609
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8830594184576487e-06,
+      "loss": 1.4509,
+      "step": 23610
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.88301201011378e-06,
+      "loss": 2.6986,
+      "step": 23611
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8829646017699117e-06,
+      "loss": 2.3986,
+      "step": 23612
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.882917193426043e-06,
+      "loss": 2.9553,
+      "step": 23613
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8828697850821745e-06,
+      "loss": 2.8724,
+      "step": 23614
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8828223767383062e-06,
+      "loss": 1.6682,
+      "step": 23615
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8827749683944374e-06,
+      "loss": 3.5571,
+      "step": 23616
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.882727560050569e-06,
+      "loss": 3.0625,
+      "step": 23617
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8826801517067002e-06,
+      "loss": 3.5386,
+      "step": 23618
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.882632743362832e-06,
+      "loss": 1.5072,
+      "step": 23619
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8825853350189634e-06,
+      "loss": 2.9151,
+      "step": 23620
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8825379266750947e-06,
+      "loss": 2.6475,
+      "step": 23621
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8824905183312264e-06,
+      "loss": 1.3313,
+      "step": 23622
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8824431099873579e-06,
+      "loss": 1.7565,
+      "step": 23623
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8823957016434892e-06,
+      "loss": 1.7349,
+      "step": 23624
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8823482932996209e-06,
+      "loss": 2.1304,
+      "step": 23625
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8823008849557522e-06,
+      "loss": 1.4449,
+      "step": 23626
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8822534766118837e-06,
+      "loss": 3.1231,
+      "step": 23627
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8822060682680154e-06,
+      "loss": 2.4074,
+      "step": 23628
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8821586599241466e-06,
+      "loss": 3.0058,
+      "step": 23629
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8821112515802781e-06,
+      "loss": 2.472,
+      "step": 23630
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8820638432364098e-06,
+      "loss": 2.3933,
+      "step": 23631
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8820164348925411e-06,
+      "loss": 3.3871,
+      "step": 23632
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8819690265486726e-06,
+      "loss": 2.2826,
+      "step": 23633
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8819216182048043e-06,
+      "loss": 2.9992,
+      "step": 23634
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8818742098609356e-06,
+      "loss": 2.2322,
+      "step": 23635
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.881826801517067e-06,
+      "loss": 2.055,
+      "step": 23636
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8817793931731988e-06,
+      "loss": 2.1039,
+      "step": 23637
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.88173198482933e-06,
+      "loss": 2.6501,
+      "step": 23638
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8816845764854613e-06,
+      "loss": 2.2312,
+      "step": 23639
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8816371681415928e-06,
+      "loss": 2.3625,
+      "step": 23640
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8815897597977245e-06,
+      "loss": 2.0855,
+      "step": 23641
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8815423514538558e-06,
+      "loss": 2.5249,
+      "step": 23642
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8814949431099873e-06,
+      "loss": 2.4443,
+      "step": 23643
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.881447534766119e-06,
+      "loss": 1.6411,
+      "step": 23644
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8814001264222503e-06,
+      "loss": 2.1931,
+      "step": 23645
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8813527180783818e-06,
+      "loss": 1.4983,
+      "step": 23646
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8813053097345135e-06,
+      "loss": 4.4584,
+      "step": 23647
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8812579013906448e-06,
+      "loss": 2.2331,
+      "step": 23648
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8812104930467763e-06,
+      "loss": 3.1509,
+      "step": 23649
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.881163084702908e-06,
+      "loss": 2.9863,
+      "step": 23650
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8811156763590393e-06,
+      "loss": 3.1267,
+      "step": 23651
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8810682680151705e-06,
+      "loss": 2.8353,
+      "step": 23652
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.8810208596713022e-06,
+      "loss": 1.5519,
+      "step": 23653
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8809734513274337e-06,
+      "loss": 4.0082,
+      "step": 23654
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.880926042983565e-06,
+      "loss": 2.5044,
+      "step": 23655
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8808786346396967e-06,
+      "loss": 3.1339,
+      "step": 23656
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8808312262958282e-06,
+      "loss": 2.6562,
+      "step": 23657
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8807838179519595e-06,
+      "loss": 2.3608,
+      "step": 23658
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8807364096080912e-06,
+      "loss": 2.9926,
+      "step": 23659
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8806890012642227e-06,
+      "loss": 3.0551,
+      "step": 23660
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.880641592920354e-06,
+      "loss": 3.7951,
+      "step": 23661
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8805941845764855e-06,
+      "loss": 2.4775,
+      "step": 23662
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8805467762326172e-06,
+      "loss": 1.544,
+      "step": 23663
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8804993678887484e-06,
+      "loss": 2.0943,
+      "step": 23664
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8804519595448797e-06,
+      "loss": 2.936,
+      "step": 23665
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8804045512010114e-06,
+      "loss": 2.7301,
+      "step": 23666
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.880357142857143e-06,
+      "loss": 3.4943,
+      "step": 23667
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8803097345132742e-06,
+      "loss": 2.6018,
+      "step": 23668
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.880262326169406e-06,
+      "loss": 2.127,
+      "step": 23669
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8802149178255374e-06,
+      "loss": 1.8051,
+      "step": 23670
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8801675094816687e-06,
+      "loss": 2.7048,
+      "step": 23671
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8801201011378004e-06,
+      "loss": 2.2386,
+      "step": 23672
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8800726927939319e-06,
+      "loss": 2.2032,
+      "step": 23673
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8800252844500631e-06,
+      "loss": 2.7688,
+      "step": 23674
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8799778761061949e-06,
+      "loss": 1.6107,
+      "step": 23675
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8799304677623263e-06,
+      "loss": 3.2073,
+      "step": 23676
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8798830594184576e-06,
+      "loss": 3.1246,
+      "step": 23677
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8798356510745893e-06,
+      "loss": 2.0682,
+      "step": 23678
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8797882427307206e-06,
+      "loss": 3.2033,
+      "step": 23679
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.879740834386852e-06,
+      "loss": 2.2608,
+      "step": 23680
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8796934260429838e-06,
+      "loss": 1.8049,
+      "step": 23681
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.879646017699115e-06,
+      "loss": 1.9321,
+      "step": 23682
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8795986093552466e-06,
+      "loss": 2.3376,
+      "step": 23683
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8795512010113779e-06,
+      "loss": 3.1006,
+      "step": 23684
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8795037926675096e-06,
+      "loss": 3.3981,
+      "step": 23685
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.879456384323641e-06,
+      "loss": 1.794,
+      "step": 23686
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8794089759797723e-06,
+      "loss": 2.8524,
+      "step": 23687
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.879361567635904e-06,
+      "loss": 2.4013,
+      "step": 23688
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8793141592920355e-06,
+      "loss": 2.2176,
+      "step": 23689
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8792667509481668e-06,
+      "loss": 3.5203,
+      "step": 23690
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8792193426042985e-06,
+      "loss": 2.7155,
+      "step": 23691
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8791719342604298e-06,
+      "loss": 2.9131,
+      "step": 23692
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8791245259165613e-06,
+      "loss": 3.8359,
+      "step": 23693
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.879077117572693e-06,
+      "loss": 3.3533,
+      "step": 23694
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8790297092288243e-06,
+      "loss": 2.7642,
+      "step": 23695
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8789823008849558e-06,
+      "loss": 1.7349,
+      "step": 23696
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8789348925410875e-06,
+      "loss": 2.3808,
+      "step": 23697
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8788874841972187e-06,
+      "loss": 2.8383,
+      "step": 23698
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8788400758533502e-06,
+      "loss": 3.0681,
+      "step": 23699
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.878792667509482e-06,
+      "loss": 3.5083,
+      "step": 23700
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8787452591656132e-06,
+      "loss": 3.2407,
+      "step": 23701
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8786978508217447e-06,
+      "loss": 3.8062,
+      "step": 23702
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8786504424778764e-06,
+      "loss": 3.0904,
+      "step": 23703
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8786030341340077e-06,
+      "loss": 2.6493,
+      "step": 23704
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.878555625790139e-06,
+      "loss": 2.2792,
+      "step": 23705
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8785082174462705e-06,
+      "loss": 3.3336,
+      "step": 23706
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8784608091024022e-06,
+      "loss": 3.133,
+      "step": 23707
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8784134007585335e-06,
+      "loss": 3.1376,
+      "step": 23708
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.878365992414665e-06,
+      "loss": 2.3672,
+      "step": 23709
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8783185840707967e-06,
+      "loss": 2.9339,
+      "step": 23710
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.878271175726928e-06,
+      "loss": 1.2524,
+      "step": 23711
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8782237673830594e-06,
+      "loss": 2.8364,
+      "step": 23712
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8781763590391911e-06,
+      "loss": 1.8754,
+      "step": 23713
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8781289506953224e-06,
+      "loss": 2.7982,
+      "step": 23714
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8780815423514537e-06,
+      "loss": 1.9304,
+      "step": 23715
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8780341340075856e-06,
+      "loss": 3.1155,
+      "step": 23716
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.8779867256637169e-06,
+      "loss": 2.8756,
+      "step": 23717
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8779393173198482e-06,
+      "loss": 2.5704,
+      "step": 23718
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8778919089759799e-06,
+      "loss": 1.9165,
+      "step": 23719
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8778445006321114e-06,
+      "loss": 3.0371,
+      "step": 23720
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8777970922882426e-06,
+      "loss": 2.9965,
+      "step": 23721
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8777496839443743e-06,
+      "loss": 1.378,
+      "step": 23722
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8777022756005058e-06,
+      "loss": 2.8668,
+      "step": 23723
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8776548672566371e-06,
+      "loss": 2.6951,
+      "step": 23724
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8776074589127688e-06,
+      "loss": 2.013,
+      "step": 23725
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8775600505689003e-06,
+      "loss": 4.0941,
+      "step": 23726
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8775126422250316e-06,
+      "loss": 2.0634,
+      "step": 23727
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8774652338811629e-06,
+      "loss": 2.8297,
+      "step": 23728
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8774178255372946e-06,
+      "loss": 2.923,
+      "step": 23729
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.877370417193426e-06,
+      "loss": 3.1235,
+      "step": 23730
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8773230088495574e-06,
+      "loss": 2.4285,
+      "step": 23731
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.877275600505689e-06,
+      "loss": 2.0418,
+      "step": 23732
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8772281921618206e-06,
+      "loss": 3.0188,
+      "step": 23733
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8771807838179518e-06,
+      "loss": 3.2515,
+      "step": 23734
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8771333754740835e-06,
+      "loss": 1.8857,
+      "step": 23735
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.877085967130215e-06,
+      "loss": 1.899,
+      "step": 23736
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8770385587863463e-06,
+      "loss": 2.9617,
+      "step": 23737
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.876991150442478e-06,
+      "loss": 0.8326,
+      "step": 23738
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8769437420986095e-06,
+      "loss": 3.4912,
+      "step": 23739
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8768963337547408e-06,
+      "loss": 3.4934,
+      "step": 23740
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8768489254108725e-06,
+      "loss": 1.5732,
+      "step": 23741
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8768015170670038e-06,
+      "loss": 2.6857,
+      "step": 23742
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8767541087231353e-06,
+      "loss": 1.7378,
+      "step": 23743
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.876706700379267e-06,
+      "loss": 3.1034,
+      "step": 23744
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8766592920353982e-06,
+      "loss": 3.0901,
+      "step": 23745
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8766118836915297e-06,
+      "loss": 3.0626,
+      "step": 23746
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8765644753476614e-06,
+      "loss": 2.3315,
+      "step": 23747
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8765170670037927e-06,
+      "loss": 2.4692,
+      "step": 23748
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8764696586599242e-06,
+      "loss": 2.4859,
+      "step": 23749
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8764222503160555e-06,
+      "loss": 1.2748,
+      "step": 23750
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8763748419721872e-06,
+      "loss": 3.0909,
+      "step": 23751
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8763274336283187e-06,
+      "loss": 2.4134,
+      "step": 23752
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.87628002528445e-06,
+      "loss": 1.782,
+      "step": 23753
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8762326169405817e-06,
+      "loss": 1.7781,
+      "step": 23754
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.876185208596713e-06,
+      "loss": 2.5249,
+      "step": 23755
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8761378002528444e-06,
+      "loss": 2.1865,
+      "step": 23756
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8760903919089762e-06,
+      "loss": 2.8948,
+      "step": 23757
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8760429835651074e-06,
+      "loss": 2.358,
+      "step": 23758
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.875995575221239e-06,
+      "loss": 1.9791,
+      "step": 23759
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8759481668773706e-06,
+      "loss": 2.3109,
+      "step": 23760
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.875900758533502e-06,
+      "loss": 2.5528,
+      "step": 23761
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8758533501896334e-06,
+      "loss": 2.0417,
+      "step": 23762
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.875805941845765e-06,
+      "loss": 1.7605,
+      "step": 23763
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8757585335018964e-06,
+      "loss": 2.474,
+      "step": 23764
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8757111251580279e-06,
+      "loss": 2.8645,
+      "step": 23765
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8756637168141596e-06,
+      "loss": 2.0748,
+      "step": 23766
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8756163084702909e-06,
+      "loss": 1.3776,
+      "step": 23767
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8755689001264221e-06,
+      "loss": 3.2044,
+      "step": 23768
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8755214917825538e-06,
+      "loss": 3.1578,
+      "step": 23769
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8754740834386853e-06,
+      "loss": 2.6912,
+      "step": 23770
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8754266750948166e-06,
+      "loss": 2.534,
+      "step": 23771
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8753792667509481e-06,
+      "loss": 3.3072,
+      "step": 23772
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8753318584070798e-06,
+      "loss": 1.8246,
+      "step": 23773
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.875284450063211e-06,
+      "loss": 2.886,
+      "step": 23774
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8752370417193426e-06,
+      "loss": 2.9269,
+      "step": 23775
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8751896333754743e-06,
+      "loss": 2.5685,
+      "step": 23776
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8751422250316056e-06,
+      "loss": 1.5518,
+      "step": 23777
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.875094816687737e-06,
+      "loss": 2.4828,
+      "step": 23778
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8750474083438688e-06,
+      "loss": 2.0823,
+      "step": 23779
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.875e-06,
+      "loss": 2.6854,
+      "step": 23780
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8749525916561313e-06,
+      "loss": 2.4212,
+      "step": 23781
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.874905183312263e-06,
+      "loss": 2.9649,
+      "step": 23782
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8748577749683945e-06,
+      "loss": 2.3037,
+      "step": 23783
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8748103666245258e-06,
+      "loss": 3.5683,
+      "step": 23784
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8747629582806575e-06,
+      "loss": 2.2867,
+      "step": 23785
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.874715549936789e-06,
+      "loss": 2.5126,
+      "step": 23786
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8746681415929203e-06,
+      "loss": 3.1969,
+      "step": 23787
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.874620733249052e-06,
+      "loss": 2.7874,
+      "step": 23788
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8745733249051835e-06,
+      "loss": 2.8249,
+      "step": 23789
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8745259165613148e-06,
+      "loss": 3.4121,
+      "step": 23790
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8744785082174462e-06,
+      "loss": 2.845,
+      "step": 23791
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.874431099873578e-06,
+      "loss": 3.1672,
+      "step": 23792
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8743836915297092e-06,
+      "loss": 1.5501,
+      "step": 23793
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8743362831858405e-06,
+      "loss": 2.7766,
+      "step": 23794
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8742888748419722e-06,
+      "loss": 2.4354,
+      "step": 23795
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8742414664981037e-06,
+      "loss": 1.8339,
+      "step": 23796
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.874194058154235e-06,
+      "loss": 2.1628,
+      "step": 23797
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8741466498103667e-06,
+      "loss": 2.1077,
+      "step": 23798
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8740992414664982e-06,
+      "loss": 2.3806,
+      "step": 23799
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8740518331226295e-06,
+      "loss": 2.682,
+      "step": 23800
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8740044247787612e-06,
+      "loss": 1.8515,
+      "step": 23801
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8739570164348927e-06,
+      "loss": 2.7989,
+      "step": 23802
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.873909608091024e-06,
+      "loss": 2.3166,
+      "step": 23803
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8738621997471556e-06,
+      "loss": 1.0931,
+      "step": 23804
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8738147914032871e-06,
+      "loss": 3.5202,
+      "step": 23805
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8737673830594184e-06,
+      "loss": 2.462,
+      "step": 23806
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8737199747155501e-06,
+      "loss": 3.5737,
+      "step": 23807
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8736725663716814e-06,
+      "loss": 1.955,
+      "step": 23808
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.873625158027813e-06,
+      "loss": 3.8411,
+      "step": 23809
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8735777496839446e-06,
+      "loss": 2.5461,
+      "step": 23810
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8735303413400759e-06,
+      "loss": 2.131,
+      "step": 23811
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8734829329962074e-06,
+      "loss": 3.704,
+      "step": 23812
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8734355246523387e-06,
+      "loss": 2.2309,
+      "step": 23813
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8733881163084704e-06,
+      "loss": 2.8151,
+      "step": 23814
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8733407079646019e-06,
+      "loss": 3.1301,
+      "step": 23815
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8732932996207331e-06,
+      "loss": 3.1726,
+      "step": 23816
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8732458912768648e-06,
+      "loss": 3.4182,
+      "step": 23817
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8731984829329963e-06,
+      "loss": 2.3702,
+      "step": 23818
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8731510745891276e-06,
+      "loss": 1.3415,
+      "step": 23819
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8731036662452593e-06,
+      "loss": 2.6402,
+      "step": 23820
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8730562579013906e-06,
+      "loss": 2.5557,
+      "step": 23821
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.873008849557522e-06,
+      "loss": 2.4749,
+      "step": 23822
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8729614412136538e-06,
+      "loss": 3.2571,
+      "step": 23823
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.872914032869785e-06,
+      "loss": 2.2986,
+      "step": 23824
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8728666245259166e-06,
+      "loss": 2.3354,
+      "step": 23825
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8728192161820483e-06,
+      "loss": 1.4874,
+      "step": 23826
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8727718078381795e-06,
+      "loss": 3.1115,
+      "step": 23827
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.872724399494311e-06,
+      "loss": 3.0148,
+      "step": 23828
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8726769911504427e-06,
+      "loss": 1.9266,
+      "step": 23829
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.872629582806574e-06,
+      "loss": 3.1688,
+      "step": 23830
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8725821744627055e-06,
+      "loss": 2.7305,
+      "step": 23831
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8725347661188372e-06,
+      "loss": 2.5496,
+      "step": 23832
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8724873577749685e-06,
+      "loss": 1.8296,
+      "step": 23833
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8724399494310998e-06,
+      "loss": 2.7257,
+      "step": 23834
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8723925410872313e-06,
+      "loss": 2.7324,
+      "step": 23835
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.872345132743363e-06,
+      "loss": 3.2438,
+      "step": 23836
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8722977243994943e-06,
+      "loss": 3.2431,
+      "step": 23837
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8722503160556257e-06,
+      "loss": 2.6145,
+      "step": 23838
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8722029077117575e-06,
+      "loss": 3.2497,
+      "step": 23839
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8721554993678887e-06,
+      "loss": 2.5038,
+      "step": 23840
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8721080910240202e-06,
+      "loss": 2.9823,
+      "step": 23841
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.872060682680152e-06,
+      "loss": 3.1328,
+      "step": 23842
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8720132743362832e-06,
+      "loss": 2.7057,
+      "step": 23843
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8719658659924145e-06,
+      "loss": 2.0488,
+      "step": 23844
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8719184576485464e-06,
+      "loss": 2.5865,
+      "step": 23845
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8718710493046777e-06,
+      "loss": 3.6584,
+      "step": 23846
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.871823640960809e-06,
+      "loss": 2.7552,
+      "step": 23847
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8717762326169407e-06,
+      "loss": 2.6295,
+      "step": 23848
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8717288242730722e-06,
+      "loss": 0.8243,
+      "step": 23849
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8716814159292034e-06,
+      "loss": 2.1543,
+      "step": 23850
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8716340075853351e-06,
+      "loss": 2.9271,
+      "step": 23851
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8715865992414666e-06,
+      "loss": 4.3882,
+      "step": 23852
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.871539190897598e-06,
+      "loss": 2.604,
+      "step": 23853
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8714917825537296e-06,
+      "loss": 1.5272,
+      "step": 23854
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8714443742098611e-06,
+      "loss": 2.9132,
+      "step": 23855
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8713969658659924e-06,
+      "loss": 2.8804,
+      "step": 23856
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8713495575221237e-06,
+      "loss": 3.7597,
+      "step": 23857
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8713021491782554e-06,
+      "loss": 2.7338,
+      "step": 23858
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8712547408343869e-06,
+      "loss": 1.4522,
+      "step": 23859
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8712073324905182e-06,
+      "loss": 1.6674,
+      "step": 23860
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8711599241466499e-06,
+      "loss": 2.9263,
+      "step": 23861
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8711125158027813e-06,
+      "loss": 2.6431,
+      "step": 23862
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8710651074589126e-06,
+      "loss": 2.9747,
+      "step": 23863
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8710176991150443e-06,
+      "loss": 1.845,
+      "step": 23864
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8709702907711758e-06,
+      "loss": 2.3624,
+      "step": 23865
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.870922882427307e-06,
+      "loss": 2.9365,
+      "step": 23866
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8708754740834388e-06,
+      "loss": 1.7342,
+      "step": 23867
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8708280657395703e-06,
+      "loss": 3.3618,
+      "step": 23868
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8707806573957016e-06,
+      "loss": 0.7157,
+      "step": 23869
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8707332490518333e-06,
+      "loss": 2.2609,
+      "step": 23870
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8706858407079646e-06,
+      "loss": 2.6216,
+      "step": 23871
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.870638432364096e-06,
+      "loss": 2.5992,
+      "step": 23872
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8705910240202278e-06,
+      "loss": 2.7023,
+      "step": 23873
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.870543615676359e-06,
+      "loss": 2.4502,
+      "step": 23874
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8704962073324905e-06,
+      "loss": 2.574,
+      "step": 23875
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8704487989886222e-06,
+      "loss": 2.2985,
+      "step": 23876
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8704013906447535e-06,
+      "loss": 2.061,
+      "step": 23877
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.870353982300885e-06,
+      "loss": 1.3699,
+      "step": 23878
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8703065739570163e-06,
+      "loss": 2.0159,
+      "step": 23879
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.870259165613148e-06,
+      "loss": 2.3099,
+      "step": 23880
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8702117572692795e-06,
+      "loss": 2.3274,
+      "step": 23881
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8701643489254108e-06,
+      "loss": 1.5315,
+      "step": 23882
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8701169405815425e-06,
+      "loss": 3.7326,
+      "step": 23883
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8700695322376738e-06,
+      "loss": 3.4533,
+      "step": 23884
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8700221238938052e-06,
+      "loss": 1.8728,
+      "step": 23885
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.869974715549937e-06,
+      "loss": 2.6754,
+      "step": 23886
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8699273072060682e-06,
+      "loss": 2.344,
+      "step": 23887
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8698798988621997e-06,
+      "loss": 0.8632,
+      "step": 23888
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8698324905183314e-06,
+      "loss": 1.2979,
+      "step": 23889
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8697850821744627e-06,
+      "loss": 1.2187,
+      "step": 23890
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8697376738305942e-06,
+      "loss": 2.8692,
+      "step": 23891
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.869690265486726e-06,
+      "loss": 2.5559,
+      "step": 23892
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8696428571428572e-06,
+      "loss": 2.2874,
+      "step": 23893
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8695954487989887e-06,
+      "loss": 3.6463,
+      "step": 23894
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8695480404551204e-06,
+      "loss": 3.8101,
+      "step": 23895
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8695006321112517e-06,
+      "loss": 2.7831,
+      "step": 23896
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.869453223767383e-06,
+      "loss": 1.7387,
+      "step": 23897
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8694058154235146e-06,
+      "loss": 3.4046,
+      "step": 23898
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8693584070796461e-06,
+      "loss": 1.1233,
+      "step": 23899
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8693109987357774e-06,
+      "loss": 4.285,
+      "step": 23900
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.869263590391909e-06,
+      "loss": 3.2922,
+      "step": 23901
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8692161820480406e-06,
+      "loss": 3.2246,
+      "step": 23902
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8691687737041719e-06,
+      "loss": 2.5358,
+      "step": 23903
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8691213653603034e-06,
+      "loss": 2.4634,
+      "step": 23904
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.869073957016435e-06,
+      "loss": 1.9,
+      "step": 23905
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8690265486725664e-06,
+      "loss": 2.9845,
+      "step": 23906
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8689791403286979e-06,
+      "loss": 2.9682,
+      "step": 23907
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8689317319848296e-06,
+      "loss": 2.1765,
+      "step": 23908
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8688843236409608e-06,
+      "loss": 1.7646,
+      "step": 23909
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8688369152970921e-06,
+      "loss": 3.1505,
+      "step": 23910
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8687895069532238e-06,
+      "loss": 1.8686,
+      "step": 23911
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8687420986093553e-06,
+      "loss": 2.3139,
+      "step": 23912
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8686946902654866e-06,
+      "loss": 2.2534,
+      "step": 23913
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8686472819216183e-06,
+      "loss": 2.883,
+      "step": 23914
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8685998735777498e-06,
+      "loss": 2.0063,
+      "step": 23915
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.868552465233881e-06,
+      "loss": 3.28,
+      "step": 23916
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8685050568900128e-06,
+      "loss": 2.6399,
+      "step": 23917
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8684576485461443e-06,
+      "loss": 2.856,
+      "step": 23918
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8684102402022756e-06,
+      "loss": 2.7526,
+      "step": 23919
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8683628318584073e-06,
+      "loss": 3.2942,
+      "step": 23920
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8683154235145388e-06,
+      "loss": 2.7954,
+      "step": 23921
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.86826801517067e-06,
+      "loss": 2.5359,
+      "step": 23922
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8682206068268013e-06,
+      "loss": 2.8417,
+      "step": 23923
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.868173198482933e-06,
+      "loss": 3.3749,
+      "step": 23924
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8681257901390645e-06,
+      "loss": 4.2134,
+      "step": 23925
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8680783817951958e-06,
+      "loss": 1.9924,
+      "step": 23926
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8680309734513275e-06,
+      "loss": 1.6893,
+      "step": 23927
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.867983565107459e-06,
+      "loss": 3.3839,
+      "step": 23928
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8679361567635903e-06,
+      "loss": 3.0582,
+      "step": 23929
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.867888748419722e-06,
+      "loss": 2.2854,
+      "step": 23930
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8678413400758535e-06,
+      "loss": 2.6367,
+      "step": 23931
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8677939317319847e-06,
+      "loss": 1.9168,
+      "step": 23932
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8677465233881164e-06,
+      "loss": 2.7943,
+      "step": 23933
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.867699115044248e-06,
+      "loss": 2.9913,
+      "step": 23934
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8676517067003792e-06,
+      "loss": 3.46,
+      "step": 23935
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.867604298356511e-06,
+      "loss": 2.7991,
+      "step": 23936
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8675568900126422e-06,
+      "loss": 2.2792,
+      "step": 23937
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8675094816687737e-06,
+      "loss": 3.1756,
+      "step": 23938
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8674620733249054e-06,
+      "loss": 3.1501,
+      "step": 23939
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8674146649810367e-06,
+      "loss": 2.4928,
+      "step": 23940
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8673672566371682e-06,
+      "loss": 2.6396,
+      "step": 23941
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8673198482932999e-06,
+      "loss": 3.9244,
+      "step": 23942
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8672724399494312e-06,
+      "loss": 2.9455,
+      "step": 23943
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8672250316055626e-06,
+      "loss": 2.0247,
+      "step": 23944
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.867177623261694e-06,
+      "loss": 2.6876,
+      "step": 23945
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8671302149178256e-06,
+      "loss": 2.9844,
+      "step": 23946
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8670828065739571e-06,
+      "loss": 2.7795,
+      "step": 23947
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8670353982300884e-06,
+      "loss": 2.6826,
+      "step": 23948
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8669879898862201e-06,
+      "loss": 3.6303,
+      "step": 23949
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8669405815423514e-06,
+      "loss": 2.3668,
+      "step": 23950
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8668931731984829e-06,
+      "loss": 1.8614,
+      "step": 23951
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8668457648546146e-06,
+      "loss": 3.0368,
+      "step": 23952
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8667983565107459e-06,
+      "loss": 2.6712,
+      "step": 23953
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8667509481668774e-06,
+      "loss": 2.8059,
+      "step": 23954
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.866703539823009e-06,
+      "loss": 3.8007,
+      "step": 23955
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8666561314791403e-06,
+      "loss": 2.8649,
+      "step": 23956
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8666087231352718e-06,
+      "loss": 2.6992,
+      "step": 23957
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8665613147914035e-06,
+      "loss": 1.6595,
+      "step": 23958
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8665139064475348e-06,
+      "loss": 1.2144,
+      "step": 23959
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8664664981036663e-06,
+      "loss": 1.0482,
+      "step": 23960
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.866419089759798e-06,
+      "loss": 3.3444,
+      "step": 23961
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8663716814159293e-06,
+      "loss": 2.2496,
+      "step": 23962
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8663242730720606e-06,
+      "loss": 2.0817,
+      "step": 23963
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8662768647281923e-06,
+      "loss": 1.5473,
+      "step": 23964
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8662294563843238e-06,
+      "loss": 3.3621,
+      "step": 23965
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.866182048040455e-06,
+      "loss": 2.9091,
+      "step": 23966
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8661346396965865e-06,
+      "loss": 2.9819,
+      "step": 23967
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8660872313527182e-06,
+      "loss": 3.0154,
+      "step": 23968
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8660398230088495e-06,
+      "loss": 1.6591,
+      "step": 23969
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.865992414664981e-06,
+      "loss": 1.4377,
+      "step": 23970
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8659450063211127e-06,
+      "loss": 3.0466,
+      "step": 23971
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.865897597977244e-06,
+      "loss": 3.5791,
+      "step": 23972
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8658501896333755e-06,
+      "loss": 2.9365,
+      "step": 23973
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8658027812895072e-06,
+      "loss": 3.6086,
+      "step": 23974
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8657553729456385e-06,
+      "loss": 2.8922,
+      "step": 23975
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8657079646017698e-06,
+      "loss": 2.2554,
+      "step": 23976
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8656605562579015e-06,
+      "loss": 2.0139,
+      "step": 23977
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.865613147914033e-06,
+      "loss": 3.5166,
+      "step": 23978
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8655657395701642e-06,
+      "loss": 2.1559,
+      "step": 23979
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.865518331226296e-06,
+      "loss": 1.8237,
+      "step": 23980
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8654709228824274e-06,
+      "loss": 2.6392,
+      "step": 23981
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8654235145385587e-06,
+      "loss": 2.6336,
+      "step": 23982
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8653761061946904e-06,
+      "loss": 2.2189,
+      "step": 23983
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.865328697850822e-06,
+      "loss": 2.8908,
+      "step": 23984
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8652812895069532e-06,
+      "loss": 2.3606,
+      "step": 23985
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.865233881163085e-06,
+      "loss": 1.8995,
+      "step": 23986
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8651864728192164e-06,
+      "loss": 3.5953,
+      "step": 23987
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8651390644753477e-06,
+      "loss": 1.821,
+      "step": 23988
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.865091656131479e-06,
+      "loss": 1.2143,
+      "step": 23989
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8650442477876107e-06,
+      "loss": 2.0517,
+      "step": 23990
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8649968394437421e-06,
+      "loss": 1.9616,
+      "step": 23991
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8649494310998734e-06,
+      "loss": 2.9601,
+      "step": 23992
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8649020227560051e-06,
+      "loss": 2.421,
+      "step": 23993
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8648546144121366e-06,
+      "loss": 1.8791,
+      "step": 23994
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.864807206068268e-06,
+      "loss": 3.5709,
+      "step": 23995
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8647597977243996e-06,
+      "loss": 2.7607,
+      "step": 23996
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.864712389380531e-06,
+      "loss": 2.5362,
+      "step": 23997
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8646649810366624e-06,
+      "loss": 3.7307,
+      "step": 23998
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.864617572692794e-06,
+      "loss": 2.9684,
+      "step": 23999
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8645701643489254e-06,
+      "loss": 2.737,
+      "step": 24000
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8645227560050569e-06,
+      "loss": 2.6196,
+      "step": 24001
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8644753476611886e-06,
+      "loss": 2.7634,
+      "step": 24002
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8644279393173198e-06,
+      "loss": 2.6999,
+      "step": 24003
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8643805309734513e-06,
+      "loss": 2.7375,
+      "step": 24004
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.864333122629583e-06,
+      "loss": 4.2914,
+      "step": 24005
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8642857142857143e-06,
+      "loss": 2.1645,
+      "step": 24006
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8642383059418458e-06,
+      "loss": 2.9085,
+      "step": 24007
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8641908975979775e-06,
+      "loss": 3.5241,
+      "step": 24008
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8641434892541088e-06,
+      "loss": 3.5223,
+      "step": 24009
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8640960809102403e-06,
+      "loss": 1.7819,
+      "step": 24010
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8640486725663716e-06,
+      "loss": 1.3292,
+      "step": 24011
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8640012642225033e-06,
+      "loss": 3.1265,
+      "step": 24012
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8639538558786345e-06,
+      "loss": 2.6562,
+      "step": 24013
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.863906447534766e-06,
+      "loss": 4.5323,
+      "step": 24014
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8638590391908977e-06,
+      "loss": 3.2749,
+      "step": 24015
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.863811630847029e-06,
+      "loss": 2.1976,
+      "step": 24016
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8637642225031605e-06,
+      "loss": 2.4609,
+      "step": 24017
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8637168141592922e-06,
+      "loss": 2.213,
+      "step": 24018
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8636694058154235e-06,
+      "loss": 3.6855,
+      "step": 24019
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.863621997471555e-06,
+      "loss": 2.9505,
+      "step": 24020
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8635745891276867e-06,
+      "loss": 1.655,
+      "step": 24021
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.863527180783818e-06,
+      "loss": 3.5171,
+      "step": 24022
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8634797724399495e-06,
+      "loss": 3.0288,
+      "step": 24023
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8634323640960812e-06,
+      "loss": 2.7558,
+      "step": 24024
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8633849557522125e-06,
+      "loss": 1.737,
+      "step": 24025
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8633375474083437e-06,
+      "loss": 2.1617,
+      "step": 24026
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8632901390644754e-06,
+      "loss": 3.2283,
+      "step": 24027
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.863242730720607e-06,
+      "loss": 2.4114,
+      "step": 24028
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8631953223767382e-06,
+      "loss": 2.9414,
+      "step": 24029
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.86314791403287e-06,
+      "loss": 2.8615,
+      "step": 24030
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8631005056890014e-06,
+      "loss": 3.3044,
+      "step": 24031
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8630530973451327e-06,
+      "loss": 2.002,
+      "step": 24032
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8630056890012642e-06,
+      "loss": 2.9925,
+      "step": 24033
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8629582806573959e-06,
+      "loss": 2.2692,
+      "step": 24034
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8629108723135272e-06,
+      "loss": 1.8012,
+      "step": 24035
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8628634639696587e-06,
+      "loss": 2.5214,
+      "step": 24036
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8628160556257904e-06,
+      "loss": 2.3937,
+      "step": 24037
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8627686472819216e-06,
+      "loss": 2.8627,
+      "step": 24038
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.862721238938053e-06,
+      "loss": 1.2732,
+      "step": 24039
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8626738305941846e-06,
+      "loss": 3.2118,
+      "step": 24040
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8626264222503161e-06,
+      "loss": 2.7157,
+      "step": 24041
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8625790139064474e-06,
+      "loss": 3.0056,
+      "step": 24042
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.862531605562579e-06,
+      "loss": 2.6319,
+      "step": 24043
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8624841972187106e-06,
+      "loss": 2.8766,
+      "step": 24044
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8624367888748419e-06,
+      "loss": 2.6494,
+      "step": 24045
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8623893805309736e-06,
+      "loss": 3.7396,
+      "step": 24046
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.862341972187105e-06,
+      "loss": 1.6579,
+      "step": 24047
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8622945638432364e-06,
+      "loss": 2.3327,
+      "step": 24048
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.862247155499368e-06,
+      "loss": 2.919,
+      "step": 24049
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8621997471554995e-06,
+      "loss": 2.1088,
+      "step": 24050
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8621523388116308e-06,
+      "loss": 2.3663,
+      "step": 24051
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8621049304677625e-06,
+      "loss": 1.4932,
+      "step": 24052
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8620575221238938e-06,
+      "loss": 4.2343,
+      "step": 24053
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8620101137800253e-06,
+      "loss": 1.6958,
+      "step": 24054
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8619627054361566e-06,
+      "loss": 2.8732,
+      "step": 24055
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8619152970922883e-06,
+      "loss": 2.7465,
+      "step": 24056
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8618678887484198e-06,
+      "loss": 1.72,
+      "step": 24057
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.861820480404551e-06,
+      "loss": 2.3154,
+      "step": 24058
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8617730720606828e-06,
+      "loss": 2.43,
+      "step": 24059
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8617256637168143e-06,
+      "loss": 1.8619,
+      "step": 24060
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8616782553729455e-06,
+      "loss": 3.6435,
+      "step": 24061
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8616308470290772e-06,
+      "loss": 2.354,
+      "step": 24062
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8615834386852087e-06,
+      "loss": 2.0008,
+      "step": 24063
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.86153603034134e-06,
+      "loss": 1.761,
+      "step": 24064
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8614886219974717e-06,
+      "loss": 2.307,
+      "step": 24065
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.861441213653603e-06,
+      "loss": 3.3259,
+      "step": 24066
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8613938053097345e-06,
+      "loss": 1.113,
+      "step": 24067
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8613463969658662e-06,
+      "loss": 3.2541,
+      "step": 24068
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8612989886219975e-06,
+      "loss": 2.1619,
+      "step": 24069
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.861251580278129e-06,
+      "loss": 1.7481,
+      "step": 24070
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8612041719342607e-06,
+      "loss": 2.6419,
+      "step": 24071
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.861156763590392e-06,
+      "loss": 2.9187,
+      "step": 24072
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8611093552465234e-06,
+      "loss": 2.4936,
+      "step": 24073
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8610619469026551e-06,
+      "loss": 3.0998,
+      "step": 24074
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8610145385587864e-06,
+      "loss": 2.5258,
+      "step": 24075
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.860967130214918e-06,
+      "loss": 2.32,
+      "step": 24076
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8609197218710492e-06,
+      "loss": 3.113,
+      "step": 24077
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.860872313527181e-06,
+      "loss": 3.3933,
+      "step": 24078
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8608249051833122e-06,
+      "loss": 2.2218,
+      "step": 24079
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8607774968394437e-06,
+      "loss": 2.6691,
+      "step": 24080
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8607300884955754e-06,
+      "loss": 2.591,
+      "step": 24081
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8606826801517067e-06,
+      "loss": 2.361,
+      "step": 24082
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8606352718078382e-06,
+      "loss": 2.4774,
+      "step": 24083
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8605878634639699e-06,
+      "loss": 2.2112,
+      "step": 24084
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8605404551201011e-06,
+      "loss": 2.7222,
+      "step": 24085
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8604930467762326e-06,
+      "loss": 2.8472,
+      "step": 24086
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8604456384323643e-06,
+      "loss": 1.3755,
+      "step": 24087
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8603982300884956e-06,
+      "loss": 2.4711,
+      "step": 24088
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8603508217446271e-06,
+      "loss": 2.5685,
+      "step": 24089
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8603034134007588e-06,
+      "loss": 1.1733,
+      "step": 24090
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.86025600505689e-06,
+      "loss": 2.0798,
+      "step": 24091
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8602085967130214e-06,
+      "loss": 2.0157,
+      "step": 24092
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.860161188369153e-06,
+      "loss": 1.4775,
+      "step": 24093
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8601137800252846e-06,
+      "loss": 2.7843,
+      "step": 24094
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8600663716814158e-06,
+      "loss": 1.3909,
+      "step": 24095
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8600189633375476e-06,
+      "loss": 3.1985,
+      "step": 24096
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.859971554993679e-06,
+      "loss": 2.2647,
+      "step": 24097
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8599241466498103e-06,
+      "loss": 2.8009,
+      "step": 24098
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8598767383059418e-06,
+      "loss": 3.8353,
+      "step": 24099
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8598293299620735e-06,
+      "loss": 3.2622,
+      "step": 24100
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8597819216182048e-06,
+      "loss": 1.7514,
+      "step": 24101
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8597345132743363e-06,
+      "loss": 2.1887,
+      "step": 24102
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.859687104930468e-06,
+      "loss": 2.4414,
+      "step": 24103
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8596396965865993e-06,
+      "loss": 2.9024,
+      "step": 24104
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8595922882427306e-06,
+      "loss": 2.0149,
+      "step": 24105
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8595448798988623e-06,
+      "loss": 2.4001,
+      "step": 24106
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8594974715549938e-06,
+      "loss": 2.8891,
+      "step": 24107
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.859450063211125e-06,
+      "loss": 3.0245,
+      "step": 24108
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8594026548672567e-06,
+      "loss": 3.7584,
+      "step": 24109
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8593552465233882e-06,
+      "loss": 2.2593,
+      "step": 24110
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8593078381795195e-06,
+      "loss": 2.9882,
+      "step": 24111
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8592604298356512e-06,
+      "loss": 2.2459,
+      "step": 24112
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8592130214917827e-06,
+      "loss": 2.3166,
+      "step": 24113
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.859165613147914e-06,
+      "loss": 2.549,
+      "step": 24114
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8591182048040457e-06,
+      "loss": 2.7164,
+      "step": 24115
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8590707964601772e-06,
+      "loss": 4.3667,
+      "step": 24116
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8590233881163085e-06,
+      "loss": 3.3783,
+      "step": 24117
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8589759797724402e-06,
+      "loss": 2.5294,
+      "step": 24118
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8589285714285714e-06,
+      "loss": 2.2156,
+      "step": 24119
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.858881163084703e-06,
+      "loss": 3.6938,
+      "step": 24120
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8588337547408342e-06,
+      "loss": 2.7398,
+      "step": 24121
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.858786346396966e-06,
+      "loss": 1.2955,
+      "step": 24122
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8587389380530974e-06,
+      "loss": 2.8502,
+      "step": 24123
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8586915297092287e-06,
+      "loss": 3.3011,
+      "step": 24124
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8586441213653604e-06,
+      "loss": 3.1418,
+      "step": 24125
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8585967130214919e-06,
+      "loss": 2.0069,
+      "step": 24126
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8585493046776232e-06,
+      "loss": 2.6557,
+      "step": 24127
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8585018963337549e-06,
+      "loss": 2.1656,
+      "step": 24128
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8584544879898862e-06,
+      "loss": 2.0875,
+      "step": 24129
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8584070796460177e-06,
+      "loss": 2.6726,
+      "step": 24130
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8583596713021494e-06,
+      "loss": 2.6578,
+      "step": 24131
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8583122629582806e-06,
+      "loss": 1.2133,
+      "step": 24132
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8582648546144121e-06,
+      "loss": 3.9546,
+      "step": 24133
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8582174462705438e-06,
+      "loss": 2.495,
+      "step": 24134
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8581700379266751e-06,
+      "loss": 2.1411,
+      "step": 24135
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8581226295828066e-06,
+      "loss": 2.5983,
+      "step": 24136
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8580752212389383e-06,
+      "loss": 3.2498,
+      "step": 24137
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8580278128950696e-06,
+      "loss": 1.9808,
+      "step": 24138
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.857980404551201e-06,
+      "loss": 2.2896,
+      "step": 24139
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8579329962073328e-06,
+      "loss": 1.1765,
+      "step": 24140
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.857885587863464e-06,
+      "loss": 1.4366,
+      "step": 24141
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8578381795195953e-06,
+      "loss": 1.9227,
+      "step": 24142
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8577907711757268e-06,
+      "loss": 1.9415,
+      "step": 24143
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8577433628318585e-06,
+      "loss": 2.6294,
+      "step": 24144
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8576959544879898e-06,
+      "loss": 1.7963,
+      "step": 24145
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8576485461441213e-06,
+      "loss": 3.9494,
+      "step": 24146
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.857601137800253e-06,
+      "loss": 1.7505,
+      "step": 24147
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8575537294563843e-06,
+      "loss": 3.2268,
+      "step": 24148
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8575063211125158e-06,
+      "loss": 3.2619,
+      "step": 24149
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8574589127686475e-06,
+      "loss": 2.0373,
+      "step": 24150
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8574115044247788e-06,
+      "loss": 2.6469,
+      "step": 24151
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8573640960809103e-06,
+      "loss": 1.943,
+      "step": 24152
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.857316687737042e-06,
+      "loss": 1.9996,
+      "step": 24153
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8572692793931733e-06,
+      "loss": 2.4615,
+      "step": 24154
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8572218710493045e-06,
+      "loss": 2.4964,
+      "step": 24155
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8571744627054362e-06,
+      "loss": 2.9503,
+      "step": 24156
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8571270543615677e-06,
+      "loss": 2.0815,
+      "step": 24157
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.857079646017699e-06,
+      "loss": 3.7489,
+      "step": 24158
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8570322376738307e-06,
+      "loss": 1.4814,
+      "step": 24159
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8569848293299622e-06,
+      "loss": 2.2546,
+      "step": 24160
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8569374209860935e-06,
+      "loss": 1.8316,
+      "step": 24161
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8568900126422252e-06,
+      "loss": 2.5644,
+      "step": 24162
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8568426042983567e-06,
+      "loss": 2.9962,
+      "step": 24163
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.856795195954488e-06,
+      "loss": 1.9319,
+      "step": 24164
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8567477876106195e-06,
+      "loss": 2.2314,
+      "step": 24165
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8567003792667512e-06,
+      "loss": 2.4302,
+      "step": 24166
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8566529709228824e-06,
+      "loss": 1.5676,
+      "step": 24167
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8566055625790137e-06,
+      "loss": 4.1531,
+      "step": 24168
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8565581542351454e-06,
+      "loss": 2.6169,
+      "step": 24169
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.856510745891277e-06,
+      "loss": 3.6061,
+      "step": 24170
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8564633375474082e-06,
+      "loss": 1.9993,
+      "step": 24171
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.85641592920354e-06,
+      "loss": 2.375,
+      "step": 24172
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8563685208596714e-06,
+      "loss": 2.5663,
+      "step": 24173
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8563211125158027e-06,
+      "loss": 2.249,
+      "step": 24174
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8562737041719344e-06,
+      "loss": 2.7026,
+      "step": 24175
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8562262958280659e-06,
+      "loss": 3.0839,
+      "step": 24176
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8561788874841971e-06,
+      "loss": 2.5604,
+      "step": 24177
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8561314791403289e-06,
+      "loss": 1.7962,
+      "step": 24178
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8560840707964603e-06,
+      "loss": 2.2765,
+      "step": 24179
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8560366624525916e-06,
+      "loss": 1.4731,
+      "step": 24180
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8559892541087233e-06,
+      "loss": 2.3294,
+      "step": 24181
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8559418457648546e-06,
+      "loss": 4.1363,
+      "step": 24182
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.855894437420986e-06,
+      "loss": 1.163,
+      "step": 24183
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8558470290771178e-06,
+      "loss": 1.7661,
+      "step": 24184
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.855799620733249e-06,
+      "loss": 2.4448,
+      "step": 24185
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8557522123893806e-06,
+      "loss": 2.4181,
+      "step": 24186
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8557048040455119e-06,
+      "loss": 3.1128,
+      "step": 24187
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8556573957016436e-06,
+      "loss": 3.317,
+      "step": 24188
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.855609987357775e-06,
+      "loss": 2.0618,
+      "step": 24189
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8555625790139063e-06,
+      "loss": 2.0272,
+      "step": 24190
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.855515170670038e-06,
+      "loss": 2.436,
+      "step": 24191
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8554677623261695e-06,
+      "loss": 3.4684,
+      "step": 24192
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8554203539823008e-06,
+      "loss": 1.6445,
+      "step": 24193
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8553729456384325e-06,
+      "loss": 2.6819,
+      "step": 24194
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8553255372945638e-06,
+      "loss": 1.8878,
+      "step": 24195
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8552781289506953e-06,
+      "loss": 1.1819,
+      "step": 24196
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.855230720606827e-06,
+      "loss": 2.6949,
+      "step": 24197
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8551833122629583e-06,
+      "loss": 3.1427,
+      "step": 24198
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8551359039190898e-06,
+      "loss": 0.8742,
+      "step": 24199
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8550884955752215e-06,
+      "loss": 2.8509,
+      "step": 24200
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8550410872313527e-06,
+      "loss": 2.1862,
+      "step": 24201
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8549936788874842e-06,
+      "loss": 2.4817,
+      "step": 24202
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.854946270543616e-06,
+      "loss": 1.4796,
+      "step": 24203
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8548988621997472e-06,
+      "loss": 2.6998,
+      "step": 24204
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8548514538558787e-06,
+      "loss": 2.7511,
+      "step": 24205
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8548040455120104e-06,
+      "loss": 2.0715,
+      "step": 24206
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8547566371681417e-06,
+      "loss": 2.2612,
+      "step": 24207
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.854709228824273e-06,
+      "loss": 1.2276,
+      "step": 24208
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8546618204804045e-06,
+      "loss": 3.7769,
+      "step": 24209
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8546144121365362e-06,
+      "loss": 3.543,
+      "step": 24210
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8545670037926675e-06,
+      "loss": 2.544,
+      "step": 24211
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.854519595448799e-06,
+      "loss": 2.8845,
+      "step": 24212
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8544721871049307e-06,
+      "loss": 2.7033,
+      "step": 24213
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.854424778761062e-06,
+      "loss": 2.6566,
+      "step": 24214
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8543773704171934e-06,
+      "loss": 2.1749,
+      "step": 24215
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8543299620733251e-06,
+      "loss": 1.7664,
+      "step": 24216
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8542825537294564e-06,
+      "loss": 2.9753,
+      "step": 24217
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.854235145385588e-06,
+      "loss": 1.3963,
+      "step": 24218
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8541877370417196e-06,
+      "loss": 2.7646,
+      "step": 24219
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8541403286978509e-06,
+      "loss": 3.1415,
+      "step": 24220
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8540929203539822e-06,
+      "loss": 1.8509,
+      "step": 24221
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8540455120101139e-06,
+      "loss": 2.9768,
+      "step": 24222
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.8539981036662454e-06,
+      "loss": 2.8313,
+      "step": 24223
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8539506953223766e-06,
+      "loss": 2.7394,
+      "step": 24224
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8539032869785083e-06,
+      "loss": 2.72,
+      "step": 24225
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8538558786346398e-06,
+      "loss": 4.065,
+      "step": 24226
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8538084702907711e-06,
+      "loss": 2.9921,
+      "step": 24227
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8537610619469026e-06,
+      "loss": 2.9464,
+      "step": 24228
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8537136536030343e-06,
+      "loss": 2.3235,
+      "step": 24229
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8536662452591656e-06,
+      "loss": 2.4806,
+      "step": 24230
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.853618836915297e-06,
+      "loss": 3.0312,
+      "step": 24231
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8535714285714288e-06,
+      "loss": 2.1549,
+      "step": 24232
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.85352402022756e-06,
+      "loss": 3.8522,
+      "step": 24233
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8534766118836914e-06,
+      "loss": 3.1357,
+      "step": 24234
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.853429203539823e-06,
+      "loss": 1.915,
+      "step": 24235
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8533817951959546e-06,
+      "loss": 3.3396,
+      "step": 24236
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8533343868520858e-06,
+      "loss": 2.3344,
+      "step": 24237
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8532869785082175e-06,
+      "loss": 2.3262,
+      "step": 24238
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.853239570164349e-06,
+      "loss": 1.4735,
+      "step": 24239
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8531921618204803e-06,
+      "loss": 3.7654,
+      "step": 24240
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.853144753476612e-06,
+      "loss": 2.5701,
+      "step": 24241
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8530973451327435e-06,
+      "loss": 1.6458,
+      "step": 24242
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8530499367888748e-06,
+      "loss": 1.8057,
+      "step": 24243
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8530025284450065e-06,
+      "loss": 2.581,
+      "step": 24244
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.852955120101138e-06,
+      "loss": 3.7299,
+      "step": 24245
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8529077117572693e-06,
+      "loss": 2.2425,
+      "step": 24246
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.852860303413401e-06,
+      "loss": 3.1937,
+      "step": 24247
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8528128950695322e-06,
+      "loss": 2.3704,
+      "step": 24248
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8527654867256637e-06,
+      "loss": 3.1383,
+      "step": 24249
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.852718078381795e-06,
+      "loss": 2.9965,
+      "step": 24250
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8526706700379267e-06,
+      "loss": 2.2952,
+      "step": 24251
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8526232616940582e-06,
+      "loss": 2.2409,
+      "step": 24252
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8525758533501895e-06,
+      "loss": 3.2223,
+      "step": 24253
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8525284450063212e-06,
+      "loss": 2.7933,
+      "step": 24254
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8524810366624527e-06,
+      "loss": 2.4497,
+      "step": 24255
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.852433628318584e-06,
+      "loss": 2.2256,
+      "step": 24256
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8523862199747157e-06,
+      "loss": 1.3723,
+      "step": 24257
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8523388116308472e-06,
+      "loss": 3.111,
+      "step": 24258
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8522914032869784e-06,
+      "loss": 3.303,
+      "step": 24259
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8522439949431102e-06,
+      "loss": 3.5752,
+      "step": 24260
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8521965865992414e-06,
+      "loss": 0.8782,
+      "step": 24261
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.852149178255373e-06,
+      "loss": 1.8113,
+      "step": 24262
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8521017699115046e-06,
+      "loss": 1.9484,
+      "step": 24263
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.852054361567636e-06,
+      "loss": 1.322,
+      "step": 24264
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8520069532237674e-06,
+      "loss": 2.3001,
+      "step": 24265
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.851959544879899e-06,
+      "loss": 1.9498,
+      "step": 24266
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8519121365360304e-06,
+      "loss": 1.3923,
+      "step": 24267
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8518647281921619e-06,
+      "loss": 1.0324,
+      "step": 24268
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8518173198482936e-06,
+      "loss": 2.4499,
+      "step": 24269
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8517699115044249e-06,
+      "loss": 1.8099,
+      "step": 24270
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8517225031605561e-06,
+      "loss": 3.0783,
+      "step": 24271
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8516750948166876e-06,
+      "loss": 2.4034,
+      "step": 24272
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8516276864728193e-06,
+      "loss": 1.8882,
+      "step": 24273
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8515802781289506e-06,
+      "loss": 2.5759,
+      "step": 24274
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8515328697850821e-06,
+      "loss": 2.8254,
+      "step": 24275
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8514854614412138e-06,
+      "loss": 3.21,
+      "step": 24276
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.851438053097345e-06,
+      "loss": 2.2909,
+      "step": 24277
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8513906447534766e-06,
+      "loss": 2.0652,
+      "step": 24278
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8513432364096083e-06,
+      "loss": 1.4936,
+      "step": 24279
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8512958280657396e-06,
+      "loss": 3.2929,
+      "step": 24280
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.851248419721871e-06,
+      "loss": 3.3908,
+      "step": 24281
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8512010113780028e-06,
+      "loss": 2.4417,
+      "step": 24282
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.851153603034134e-06,
+      "loss": 2.541,
+      "step": 24283
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8511061946902653e-06,
+      "loss": 2.1936,
+      "step": 24284
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.851058786346397e-06,
+      "loss": 1.5339,
+      "step": 24285
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8510113780025285e-06,
+      "loss": 2.4723,
+      "step": 24286
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.8509639696586598e-06,
+      "loss": 2.264,
+      "step": 24287
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8509165613147915e-06,
+      "loss": 2.8462,
+      "step": 24288
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.850869152970923e-06,
+      "loss": 2.757,
+      "step": 24289
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8508217446270543e-06,
+      "loss": 2.9082,
+      "step": 24290
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.850774336283186e-06,
+      "loss": 2.0004,
+      "step": 24291
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8507269279393175e-06,
+      "loss": 2.7676,
+      "step": 24292
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8506795195954488e-06,
+      "loss": 2.9035,
+      "step": 24293
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8506321112515803e-06,
+      "loss": 2.5778,
+      "step": 24294
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.850584702907712e-06,
+      "loss": 2.163,
+      "step": 24295
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8505372945638432e-06,
+      "loss": 0.7462,
+      "step": 24296
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8504898862199745e-06,
+      "loss": 1.9457,
+      "step": 24297
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8504424778761062e-06,
+      "loss": 2.0768,
+      "step": 24298
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8503950695322377e-06,
+      "loss": 2.7367,
+      "step": 24299
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.850347661188369e-06,
+      "loss": 1.7764,
+      "step": 24300
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8503002528445007e-06,
+      "loss": 2.9661,
+      "step": 24301
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8502528445006322e-06,
+      "loss": 1.9531,
+      "step": 24302
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8502054361567635e-06,
+      "loss": 1.8705,
+      "step": 24303
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8501580278128952e-06,
+      "loss": 2.0172,
+      "step": 24304
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8501106194690267e-06,
+      "loss": 3.293,
+      "step": 24305
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.850063211125158e-06,
+      "loss": 2.3088,
+      "step": 24306
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8500158027812896e-06,
+      "loss": 2.697,
+      "step": 24307
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8499683944374211e-06,
+      "loss": 2.5771,
+      "step": 24308
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8499209860935524e-06,
+      "loss": 3.0333,
+      "step": 24309
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8498735777496841e-06,
+      "loss": 3.136,
+      "step": 24310
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8498261694058154e-06,
+      "loss": 1.0738,
+      "step": 24311
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.849778761061947e-06,
+      "loss": 2.0072,
+      "step": 24312
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8497313527180786e-06,
+      "loss": 2.6212,
+      "step": 24313
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8496839443742099e-06,
+      "loss": 3.1053,
+      "step": 24314
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8496365360303414e-06,
+      "loss": 2.1772,
+      "step": 24315
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8495891276864727e-06,
+      "loss": 1.7369,
+      "step": 24316
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8495417193426044e-06,
+      "loss": 1.9982,
+      "step": 24317
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8494943109987359e-06,
+      "loss": 1.8884,
+      "step": 24318
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8494469026548671e-06,
+      "loss": 2.0669,
+      "step": 24319
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8493994943109988e-06,
+      "loss": 3.4471,
+      "step": 24320
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8493520859671303e-06,
+      "loss": 2.3164,
+      "step": 24321
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8493046776232616e-06,
+      "loss": 2.0155,
+      "step": 24322
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8492572692793933e-06,
+      "loss": 3.278,
+      "step": 24323
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8492098609355246e-06,
+      "loss": 2.2688,
+      "step": 24324
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.849162452591656e-06,
+      "loss": 1.761,
+      "step": 24325
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8491150442477878e-06,
+      "loss": 2.5502,
+      "step": 24326
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.849067635903919e-06,
+      "loss": 1.2294,
+      "step": 24327
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8490202275600506e-06,
+      "loss": 2.2977,
+      "step": 24328
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8489728192161823e-06,
+      "loss": 2.4687,
+      "step": 24329
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8489254108723135e-06,
+      "loss": 2.1267,
+      "step": 24330
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.848878002528445e-06,
+      "loss": 3.6862,
+      "step": 24331
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8488305941845767e-06,
+      "loss": 4.001,
+      "step": 24332
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.848783185840708e-06,
+      "loss": 1.9146,
+      "step": 24333
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8487357774968395e-06,
+      "loss": 1.4508,
+      "step": 24334
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8486883691529712e-06,
+      "loss": 3.2855,
+      "step": 24335
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8486409608091025e-06,
+      "loss": 3.8503,
+      "step": 24336
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8485935524652338e-06,
+      "loss": 3.1818,
+      "step": 24337
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8485461441213653e-06,
+      "loss": 3.3501,
+      "step": 24338
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.848498735777497e-06,
+      "loss": 2.9881,
+      "step": 24339
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8484513274336283e-06,
+      "loss": 3.3096,
+      "step": 24340
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8484039190897597e-06,
+      "loss": 2.2271,
+      "step": 24341
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8483565107458915e-06,
+      "loss": 2.1846,
+      "step": 24342
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8483091024020227e-06,
+      "loss": 2.9066,
+      "step": 24343
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8482616940581542e-06,
+      "loss": 1.0652,
+      "step": 24344
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.848214285714286e-06,
+      "loss": 3.2611,
+      "step": 24345
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8481668773704172e-06,
+      "loss": 2.2919,
+      "step": 24346
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8481194690265487e-06,
+      "loss": 2.4772,
+      "step": 24347
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8480720606826804e-06,
+      "loss": 2.2659,
+      "step": 24348
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.8480246523388117e-06,
+      "loss": 3.6847,
+      "step": 24349
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.847977243994943e-06,
+      "loss": 2.3198,
+      "step": 24350
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8479298356510747e-06,
+      "loss": 1.9107,
+      "step": 24351
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8478824273072062e-06,
+      "loss": 2.4902,
+      "step": 24352
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8478350189633374e-06,
+      "loss": 3.3828,
+      "step": 24353
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8477876106194691e-06,
+      "loss": 2.6555,
+      "step": 24354
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8477402022756006e-06,
+      "loss": 3.5364,
+      "step": 24355
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.847692793931732e-06,
+      "loss": 1.1493,
+      "step": 24356
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8476453855878636e-06,
+      "loss": 3.3915,
+      "step": 24357
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8475979772439951e-06,
+      "loss": 2.9462,
+      "step": 24358
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8475505689001264e-06,
+      "loss": 2.4399,
+      "step": 24359
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8475031605562579e-06,
+      "loss": 2.0264,
+      "step": 24360
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8474557522123896e-06,
+      "loss": 2.8879,
+      "step": 24361
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8474083438685209e-06,
+      "loss": 3.7744,
+      "step": 24362
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8473609355246522e-06,
+      "loss": 2.1966,
+      "step": 24363
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8473135271807839e-06,
+      "loss": 2.5448,
+      "step": 24364
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8472661188369153e-06,
+      "loss": 2.1655,
+      "step": 24365
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8472187104930466e-06,
+      "loss": 2.6994,
+      "step": 24366
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8471713021491783e-06,
+      "loss": 4.1474,
+      "step": 24367
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8471238938053098e-06,
+      "loss": 2.4109,
+      "step": 24368
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.847076485461441e-06,
+      "loss": 1.1776,
+      "step": 24369
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8470290771175728e-06,
+      "loss": 3.3482,
+      "step": 24370
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8469816687737043e-06,
+      "loss": 3.1707,
+      "step": 24371
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8469342604298356e-06,
+      "loss": 2.8748,
+      "step": 24372
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8468868520859673e-06,
+      "loss": 3.8119,
+      "step": 24373
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8468394437420988e-06,
+      "loss": 2.9065,
+      "step": 24374
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.84679203539823e-06,
+      "loss": 1.7322,
+      "step": 24375
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8467446270543618e-06,
+      "loss": 2.487,
+      "step": 24376
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.846697218710493e-06,
+      "loss": 2.9325,
+      "step": 24377
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8466498103666245e-06,
+      "loss": 2.7525,
+      "step": 24378
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8466024020227562e-06,
+      "loss": 2.5957,
+      "step": 24379
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8465549936788875e-06,
+      "loss": 3.1074,
+      "step": 24380
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.846507585335019e-06,
+      "loss": 1.3761,
+      "step": 24381
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8464601769911503e-06,
+      "loss": 2.8575,
+      "step": 24382
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.846412768647282e-06,
+      "loss": 1.6365,
+      "step": 24383
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8463653603034135e-06,
+      "loss": 3.4521,
+      "step": 24384
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8463179519595448e-06,
+      "loss": 2.4359,
+      "step": 24385
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8462705436156765e-06,
+      "loss": 3.0281,
+      "step": 24386
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.846223135271808e-06,
+      "loss": 2.225,
+      "step": 24387
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8461757269279392e-06,
+      "loss": 3.2934,
+      "step": 24388
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.846128318584071e-06,
+      "loss": 1.1818,
+      "step": 24389
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8460809102402022e-06,
+      "loss": 3.3396,
+      "step": 24390
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8460335018963337e-06,
+      "loss": 1.7437,
+      "step": 24391
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8459860935524654e-06,
+      "loss": 2.4584,
+      "step": 24392
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8459386852085967e-06,
+      "loss": 2.7333,
+      "step": 24393
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8458912768647282e-06,
+      "loss": 2.2933,
+      "step": 24394
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.84584386852086e-06,
+      "loss": 3.2461,
+      "step": 24395
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8457964601769912e-06,
+      "loss": 2.5922,
+      "step": 24396
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8457490518331227e-06,
+      "loss": 2.1731,
+      "step": 24397
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8457016434892544e-06,
+      "loss": 2.5742,
+      "step": 24398
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8456542351453857e-06,
+      "loss": 2.5215,
+      "step": 24399
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.845606826801517e-06,
+      "loss": 3.4284,
+      "step": 24400
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8455594184576489e-06,
+      "loss": 2.035,
+      "step": 24401
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8455120101137801e-06,
+      "loss": 4.1362,
+      "step": 24402
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8454646017699114e-06,
+      "loss": 1.5605,
+      "step": 24403
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.845417193426043e-06,
+      "loss": 2.9933,
+      "step": 24404
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8453697850821746e-06,
+      "loss": 0.9109,
+      "step": 24405
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8453223767383059e-06,
+      "loss": 2.771,
+      "step": 24406
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8452749683944374e-06,
+      "loss": 0.6281,
+      "step": 24407
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.845227560050569e-06,
+      "loss": 4.8052,
+      "step": 24408
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8451801517067004e-06,
+      "loss": 4.0674,
+      "step": 24409
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8451327433628319e-06,
+      "loss": 2.9063,
+      "step": 24410
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8450853350189636e-06,
+      "loss": 1.1007,
+      "step": 24411
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8450379266750948e-06,
+      "loss": 1.8332,
+      "step": 24412
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.8449905183312261e-06,
+      "loss": 1.8676,
+      "step": 24413
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8449431099873578e-06,
+      "loss": 2.1877,
+      "step": 24414
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8448957016434893e-06,
+      "loss": 1.6957,
+      "step": 24415
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8448482932996206e-06,
+      "loss": 2.115,
+      "step": 24416
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8448008849557523e-06,
+      "loss": 2.7727,
+      "step": 24417
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8447534766118838e-06,
+      "loss": 2.7438,
+      "step": 24418
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.844706068268015e-06,
+      "loss": 2.4459,
+      "step": 24419
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8446586599241468e-06,
+      "loss": 2.574,
+      "step": 24420
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8446112515802783e-06,
+      "loss": 2.551,
+      "step": 24421
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8445638432364096e-06,
+      "loss": 2.1773,
+      "step": 24422
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8445164348925413e-06,
+      "loss": 2.29,
+      "step": 24423
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8444690265486728e-06,
+      "loss": 2.4228,
+      "step": 24424
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.844421618204804e-06,
+      "loss": 2.7257,
+      "step": 24425
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8443742098609353e-06,
+      "loss": 2.5466,
+      "step": 24426
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.844326801517067e-06,
+      "loss": 1.2903,
+      "step": 24427
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8442793931731985e-06,
+      "loss": 1.9973,
+      "step": 24428
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8442319848293298e-06,
+      "loss": 3.1652,
+      "step": 24429
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8441845764854615e-06,
+      "loss": 1.7442,
+      "step": 24430
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.844137168141593e-06,
+      "loss": 3.0059,
+      "step": 24431
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8440897597977243e-06,
+      "loss": 2.3703,
+      "step": 24432
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.844042351453856e-06,
+      "loss": 1.4271,
+      "step": 24433
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8439949431099875e-06,
+      "loss": 2.0648,
+      "step": 24434
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8439475347661187e-06,
+      "loss": 2.2878,
+      "step": 24435
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8439001264222504e-06,
+      "loss": 2.3106,
+      "step": 24436
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.843852718078382e-06,
+      "loss": 3.9608,
+      "step": 24437
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8438053097345132e-06,
+      "loss": 3.8162,
+      "step": 24438
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.843757901390645e-06,
+      "loss": 1.6413,
+      "step": 24439
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8437104930467762e-06,
+      "loss": 2.5165,
+      "step": 24440
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8436630847029077e-06,
+      "loss": 3.3742,
+      "step": 24441
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8436156763590394e-06,
+      "loss": 2.2602,
+      "step": 24442
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8435682680151707e-06,
+      "loss": 3.1241,
+      "step": 24443
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8435208596713022e-06,
+      "loss": 2.6146,
+      "step": 24444
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8434734513274339e-06,
+      "loss": 2.6651,
+      "step": 24445
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8434260429835652e-06,
+      "loss": 2.175,
+      "step": 24446
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8433786346396966e-06,
+      "loss": 2.7274,
+      "step": 24447
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.843331226295828e-06,
+      "loss": 4.064,
+      "step": 24448
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8432838179519596e-06,
+      "loss": 2.8265,
+      "step": 24449
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8432364096080911e-06,
+      "loss": 3.1162,
+      "step": 24450
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8431890012642224e-06,
+      "loss": 3.8744,
+      "step": 24451
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8431415929203541e-06,
+      "loss": 2.5114,
+      "step": 24452
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8430941845764854e-06,
+      "loss": 2.4243,
+      "step": 24453
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8430467762326169e-06,
+      "loss": 2.6965,
+      "step": 24454
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8429993678887486e-06,
+      "loss": 3.044,
+      "step": 24455
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8429519595448799e-06,
+      "loss": 3.1034,
+      "step": 24456
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8429045512010114e-06,
+      "loss": 3.0719,
+      "step": 24457
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.842857142857143e-06,
+      "loss": 1.9607,
+      "step": 24458
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8428097345132743e-06,
+      "loss": 3.7593,
+      "step": 24459
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8427623261694058e-06,
+      "loss": 3.0818,
+      "step": 24460
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8427149178255375e-06,
+      "loss": 1.9415,
+      "step": 24461
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8426675094816688e-06,
+      "loss": 2.4681,
+      "step": 24462
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8426201011378003e-06,
+      "loss": 2.9244,
+      "step": 24463
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.842572692793932e-06,
+      "loss": 3.3275,
+      "step": 24464
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8425252844500633e-06,
+      "loss": 2.0813,
+      "step": 24465
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8424778761061946e-06,
+      "loss": 3.1309,
+      "step": 24466
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8424304677623263e-06,
+      "loss": 2.6327,
+      "step": 24467
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8423830594184578e-06,
+      "loss": 4.0306,
+      "step": 24468
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.842335651074589e-06,
+      "loss": 2.1463,
+      "step": 24469
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8422882427307205e-06,
+      "loss": 3.1467,
+      "step": 24470
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8422408343868522e-06,
+      "loss": 3.9244,
+      "step": 24471
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8421934260429835e-06,
+      "loss": 2.7695,
+      "step": 24472
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.842146017699115e-06,
+      "loss": 2.4408,
+      "step": 24473
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8420986093552467e-06,
+      "loss": 1.6091,
+      "step": 24474
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.842051201011378e-06,
+      "loss": 1.6056,
+      "step": 24475
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8420037926675095e-06,
+      "loss": 1.3235,
+      "step": 24476
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.8419563843236412e-06,
+      "loss": 2.3676,
+      "step": 24477
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8419089759797725e-06,
+      "loss": 2.2663,
+      "step": 24478
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8418615676359038e-06,
+      "loss": 3.224,
+      "step": 24479
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8418141592920355e-06,
+      "loss": 2.1554,
+      "step": 24480
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.841766750948167e-06,
+      "loss": 2.0324,
+      "step": 24481
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8417193426042982e-06,
+      "loss": 2.9504,
+      "step": 24482
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.84167193426043e-06,
+      "loss": 3.0516,
+      "step": 24483
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8416245259165614e-06,
+      "loss": 3.0587,
+      "step": 24484
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8415771175726927e-06,
+      "loss": 2.5113,
+      "step": 24485
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8415297092288244e-06,
+      "loss": 3.9026,
+      "step": 24486
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.841482300884956e-06,
+      "loss": 2.8881,
+      "step": 24487
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8414348925410872e-06,
+      "loss": 3.1306,
+      "step": 24488
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.841387484197219e-06,
+      "loss": 3.8707,
+      "step": 24489
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8413400758533504e-06,
+      "loss": 2.6191,
+      "step": 24490
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8412926675094817e-06,
+      "loss": 1.8205,
+      "step": 24491
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.841245259165613e-06,
+      "loss": 2.8908,
+      "step": 24492
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8411978508217447e-06,
+      "loss": 3.2845,
+      "step": 24493
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8411504424778761e-06,
+      "loss": 2.5938,
+      "step": 24494
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8411030341340074e-06,
+      "loss": 2.5202,
+      "step": 24495
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8410556257901391e-06,
+      "loss": 3.5954,
+      "step": 24496
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8410082174462706e-06,
+      "loss": 2.051,
+      "step": 24497
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.840960809102402e-06,
+      "loss": 2.6884,
+      "step": 24498
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8409134007585336e-06,
+      "loss": 2.6296,
+      "step": 24499
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.840865992414665e-06,
+      "loss": 2.8081,
+      "step": 24500
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8408185840707964e-06,
+      "loss": 3.7577,
+      "step": 24501
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.840771175726928e-06,
+      "loss": 2.5973,
+      "step": 24502
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8407237673830596e-06,
+      "loss": 1.9575,
+      "step": 24503
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8406763590391909e-06,
+      "loss": 1.638,
+      "step": 24504
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8406289506953226e-06,
+      "loss": 1.6247,
+      "step": 24505
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8405815423514538e-06,
+      "loss": 3.0668,
+      "step": 24506
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8405341340075853e-06,
+      "loss": 2.3467,
+      "step": 24507
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.840486725663717e-06,
+      "loss": 2.1265,
+      "step": 24508
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8404393173198483e-06,
+      "loss": 1.7356,
+      "step": 24509
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8403919089759798e-06,
+      "loss": 3.8432,
+      "step": 24510
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8403445006321115e-06,
+      "loss": 2.6655,
+      "step": 24511
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8402970922882428e-06,
+      "loss": 3.418,
+      "step": 24512
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8402496839443743e-06,
+      "loss": 2.9092,
+      "step": 24513
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8402022756005056e-06,
+      "loss": 3.1432,
+      "step": 24514
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8401548672566373e-06,
+      "loss": 2.5561,
+      "step": 24515
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8401074589127688e-06,
+      "loss": 3.0526,
+      "step": 24516
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8400600505689e-06,
+      "loss": 2.8418,
+      "step": 24517
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8400126422250317e-06,
+      "loss": 1.8473,
+      "step": 24518
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.839965233881163e-06,
+      "loss": 2.0731,
+      "step": 24519
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8399178255372945e-06,
+      "loss": 3.2764,
+      "step": 24520
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8398704171934262e-06,
+      "loss": 3.2113,
+      "step": 24521
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8398230088495575e-06,
+      "loss": 3.1816,
+      "step": 24522
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.839775600505689e-06,
+      "loss": 2.4843,
+      "step": 24523
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8397281921618207e-06,
+      "loss": 3.3619,
+      "step": 24524
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.839680783817952e-06,
+      "loss": 2.8573,
+      "step": 24525
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8396333754740835e-06,
+      "loss": 3.1147,
+      "step": 24526
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8395859671302152e-06,
+      "loss": 2.7274,
+      "step": 24527
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8395385587863465e-06,
+      "loss": 2.7501,
+      "step": 24528
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.839491150442478e-06,
+      "loss": 2.6922,
+      "step": 24529
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8394437420986097e-06,
+      "loss": 3.388,
+      "step": 24530
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.839396333754741e-06,
+      "loss": 1.6589,
+      "step": 24531
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8393489254108722e-06,
+      "loss": 0.9878,
+      "step": 24532
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.839301517067004e-06,
+      "loss": 2.5926,
+      "step": 24533
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8392541087231354e-06,
+      "loss": 2.8496,
+      "step": 24534
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8392067003792667e-06,
+      "loss": 2.8631,
+      "step": 24535
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8391592920353982e-06,
+      "loss": 2.1101,
+      "step": 24536
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8391118836915299e-06,
+      "loss": 3.0506,
+      "step": 24537
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8390644753476612e-06,
+      "loss": 3.1053,
+      "step": 24538
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8390170670037927e-06,
+      "loss": 2.3858,
+      "step": 24539
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.8389696586599244e-06,
+      "loss": 2.2245,
+      "step": 24540
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8389222503160556e-06,
+      "loss": 2.8018,
+      "step": 24541
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.838874841972187e-06,
+      "loss": 2.6255,
+      "step": 24542
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8388274336283188e-06,
+      "loss": 2.2575,
+      "step": 24543
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8387800252844501e-06,
+      "loss": 2.2403,
+      "step": 24544
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8387326169405814e-06,
+      "loss": 2.9075,
+      "step": 24545
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.838685208596713e-06,
+      "loss": 1.4157,
+      "step": 24546
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8386378002528446e-06,
+      "loss": 1.8342,
+      "step": 24547
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8385903919089759e-06,
+      "loss": 2.7787,
+      "step": 24548
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8385429835651076e-06,
+      "loss": 1.6861,
+      "step": 24549
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.838495575221239e-06,
+      "loss": 3.3075,
+      "step": 24550
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8384481668773704e-06,
+      "loss": 3.4172,
+      "step": 24551
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.838400758533502e-06,
+      "loss": 2.1012,
+      "step": 24552
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8383533501896335e-06,
+      "loss": 2.5396,
+      "step": 24553
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8383059418457648e-06,
+      "loss": 2.7823,
+      "step": 24554
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8382585335018965e-06,
+      "loss": 2.0235,
+      "step": 24555
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8382111251580278e-06,
+      "loss": 2.1569,
+      "step": 24556
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8381637168141593e-06,
+      "loss": 1.2856,
+      "step": 24557
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8381163084702906e-06,
+      "loss": 2.8522,
+      "step": 24558
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8380689001264223e-06,
+      "loss": 2.1127,
+      "step": 24559
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8380214917825538e-06,
+      "loss": 2.0243,
+      "step": 24560
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837974083438685e-06,
+      "loss": 3.8889,
+      "step": 24561
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8379266750948168e-06,
+      "loss": 3.8044,
+      "step": 24562
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8378792667509483e-06,
+      "loss": 3.0882,
+      "step": 24563
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8378318584070795e-06,
+      "loss": 2.3288,
+      "step": 24564
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8377844500632112e-06,
+      "loss": 3.0468,
+      "step": 24565
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8377370417193427e-06,
+      "loss": 3.0002,
+      "step": 24566
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837689633375474e-06,
+      "loss": 2.9927,
+      "step": 24567
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8376422250316057e-06,
+      "loss": 2.3002,
+      "step": 24568
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837594816687737e-06,
+      "loss": 2.1181,
+      "step": 24569
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8375474083438685e-06,
+      "loss": 2.4451,
+      "step": 24570
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8375000000000002e-06,
+      "loss": 2.6854,
+      "step": 24571
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8374525916561315e-06,
+      "loss": 2.9631,
+      "step": 24572
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837405183312263e-06,
+      "loss": 2.5574,
+      "step": 24573
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8373577749683947e-06,
+      "loss": 1.2112,
+      "step": 24574
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837310366624526e-06,
+      "loss": 1.9325,
+      "step": 24575
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8372629582806574e-06,
+      "loss": 3.6993,
+      "step": 24576
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8372155499367891e-06,
+      "loss": 2.7175,
+      "step": 24577
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8371681415929204e-06,
+      "loss": 1.9887,
+      "step": 24578
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837120733249052e-06,
+      "loss": 1.5323,
+      "step": 24579
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8370733249051832e-06,
+      "loss": 2.7961,
+      "step": 24580
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.837025916561315e-06,
+      "loss": 1.9945,
+      "step": 24581
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8369785082174462e-06,
+      "loss": 1.6105,
+      "step": 24582
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8369310998735777e-06,
+      "loss": 3.2859,
+      "step": 24583
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8368836915297094e-06,
+      "loss": 3.5598,
+      "step": 24584
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8368362831858407e-06,
+      "loss": 2.9476,
+      "step": 24585
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8367888748419722e-06,
+      "loss": 2.5437,
+      "step": 24586
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8367414664981039e-06,
+      "loss": 1.965,
+      "step": 24587
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8366940581542351e-06,
+      "loss": 2.146,
+      "step": 24588
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8366466498103666e-06,
+      "loss": 1.7558,
+      "step": 24589
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8365992414664983e-06,
+      "loss": 2.3911,
+      "step": 24590
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8365518331226296e-06,
+      "loss": 2.0138,
+      "step": 24591
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8365044247787611e-06,
+      "loss": 2.9333,
+      "step": 24592
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8364570164348928e-06,
+      "loss": 2.2573,
+      "step": 24593
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.836409608091024e-06,
+      "loss": 3.7248,
+      "step": 24594
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8363621997471554e-06,
+      "loss": 3.0987,
+      "step": 24595
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.836314791403287e-06,
+      "loss": 2.3443,
+      "step": 24596
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8362673830594186e-06,
+      "loss": 2.4074,
+      "step": 24597
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8362199747155498e-06,
+      "loss": 3.049,
+      "step": 24598
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8361725663716816e-06,
+      "loss": 2.9924,
+      "step": 24599
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.836125158027813e-06,
+      "loss": 1.4804,
+      "step": 24600
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8360777496839443e-06,
+      "loss": 1.8371,
+      "step": 24601
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8360303413400758e-06,
+      "loss": 3.112,
+      "step": 24602
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.8359829329962075e-06,
+      "loss": 3.8479,
+      "step": 24603
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8359355246523388e-06,
+      "loss": 3.5788,
+      "step": 24604
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8358881163084703e-06,
+      "loss": 3.1413,
+      "step": 24605
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.835840707964602e-06,
+      "loss": 2.1642,
+      "step": 24606
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8357932996207333e-06,
+      "loss": 2.8055,
+      "step": 24607
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8357458912768646e-06,
+      "loss": 2.1307,
+      "step": 24608
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8356984829329963e-06,
+      "loss": 2.0122,
+      "step": 24609
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8356510745891278e-06,
+      "loss": 2.6335,
+      "step": 24610
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.835603666245259e-06,
+      "loss": 2.3773,
+      "step": 24611
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8355562579013907e-06,
+      "loss": 2.5601,
+      "step": 24612
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8355088495575222e-06,
+      "loss": 2.0991,
+      "step": 24613
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8354614412136535e-06,
+      "loss": 2.2857,
+      "step": 24614
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8354140328697852e-06,
+      "loss": 2.9169,
+      "step": 24615
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8353666245259167e-06,
+      "loss": 2.9242,
+      "step": 24616
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.835319216182048e-06,
+      "loss": 2.4296,
+      "step": 24617
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8352718078381797e-06,
+      "loss": 3.522,
+      "step": 24618
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8352243994943112e-06,
+      "loss": 3.8653,
+      "step": 24619
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8351769911504425e-06,
+      "loss": 2.4955,
+      "step": 24620
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8351295828065742e-06,
+      "loss": 1.3884,
+      "step": 24621
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8350821744627054e-06,
+      "loss": 1.984,
+      "step": 24622
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.835034766118837e-06,
+      "loss": 3.0674,
+      "step": 24623
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8349873577749682e-06,
+      "loss": 1.9979,
+      "step": 24624
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8349399494311e-06,
+      "loss": 1.9242,
+      "step": 24625
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8348925410872314e-06,
+      "loss": 0.5823,
+      "step": 24626
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8348451327433627e-06,
+      "loss": 3.5685,
+      "step": 24627
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8347977243994944e-06,
+      "loss": 3.2199,
+      "step": 24628
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8347503160556259e-06,
+      "loss": 1.7397,
+      "step": 24629
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8347029077117572e-06,
+      "loss": 3.1151,
+      "step": 24630
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8346554993678889e-06,
+      "loss": 2.6981,
+      "step": 24631
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8346080910240204e-06,
+      "loss": 2.3808,
+      "step": 24632
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8345606826801517e-06,
+      "loss": 3.2213,
+      "step": 24633
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8345132743362834e-06,
+      "loss": 2.4755,
+      "step": 24634
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8344658659924146e-06,
+      "loss": 2.8318,
+      "step": 24635
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8344184576485461e-06,
+      "loss": 2.6161,
+      "step": 24636
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8343710493046778e-06,
+      "loss": 2.6437,
+      "step": 24637
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8343236409608091e-06,
+      "loss": 2.4728,
+      "step": 24638
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8342762326169406e-06,
+      "loss": 2.287,
+      "step": 24639
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8342288242730723e-06,
+      "loss": 1.44,
+      "step": 24640
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8341814159292036e-06,
+      "loss": 1.5848,
+      "step": 24641
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.834134007585335e-06,
+      "loss": 3.4198,
+      "step": 24642
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8340865992414668e-06,
+      "loss": 4.2125,
+      "step": 24643
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.834039190897598e-06,
+      "loss": 2.1067,
+      "step": 24644
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8339917825537296e-06,
+      "loss": 2.913,
+      "step": 24645
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8339443742098608e-06,
+      "loss": 3.2054,
+      "step": 24646
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8338969658659925e-06,
+      "loss": 3.0827,
+      "step": 24647
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8338495575221238e-06,
+      "loss": 2.2206,
+      "step": 24648
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8338021491782553e-06,
+      "loss": 2.7386,
+      "step": 24649
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.833754740834387e-06,
+      "loss": 2.1555,
+      "step": 24650
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8337073324905183e-06,
+      "loss": 2.2148,
+      "step": 24651
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8336599241466498e-06,
+      "loss": 3.1544,
+      "step": 24652
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8336125158027815e-06,
+      "loss": 2.1679,
+      "step": 24653
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8335651074589128e-06,
+      "loss": 3.8316,
+      "step": 24654
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8335176991150443e-06,
+      "loss": 2.4192,
+      "step": 24655
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.833470290771176e-06,
+      "loss": 3.1588,
+      "step": 24656
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8334228824273073e-06,
+      "loss": 2.1286,
+      "step": 24657
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8333754740834387e-06,
+      "loss": 1.8503,
+      "step": 24658
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8333280657395704e-06,
+      "loss": 2.3937,
+      "step": 24659
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8332806573957017e-06,
+      "loss": 2.6174,
+      "step": 24660
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.833233249051833e-06,
+      "loss": 2.1867,
+      "step": 24661
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8331858407079647e-06,
+      "loss": 2.0615,
+      "step": 24662
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8331384323640962e-06,
+      "loss": 2.8644,
+      "step": 24663
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8330910240202275e-06,
+      "loss": 1.9194,
+      "step": 24664
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.833043615676359e-06,
+      "loss": 2.2795,
+      "step": 24665
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.8329962073324907e-06,
+      "loss": 1.9705,
+      "step": 24666
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.832948798988622e-06,
+      "loss": 1.9912,
+      "step": 24667
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8329013906447535e-06,
+      "loss": 3.9123,
+      "step": 24668
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8328539823008852e-06,
+      "loss": 1.0628,
+      "step": 24669
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8328065739570164e-06,
+      "loss": 0.8766,
+      "step": 24670
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.832759165613148e-06,
+      "loss": 2.711,
+      "step": 24671
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8327117572692796e-06,
+      "loss": 2.6447,
+      "step": 24672
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.832664348925411e-06,
+      "loss": 1.8367,
+      "step": 24673
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8326169405815422e-06,
+      "loss": 2.7964,
+      "step": 24674
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.832569532237674e-06,
+      "loss": 3.1781,
+      "step": 24675
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8325221238938054e-06,
+      "loss": 3.1087,
+      "step": 24676
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8324747155499367e-06,
+      "loss": 3.2235,
+      "step": 24677
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8324273072060684e-06,
+      "loss": 3.2865,
+      "step": 24678
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8323798988621999e-06,
+      "loss": 3.6348,
+      "step": 24679
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8323324905183311e-06,
+      "loss": 2.3736,
+      "step": 24680
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8322850821744629e-06,
+      "loss": 1.7262,
+      "step": 24681
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8322376738305943e-06,
+      "loss": 2.315,
+      "step": 24682
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8321902654867256e-06,
+      "loss": 2.039,
+      "step": 24683
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8321428571428573e-06,
+      "loss": 2.3757,
+      "step": 24684
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8320954487989888e-06,
+      "loss": 1.479,
+      "step": 24685
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.83204804045512e-06,
+      "loss": 3.2166,
+      "step": 24686
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8320006321112514e-06,
+      "loss": 3.5317,
+      "step": 24687
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.831953223767383e-06,
+      "loss": 3.4657,
+      "step": 24688
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8319058154235146e-06,
+      "loss": 1.5155,
+      "step": 24689
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8318584070796459e-06,
+      "loss": 0.9916,
+      "step": 24690
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8318109987357776e-06,
+      "loss": 3.6789,
+      "step": 24691
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.831763590391909e-06,
+      "loss": 3.5847,
+      "step": 24692
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8317161820480403e-06,
+      "loss": 2.6811,
+      "step": 24693
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.831668773704172e-06,
+      "loss": 2.9349,
+      "step": 24694
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8316213653603035e-06,
+      "loss": 2.8044,
+      "step": 24695
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8315739570164348e-06,
+      "loss": 1.5684,
+      "step": 24696
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8315265486725665e-06,
+      "loss": 2.9395,
+      "step": 24697
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8314791403286978e-06,
+      "loss": 2.2743,
+      "step": 24698
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8314317319848293e-06,
+      "loss": 3.0311,
+      "step": 24699
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.831384323640961e-06,
+      "loss": 1.7277,
+      "step": 24700
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8313369152970923e-06,
+      "loss": 2.6297,
+      "step": 24701
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8312895069532238e-06,
+      "loss": 3.0556,
+      "step": 24702
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8312420986093555e-06,
+      "loss": 1.5083,
+      "step": 24703
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8311946902654867e-06,
+      "loss": 3.3877,
+      "step": 24704
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8311472819216182e-06,
+      "loss": 4.0809,
+      "step": 24705
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.83109987357775e-06,
+      "loss": 2.4147,
+      "step": 24706
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8310524652338812e-06,
+      "loss": 3.3428,
+      "step": 24707
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8310050568900127e-06,
+      "loss": 2.9094,
+      "step": 24708
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.830957648546144e-06,
+      "loss": 1.5993,
+      "step": 24709
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8309102402022757e-06,
+      "loss": 2.8902,
+      "step": 24710
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.830862831858407e-06,
+      "loss": 3.113,
+      "step": 24711
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8308154235145385e-06,
+      "loss": 3.0207,
+      "step": 24712
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8307680151706702e-06,
+      "loss": 1.9734,
+      "step": 24713
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8307206068268015e-06,
+      "loss": 3.5364,
+      "step": 24714
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.830673198482933e-06,
+      "loss": 2.6861,
+      "step": 24715
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8306257901390647e-06,
+      "loss": 3.031,
+      "step": 24716
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.830578381795196e-06,
+      "loss": 2.812,
+      "step": 24717
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8305309734513274e-06,
+      "loss": 1.977,
+      "step": 24718
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8304835651074591e-06,
+      "loss": 1.4956,
+      "step": 24719
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8304361567635904e-06,
+      "loss": 2.7066,
+      "step": 24720
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.830388748419722e-06,
+      "loss": 2.2641,
+      "step": 24721
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8303413400758536e-06,
+      "loss": 2.5221,
+      "step": 24722
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8302939317319849e-06,
+      "loss": 2.8385,
+      "step": 24723
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8302465233881162e-06,
+      "loss": 1.2176,
+      "step": 24724
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8301991150442479e-06,
+      "loss": 2.9316,
+      "step": 24725
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8301517067003794e-06,
+      "loss": 1.6446,
+      "step": 24726
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8301042983565106e-06,
+      "loss": 3.6528,
+      "step": 24727
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8300568900126423e-06,
+      "loss": 2.0002,
+      "step": 24728
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8300094816687738e-06,
+      "loss": 3.2505,
+      "step": 24729
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.8299620733249051e-06,
+      "loss": 2.2225,
+      "step": 24730
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8299146649810366e-06,
+      "loss": 2.7805,
+      "step": 24731
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8298672566371683e-06,
+      "loss": 3.4576,
+      "step": 24732
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8298198482932996e-06,
+      "loss": 2.9496,
+      "step": 24733
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829772439949431e-06,
+      "loss": 3.1087,
+      "step": 24734
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8297250316055628e-06,
+      "loss": 2.6169,
+      "step": 24735
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829677623261694e-06,
+      "loss": 1.5714,
+      "step": 24736
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8296302149178254e-06,
+      "loss": 2.8659,
+      "step": 24737
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829582806573957e-06,
+      "loss": 2.6198,
+      "step": 24738
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8295353982300886e-06,
+      "loss": 3.6359,
+      "step": 24739
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8294879898862198e-06,
+      "loss": 1.3936,
+      "step": 24740
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8294405815423515e-06,
+      "loss": 2.2718,
+      "step": 24741
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829393173198483e-06,
+      "loss": 2.2488,
+      "step": 24742
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8293457648546143e-06,
+      "loss": 2.1079,
+      "step": 24743
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829298356510746e-06,
+      "loss": 1.8765,
+      "step": 24744
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8292509481668775e-06,
+      "loss": 3.03,
+      "step": 24745
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8292035398230088e-06,
+      "loss": 2.1498,
+      "step": 24746
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8291561314791405e-06,
+      "loss": 2.8576,
+      "step": 24747
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829108723135272e-06,
+      "loss": 2.821,
+      "step": 24748
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8290613147914033e-06,
+      "loss": 2.3635,
+      "step": 24749
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.829013906447535e-06,
+      "loss": 2.6882,
+      "step": 24750
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8289664981036662e-06,
+      "loss": 3.2835,
+      "step": 24751
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8289190897597977e-06,
+      "loss": 2.0992,
+      "step": 24752
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.828871681415929e-06,
+      "loss": 3.1115,
+      "step": 24753
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8288242730720607e-06,
+      "loss": 3.5091,
+      "step": 24754
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8287768647281922e-06,
+      "loss": 2.7186,
+      "step": 24755
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8287294563843235e-06,
+      "loss": 3.412,
+      "step": 24756
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8286820480404552e-06,
+      "loss": 3.723,
+      "step": 24757
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8286346396965867e-06,
+      "loss": 3.6593,
+      "step": 24758
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.828587231352718e-06,
+      "loss": 2.2747,
+      "step": 24759
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8285398230088497e-06,
+      "loss": 2.0736,
+      "step": 24760
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8284924146649812e-06,
+      "loss": 1.162,
+      "step": 24761
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8284450063211124e-06,
+      "loss": 1.4619,
+      "step": 24762
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8283975979772442e-06,
+      "loss": 1.9787,
+      "step": 24763
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8283501896333754e-06,
+      "loss": 2.9124,
+      "step": 24764
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.828302781289507e-06,
+      "loss": 2.3338,
+      "step": 24765
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8282553729456386e-06,
+      "loss": 3.2634,
+      "step": 24766
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.82820796460177e-06,
+      "loss": 3.1773,
+      "step": 24767
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8281605562579014e-06,
+      "loss": 2.3233,
+      "step": 24768
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.828113147914033e-06,
+      "loss": 1.17,
+      "step": 24769
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8280657395701644e-06,
+      "loss": 2.9771,
+      "step": 24770
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8280183312262959e-06,
+      "loss": 2.842,
+      "step": 24771
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8279709228824276e-06,
+      "loss": 2.1188,
+      "step": 24772
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8279235145385589e-06,
+      "loss": 1.331,
+      "step": 24773
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8278761061946904e-06,
+      "loss": 2.327,
+      "step": 24774
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8278286978508216e-06,
+      "loss": 1.9484,
+      "step": 24775
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8277812895069533e-06,
+      "loss": 2.9154,
+      "step": 24776
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8277338811630846e-06,
+      "loss": 2.7176,
+      "step": 24777
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8276864728192161e-06,
+      "loss": 1.0766,
+      "step": 24778
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8276390644753478e-06,
+      "loss": 3.031,
+      "step": 24779
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.827591656131479e-06,
+      "loss": 3.6133,
+      "step": 24780
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8275442477876106e-06,
+      "loss": 2.7467,
+      "step": 24781
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8274968394437423e-06,
+      "loss": 3.7309,
+      "step": 24782
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8274494310998736e-06,
+      "loss": 1.2681,
+      "step": 24783
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.827402022756005e-06,
+      "loss": 2.1711,
+      "step": 24784
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8273546144121368e-06,
+      "loss": 3.3507,
+      "step": 24785
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.827307206068268e-06,
+      "loss": 1.6385,
+      "step": 24786
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8272597977243995e-06,
+      "loss": 2.6535,
+      "step": 24787
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8272123893805312e-06,
+      "loss": 3.5731,
+      "step": 24788
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8271649810366625e-06,
+      "loss": 1.0111,
+      "step": 24789
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8271175726927938e-06,
+      "loss": 1.7646,
+      "step": 24790
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8270701643489255e-06,
+      "loss": 1.0214,
+      "step": 24791
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.827022756005057e-06,
+      "loss": 2.8489,
+      "step": 24792
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.8269753476611883e-06,
+      "loss": 2.151,
+      "step": 24793
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.82692793931732e-06,
+      "loss": 2.5845,
+      "step": 24794
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8268805309734515e-06,
+      "loss": 2.4411,
+      "step": 24795
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8268331226295828e-06,
+      "loss": 3.4404,
+      "step": 24796
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8267857142857143e-06,
+      "loss": 3.1024,
+      "step": 24797
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.826738305941846e-06,
+      "loss": 1.8548,
+      "step": 24798
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8266908975979772e-06,
+      "loss": 1.9111,
+      "step": 24799
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8266434892541087e-06,
+      "loss": 3.0152,
+      "step": 24800
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8265960809102404e-06,
+      "loss": 1.4031,
+      "step": 24801
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8265486725663717e-06,
+      "loss": 2.7116,
+      "step": 24802
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.826501264222503e-06,
+      "loss": 3.3806,
+      "step": 24803
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8264538558786347e-06,
+      "loss": 2.6852,
+      "step": 24804
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8264064475347662e-06,
+      "loss": 1.8566,
+      "step": 24805
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8263590391908975e-06,
+      "loss": 2.2819,
+      "step": 24806
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8263116308470292e-06,
+      "loss": 2.1473,
+      "step": 24807
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8262642225031607e-06,
+      "loss": 2.3957,
+      "step": 24808
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.826216814159292e-06,
+      "loss": 1.5271,
+      "step": 24809
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8261694058154236e-06,
+      "loss": 2.1175,
+      "step": 24810
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8261219974715551e-06,
+      "loss": 2.6311,
+      "step": 24811
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8260745891276864e-06,
+      "loss": 3.1326,
+      "step": 24812
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8260271807838181e-06,
+      "loss": 2.7985,
+      "step": 24813
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8259797724399496e-06,
+      "loss": 3.0698,
+      "step": 24814
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.825932364096081e-06,
+      "loss": 2.6463,
+      "step": 24815
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8258849557522126e-06,
+      "loss": 1.6647,
+      "step": 24816
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8258375474083439e-06,
+      "loss": 1.799,
+      "step": 24817
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8257901390644754e-06,
+      "loss": 3.5303,
+      "step": 24818
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8257427307206067e-06,
+      "loss": 2.7076,
+      "step": 24819
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8256953223767384e-06,
+      "loss": 2.8558,
+      "step": 24820
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8256479140328699e-06,
+      "loss": 2.7486,
+      "step": 24821
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8256005056890011e-06,
+      "loss": 3.5724,
+      "step": 24822
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8255530973451328e-06,
+      "loss": 1.6894,
+      "step": 24823
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8255056890012643e-06,
+      "loss": 3.3536,
+      "step": 24824
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8254582806573956e-06,
+      "loss": 2.6579,
+      "step": 24825
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8254108723135273e-06,
+      "loss": 1.8669,
+      "step": 24826
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8253634639696586e-06,
+      "loss": 2.5114,
+      "step": 24827
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.82531605562579e-06,
+      "loss": 2.0575,
+      "step": 24828
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8252686472819218e-06,
+      "loss": 3.368,
+      "step": 24829
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.825221238938053e-06,
+      "loss": 2.6522,
+      "step": 24830
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8251738305941846e-06,
+      "loss": 3.4154,
+      "step": 24831
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8251264222503163e-06,
+      "loss": 2.0717,
+      "step": 24832
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8250790139064475e-06,
+      "loss": 2.4758,
+      "step": 24833
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.825031605562579e-06,
+      "loss": 3.7626,
+      "step": 24834
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8249841972187107e-06,
+      "loss": 2.3332,
+      "step": 24835
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.824936788874842e-06,
+      "loss": 3.849,
+      "step": 24836
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8248893805309735e-06,
+      "loss": 2.1174,
+      "step": 24837
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8248419721871052e-06,
+      "loss": 1.3123,
+      "step": 24838
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8247945638432365e-06,
+      "loss": 3.2243,
+      "step": 24839
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8247471554993678e-06,
+      "loss": 2.7023,
+      "step": 24840
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8246997471554993e-06,
+      "loss": 2.8234,
+      "step": 24841
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.824652338811631e-06,
+      "loss": 2.8083,
+      "step": 24842
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8246049304677623e-06,
+      "loss": 2.7807,
+      "step": 24843
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8245575221238937e-06,
+      "loss": 3.6846,
+      "step": 24844
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8245101137800255e-06,
+      "loss": 3.0557,
+      "step": 24845
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8244627054361567e-06,
+      "loss": 2.5273,
+      "step": 24846
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8244152970922882e-06,
+      "loss": 2.6391,
+      "step": 24847
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.82436788874842e-06,
+      "loss": 3.4407,
+      "step": 24848
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8243204804045512e-06,
+      "loss": 2.747,
+      "step": 24849
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8242730720606827e-06,
+      "loss": 3.8489,
+      "step": 24850
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8242256637168144e-06,
+      "loss": 1.8082,
+      "step": 24851
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8241782553729457e-06,
+      "loss": 2.6671,
+      "step": 24852
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.824130847029077e-06,
+      "loss": 1.9068,
+      "step": 24853
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8240834386852087e-06,
+      "loss": 2.5324,
+      "step": 24854
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8240360303413402e-06,
+      "loss": 2.1877,
+      "step": 24855
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8239886219974714e-06,
+      "loss": 3.3558,
+      "step": 24856
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.8239412136536031e-06,
+      "loss": 2.8104,
+      "step": 24857
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8238938053097346e-06,
+      "loss": 3.4737,
+      "step": 24858
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.823846396965866e-06,
+      "loss": 3.7335,
+      "step": 24859
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8237989886219976e-06,
+      "loss": 3.6473,
+      "step": 24860
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8237515802781291e-06,
+      "loss": 2.4786,
+      "step": 24861
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8237041719342604e-06,
+      "loss": 2.764,
+      "step": 24862
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8236567635903919e-06,
+      "loss": 2.4823,
+      "step": 24863
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8236093552465236e-06,
+      "loss": 3.6875,
+      "step": 24864
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8235619469026549e-06,
+      "loss": 2.1071,
+      "step": 24865
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8235145385587862e-06,
+      "loss": 2.2752,
+      "step": 24866
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8234671302149179e-06,
+      "loss": 1.8341,
+      "step": 24867
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8234197218710493e-06,
+      "loss": 2.438,
+      "step": 24868
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8233723135271806e-06,
+      "loss": 1.5763,
+      "step": 24869
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8233249051833123e-06,
+      "loss": 2.6539,
+      "step": 24870
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8232774968394438e-06,
+      "loss": 2.4895,
+      "step": 24871
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.823230088495575e-06,
+      "loss": 2.8074,
+      "step": 24872
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8231826801517068e-06,
+      "loss": 2.5205,
+      "step": 24873
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8231352718078383e-06,
+      "loss": 2.8002,
+      "step": 24874
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8230878634639696e-06,
+      "loss": 3.3259,
+      "step": 24875
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8230404551201013e-06,
+      "loss": 2.7751,
+      "step": 24876
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8229930467762328e-06,
+      "loss": 3.2109,
+      "step": 24877
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.822945638432364e-06,
+      "loss": 2.993,
+      "step": 24878
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8228982300884958e-06,
+      "loss": 2.3761,
+      "step": 24879
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.822850821744627e-06,
+      "loss": 2.8142,
+      "step": 24880
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8228034134007585e-06,
+      "loss": 3.0561,
+      "step": 24881
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8227560050568902e-06,
+      "loss": 2.0529,
+      "step": 24882
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8227085967130215e-06,
+      "loss": 2.9534,
+      "step": 24883
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.822661188369153e-06,
+      "loss": 2.6514,
+      "step": 24884
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8226137800252843e-06,
+      "loss": 2.81,
+      "step": 24885
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.822566371681416e-06,
+      "loss": 2.2931,
+      "step": 24886
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8225189633375475e-06,
+      "loss": 3.0411,
+      "step": 24887
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8224715549936788e-06,
+      "loss": 1.6691,
+      "step": 24888
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8224241466498105e-06,
+      "loss": 3.1642,
+      "step": 24889
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.822376738305942e-06,
+      "loss": 2.4781,
+      "step": 24890
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8223293299620732e-06,
+      "loss": 2.3699,
+      "step": 24891
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.822281921618205e-06,
+      "loss": 3.1947,
+      "step": 24892
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8222345132743362e-06,
+      "loss": 2.2551,
+      "step": 24893
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8221871049304677e-06,
+      "loss": 4.0389,
+      "step": 24894
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8221396965865994e-06,
+      "loss": 3.3792,
+      "step": 24895
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8220922882427307e-06,
+      "loss": 2.8531,
+      "step": 24896
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8220448798988622e-06,
+      "loss": 3.134,
+      "step": 24897
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.821997471554994e-06,
+      "loss": 1.8906,
+      "step": 24898
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8219500632111252e-06,
+      "loss": 4.058,
+      "step": 24899
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8219026548672567e-06,
+      "loss": 2.1634,
+      "step": 24900
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8218552465233884e-06,
+      "loss": 3.8975,
+      "step": 24901
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8218078381795197e-06,
+      "loss": 3.6013,
+      "step": 24902
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8217604298356512e-06,
+      "loss": 3.028,
+      "step": 24903
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8217130214917829e-06,
+      "loss": 0.8619,
+      "step": 24904
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8216656131479141e-06,
+      "loss": 2.4283,
+      "step": 24905
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8216182048040454e-06,
+      "loss": 2.5588,
+      "step": 24906
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.821570796460177e-06,
+      "loss": 2.0725,
+      "step": 24907
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8215233881163086e-06,
+      "loss": 2.3232,
+      "step": 24908
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8214759797724399e-06,
+      "loss": 2.3479,
+      "step": 24909
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8214285714285714e-06,
+      "loss": 2.255,
+      "step": 24910
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.821381163084703e-06,
+      "loss": 2.2615,
+      "step": 24911
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8213337547408344e-06,
+      "loss": 3.1901,
+      "step": 24912
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8212863463969659e-06,
+      "loss": 2.8537,
+      "step": 24913
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8212389380530976e-06,
+      "loss": 1.9119,
+      "step": 24914
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8211915297092288e-06,
+      "loss": 2.4463,
+      "step": 24915
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8211441213653603e-06,
+      "loss": 2.5016,
+      "step": 24916
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.821096713021492e-06,
+      "loss": 2.8054,
+      "step": 24917
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8210493046776233e-06,
+      "loss": 2.5503,
+      "step": 24918
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8210018963337546e-06,
+      "loss": 2.1799,
+      "step": 24919
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.8209544879898863e-06,
+      "loss": 2.5417,
+      "step": 24920
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8209070796460178e-06,
+      "loss": 2.7095,
+      "step": 24921
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.820859671302149e-06,
+      "loss": 2.5659,
+      "step": 24922
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8208122629582808e-06,
+      "loss": 1.7299,
+      "step": 24923
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8207648546144123e-06,
+      "loss": 1.9425,
+      "step": 24924
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8207174462705436e-06,
+      "loss": 2.1706,
+      "step": 24925
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8206700379266753e-06,
+      "loss": 1.1588,
+      "step": 24926
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8206226295828068e-06,
+      "loss": 3.0629,
+      "step": 24927
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.820575221238938e-06,
+      "loss": 2.6979,
+      "step": 24928
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8205278128950695e-06,
+      "loss": 1.2798,
+      "step": 24929
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8204804045512012e-06,
+      "loss": 1.282,
+      "step": 24930
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8204329962073325e-06,
+      "loss": 2.2996,
+      "step": 24931
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8203855878634638e-06,
+      "loss": 3.4791,
+      "step": 24932
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8203381795195955e-06,
+      "loss": 1.5808,
+      "step": 24933
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.820290771175727e-06,
+      "loss": 2.7069,
+      "step": 24934
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8202433628318583e-06,
+      "loss": 2.1278,
+      "step": 24935
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.82019595448799e-06,
+      "loss": 2.567,
+      "step": 24936
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8201485461441215e-06,
+      "loss": 3.4542,
+      "step": 24937
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8201011378002527e-06,
+      "loss": 2.0403,
+      "step": 24938
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8200537294563844e-06,
+      "loss": 2.4451,
+      "step": 24939
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.820006321112516e-06,
+      "loss": 3.0275,
+      "step": 24940
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8199589127686472e-06,
+      "loss": 1.683,
+      "step": 24941
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.819911504424779e-06,
+      "loss": 2.5238,
+      "step": 24942
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8198640960809104e-06,
+      "loss": 3.9137,
+      "step": 24943
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8198166877370417e-06,
+      "loss": 2.017,
+      "step": 24944
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8197692793931734e-06,
+      "loss": 3.4639,
+      "step": 24945
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8197218710493047e-06,
+      "loss": 3.1035,
+      "step": 24946
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8196744627054362e-06,
+      "loss": 2.4867,
+      "step": 24947
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8196270543615679e-06,
+      "loss": 2.7384,
+      "step": 24948
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8195796460176992e-06,
+      "loss": 4.2301,
+      "step": 24949
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8195322376738306e-06,
+      "loss": 3.6809,
+      "step": 24950
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.819484829329962e-06,
+      "loss": 2.4021,
+      "step": 24951
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8194374209860936e-06,
+      "loss": 1.7461,
+      "step": 24952
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8193900126422251e-06,
+      "loss": 4.6659,
+      "step": 24953
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8193426042983564e-06,
+      "loss": 1.8563,
+      "step": 24954
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8192951959544881e-06,
+      "loss": 3.2554,
+      "step": 24955
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8192477876106196e-06,
+      "loss": 2.9677,
+      "step": 24956
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8192003792667509e-06,
+      "loss": 3.6908,
+      "step": 24957
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8191529709228826e-06,
+      "loss": 2.9622,
+      "step": 24958
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8191055625790139e-06,
+      "loss": 3.1511,
+      "step": 24959
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8190581542351454e-06,
+      "loss": 2.7703,
+      "step": 24960
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.819010745891277e-06,
+      "loss": 3.0279,
+      "step": 24961
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8189633375474083e-06,
+      "loss": 2.8288,
+      "step": 24962
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8189159292035398e-06,
+      "loss": 0.8956,
+      "step": 24963
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8188685208596715e-06,
+      "loss": 1.4322,
+      "step": 24964
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8188211125158028e-06,
+      "loss": 2.5065,
+      "step": 24965
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8187737041719343e-06,
+      "loss": 1.2963,
+      "step": 24966
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.818726295828066e-06,
+      "loss": 3.3916,
+      "step": 24967
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8186788874841973e-06,
+      "loss": 2.8023,
+      "step": 24968
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8186314791403286e-06,
+      "loss": 3.467,
+      "step": 24969
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8185840707964605e-06,
+      "loss": 2.0107,
+      "step": 24970
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8185366624525918e-06,
+      "loss": 2.3964,
+      "step": 24971
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.818489254108723e-06,
+      "loss": 3.1921,
+      "step": 24972
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8184418457648545e-06,
+      "loss": 1.1623,
+      "step": 24973
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8183944374209862e-06,
+      "loss": 2.6536,
+      "step": 24974
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8183470290771175e-06,
+      "loss": 1.6858,
+      "step": 24975
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.818299620733249e-06,
+      "loss": 3.1992,
+      "step": 24976
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8182522123893807e-06,
+      "loss": 2.5071,
+      "step": 24977
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.818204804045512e-06,
+      "loss": 3.4054,
+      "step": 24978
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8181573957016435e-06,
+      "loss": 3.2092,
+      "step": 24979
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8181099873577752e-06,
+      "loss": 3.6592,
+      "step": 24980
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8180625790139065e-06,
+      "loss": 2.2303,
+      "step": 24981
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8180151706700378e-06,
+      "loss": 2.0239,
+      "step": 24982
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.8179677623261695e-06,
+      "loss": 2.667,
+      "step": 24983
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.817920353982301e-06,
+      "loss": 2.7798,
+      "step": 24984
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8178729456384322e-06,
+      "loss": 2.954,
+      "step": 24985
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.817825537294564e-06,
+      "loss": 2.8414,
+      "step": 24986
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8177781289506954e-06,
+      "loss": 3.218,
+      "step": 24987
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8177307206068267e-06,
+      "loss": 2.2965,
+      "step": 24988
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8176833122629584e-06,
+      "loss": 3.2187,
+      "step": 24989
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.81763590391909e-06,
+      "loss": 1.8525,
+      "step": 24990
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8175884955752212e-06,
+      "loss": 2.0721,
+      "step": 24991
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.817541087231353e-06,
+      "loss": 1.6989,
+      "step": 24992
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8174936788874844e-06,
+      "loss": 2.4717,
+      "step": 24993
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8174462705436157e-06,
+      "loss": 3.0222,
+      "step": 24994
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.817398862199747e-06,
+      "loss": 2.2433,
+      "step": 24995
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8173514538558787e-06,
+      "loss": 2.356,
+      "step": 24996
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8173040455120101e-06,
+      "loss": 1.4692,
+      "step": 24997
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8172566371681414e-06,
+      "loss": 2.8251,
+      "step": 24998
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8172092288242731e-06,
+      "loss": 2.7627,
+      "step": 24999
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8171618204804046e-06,
+      "loss": 2.7348,
+      "step": 25000
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.817114412136536e-06,
+      "loss": 2.0381,
+      "step": 25001
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8170670037926676e-06,
+      "loss": 1.7256,
+      "step": 25002
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.817019595448799e-06,
+      "loss": 2.8079,
+      "step": 25003
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8169721871049304e-06,
+      "loss": 3.5659,
+      "step": 25004
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.816924778761062e-06,
+      "loss": 1.6984,
+      "step": 25005
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8168773704171936e-06,
+      "loss": 1.6522,
+      "step": 25006
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8168299620733249e-06,
+      "loss": 2.5666,
+      "step": 25007
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8167825537294566e-06,
+      "loss": 2.0466,
+      "step": 25008
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8167351453855878e-06,
+      "loss": 0.6582,
+      "step": 25009
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8166877370417193e-06,
+      "loss": 1.9517,
+      "step": 25010
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.816640328697851e-06,
+      "loss": 3.0776,
+      "step": 25011
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8165929203539823e-06,
+      "loss": 1.6924,
+      "step": 25012
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8165455120101138e-06,
+      "loss": 2.1547,
+      "step": 25013
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8164981036662455e-06,
+      "loss": 1.5897,
+      "step": 25014
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8164506953223768e-06,
+      "loss": 3.182,
+      "step": 25015
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8164032869785083e-06,
+      "loss": 2.8582,
+      "step": 25016
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8163558786346396e-06,
+      "loss": 2.5932,
+      "step": 25017
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8163084702907713e-06,
+      "loss": 2.2574,
+      "step": 25018
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8162610619469028e-06,
+      "loss": 2.9719,
+      "step": 25019
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.816213653603034e-06,
+      "loss": 3.893,
+      "step": 25020
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8161662452591657e-06,
+      "loss": 3.7484,
+      "step": 25021
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.816118836915297e-06,
+      "loss": 2.4004,
+      "step": 25022
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8160714285714285e-06,
+      "loss": 2.8268,
+      "step": 25023
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8160240202275602e-06,
+      "loss": 2.3235,
+      "step": 25024
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8159766118836915e-06,
+      "loss": 2.273,
+      "step": 25025
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.815929203539823e-06,
+      "loss": 1.3434,
+      "step": 25026
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8158817951959547e-06,
+      "loss": 2.1526,
+      "step": 25027
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.815834386852086e-06,
+      "loss": 2.6536,
+      "step": 25028
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8157869785082175e-06,
+      "loss": 3.3287,
+      "step": 25029
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8157395701643492e-06,
+      "loss": 2.6639,
+      "step": 25030
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8156921618204805e-06,
+      "loss": 2.7755,
+      "step": 25031
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.815644753476612e-06,
+      "loss": 4.2045,
+      "step": 25032
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8155973451327437e-06,
+      "loss": 1.8213,
+      "step": 25033
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.815549936788875e-06,
+      "loss": 2.2147,
+      "step": 25034
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8155025284450062e-06,
+      "loss": 1.1986,
+      "step": 25035
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.815455120101138e-06,
+      "loss": 1.8414,
+      "step": 25036
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8154077117572694e-06,
+      "loss": 2.5512,
+      "step": 25037
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8153603034134007e-06,
+      "loss": 2.8072,
+      "step": 25038
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8153128950695322e-06,
+      "loss": 1.4437,
+      "step": 25039
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8152654867256639e-06,
+      "loss": 1.0013,
+      "step": 25040
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8152180783817952e-06,
+      "loss": 2.3947,
+      "step": 25041
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8151706700379267e-06,
+      "loss": 2.5663,
+      "step": 25042
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8151232616940584e-06,
+      "loss": 3.0049,
+      "step": 25043
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8150758533501896e-06,
+      "loss": 1.8794,
+      "step": 25044
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8150284450063211e-06,
+      "loss": 3.0569,
+      "step": 25045
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8149810366624528e-06,
+      "loss": 2.3162,
+      "step": 25046
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.8149336283185841e-06,
+      "loss": 3.0903,
+      "step": 25047
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8148862199747154e-06,
+      "loss": 1.9059,
+      "step": 25048
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.814838811630847e-06,
+      "loss": 2.0808,
+      "step": 25049
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8147914032869786e-06,
+      "loss": 2.5114,
+      "step": 25050
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8147439949431099e-06,
+      "loss": 3.5057,
+      "step": 25051
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8146965865992416e-06,
+      "loss": 3.6223,
+      "step": 25052
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.814649178255373e-06,
+      "loss": 2.4378,
+      "step": 25053
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8146017699115044e-06,
+      "loss": 1.9279,
+      "step": 25054
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.814554361567636e-06,
+      "loss": 1.822,
+      "step": 25055
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8145069532237675e-06,
+      "loss": 3.2355,
+      "step": 25056
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8144595448798988e-06,
+      "loss": 1.1528,
+      "step": 25057
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8144121365360305e-06,
+      "loss": 2.5667,
+      "step": 25058
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.814364728192162e-06,
+      "loss": 3.2721,
+      "step": 25059
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8143173198482933e-06,
+      "loss": 3.7878,
+      "step": 25060
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8142699115044246e-06,
+      "loss": 2.5851,
+      "step": 25061
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8142225031605563e-06,
+      "loss": 2.9789,
+      "step": 25062
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8141750948166878e-06,
+      "loss": 1.9053,
+      "step": 25063
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.814127686472819e-06,
+      "loss": 2.9015,
+      "step": 25064
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8140802781289508e-06,
+      "loss": 2.1511,
+      "step": 25065
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8140328697850823e-06,
+      "loss": 3.3045,
+      "step": 25066
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8139854614412135e-06,
+      "loss": 3.0673,
+      "step": 25067
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8139380530973452e-06,
+      "loss": 1.8991,
+      "step": 25068
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8138906447534767e-06,
+      "loss": 2.3938,
+      "step": 25069
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.813843236409608e-06,
+      "loss": 1.58,
+      "step": 25070
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8137958280657397e-06,
+      "loss": 1.2507,
+      "step": 25071
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8137484197218712e-06,
+      "loss": 2.4191,
+      "step": 25072
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8137010113780025e-06,
+      "loss": 2.849,
+      "step": 25073
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8136536030341342e-06,
+      "loss": 3.1905,
+      "step": 25074
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8136061946902655e-06,
+      "loss": 3.2152,
+      "step": 25075
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.813558786346397e-06,
+      "loss": 2.3484,
+      "step": 25076
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8135113780025287e-06,
+      "loss": 2.0492,
+      "step": 25077
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.81346396965866e-06,
+      "loss": 2.0446,
+      "step": 25078
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8134165613147914e-06,
+      "loss": 3.979,
+      "step": 25079
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8133691529709231e-06,
+      "loss": 1.549,
+      "step": 25080
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8133217446270544e-06,
+      "loss": 2.9229,
+      "step": 25081
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.813274336283186e-06,
+      "loss": 3.471,
+      "step": 25082
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8132269279393172e-06,
+      "loss": 2.4303,
+      "step": 25083
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.813179519595449e-06,
+      "loss": 1.8771,
+      "step": 25084
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8131321112515804e-06,
+      "loss": 2.765,
+      "step": 25085
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8130847029077117e-06,
+      "loss": 2.606,
+      "step": 25086
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8130372945638434e-06,
+      "loss": 2.1528,
+      "step": 25087
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8129898862199747e-06,
+      "loss": 1.2824,
+      "step": 25088
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8129424778761062e-06,
+      "loss": 2.0465,
+      "step": 25089
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8128950695322379e-06,
+      "loss": 3.1401,
+      "step": 25090
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8128476611883691e-06,
+      "loss": 1.1311,
+      "step": 25091
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8128002528445006e-06,
+      "loss": 1.9838,
+      "step": 25092
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8127528445006323e-06,
+      "loss": 2.9057,
+      "step": 25093
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8127054361567636e-06,
+      "loss": 1.5521,
+      "step": 25094
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8126580278128951e-06,
+      "loss": 3.7089,
+      "step": 25095
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8126106194690268e-06,
+      "loss": 3.1478,
+      "step": 25096
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.812563211125158e-06,
+      "loss": 1.5289,
+      "step": 25097
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8125158027812894e-06,
+      "loss": 2.612,
+      "step": 25098
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8124683944374213e-06,
+      "loss": 3.098,
+      "step": 25099
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8124209860935526e-06,
+      "loss": 3.1402,
+      "step": 25100
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8123735777496838e-06,
+      "loss": 1.7638,
+      "step": 25101
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8123261694058153e-06,
+      "loss": 2.7114,
+      "step": 25102
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.812278761061947e-06,
+      "loss": 3.1993,
+      "step": 25103
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8122313527180783e-06,
+      "loss": 1.4974,
+      "step": 25104
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8121839443742098e-06,
+      "loss": 2.3061,
+      "step": 25105
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8121365360303415e-06,
+      "loss": 1.5683,
+      "step": 25106
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8120891276864728e-06,
+      "loss": 1.6074,
+      "step": 25107
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8120417193426043e-06,
+      "loss": 1.8787,
+      "step": 25108
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.811994310998736e-06,
+      "loss": 1.9377,
+      "step": 25109
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.8119469026548673e-06,
+      "loss": 3.3449,
+      "step": 25110
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8118994943109986e-06,
+      "loss": 1.6958,
+      "step": 25111
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8118520859671303e-06,
+      "loss": 2.3295,
+      "step": 25112
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8118046776232618e-06,
+      "loss": 2.887,
+      "step": 25113
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.811757269279393e-06,
+      "loss": 2.7958,
+      "step": 25114
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8117098609355247e-06,
+      "loss": 4.1185,
+      "step": 25115
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8116624525916562e-06,
+      "loss": 2.9859,
+      "step": 25116
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8116150442477875e-06,
+      "loss": 4.3775,
+      "step": 25117
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8115676359039192e-06,
+      "loss": 2.4761,
+      "step": 25118
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8115202275600507e-06,
+      "loss": 3.3949,
+      "step": 25119
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.811472819216182e-06,
+      "loss": 2.1398,
+      "step": 25120
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8114254108723137e-06,
+      "loss": 2.4551,
+      "step": 25121
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8113780025284452e-06,
+      "loss": 2.4823,
+      "step": 25122
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8113305941845765e-06,
+      "loss": 3.6738,
+      "step": 25123
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8112831858407077e-06,
+      "loss": 2.9146,
+      "step": 25124
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8112357774968394e-06,
+      "loss": 2.8622,
+      "step": 25125
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.811188369152971e-06,
+      "loss": 2.5946,
+      "step": 25126
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8111409608091022e-06,
+      "loss": 3.4721,
+      "step": 25127
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.811093552465234e-06,
+      "loss": 1.3866,
+      "step": 25128
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8110461441213654e-06,
+      "loss": 2.896,
+      "step": 25129
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8109987357774967e-06,
+      "loss": 1.7447,
+      "step": 25130
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8109513274336284e-06,
+      "loss": 2.8343,
+      "step": 25131
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.81090391908976e-06,
+      "loss": 3.8687,
+      "step": 25132
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8108565107458912e-06,
+      "loss": 3.4798,
+      "step": 25133
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8108091024020229e-06,
+      "loss": 2.2145,
+      "step": 25134
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8107616940581544e-06,
+      "loss": 2.5553,
+      "step": 25135
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8107142857142857e-06,
+      "loss": 3.0976,
+      "step": 25136
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8106668773704174e-06,
+      "loss": 1.7393,
+      "step": 25137
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8106194690265486e-06,
+      "loss": 2.4116,
+      "step": 25138
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8105720606826801e-06,
+      "loss": 2.4377,
+      "step": 25139
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8105246523388118e-06,
+      "loss": 1.3406,
+      "step": 25140
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8104772439949431e-06,
+      "loss": 3.1374,
+      "step": 25141
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8104298356510746e-06,
+      "loss": 3.1437,
+      "step": 25142
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8103824273072063e-06,
+      "loss": 3.2725,
+      "step": 25143
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8103350189633376e-06,
+      "loss": 3.0798,
+      "step": 25144
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.810287610619469e-06,
+      "loss": 3.8193,
+      "step": 25145
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8102402022756004e-06,
+      "loss": 1.728,
+      "step": 25146
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.810192793931732e-06,
+      "loss": 3.0516,
+      "step": 25147
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8101453855878636e-06,
+      "loss": 2.3818,
+      "step": 25148
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8100979772439948e-06,
+      "loss": 2.5918,
+      "step": 25149
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8100505689001265e-06,
+      "loss": 3.5609,
+      "step": 25150
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8100031605562578e-06,
+      "loss": 1.6348,
+      "step": 25151
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8099557522123893e-06,
+      "loss": 1.4085,
+      "step": 25152
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.809908343868521e-06,
+      "loss": 2.5255,
+      "step": 25153
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8098609355246523e-06,
+      "loss": 3.0122,
+      "step": 25154
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8098135271807838e-06,
+      "loss": 1.6624,
+      "step": 25155
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8097661188369155e-06,
+      "loss": 1.7685,
+      "step": 25156
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8097187104930468e-06,
+      "loss": 3.1932,
+      "step": 25157
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8096713021491783e-06,
+      "loss": 3.2675,
+      "step": 25158
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.80962389380531e-06,
+      "loss": 2.6798,
+      "step": 25159
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8095764854614413e-06,
+      "loss": 2.2749,
+      "step": 25160
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8095290771175727e-06,
+      "loss": 2.5666,
+      "step": 25161
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8094816687737044e-06,
+      "loss": 2.5023,
+      "step": 25162
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8094342604298357e-06,
+      "loss": 3.4603,
+      "step": 25163
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.809386852085967e-06,
+      "loss": 2.0492,
+      "step": 25164
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8093394437420987e-06,
+      "loss": 3.7839,
+      "step": 25165
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8092920353982302e-06,
+      "loss": 1.2449,
+      "step": 25166
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8092446270543615e-06,
+      "loss": 2.6882,
+      "step": 25167
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.809197218710493e-06,
+      "loss": 2.7893,
+      "step": 25168
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8091498103666247e-06,
+      "loss": 2.8557,
+      "step": 25169
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.809102402022756e-06,
+      "loss": 2.182,
+      "step": 25170
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8090549936788875e-06,
+      "loss": 1.8022,
+      "step": 25171
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8090075853350192e-06,
+      "loss": 2.8515,
+      "step": 25172
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.8089601769911504e-06,
+      "loss": 2.9028,
+      "step": 25173
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.808912768647282e-06,
+      "loss": 1.7619,
+      "step": 25174
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8088653603034136e-06,
+      "loss": 2.4545,
+      "step": 25175
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.808817951959545e-06,
+      "loss": 2.3134,
+      "step": 25176
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8087705436156762e-06,
+      "loss": 2.9574,
+      "step": 25177
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.808723135271808e-06,
+      "loss": 2.9172,
+      "step": 25178
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8086757269279394e-06,
+      "loss": 2.7074,
+      "step": 25179
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8086283185840707e-06,
+      "loss": 2.7959,
+      "step": 25180
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8085809102402024e-06,
+      "loss": 2.9039,
+      "step": 25181
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8085335018963339e-06,
+      "loss": 2.8799,
+      "step": 25182
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8084860935524651e-06,
+      "loss": 1.7204,
+      "step": 25183
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8084386852085969e-06,
+      "loss": 1.9235,
+      "step": 25184
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8083912768647283e-06,
+      "loss": 2.9735,
+      "step": 25185
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8083438685208596e-06,
+      "loss": 3.2149,
+      "step": 25186
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8082964601769913e-06,
+      "loss": 1.732,
+      "step": 25187
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8082490518331228e-06,
+      "loss": 3.9594,
+      "step": 25188
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.808201643489254e-06,
+      "loss": 3.3316,
+      "step": 25189
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8081542351453854e-06,
+      "loss": 2.3688,
+      "step": 25190
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.808106826801517e-06,
+      "loss": 2.1676,
+      "step": 25191
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8080594184576486e-06,
+      "loss": 2.511,
+      "step": 25192
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8080120101137799e-06,
+      "loss": 2.999,
+      "step": 25193
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8079646017699116e-06,
+      "loss": 1.6829,
+      "step": 25194
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.807917193426043e-06,
+      "loss": 1.7097,
+      "step": 25195
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8078697850821743e-06,
+      "loss": 2.3025,
+      "step": 25196
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.807822376738306e-06,
+      "loss": 4.0062,
+      "step": 25197
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8077749683944375e-06,
+      "loss": 2.4156,
+      "step": 25198
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8077275600505688e-06,
+      "loss": 1.3159,
+      "step": 25199
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8076801517067005e-06,
+      "loss": 2.5433,
+      "step": 25200
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.807632743362832e-06,
+      "loss": 3.3637,
+      "step": 25201
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8075853350189633e-06,
+      "loss": 2.4709,
+      "step": 25202
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.807537926675095e-06,
+      "loss": 2.6159,
+      "step": 25203
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8074905183312263e-06,
+      "loss": 3.0462,
+      "step": 25204
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8074431099873578e-06,
+      "loss": 2.6439,
+      "step": 25205
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8073957016434895e-06,
+      "loss": 2.32,
+      "step": 25206
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8073482932996207e-06,
+      "loss": 2.6665,
+      "step": 25207
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8073008849557522e-06,
+      "loss": 1.4944,
+      "step": 25208
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.807253476611884e-06,
+      "loss": 2.0103,
+      "step": 25209
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8072060682680152e-06,
+      "loss": 1.7523,
+      "step": 25210
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8071586599241467e-06,
+      "loss": 3.3954,
+      "step": 25211
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.807111251580278e-06,
+      "loss": 2.6812,
+      "step": 25212
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8070638432364097e-06,
+      "loss": 3.8889,
+      "step": 25213
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8070164348925412e-06,
+      "loss": 1.975,
+      "step": 25214
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8069690265486725e-06,
+      "loss": 2.0015,
+      "step": 25215
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8069216182048042e-06,
+      "loss": 3.1928,
+      "step": 25216
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8068742098609355e-06,
+      "loss": 4.3441,
+      "step": 25217
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.806826801517067e-06,
+      "loss": 2.226,
+      "step": 25218
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8067793931731987e-06,
+      "loss": 2.3437,
+      "step": 25219
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.80673198482933e-06,
+      "loss": 2.4859,
+      "step": 25220
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8066845764854614e-06,
+      "loss": 1.6763,
+      "step": 25221
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8066371681415931e-06,
+      "loss": 1.9109,
+      "step": 25222
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8065897597977244e-06,
+      "loss": 2.5411,
+      "step": 25223
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.806542351453856e-06,
+      "loss": 1.55,
+      "step": 25224
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8064949431099876e-06,
+      "loss": 2.5053,
+      "step": 25225
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8064475347661189e-06,
+      "loss": 2.3497,
+      "step": 25226
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8064001264222504e-06,
+      "loss": 2.5701,
+      "step": 25227
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.806352718078382e-06,
+      "loss": 1.8156,
+      "step": 25228
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8063053097345134e-06,
+      "loss": 2.7398,
+      "step": 25229
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8062579013906446e-06,
+      "loss": 3.3005,
+      "step": 25230
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8062104930467763e-06,
+      "loss": 2.6394,
+      "step": 25231
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8061630847029078e-06,
+      "loss": 2.9986,
+      "step": 25232
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8061156763590391e-06,
+      "loss": 1.632,
+      "step": 25233
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8060682680151706e-06,
+      "loss": 2.364,
+      "step": 25234
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8060208596713023e-06,
+      "loss": 4.3508,
+      "step": 25235
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.8059734513274336e-06,
+      "loss": 2.5155,
+      "step": 25236
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.805926042983565e-06,
+      "loss": 3.8819,
+      "step": 25237
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8058786346396968e-06,
+      "loss": 3.3442,
+      "step": 25238
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.805831226295828e-06,
+      "loss": 3.3261,
+      "step": 25239
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8057838179519594e-06,
+      "loss": 1.8203,
+      "step": 25240
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8057364096080913e-06,
+      "loss": 2.8911,
+      "step": 25241
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8056890012642226e-06,
+      "loss": 1.8646,
+      "step": 25242
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8056415929203538e-06,
+      "loss": 2.9175,
+      "step": 25243
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8055941845764855e-06,
+      "loss": 2.1966,
+      "step": 25244
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.805546776232617e-06,
+      "loss": 3.1214,
+      "step": 25245
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8054993678887483e-06,
+      "loss": 2.9234,
+      "step": 25246
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.80545195954488e-06,
+      "loss": 2.0673,
+      "step": 25247
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8054045512010115e-06,
+      "loss": 1.662,
+      "step": 25248
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8053571428571428e-06,
+      "loss": 2.7614,
+      "step": 25249
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8053097345132745e-06,
+      "loss": 2.1871,
+      "step": 25250
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.805262326169406e-06,
+      "loss": 3.6451,
+      "step": 25251
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8052149178255373e-06,
+      "loss": 3.2957,
+      "step": 25252
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.805167509481669e-06,
+      "loss": 2.7497,
+      "step": 25253
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8051201011378002e-06,
+      "loss": 2.1765,
+      "step": 25254
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8050726927939317e-06,
+      "loss": 3.0833,
+      "step": 25255
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.805025284450063e-06,
+      "loss": 1.8594,
+      "step": 25256
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8049778761061947e-06,
+      "loss": 1.106,
+      "step": 25257
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8049304677623262e-06,
+      "loss": 3.38,
+      "step": 25258
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8048830594184575e-06,
+      "loss": 2.6445,
+      "step": 25259
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8048356510745892e-06,
+      "loss": 2.2232,
+      "step": 25260
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8047882427307207e-06,
+      "loss": 2.7929,
+      "step": 25261
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.804740834386852e-06,
+      "loss": 2.5983,
+      "step": 25262
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8046934260429837e-06,
+      "loss": 3.0748,
+      "step": 25263
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8046460176991152e-06,
+      "loss": 1.946,
+      "step": 25264
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8045986093552464e-06,
+      "loss": 1.5463,
+      "step": 25265
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8045512010113782e-06,
+      "loss": 3.7594,
+      "step": 25266
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8045037926675094e-06,
+      "loss": 1.5539,
+      "step": 25267
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.804456384323641e-06,
+      "loss": 1.9256,
+      "step": 25268
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8044089759797726e-06,
+      "loss": 1.5282,
+      "step": 25269
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.804361567635904e-06,
+      "loss": 2.6394,
+      "step": 25270
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8043141592920354e-06,
+      "loss": 2.6462,
+      "step": 25271
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.804266750948167e-06,
+      "loss": 1.6622,
+      "step": 25272
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8042193426042984e-06,
+      "loss": 3.0097,
+      "step": 25273
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8041719342604299e-06,
+      "loss": 2.3126,
+      "step": 25274
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8041245259165616e-06,
+      "loss": 2.7969,
+      "step": 25275
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8040771175726929e-06,
+      "loss": 3.6585,
+      "step": 25276
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8040297092288244e-06,
+      "loss": 2.5934,
+      "step": 25277
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8039823008849556e-06,
+      "loss": 2.579,
+      "step": 25278
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8039348925410873e-06,
+      "loss": 1.8669,
+      "step": 25279
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8038874841972186e-06,
+      "loss": 1.3668,
+      "step": 25280
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8038400758533501e-06,
+      "loss": 1.8465,
+      "step": 25281
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8037926675094818e-06,
+      "loss": 1.7768,
+      "step": 25282
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.803745259165613e-06,
+      "loss": 2.5113,
+      "step": 25283
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8036978508217446e-06,
+      "loss": 2.2653,
+      "step": 25284
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8036504424778763e-06,
+      "loss": 2.741,
+      "step": 25285
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8036030341340076e-06,
+      "loss": 3.7684,
+      "step": 25286
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.803555625790139e-06,
+      "loss": 4.1713,
+      "step": 25287
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8035082174462708e-06,
+      "loss": 2.5954,
+      "step": 25288
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.803460809102402e-06,
+      "loss": 3.3559,
+      "step": 25289
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8034134007585335e-06,
+      "loss": 2.8986,
+      "step": 25290
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8033659924146652e-06,
+      "loss": 2.429,
+      "step": 25291
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8033185840707965e-06,
+      "loss": 3.1004,
+      "step": 25292
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8032711757269278e-06,
+      "loss": 2.1897,
+      "step": 25293
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8032237673830595e-06,
+      "loss": 2.1861,
+      "step": 25294
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.803176359039191e-06,
+      "loss": 2.4076,
+      "step": 25295
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8031289506953223e-06,
+      "loss": 1.5774,
+      "step": 25296
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.803081542351454e-06,
+      "loss": 1.7681,
+      "step": 25297
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8030341340075855e-06,
+      "loss": 2.0877,
+      "step": 25298
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8029867256637168e-06,
+      "loss": 2.7006,
+      "step": 25299
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.8029393173198483e-06,
+      "loss": 2.2775,
+      "step": 25300
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.80289190897598e-06,
+      "loss": 1.6636,
+      "step": 25301
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8028445006321112e-06,
+      "loss": 1.9926,
+      "step": 25302
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8027970922882427e-06,
+      "loss": 2.5259,
+      "step": 25303
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8027496839443744e-06,
+      "loss": 2.511,
+      "step": 25304
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8027022756005057e-06,
+      "loss": 2.906,
+      "step": 25305
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.802654867256637e-06,
+      "loss": 2.4569,
+      "step": 25306
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8026074589127687e-06,
+      "loss": 3.3984,
+      "step": 25307
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8025600505689002e-06,
+      "loss": 3.2627,
+      "step": 25308
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8025126422250315e-06,
+      "loss": 1.8049,
+      "step": 25309
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8024652338811632e-06,
+      "loss": 4.049,
+      "step": 25310
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8024178255372947e-06,
+      "loss": 1.5778,
+      "step": 25311
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.802370417193426e-06,
+      "loss": 3.646,
+      "step": 25312
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8023230088495576e-06,
+      "loss": 1.9223,
+      "step": 25313
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8022756005056891e-06,
+      "loss": 3.6551,
+      "step": 25314
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8022281921618204e-06,
+      "loss": 2.0739,
+      "step": 25315
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8021807838179521e-06,
+      "loss": 2.5462,
+      "step": 25316
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8021333754740836e-06,
+      "loss": 1.1023,
+      "step": 25317
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.802085967130215e-06,
+      "loss": 3.6696,
+      "step": 25318
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8020385587863466e-06,
+      "loss": 1.924,
+      "step": 25319
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8019911504424779e-06,
+      "loss": 3.0534,
+      "step": 25320
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8019437420986094e-06,
+      "loss": 3.2444,
+      "step": 25321
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8018963337547407e-06,
+      "loss": 2.6884,
+      "step": 25322
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8018489254108724e-06,
+      "loss": 2.927,
+      "step": 25323
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8018015170670039e-06,
+      "loss": 3.2453,
+      "step": 25324
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8017541087231351e-06,
+      "loss": 2.6304,
+      "step": 25325
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8017067003792668e-06,
+      "loss": 3.33,
+      "step": 25326
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8016592920353983e-06,
+      "loss": 1.8168,
+      "step": 25327
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8016118836915296e-06,
+      "loss": 1.8751,
+      "step": 25328
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8015644753476613e-06,
+      "loss": 1.9134,
+      "step": 25329
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8015170670037928e-06,
+      "loss": 2.4279,
+      "step": 25330
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.801469658659924e-06,
+      "loss": 2.5546,
+      "step": 25331
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8014222503160558e-06,
+      "loss": 2.1365,
+      "step": 25332
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.4433392539964476,
+      "eval_loss": 2.4061262607574463,
+      "eval_runtime": 154.3447,
+      "eval_samples_per_second": 18.238,
+      "eval_steps_per_second": 4.561,
+      "step": 25332
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.801374841972187e-06,
+      "loss": 3.7554,
+      "step": 25333
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8013274336283186e-06,
+      "loss": 2.0185,
+      "step": 25334
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8012800252844503e-06,
+      "loss": 3.1219,
+      "step": 25335
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8012326169405815e-06,
+      "loss": 2.191,
+      "step": 25336
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.801185208596713e-06,
+      "loss": 2.4091,
+      "step": 25337
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8011378002528447e-06,
+      "loss": 1.4867,
+      "step": 25338
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.801090391908976e-06,
+      "loss": 2.0413,
+      "step": 25339
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8010429835651075e-06,
+      "loss": 2.36,
+      "step": 25340
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8009955752212392e-06,
+      "loss": 3.366,
+      "step": 25341
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8009481668773705e-06,
+      "loss": 3.3216,
+      "step": 25342
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.800900758533502e-06,
+      "loss": 3.7895,
+      "step": 25343
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8008533501896333e-06,
+      "loss": 1.8529,
+      "step": 25344
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.800805941845765e-06,
+      "loss": 1.4338,
+      "step": 25345
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8007585335018963e-06,
+      "loss": 1.0611,
+      "step": 25346
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8007111251580277e-06,
+      "loss": 1.7049,
+      "step": 25347
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8006637168141595e-06,
+      "loss": 1.7286,
+      "step": 25348
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8006163084702907e-06,
+      "loss": 2.6608,
+      "step": 25349
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8005689001264222e-06,
+      "loss": 2.0056,
+      "step": 25350
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.800521491782554e-06,
+      "loss": 2.6068,
+      "step": 25351
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8004740834386852e-06,
+      "loss": 2.7434,
+      "step": 25352
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8004266750948167e-06,
+      "loss": 3.88,
+      "step": 25353
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8003792667509484e-06,
+      "loss": 2.7657,
+      "step": 25354
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8003318584070797e-06,
+      "loss": 2.4538,
+      "step": 25355
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8002844500632112e-06,
+      "loss": 3.5392,
+      "step": 25356
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8002370417193429e-06,
+      "loss": 1.9687,
+      "step": 25357
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8001896333754742e-06,
+      "loss": 2.5325,
+      "step": 25358
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8001422250316054e-06,
+      "loss": 2.0141,
+      "step": 25359
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8000948166877371e-06,
+      "loss": 2.3055,
+      "step": 25360
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8000474083438686e-06,
+      "loss": 2.5404,
+      "step": 25361
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.8e-06,
+      "loss": 3.0104,
+      "step": 25362
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.7999525916561316e-06,
+      "loss": 2.7227,
+      "step": 25363
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7999051833122631e-06,
+      "loss": 2.6241,
+      "step": 25364
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7998577749683944e-06,
+      "loss": 3.1295,
+      "step": 25365
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7998103666245259e-06,
+      "loss": 2.3419,
+      "step": 25366
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7997629582806576e-06,
+      "loss": 3.6717,
+      "step": 25367
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7997155499367889e-06,
+      "loss": 2.7033,
+      "step": 25368
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7996681415929202e-06,
+      "loss": 2.3197,
+      "step": 25369
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.799620733249052e-06,
+      "loss": 2.7742,
+      "step": 25370
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7995733249051833e-06,
+      "loss": 1.7184,
+      "step": 25371
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7995259165613146e-06,
+      "loss": 2.6387,
+      "step": 25372
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7994785082174463e-06,
+      "loss": 3.1394,
+      "step": 25373
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7994310998735778e-06,
+      "loss": 3.6469,
+      "step": 25374
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.799383691529709e-06,
+      "loss": 1.7499,
+      "step": 25375
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7993362831858408e-06,
+      "loss": 3.7326,
+      "step": 25376
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7992888748419723e-06,
+      "loss": 1.9243,
+      "step": 25377
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7992414664981036e-06,
+      "loss": 3.7158,
+      "step": 25378
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7991940581542353e-06,
+      "loss": 1.6853,
+      "step": 25379
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7991466498103668e-06,
+      "loss": 2.5353,
+      "step": 25380
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.799099241466498e-06,
+      "loss": 1.3168,
+      "step": 25381
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7990518331226298e-06,
+      "loss": 2.6993,
+      "step": 25382
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.799004424778761e-06,
+      "loss": 3.5378,
+      "step": 25383
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7989570164348925e-06,
+      "loss": 1.6395,
+      "step": 25384
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7989096080910242e-06,
+      "loss": 3.1979,
+      "step": 25385
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7988621997471555e-06,
+      "loss": 3.1709,
+      "step": 25386
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.798814791403287e-06,
+      "loss": 2.6393,
+      "step": 25387
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7987673830594183e-06,
+      "loss": 3.7963,
+      "step": 25388
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.79871997471555e-06,
+      "loss": 3.0508,
+      "step": 25389
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7986725663716815e-06,
+      "loss": 3.1558,
+      "step": 25390
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7986251580278128e-06,
+      "loss": 3.7443,
+      "step": 25391
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7985777496839445e-06,
+      "loss": 3.0562,
+      "step": 25392
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.798530341340076e-06,
+      "loss": 2.4587,
+      "step": 25393
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7984829329962072e-06,
+      "loss": 2.6104,
+      "step": 25394
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.798435524652339e-06,
+      "loss": 2.4738,
+      "step": 25395
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7983881163084702e-06,
+      "loss": 3.0882,
+      "step": 25396
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7983407079646017e-06,
+      "loss": 3.9593,
+      "step": 25397
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7982932996207334e-06,
+      "loss": 3.1532,
+      "step": 25398
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7982458912768647e-06,
+      "loss": 3.8706,
+      "step": 25399
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7981984829329962e-06,
+      "loss": 2.0204,
+      "step": 25400
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.798151074589128e-06,
+      "loss": 2.6869,
+      "step": 25401
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7981036662452592e-06,
+      "loss": 3.3249,
+      "step": 25402
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7980562579013907e-06,
+      "loss": 1.725,
+      "step": 25403
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7980088495575224e-06,
+      "loss": 2.3409,
+      "step": 25404
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7979614412136537e-06,
+      "loss": 2.2675,
+      "step": 25405
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7979140328697852e-06,
+      "loss": 2.3777,
+      "step": 25406
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7978666245259169e-06,
+      "loss": 3.596,
+      "step": 25407
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7978192161820481e-06,
+      "loss": 2.8786,
+      "step": 25408
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7977718078381794e-06,
+      "loss": 2.9272,
+      "step": 25409
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.797724399494311e-06,
+      "loss": 3.7335,
+      "step": 25410
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7976769911504426e-06,
+      "loss": 1.9785,
+      "step": 25411
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7976295828065739e-06,
+      "loss": 1.9911,
+      "step": 25412
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7975821744627054e-06,
+      "loss": 2.8423,
+      "step": 25413
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.797534766118837e-06,
+      "loss": 3.6381,
+      "step": 25414
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7974873577749684e-06,
+      "loss": 2.8677,
+      "step": 25415
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7974399494310999e-06,
+      "loss": 3.9937,
+      "step": 25416
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7973925410872316e-06,
+      "loss": 3.4266,
+      "step": 25417
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7973451327433628e-06,
+      "loss": 4.2645,
+      "step": 25418
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7972977243994943e-06,
+      "loss": 2.3387,
+      "step": 25419
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.797250316055626e-06,
+      "loss": 4.3327,
+      "step": 25420
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7972029077117573e-06,
+      "loss": 2.4918,
+      "step": 25421
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7971554993678886e-06,
+      "loss": 3.1556,
+      "step": 25422
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7971080910240203e-06,
+      "loss": 2.5241,
+      "step": 25423
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7970606826801518e-06,
+      "loss": 2.2084,
+      "step": 25424
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.797013274336283e-06,
+      "loss": 1.6801,
+      "step": 25425
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.7969658659924148e-06,
+      "loss": 2.7815,
+      "step": 25426
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7969184576485463e-06,
+      "loss": 2.0015,
+      "step": 25427
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7968710493046776e-06,
+      "loss": 2.3821,
+      "step": 25428
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7968236409608093e-06,
+      "loss": 2.8899,
+      "step": 25429
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7967762326169408e-06,
+      "loss": 1.7577,
+      "step": 25430
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.796728824273072e-06,
+      "loss": 4.5806,
+      "step": 25431
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7966814159292035e-06,
+      "loss": 2.3093,
+      "step": 25432
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7966340075853352e-06,
+      "loss": 2.3191,
+      "step": 25433
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7965865992414665e-06,
+      "loss": 3.4522,
+      "step": 25434
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7965391908975978e-06,
+      "loss": 3.4132,
+      "step": 25435
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7964917825537295e-06,
+      "loss": 2.5052,
+      "step": 25436
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.796444374209861e-06,
+      "loss": 3.3063,
+      "step": 25437
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7963969658659923e-06,
+      "loss": 2.9392,
+      "step": 25438
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.796349557522124e-06,
+      "loss": 2.5495,
+      "step": 25439
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7963021491782555e-06,
+      "loss": 3.4317,
+      "step": 25440
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7962547408343867e-06,
+      "loss": 1.4375,
+      "step": 25441
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7962073324905184e-06,
+      "loss": 3.858,
+      "step": 25442
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.79615992414665e-06,
+      "loss": 2.3774,
+      "step": 25443
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7961125158027812e-06,
+      "loss": 2.4024,
+      "step": 25444
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.796065107458913e-06,
+      "loss": 1.5986,
+      "step": 25445
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7960176991150444e-06,
+      "loss": 3.2911,
+      "step": 25446
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7959702907711757e-06,
+      "loss": 4.0275,
+      "step": 25447
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7959228824273074e-06,
+      "loss": 2.4971,
+      "step": 25448
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7958754740834387e-06,
+      "loss": 3.2626,
+      "step": 25449
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7958280657395702e-06,
+      "loss": 2.6247,
+      "step": 25450
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7957806573957019e-06,
+      "loss": 1.9449,
+      "step": 25451
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7957332490518332e-06,
+      "loss": 2.2553,
+      "step": 25452
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7956858407079646e-06,
+      "loss": 3.1309,
+      "step": 25453
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.795638432364096e-06,
+      "loss": 3.2316,
+      "step": 25454
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7955910240202276e-06,
+      "loss": 2.5789,
+      "step": 25455
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7955436156763591e-06,
+      "loss": 2.7417,
+      "step": 25456
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7954962073324904e-06,
+      "loss": 2.8303,
+      "step": 25457
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7954487989886221e-06,
+      "loss": 2.9651,
+      "step": 25458
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7954013906447536e-06,
+      "loss": 3.0926,
+      "step": 25459
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7953539823008849e-06,
+      "loss": 1.5922,
+      "step": 25460
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7953065739570166e-06,
+      "loss": 1.5737,
+      "step": 25461
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7952591656131479e-06,
+      "loss": 1.6384,
+      "step": 25462
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7952117572692794e-06,
+      "loss": 1.2655,
+      "step": 25463
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.795164348925411e-06,
+      "loss": 1.4155,
+      "step": 25464
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7951169405815423e-06,
+      "loss": 2.3003,
+      "step": 25465
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7950695322376738e-06,
+      "loss": 2.3661,
+      "step": 25466
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7950221238938055e-06,
+      "loss": 2.6457,
+      "step": 25467
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7949747155499368e-06,
+      "loss": 3.3048,
+      "step": 25468
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7949273072060683e-06,
+      "loss": 2.7289,
+      "step": 25469
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7948798988622e-06,
+      "loss": 3.4584,
+      "step": 25470
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7948324905183313e-06,
+      "loss": 1.8353,
+      "step": 25471
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7947850821744628e-06,
+      "loss": 2.0863,
+      "step": 25472
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7947376738305945e-06,
+      "loss": 1.5677,
+      "step": 25473
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7946902654867258e-06,
+      "loss": 2.3588,
+      "step": 25474
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.794642857142857e-06,
+      "loss": 1.9685,
+      "step": 25475
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7945954487989885e-06,
+      "loss": 3.2941,
+      "step": 25476
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7945480404551202e-06,
+      "loss": 3.5995,
+      "step": 25477
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7945006321112515e-06,
+      "loss": 1.6117,
+      "step": 25478
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.794453223767383e-06,
+      "loss": 2.5651,
+      "step": 25479
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7944058154235147e-06,
+      "loss": 1.3997,
+      "step": 25480
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.794358407079646e-06,
+      "loss": 1.3478,
+      "step": 25481
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7943109987357775e-06,
+      "loss": 1.9765,
+      "step": 25482
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7942635903919092e-06,
+      "loss": 2.1881,
+      "step": 25483
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7942161820480405e-06,
+      "loss": 2.8268,
+      "step": 25484
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.794168773704172e-06,
+      "loss": 2.078,
+      "step": 25485
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7941213653603037e-06,
+      "loss": 1.7814,
+      "step": 25486
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.794073957016435e-06,
+      "loss": 2.0971,
+      "step": 25487
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7940265486725662e-06,
+      "loss": 2.0872,
+      "step": 25488
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.793979140328698e-06,
+      "loss": 3.7058,
+      "step": 25489
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.7939317319848294e-06,
+      "loss": 2.0568,
+      "step": 25490
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7938843236409607e-06,
+      "loss": 1.5697,
+      "step": 25491
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7938369152970924e-06,
+      "loss": 2.0101,
+      "step": 25492
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.793789506953224e-06,
+      "loss": 2.83,
+      "step": 25493
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7937420986093552e-06,
+      "loss": 2.5651,
+      "step": 25494
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.793694690265487e-06,
+      "loss": 3.3954,
+      "step": 25495
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7936472819216184e-06,
+      "loss": 2.7192,
+      "step": 25496
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7935998735777497e-06,
+      "loss": 2.0293,
+      "step": 25497
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7935524652338812e-06,
+      "loss": 2.6164,
+      "step": 25498
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7935050568900129e-06,
+      "loss": 2.7264,
+      "step": 25499
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7934576485461441e-06,
+      "loss": 0.8063,
+      "step": 25500
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7934102402022754e-06,
+      "loss": 3.5154,
+      "step": 25501
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7933628318584071e-06,
+      "loss": 1.9587,
+      "step": 25502
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7933154235145386e-06,
+      "loss": 2.1257,
+      "step": 25503
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.79326801517067e-06,
+      "loss": 1.6457,
+      "step": 25504
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7932206068268016e-06,
+      "loss": 2.1394,
+      "step": 25505
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.793173198482933e-06,
+      "loss": 2.6511,
+      "step": 25506
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7931257901390644e-06,
+      "loss": 3.3395,
+      "step": 25507
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.793078381795196e-06,
+      "loss": 2.7891,
+      "step": 25508
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7930309734513276e-06,
+      "loss": 2.5436,
+      "step": 25509
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7929835651074589e-06,
+      "loss": 2.2071,
+      "step": 25510
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7929361567635906e-06,
+      "loss": 2.45,
+      "step": 25511
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.792888748419722e-06,
+      "loss": 3.3717,
+      "step": 25512
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7928413400758533e-06,
+      "loss": 2.5935,
+      "step": 25513
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.792793931731985e-06,
+      "loss": 2.1869,
+      "step": 25514
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7927465233881163e-06,
+      "loss": 4.7382,
+      "step": 25515
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7926991150442478e-06,
+      "loss": 2.8492,
+      "step": 25516
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7926517067003795e-06,
+      "loss": 1.3315,
+      "step": 25517
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7926042983565108e-06,
+      "loss": 3.7118,
+      "step": 25518
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7925568900126423e-06,
+      "loss": 3.6246,
+      "step": 25519
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7925094816687736e-06,
+      "loss": 1.7408,
+      "step": 25520
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7924620733249053e-06,
+      "loss": 1.4224,
+      "step": 25521
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7924146649810368e-06,
+      "loss": 4.1447,
+      "step": 25522
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.792367256637168e-06,
+      "loss": 3.4383,
+      "step": 25523
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7923198482932997e-06,
+      "loss": 2.4425,
+      "step": 25524
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.792272439949431e-06,
+      "loss": 2.6378,
+      "step": 25525
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7922250316055625e-06,
+      "loss": 2.1703,
+      "step": 25526
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7921776232616942e-06,
+      "loss": 3.6853,
+      "step": 25527
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7921302149178255e-06,
+      "loss": 2.5773,
+      "step": 25528
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.792082806573957e-06,
+      "loss": 3.4291,
+      "step": 25529
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7920353982300887e-06,
+      "loss": 1.3444,
+      "step": 25530
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.79198798988622e-06,
+      "loss": 2.8679,
+      "step": 25531
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7919405815423515e-06,
+      "loss": 2.5086,
+      "step": 25532
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7918931731984832e-06,
+      "loss": 2.7509,
+      "step": 25533
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7918457648546145e-06,
+      "loss": 2.0492,
+      "step": 25534
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.791798356510746e-06,
+      "loss": 1.5656,
+      "step": 25535
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7917509481668777e-06,
+      "loss": 2.6941,
+      "step": 25536
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.791703539823009e-06,
+      "loss": 2.5218,
+      "step": 25537
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7916561314791402e-06,
+      "loss": 3.3639,
+      "step": 25538
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7916087231352717e-06,
+      "loss": 2.0626,
+      "step": 25539
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7915613147914034e-06,
+      "loss": 2.6969,
+      "step": 25540
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7915139064475347e-06,
+      "loss": 2.304,
+      "step": 25541
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7914664981036662e-06,
+      "loss": 3.2002,
+      "step": 25542
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7914190897597979e-06,
+      "loss": 2.1158,
+      "step": 25543
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7913716814159292e-06,
+      "loss": 3.8923,
+      "step": 25544
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7913242730720607e-06,
+      "loss": 2.5088,
+      "step": 25545
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7912768647281924e-06,
+      "loss": 3.1308,
+      "step": 25546
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7912294563843236e-06,
+      "loss": 3.1005,
+      "step": 25547
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7911820480404551e-06,
+      "loss": 2.7007,
+      "step": 25548
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7911346396965868e-06,
+      "loss": 2.5921,
+      "step": 25549
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7910872313527181e-06,
+      "loss": 1.8822,
+      "step": 25550
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7910398230088494e-06,
+      "loss": 1.8937,
+      "step": 25551
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.790992414664981e-06,
+      "loss": 2.5777,
+      "step": 25552
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.7909450063211126e-06,
+      "loss": 2.5533,
+      "step": 25553
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7908975979772439e-06,
+      "loss": 1.161,
+      "step": 25554
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7908501896333756e-06,
+      "loss": 2.8843,
+      "step": 25555
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.790802781289507e-06,
+      "loss": 2.181,
+      "step": 25556
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7907553729456384e-06,
+      "loss": 3.9541,
+      "step": 25557
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.79070796460177e-06,
+      "loss": 1.4989,
+      "step": 25558
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7906605562579015e-06,
+      "loss": 2.721,
+      "step": 25559
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7906131479140328e-06,
+      "loss": 0.8239,
+      "step": 25560
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7905657395701643e-06,
+      "loss": 3.6291,
+      "step": 25561
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.790518331226296e-06,
+      "loss": 2.5181,
+      "step": 25562
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7904709228824273e-06,
+      "loss": 3.4515,
+      "step": 25563
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7904235145385586e-06,
+      "loss": 3.4007,
+      "step": 25564
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7903761061946903e-06,
+      "loss": 3.9851,
+      "step": 25565
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7903286978508218e-06,
+      "loss": 2.1318,
+      "step": 25566
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.790281289506953e-06,
+      "loss": 3.132,
+      "step": 25567
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7902338811630848e-06,
+      "loss": 1.533,
+      "step": 25568
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7901864728192163e-06,
+      "loss": 1.3258,
+      "step": 25569
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7901390644753475e-06,
+      "loss": 3.8374,
+      "step": 25570
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7900916561314792e-06,
+      "loss": 3.5093,
+      "step": 25571
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7900442477876107e-06,
+      "loss": 2.9218,
+      "step": 25572
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.789996839443742e-06,
+      "loss": 2.9748,
+      "step": 25573
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7899494310998737e-06,
+      "loss": 2.2246,
+      "step": 25574
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7899020227560052e-06,
+      "loss": 2.2092,
+      "step": 25575
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7898546144121365e-06,
+      "loss": 2.3302,
+      "step": 25576
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7898072060682682e-06,
+      "loss": 0.8343,
+      "step": 25577
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7897597977243995e-06,
+      "loss": 2.0665,
+      "step": 25578
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.789712389380531e-06,
+      "loss": 2.6026,
+      "step": 25579
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7896649810366627e-06,
+      "loss": 3.4481,
+      "step": 25580
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.789617572692794e-06,
+      "loss": 2.5773,
+      "step": 25581
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7895701643489254e-06,
+      "loss": 2.7083,
+      "step": 25582
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7895227560050567e-06,
+      "loss": 2.5488,
+      "step": 25583
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7894753476611884e-06,
+      "loss": 2.4012,
+      "step": 25584
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.78942793931732e-06,
+      "loss": 2.8371,
+      "step": 25585
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7893805309734512e-06,
+      "loss": 3.1687,
+      "step": 25586
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.789333122629583e-06,
+      "loss": 3.7413,
+      "step": 25587
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7892857142857144e-06,
+      "loss": 2.287,
+      "step": 25588
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7892383059418457e-06,
+      "loss": 2.8101,
+      "step": 25589
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7891908975979774e-06,
+      "loss": 2.8714,
+      "step": 25590
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7891434892541087e-06,
+      "loss": 1.9506,
+      "step": 25591
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7890960809102402e-06,
+      "loss": 2.6421,
+      "step": 25592
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7890486725663719e-06,
+      "loss": 2.1235,
+      "step": 25593
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7890012642225031e-06,
+      "loss": 2.788,
+      "step": 25594
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7889538558786346e-06,
+      "loss": 3.581,
+      "step": 25595
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7889064475347663e-06,
+      "loss": 2.0686,
+      "step": 25596
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7888590391908976e-06,
+      "loss": 2.7949,
+      "step": 25597
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7888116308470291e-06,
+      "loss": 2.1464,
+      "step": 25598
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7887642225031608e-06,
+      "loss": 1.941,
+      "step": 25599
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.788716814159292e-06,
+      "loss": 3.2936,
+      "step": 25600
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7886694058154236e-06,
+      "loss": 1.3152,
+      "step": 25601
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7886219974715553e-06,
+      "loss": 1.7124,
+      "step": 25602
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7885745891276866e-06,
+      "loss": 2.2401,
+      "step": 25603
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7885271807838178e-06,
+      "loss": 1.9889,
+      "step": 25604
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7884797724399493e-06,
+      "loss": 3.0392,
+      "step": 25605
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.788432364096081e-06,
+      "loss": 3.0449,
+      "step": 25606
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7883849557522123e-06,
+      "loss": 2.5522,
+      "step": 25607
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7883375474083438e-06,
+      "loss": 2.2115,
+      "step": 25608
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7882901390644755e-06,
+      "loss": 1.5201,
+      "step": 25609
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7882427307206068e-06,
+      "loss": 2.0677,
+      "step": 25610
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7881953223767383e-06,
+      "loss": 3.1107,
+      "step": 25611
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.78814791403287e-06,
+      "loss": 3.4404,
+      "step": 25612
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7881005056890013e-06,
+      "loss": 3.5395,
+      "step": 25613
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7880530973451328e-06,
+      "loss": 2.8391,
+      "step": 25614
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7880056890012645e-06,
+      "loss": 3.7304,
+      "step": 25615
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.7879582806573958e-06,
+      "loss": 1.9716,
+      "step": 25616
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.787910872313527e-06,
+      "loss": 2.5604,
+      "step": 25617
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7878634639696587e-06,
+      "loss": 2.4956,
+      "step": 25618
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7878160556257902e-06,
+      "loss": 2.9714,
+      "step": 25619
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7877686472819215e-06,
+      "loss": 3.5182,
+      "step": 25620
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7877212389380532e-06,
+      "loss": 1.8976,
+      "step": 25621
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7876738305941847e-06,
+      "loss": 2.4549,
+      "step": 25622
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.787626422250316e-06,
+      "loss": 2.59,
+      "step": 25623
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7875790139064477e-06,
+      "loss": 1.995,
+      "step": 25624
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7875316055625792e-06,
+      "loss": 1.5295,
+      "step": 25625
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7874841972187105e-06,
+      "loss": 2.4497,
+      "step": 25626
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.787436788874842e-06,
+      "loss": 3.1256,
+      "step": 25627
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7873893805309737e-06,
+      "loss": 3.7405,
+      "step": 25628
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.787341972187105e-06,
+      "loss": 1.8183,
+      "step": 25629
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7872945638432362e-06,
+      "loss": 3.2694,
+      "step": 25630
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.787247155499368e-06,
+      "loss": 2.2052,
+      "step": 25631
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7871997471554994e-06,
+      "loss": 1.6222,
+      "step": 25632
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7871523388116307e-06,
+      "loss": 4.0197,
+      "step": 25633
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7871049304677624e-06,
+      "loss": 0.9709,
+      "step": 25634
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.787057522123894e-06,
+      "loss": 3.1327,
+      "step": 25635
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7870101137800252e-06,
+      "loss": 2.5657,
+      "step": 25636
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7869627054361569e-06,
+      "loss": 1.7681,
+      "step": 25637
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7869152970922884e-06,
+      "loss": 3.2714,
+      "step": 25638
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7868678887484197e-06,
+      "loss": 1.6322,
+      "step": 25639
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7868204804045514e-06,
+      "loss": 2.5225,
+      "step": 25640
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7867730720606828e-06,
+      "loss": 3.3521,
+      "step": 25641
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7867256637168141e-06,
+      "loss": 2.0328,
+      "step": 25642
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7866782553729458e-06,
+      "loss": 1.4469,
+      "step": 25643
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7866308470290771e-06,
+      "loss": 1.5993,
+      "step": 25644
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7865834386852086e-06,
+      "loss": 2.6963,
+      "step": 25645
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7865360303413403e-06,
+      "loss": 2.6417,
+      "step": 25646
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7864886219974716e-06,
+      "loss": 3.0959,
+      "step": 25647
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.786441213653603e-06,
+      "loss": 1.9107,
+      "step": 25648
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7863938053097344e-06,
+      "loss": 1.7332,
+      "step": 25649
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.786346396965866e-06,
+      "loss": 1.571,
+      "step": 25650
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7862989886219976e-06,
+      "loss": 3.0937,
+      "step": 25651
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7862515802781288e-06,
+      "loss": 1.302,
+      "step": 25652
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7862041719342605e-06,
+      "loss": 1.8459,
+      "step": 25653
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7861567635903918e-06,
+      "loss": 1.5478,
+      "step": 25654
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7861093552465233e-06,
+      "loss": 2.5991,
+      "step": 25655
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.786061946902655e-06,
+      "loss": 3.0952,
+      "step": 25656
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7860145385587863e-06,
+      "loss": 2.7873,
+      "step": 25657
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7859671302149178e-06,
+      "loss": 3.6055,
+      "step": 25658
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7859197218710495e-06,
+      "loss": 1.4057,
+      "step": 25659
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7858723135271808e-06,
+      "loss": 2.3736,
+      "step": 25660
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7858249051833123e-06,
+      "loss": 1.6789,
+      "step": 25661
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.785777496839444e-06,
+      "loss": 3.6384,
+      "step": 25662
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7857300884955753e-06,
+      "loss": 1.5999,
+      "step": 25663
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7856826801517067e-06,
+      "loss": 2.1445,
+      "step": 25664
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7856352718078384e-06,
+      "loss": 2.3583,
+      "step": 25665
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7855878634639697e-06,
+      "loss": 2.169,
+      "step": 25666
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.785540455120101e-06,
+      "loss": 2.0341,
+      "step": 25667
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7854930467762327e-06,
+      "loss": 3.0898,
+      "step": 25668
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7854456384323642e-06,
+      "loss": 1.4037,
+      "step": 25669
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7853982300884955e-06,
+      "loss": 2.0476,
+      "step": 25670
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.785350821744627e-06,
+      "loss": 2.9211,
+      "step": 25671
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7853034134007587e-06,
+      "loss": 3.0202,
+      "step": 25672
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.78525600505689e-06,
+      "loss": 1.2868,
+      "step": 25673
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7852085967130215e-06,
+      "loss": 2.3081,
+      "step": 25674
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7851611883691532e-06,
+      "loss": 2.1041,
+      "step": 25675
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7851137800252844e-06,
+      "loss": 2.2937,
+      "step": 25676
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.785066371681416e-06,
+      "loss": 2.7317,
+      "step": 25677
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7850189633375476e-06,
+      "loss": 3.2521,
+      "step": 25678
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.784971554993679e-06,
+      "loss": 3.8831,
+      "step": 25679
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.7849241466498102e-06,
+      "loss": 2.409,
+      "step": 25680
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.784876738305942e-06,
+      "loss": 1.8699,
+      "step": 25681
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7848293299620734e-06,
+      "loss": 3.4318,
+      "step": 25682
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7847819216182047e-06,
+      "loss": 2.1836,
+      "step": 25683
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7847345132743364e-06,
+      "loss": 1.6061,
+      "step": 25684
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7846871049304679e-06,
+      "loss": 1.9197,
+      "step": 25685
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7846396965865991e-06,
+      "loss": 4.3269,
+      "step": 25686
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7845922882427309e-06,
+      "loss": 2.9934,
+      "step": 25687
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7845448798988623e-06,
+      "loss": 2.9015,
+      "step": 25688
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7844974715549936e-06,
+      "loss": 1.0254,
+      "step": 25689
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7844500632111253e-06,
+      "loss": 2.4601,
+      "step": 25690
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7844026548672568e-06,
+      "loss": 2.0851,
+      "step": 25691
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.784355246523388e-06,
+      "loss": 2.398,
+      "step": 25692
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7843078381795194e-06,
+      "loss": 2.1116,
+      "step": 25693
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.784260429835651e-06,
+      "loss": 2.5273,
+      "step": 25694
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7842130214917826e-06,
+      "loss": 1.192,
+      "step": 25695
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7841656131479139e-06,
+      "loss": 2.064,
+      "step": 25696
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7841182048040456e-06,
+      "loss": 2.8603,
+      "step": 25697
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.784070796460177e-06,
+      "loss": 2.6187,
+      "step": 25698
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7840233881163083e-06,
+      "loss": 3.3341,
+      "step": 25699
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.78397597977244e-06,
+      "loss": 2.0089,
+      "step": 25700
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7839285714285715e-06,
+      "loss": 2.1077,
+      "step": 25701
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7838811630847028e-06,
+      "loss": 1.4557,
+      "step": 25702
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7838337547408345e-06,
+      "loss": 2.8551,
+      "step": 25703
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.783786346396966e-06,
+      "loss": 1.9205,
+      "step": 25704
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7837389380530973e-06,
+      "loss": 3.8132,
+      "step": 25705
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.783691529709229e-06,
+      "loss": 3.5687,
+      "step": 25706
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7836441213653603e-06,
+      "loss": 2.9378,
+      "step": 25707
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7835967130214918e-06,
+      "loss": 2.5837,
+      "step": 25708
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7835493046776235e-06,
+      "loss": 3.1194,
+      "step": 25709
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7835018963337547e-06,
+      "loss": 3.5758,
+      "step": 25710
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7834544879898862e-06,
+      "loss": 1.8415,
+      "step": 25711
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.783407079646018e-06,
+      "loss": 2.1685,
+      "step": 25712
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7833596713021492e-06,
+      "loss": 4.307,
+      "step": 25713
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7833122629582807e-06,
+      "loss": 2.2959,
+      "step": 25714
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.783264854614412e-06,
+      "loss": 3.3625,
+      "step": 25715
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7832174462705437e-06,
+      "loss": 3.0369,
+      "step": 25716
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7831700379266752e-06,
+      "loss": 1.3872,
+      "step": 25717
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7831226295828065e-06,
+      "loss": 2.7116,
+      "step": 25718
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7830752212389382e-06,
+      "loss": 0.6604,
+      "step": 25719
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7830278128950695e-06,
+      "loss": 0.9959,
+      "step": 25720
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.782980404551201e-06,
+      "loss": 1.5468,
+      "step": 25721
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7829329962073327e-06,
+      "loss": 2.0001,
+      "step": 25722
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.782885587863464e-06,
+      "loss": 3.4833,
+      "step": 25723
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7828381795195954e-06,
+      "loss": 2.7273,
+      "step": 25724
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7827907711757271e-06,
+      "loss": 2.6176,
+      "step": 25725
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7827433628318584e-06,
+      "loss": 2.1649,
+      "step": 25726
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.78269595448799e-06,
+      "loss": 2.89,
+      "step": 25727
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7826485461441216e-06,
+      "loss": 2.0479,
+      "step": 25728
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7826011378002529e-06,
+      "loss": 1.6054,
+      "step": 25729
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7825537294563844e-06,
+      "loss": 1.6626,
+      "step": 25730
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.782506321112516e-06,
+      "loss": 2.6729,
+      "step": 25731
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7824589127686474e-06,
+      "loss": 3.5254,
+      "step": 25732
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7824115044247786e-06,
+      "loss": 2.5208,
+      "step": 25733
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7823640960809103e-06,
+      "loss": 2.7325,
+      "step": 25734
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7823166877370418e-06,
+      "loss": 2.189,
+      "step": 25735
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7822692793931731e-06,
+      "loss": 2.574,
+      "step": 25736
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7822218710493046e-06,
+      "loss": 2.3709,
+      "step": 25737
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7821744627054363e-06,
+      "loss": 1.7548,
+      "step": 25738
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7821270543615676e-06,
+      "loss": 2.7666,
+      "step": 25739
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.782079646017699e-06,
+      "loss": 3.2732,
+      "step": 25740
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7820322376738308e-06,
+      "loss": 1.6257,
+      "step": 25741
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.781984829329962e-06,
+      "loss": 2.8131,
+      "step": 25742
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.7819374209860936e-06,
+      "loss": 2.7056,
+      "step": 25743
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7818900126422253e-06,
+      "loss": 2.8025,
+      "step": 25744
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7818426042983566e-06,
+      "loss": 2.749,
+      "step": 25745
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7817951959544878e-06,
+      "loss": 2.9233,
+      "step": 25746
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7817477876106195e-06,
+      "loss": 2.4559,
+      "step": 25747
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.781700379266751e-06,
+      "loss": 2.6953,
+      "step": 25748
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7816529709228823e-06,
+      "loss": 2.6143,
+      "step": 25749
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.781605562579014e-06,
+      "loss": 3.2722,
+      "step": 25750
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7815581542351455e-06,
+      "loss": 2.0245,
+      "step": 25751
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7815107458912768e-06,
+      "loss": 2.0634,
+      "step": 25752
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7814633375474085e-06,
+      "loss": 2.4798,
+      "step": 25753
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.78141592920354e-06,
+      "loss": 2.0141,
+      "step": 25754
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7813685208596713e-06,
+      "loss": 1.9339,
+      "step": 25755
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.781321112515803e-06,
+      "loss": 2.3997,
+      "step": 25756
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7812737041719345e-06,
+      "loss": 3.4957,
+      "step": 25757
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7812262958280657e-06,
+      "loss": 3.2645,
+      "step": 25758
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.781178887484197e-06,
+      "loss": 1.4565,
+      "step": 25759
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7811314791403287e-06,
+      "loss": 1.7637,
+      "step": 25760
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7810840707964602e-06,
+      "loss": 2.8619,
+      "step": 25761
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7810366624525915e-06,
+      "loss": 2.8886,
+      "step": 25762
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7809892541087232e-06,
+      "loss": 2.0126,
+      "step": 25763
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7809418457648547e-06,
+      "loss": 2.1942,
+      "step": 25764
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.780894437420986e-06,
+      "loss": 1.9569,
+      "step": 25765
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7808470290771177e-06,
+      "loss": 2.7453,
+      "step": 25766
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7807996207332492e-06,
+      "loss": 2.4397,
+      "step": 25767
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7807522123893804e-06,
+      "loss": 3.0363,
+      "step": 25768
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7807048040455122e-06,
+      "loss": 2.8964,
+      "step": 25769
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7806573957016436e-06,
+      "loss": 2.9609,
+      "step": 25770
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.780609987357775e-06,
+      "loss": 3.3007,
+      "step": 25771
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7805625790139066e-06,
+      "loss": 1.4253,
+      "step": 25772
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.780515170670038e-06,
+      "loss": 3.1989,
+      "step": 25773
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7804677623261694e-06,
+      "loss": 2.7463,
+      "step": 25774
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.780420353982301e-06,
+      "loss": 1.9174,
+      "step": 25775
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7803729456384324e-06,
+      "loss": 1.9192,
+      "step": 25776
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7803255372945639e-06,
+      "loss": 3.3454,
+      "step": 25777
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7802781289506956e-06,
+      "loss": 3.3511,
+      "step": 25778
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7802307206068269e-06,
+      "loss": 2.3558,
+      "step": 25779
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7801833122629584e-06,
+      "loss": 3.4102,
+      "step": 25780
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7801359039190896e-06,
+      "loss": 2.434,
+      "step": 25781
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7800884955752213e-06,
+      "loss": 2.3555,
+      "step": 25782
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7800410872313528e-06,
+      "loss": 2.3302,
+      "step": 25783
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7799936788874841e-06,
+      "loss": 2.5549,
+      "step": 25784
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7799462705436158e-06,
+      "loss": 3.307,
+      "step": 25785
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.779898862199747e-06,
+      "loss": 2.301,
+      "step": 25786
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7798514538558786e-06,
+      "loss": 3.435,
+      "step": 25787
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7798040455120103e-06,
+      "loss": 2.3733,
+      "step": 25788
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7797566371681416e-06,
+      "loss": 2.9903,
+      "step": 25789
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.779709228824273e-06,
+      "loss": 2.5416,
+      "step": 25790
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7796618204804048e-06,
+      "loss": 2.5885,
+      "step": 25791
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.779614412136536e-06,
+      "loss": 2.4329,
+      "step": 25792
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7795670037926675e-06,
+      "loss": 2.2498,
+      "step": 25793
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7795195954487992e-06,
+      "loss": 0.6079,
+      "step": 25794
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7794721871049305e-06,
+      "loss": 1.6395,
+      "step": 25795
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7794247787610618e-06,
+      "loss": 2.7784,
+      "step": 25796
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7793773704171937e-06,
+      "loss": 2.2221,
+      "step": 25797
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.779329962073325e-06,
+      "loss": 1.5843,
+      "step": 25798
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7792825537294563e-06,
+      "loss": 2.998,
+      "step": 25799
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.779235145385588e-06,
+      "loss": 3.0941,
+      "step": 25800
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7791877370417195e-06,
+      "loss": 1.8042,
+      "step": 25801
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7791403286978508e-06,
+      "loss": 1.8978,
+      "step": 25802
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7790929203539823e-06,
+      "loss": 2.1298,
+      "step": 25803
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.779045512010114e-06,
+      "loss": 1.6708,
+      "step": 25804
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7789981036662452e-06,
+      "loss": 2.0098,
+      "step": 25805
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.7789506953223767e-06,
+      "loss": 3.0843,
+      "step": 25806
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7789032869785084e-06,
+      "loss": 4.126,
+      "step": 25807
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7788558786346397e-06,
+      "loss": 2.3569,
+      "step": 25808
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.778808470290771e-06,
+      "loss": 1.6757,
+      "step": 25809
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7787610619469027e-06,
+      "loss": 2.6527,
+      "step": 25810
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7787136536030342e-06,
+      "loss": 1.7407,
+      "step": 25811
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7786662452591655e-06,
+      "loss": 3.287,
+      "step": 25812
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7786188369152972e-06,
+      "loss": 3.4386,
+      "step": 25813
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7785714285714287e-06,
+      "loss": 1.8634,
+      "step": 25814
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.77852402022756e-06,
+      "loss": 2.5226,
+      "step": 25815
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7784766118836916e-06,
+      "loss": 0.9356,
+      "step": 25816
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7784292035398231e-06,
+      "loss": 2.6334,
+      "step": 25817
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7783817951959544e-06,
+      "loss": 2.5445,
+      "step": 25818
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7783343868520861e-06,
+      "loss": 1.6886,
+      "step": 25819
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7782869785082176e-06,
+      "loss": 2.7668,
+      "step": 25820
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.778239570164349e-06,
+      "loss": 1.5465,
+      "step": 25821
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7781921618204806e-06,
+      "loss": 1.5931,
+      "step": 25822
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7781447534766119e-06,
+      "loss": 3.0036,
+      "step": 25823
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7780973451327434e-06,
+      "loss": 2.4999,
+      "step": 25824
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7780499367888747e-06,
+      "loss": 2.275,
+      "step": 25825
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7780025284450064e-06,
+      "loss": 1.9912,
+      "step": 25826
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7779551201011379e-06,
+      "loss": 1.8262,
+      "step": 25827
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7779077117572691e-06,
+      "loss": 2.5782,
+      "step": 25828
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7778603034134008e-06,
+      "loss": 2.804,
+      "step": 25829
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7778128950695323e-06,
+      "loss": 2.012,
+      "step": 25830
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7777654867256636e-06,
+      "loss": 4.0243,
+      "step": 25831
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7777180783817953e-06,
+      "loss": 1.387,
+      "step": 25832
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7776706700379268e-06,
+      "loss": 2.3849,
+      "step": 25833
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.777623261694058e-06,
+      "loss": 2.2091,
+      "step": 25834
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7775758533501898e-06,
+      "loss": 2.167,
+      "step": 25835
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.777528445006321e-06,
+      "loss": 3.4,
+      "step": 25836
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7774810366624526e-06,
+      "loss": 3.3912,
+      "step": 25837
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7774336283185843e-06,
+      "loss": 2.5429,
+      "step": 25838
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7773862199747155e-06,
+      "loss": 2.1488,
+      "step": 25839
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.777338811630847e-06,
+      "loss": 2.1812,
+      "step": 25840
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7772914032869787e-06,
+      "loss": 2.5031,
+      "step": 25841
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.77724399494311e-06,
+      "loss": 1.2063,
+      "step": 25842
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7771965865992415e-06,
+      "loss": 1.9812,
+      "step": 25843
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7771491782553732e-06,
+      "loss": 2.4963,
+      "step": 25844
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7771017699115045e-06,
+      "loss": 2.9514,
+      "step": 25845
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.777054361567636e-06,
+      "loss": 2.4671,
+      "step": 25846
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7770069532237673e-06,
+      "loss": 1.9555,
+      "step": 25847
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.776959544879899e-06,
+      "loss": 1.1427,
+      "step": 25848
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7769121365360303e-06,
+      "loss": 2.1271,
+      "step": 25849
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7768647281921617e-06,
+      "loss": 3.2588,
+      "step": 25850
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7768173198482935e-06,
+      "loss": 4.0495,
+      "step": 25851
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7767699115044247e-06,
+      "loss": 2.0317,
+      "step": 25852
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7767225031605562e-06,
+      "loss": 2.1486,
+      "step": 25853
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.776675094816688e-06,
+      "loss": 2.9123,
+      "step": 25854
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7766276864728192e-06,
+      "loss": 1.713,
+      "step": 25855
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7765802781289507e-06,
+      "loss": 3.0536,
+      "step": 25856
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7765328697850824e-06,
+      "loss": 1.9596,
+      "step": 25857
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7764854614412137e-06,
+      "loss": 2.3107,
+      "step": 25858
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7764380530973452e-06,
+      "loss": 2.6553,
+      "step": 25859
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7763906447534769e-06,
+      "loss": 1.8007,
+      "step": 25860
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7763432364096082e-06,
+      "loss": 1.1444,
+      "step": 25861
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7762958280657394e-06,
+      "loss": 1.8062,
+      "step": 25862
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7762484197218711e-06,
+      "loss": 1.7844,
+      "step": 25863
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7762010113780026e-06,
+      "loss": 3.5611,
+      "step": 25864
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.776153603034134e-06,
+      "loss": 1.8223,
+      "step": 25865
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7761061946902656e-06,
+      "loss": 2.4417,
+      "step": 25866
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7760587863463971e-06,
+      "loss": 1.9819,
+      "step": 25867
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7760113780025284e-06,
+      "loss": 1.6405,
+      "step": 25868
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7759639696586599e-06,
+      "loss": 2.5974,
+      "step": 25869
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.7759165613147916e-06,
+      "loss": 3.3781,
+      "step": 25870
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7758691529709229e-06,
+      "loss": 3.5328,
+      "step": 25871
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7758217446270544e-06,
+      "loss": 0.8716,
+      "step": 25872
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.775774336283186e-06,
+      "loss": 1.1294,
+      "step": 25873
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7757269279393173e-06,
+      "loss": 3.4233,
+      "step": 25874
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7756795195954486e-06,
+      "loss": 4.0846,
+      "step": 25875
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7756321112515803e-06,
+      "loss": 2.081,
+      "step": 25876
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7755847029077118e-06,
+      "loss": 2.0574,
+      "step": 25877
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.775537294563843e-06,
+      "loss": 2.64,
+      "step": 25878
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7754898862199748e-06,
+      "loss": 1.601,
+      "step": 25879
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7754424778761063e-06,
+      "loss": 2.7998,
+      "step": 25880
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7753950695322376e-06,
+      "loss": 2.9489,
+      "step": 25881
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7753476611883693e-06,
+      "loss": 2.4657,
+      "step": 25882
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7753002528445008e-06,
+      "loss": 2.9195,
+      "step": 25883
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.775252844500632e-06,
+      "loss": 1.4844,
+      "step": 25884
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7752054361567638e-06,
+      "loss": 1.8294,
+      "step": 25885
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7751580278128953e-06,
+      "loss": 1.1855,
+      "step": 25886
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7751106194690265e-06,
+      "loss": 1.9634,
+      "step": 25887
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7750632111251582e-06,
+      "loss": 2.5008,
+      "step": 25888
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7750158027812895e-06,
+      "loss": 2.6545,
+      "step": 25889
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.774968394437421e-06,
+      "loss": 2.0084,
+      "step": 25890
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7749209860935523e-06,
+      "loss": 3.0384,
+      "step": 25891
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.774873577749684e-06,
+      "loss": 2.74,
+      "step": 25892
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7748261694058155e-06,
+      "loss": 1.3795,
+      "step": 25893
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7747787610619468e-06,
+      "loss": 2.6046,
+      "step": 25894
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7747313527180785e-06,
+      "loss": 1.4095,
+      "step": 25895
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.77468394437421e-06,
+      "loss": 1.7126,
+      "step": 25896
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7746365360303412e-06,
+      "loss": 2.9988,
+      "step": 25897
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.774589127686473e-06,
+      "loss": 2.6311,
+      "step": 25898
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7745417193426044e-06,
+      "loss": 3.0477,
+      "step": 25899
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7744943109987357e-06,
+      "loss": 1.4456,
+      "step": 25900
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7744469026548674e-06,
+      "loss": 2.5063,
+      "step": 25901
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7743994943109987e-06,
+      "loss": 1.79,
+      "step": 25902
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7743520859671302e-06,
+      "loss": 2.5053,
+      "step": 25903
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.774304677623262e-06,
+      "loss": 2.451,
+      "step": 25904
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7742572692793932e-06,
+      "loss": 1.85,
+      "step": 25905
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7742098609355247e-06,
+      "loss": 1.3777,
+      "step": 25906
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7741624525916564e-06,
+      "loss": 3.3941,
+      "step": 25907
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7741150442477877e-06,
+      "loss": 2.715,
+      "step": 25908
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7740676359039192e-06,
+      "loss": 1.1044,
+      "step": 25909
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7740202275600509e-06,
+      "loss": 3.09,
+      "step": 25910
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7739728192161821e-06,
+      "loss": 2.0129,
+      "step": 25911
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7739254108723136e-06,
+      "loss": 3.0532,
+      "step": 25912
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.773878002528445e-06,
+      "loss": 1.7469,
+      "step": 25913
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7738305941845766e-06,
+      "loss": 2.4491,
+      "step": 25914
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7737831858407079e-06,
+      "loss": 2.8411,
+      "step": 25915
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7737357774968394e-06,
+      "loss": 2.6326,
+      "step": 25916
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.773688369152971e-06,
+      "loss": 1.7072,
+      "step": 25917
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7736409608091024e-06,
+      "loss": 1.9955,
+      "step": 25918
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7735935524652339e-06,
+      "loss": 2.2214,
+      "step": 25919
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7735461441213656e-06,
+      "loss": 2.3588,
+      "step": 25920
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7734987357774968e-06,
+      "loss": 3.578,
+      "step": 25921
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7734513274336283e-06,
+      "loss": 3.2861,
+      "step": 25922
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.77340391908976e-06,
+      "loss": 3.0029,
+      "step": 25923
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7733565107458913e-06,
+      "loss": 2.859,
+      "step": 25924
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7733091024020226e-06,
+      "loss": 3.7391,
+      "step": 25925
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7732616940581545e-06,
+      "loss": 2.7077,
+      "step": 25926
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7732142857142858e-06,
+      "loss": 2.8535,
+      "step": 25927
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.773166877370417e-06,
+      "loss": 1.6529,
+      "step": 25928
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7731194690265488e-06,
+      "loss": 3.273,
+      "step": 25929
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7730720606826803e-06,
+      "loss": 3.7779,
+      "step": 25930
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7730246523388116e-06,
+      "loss": 1.5374,
+      "step": 25931
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7729772439949433e-06,
+      "loss": 0.9801,
+      "step": 25932
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.7729298356510748e-06,
+      "loss": 2.7525,
+      "step": 25933
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.772882427307206e-06,
+      "loss": 3.1141,
+      "step": 25934
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7728350189633375e-06,
+      "loss": 2.0603,
+      "step": 25935
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7727876106194692e-06,
+      "loss": 1.7708,
+      "step": 25936
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7727402022756005e-06,
+      "loss": 1.0145,
+      "step": 25937
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7726927939317318e-06,
+      "loss": 2.3443,
+      "step": 25938
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7726453855878635e-06,
+      "loss": 2.1081,
+      "step": 25939
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.772597977243995e-06,
+      "loss": 2.8722,
+      "step": 25940
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7725505689001263e-06,
+      "loss": 2.2299,
+      "step": 25941
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.772503160556258e-06,
+      "loss": 2.0634,
+      "step": 25942
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7724557522123895e-06,
+      "loss": 3.2422,
+      "step": 25943
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7724083438685207e-06,
+      "loss": 2.4037,
+      "step": 25944
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7723609355246524e-06,
+      "loss": 3.2838,
+      "step": 25945
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.772313527180784e-06,
+      "loss": 2.1611,
+      "step": 25946
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7722661188369152e-06,
+      "loss": 2.2912,
+      "step": 25947
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.772218710493047e-06,
+      "loss": 3.422,
+      "step": 25948
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7721713021491784e-06,
+      "loss": 3.6835,
+      "step": 25949
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7721238938053097e-06,
+      "loss": 1.6872,
+      "step": 25950
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7720764854614414e-06,
+      "loss": 1.9338,
+      "step": 25951
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7720290771175727e-06,
+      "loss": 1.3417,
+      "step": 25952
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7719816687737042e-06,
+      "loss": 2.5174,
+      "step": 25953
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7719342604298359e-06,
+      "loss": 1.3226,
+      "step": 25954
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7718868520859672e-06,
+      "loss": 2.6795,
+      "step": 25955
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7718394437420986e-06,
+      "loss": 2.2815,
+      "step": 25956
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.77179203539823e-06,
+      "loss": 2.2185,
+      "step": 25957
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7717446270543616e-06,
+      "loss": 1.3541,
+      "step": 25958
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7716972187104931e-06,
+      "loss": 2.6117,
+      "step": 25959
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7716498103666244e-06,
+      "loss": 2.0699,
+      "step": 25960
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7716024020227561e-06,
+      "loss": 2.3003,
+      "step": 25961
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7715549936788876e-06,
+      "loss": 3.6806,
+      "step": 25962
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7715075853350189e-06,
+      "loss": 2.0985,
+      "step": 25963
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7714601769911506e-06,
+      "loss": 2.9001,
+      "step": 25964
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7714127686472819e-06,
+      "loss": 2.5033,
+      "step": 25965
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7713653603034134e-06,
+      "loss": 2.6568,
+      "step": 25966
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.771317951959545e-06,
+      "loss": 3.3793,
+      "step": 25967
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7712705436156763e-06,
+      "loss": 2.5623,
+      "step": 25968
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7712231352718078e-06,
+      "loss": 2.0108,
+      "step": 25969
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7711757269279395e-06,
+      "loss": 3.8315,
+      "step": 25970
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7711283185840708e-06,
+      "loss": 2.3203,
+      "step": 25971
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7710809102402023e-06,
+      "loss": 1.8054,
+      "step": 25972
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.771033501896334e-06,
+      "loss": 1.6322,
+      "step": 25973
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7709860935524653e-06,
+      "loss": 2.8678,
+      "step": 25974
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7709386852085968e-06,
+      "loss": 2.8783,
+      "step": 25975
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7708912768647285e-06,
+      "loss": 2.5239,
+      "step": 25976
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7708438685208598e-06,
+      "loss": 2.8812,
+      "step": 25977
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.770796460176991e-06,
+      "loss": 1.7043,
+      "step": 25978
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7707490518331225e-06,
+      "loss": 3.1459,
+      "step": 25979
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7707016434892542e-06,
+      "loss": 2.1731,
+      "step": 25980
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7706542351453855e-06,
+      "loss": 2.9499,
+      "step": 25981
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.770606826801517e-06,
+      "loss": 2.0795,
+      "step": 25982
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7705594184576487e-06,
+      "loss": 2.9758,
+      "step": 25983
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.77051201011378e-06,
+      "loss": 1.082,
+      "step": 25984
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7704646017699115e-06,
+      "loss": 3.7216,
+      "step": 25985
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7704171934260432e-06,
+      "loss": 2.7033,
+      "step": 25986
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7703697850821745e-06,
+      "loss": 2.3901,
+      "step": 25987
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.770322376738306e-06,
+      "loss": 2.2878,
+      "step": 25988
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7702749683944377e-06,
+      "loss": 1.7927,
+      "step": 25989
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.770227560050569e-06,
+      "loss": 2.6544,
+      "step": 25990
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7701801517067002e-06,
+      "loss": 1.4974,
+      "step": 25991
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.770132743362832e-06,
+      "loss": 1.5623,
+      "step": 25992
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7700853350189634e-06,
+      "loss": 1.7535,
+      "step": 25993
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7700379266750947e-06,
+      "loss": 2.6439,
+      "step": 25994
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.7699905183312264e-06,
+      "loss": 2.5416,
+      "step": 25995
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.769943109987358e-06,
+      "loss": 2.1253,
+      "step": 25996
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7698957016434892e-06,
+      "loss": 3.7591,
+      "step": 25997
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7698482932996207e-06,
+      "loss": 3.4031,
+      "step": 25998
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7698008849557524e-06,
+      "loss": 3.6044,
+      "step": 25999
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7697534766118837e-06,
+      "loss": 1.6317,
+      "step": 26000
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7697060682680152e-06,
+      "loss": 2.8768,
+      "step": 26001
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7696586599241469e-06,
+      "loss": 2.972,
+      "step": 26002
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7696112515802781e-06,
+      "loss": 2.9093,
+      "step": 26003
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7695638432364094e-06,
+      "loss": 3.2697,
+      "step": 26004
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7695164348925411e-06,
+      "loss": 2.7148,
+      "step": 26005
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7694690265486726e-06,
+      "loss": 4.4795,
+      "step": 26006
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.769421618204804e-06,
+      "loss": 3.3272,
+      "step": 26007
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7693742098609356e-06,
+      "loss": 2.4495,
+      "step": 26008
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.769326801517067e-06,
+      "loss": 1.8268,
+      "step": 26009
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7692793931731984e-06,
+      "loss": 2.0824,
+      "step": 26010
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.76923198482933e-06,
+      "loss": 1.8091,
+      "step": 26011
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7691845764854616e-06,
+      "loss": 2.1737,
+      "step": 26012
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7691371681415929e-06,
+      "loss": 2.8737,
+      "step": 26013
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7690897597977246e-06,
+      "loss": 3.5344,
+      "step": 26014
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.769042351453856e-06,
+      "loss": 3.0176,
+      "step": 26015
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7689949431099873e-06,
+      "loss": 2.4951,
+      "step": 26016
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.768947534766119e-06,
+      "loss": 3.4247,
+      "step": 26017
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7689001264222503e-06,
+      "loss": 2.2186,
+      "step": 26018
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7688527180783818e-06,
+      "loss": 2.3843,
+      "step": 26019
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.768805309734513e-06,
+      "loss": 3.2816,
+      "step": 26020
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7687579013906448e-06,
+      "loss": 2.2487,
+      "step": 26021
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7687104930467763e-06,
+      "loss": 1.4323,
+      "step": 26022
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7686630847029076e-06,
+      "loss": 2.4438,
+      "step": 26023
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7686156763590393e-06,
+      "loss": 3.1762,
+      "step": 26024
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7685682680151708e-06,
+      "loss": 1.6856,
+      "step": 26025
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.768520859671302e-06,
+      "loss": 2.6093,
+      "step": 26026
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7684734513274337e-06,
+      "loss": 1.746,
+      "step": 26027
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7684260429835652e-06,
+      "loss": 3.213,
+      "step": 26028
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7683786346396965e-06,
+      "loss": 2.4724,
+      "step": 26029
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7683312262958282e-06,
+      "loss": 1.4097,
+      "step": 26030
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7682838179519595e-06,
+      "loss": 2.6585,
+      "step": 26031
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.768236409608091e-06,
+      "loss": 1.7787,
+      "step": 26032
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7681890012642227e-06,
+      "loss": 2.2404,
+      "step": 26033
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.768141592920354e-06,
+      "loss": 2.7887,
+      "step": 26034
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7680941845764855e-06,
+      "loss": 2.4387,
+      "step": 26035
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7680467762326172e-06,
+      "loss": 2.634,
+      "step": 26036
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7679993678887485e-06,
+      "loss": 2.7773,
+      "step": 26037
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.76795195954488e-06,
+      "loss": 2.6349,
+      "step": 26038
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7679045512010117e-06,
+      "loss": 2.966,
+      "step": 26039
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.767857142857143e-06,
+      "loss": 2.8506,
+      "step": 26040
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7678097345132744e-06,
+      "loss": 3.1196,
+      "step": 26041
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7677623261694057e-06,
+      "loss": 2.3596,
+      "step": 26042
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7677149178255374e-06,
+      "loss": 3.898,
+      "step": 26043
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7676675094816687e-06,
+      "loss": 3.1787,
+      "step": 26044
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7676201011378002e-06,
+      "loss": 1.9929,
+      "step": 26045
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7675726927939319e-06,
+      "loss": 1.1905,
+      "step": 26046
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7675252844500632e-06,
+      "loss": 3.179,
+      "step": 26047
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7674778761061947e-06,
+      "loss": 1.6385,
+      "step": 26048
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7674304677623264e-06,
+      "loss": 2.1741,
+      "step": 26049
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7673830594184576e-06,
+      "loss": 2.4205,
+      "step": 26050
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7673356510745891e-06,
+      "loss": 2.4624,
+      "step": 26051
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7672882427307208e-06,
+      "loss": 2.1453,
+      "step": 26052
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7672408343868521e-06,
+      "loss": 3.2374,
+      "step": 26053
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7671934260429836e-06,
+      "loss": 2.1683,
+      "step": 26054
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7671460176991153e-06,
+      "loss": 1.9917,
+      "step": 26055
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7670986093552466e-06,
+      "loss": 2.4732,
+      "step": 26056
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7670512010113779e-06,
+      "loss": 2.1067,
+      "step": 26057
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7670037926675096e-06,
+      "loss": 2.1264,
+      "step": 26058
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.766956384323641e-06,
+      "loss": 2.0637,
+      "step": 26059
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.7669089759797724e-06,
+      "loss": 1.3339,
+      "step": 26060
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.766861567635904e-06,
+      "loss": 2.7543,
+      "step": 26061
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7668141592920355e-06,
+      "loss": 2.8471,
+      "step": 26062
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7667667509481668e-06,
+      "loss": 2.2531,
+      "step": 26063
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7667193426042983e-06,
+      "loss": 3.6329,
+      "step": 26064
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.76667193426043e-06,
+      "loss": 3.9716,
+      "step": 26065
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7666245259165613e-06,
+      "loss": 4.8327,
+      "step": 26066
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7665771175726926e-06,
+      "loss": 3.0409,
+      "step": 26067
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7665297092288245e-06,
+      "loss": 2.0264,
+      "step": 26068
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7664823008849558e-06,
+      "loss": 3.0448,
+      "step": 26069
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.766434892541087e-06,
+      "loss": 2.7845,
+      "step": 26070
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7663874841972188e-06,
+      "loss": 2.8654,
+      "step": 26071
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7663400758533503e-06,
+      "loss": 2.3299,
+      "step": 26072
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7662926675094815e-06,
+      "loss": 3.0178,
+      "step": 26073
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7662452591656132e-06,
+      "loss": 1.7409,
+      "step": 26074
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7661978508217447e-06,
+      "loss": 2.083,
+      "step": 26075
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.766150442477876e-06,
+      "loss": 3.3567,
+      "step": 26076
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7661030341340077e-06,
+      "loss": 1.6676,
+      "step": 26077
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7660556257901392e-06,
+      "loss": 1.8443,
+      "step": 26078
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7660082174462705e-06,
+      "loss": 3.1767,
+      "step": 26079
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7659608091024022e-06,
+      "loss": 2.0064,
+      "step": 26080
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7659134007585335e-06,
+      "loss": 2.1439,
+      "step": 26081
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.765865992414665e-06,
+      "loss": 3.1828,
+      "step": 26082
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7658185840707967e-06,
+      "loss": 1.6462,
+      "step": 26083
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.765771175726928e-06,
+      "loss": 2.7677,
+      "step": 26084
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7657237673830594e-06,
+      "loss": 1.7855,
+      "step": 26085
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7656763590391907e-06,
+      "loss": 3.1158,
+      "step": 26086
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7656289506953224e-06,
+      "loss": 3.7599,
+      "step": 26087
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.765581542351454e-06,
+      "loss": 2.8619,
+      "step": 26088
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7655341340075852e-06,
+      "loss": 1.9324,
+      "step": 26089
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.765486725663717e-06,
+      "loss": 2.6318,
+      "step": 26090
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7654393173198484e-06,
+      "loss": 1.3096,
+      "step": 26091
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7653919089759797e-06,
+      "loss": 1.8247,
+      "step": 26092
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7653445006321114e-06,
+      "loss": 3.087,
+      "step": 26093
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7652970922882427e-06,
+      "loss": 1.5964,
+      "step": 26094
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7652496839443742e-06,
+      "loss": 2.2904,
+      "step": 26095
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7652022756005059e-06,
+      "loss": 2.1652,
+      "step": 26096
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7651548672566371e-06,
+      "loss": 2.0713,
+      "step": 26097
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7651074589127686e-06,
+      "loss": 2.4764,
+      "step": 26098
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7650600505689003e-06,
+      "loss": 1.9239,
+      "step": 26099
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7650126422250316e-06,
+      "loss": 2.8288,
+      "step": 26100
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7649652338811631e-06,
+      "loss": 2.6291,
+      "step": 26101
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7649178255372948e-06,
+      "loss": 1.7584,
+      "step": 26102
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.764870417193426e-06,
+      "loss": 0.9399,
+      "step": 26103
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7648230088495576e-06,
+      "loss": 2.047,
+      "step": 26104
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7647756005056893e-06,
+      "loss": 1.6086,
+      "step": 26105
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7647281921618206e-06,
+      "loss": 2.4159,
+      "step": 26106
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7646807838179518e-06,
+      "loss": 3.4435,
+      "step": 26107
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7646333754740833e-06,
+      "loss": 2.035,
+      "step": 26108
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.764585967130215e-06,
+      "loss": 1.7273,
+      "step": 26109
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7645385587863463e-06,
+      "loss": 3.4417,
+      "step": 26110
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7644911504424778e-06,
+      "loss": 3.1168,
+      "step": 26111
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7644437420986095e-06,
+      "loss": 1.0433,
+      "step": 26112
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7643963337547408e-06,
+      "loss": 3.8997,
+      "step": 26113
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7643489254108723e-06,
+      "loss": 1.9451,
+      "step": 26114
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.764301517067004e-06,
+      "loss": 1.7198,
+      "step": 26115
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7642541087231353e-06,
+      "loss": 2.6734,
+      "step": 26116
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7642067003792668e-06,
+      "loss": 3.1826,
+      "step": 26117
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7641592920353985e-06,
+      "loss": 1.5897,
+      "step": 26118
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7641118836915298e-06,
+      "loss": 2.666,
+      "step": 26119
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.764064475347661e-06,
+      "loss": 1.8459,
+      "step": 26120
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7640170670037927e-06,
+      "loss": 3.99,
+      "step": 26121
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7639696586599242e-06,
+      "loss": 2.5565,
+      "step": 26122
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.7639222503160555e-06,
+      "loss": 2.8118,
+      "step": 26123
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7638748419721872e-06,
+      "loss": 2.2957,
+      "step": 26124
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7638274336283187e-06,
+      "loss": 2.9701,
+      "step": 26125
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.76378002528445e-06,
+      "loss": 2.0556,
+      "step": 26126
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7637326169405817e-06,
+      "loss": 3.253,
+      "step": 26127
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7636852085967132e-06,
+      "loss": 3.0845,
+      "step": 26128
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7636378002528445e-06,
+      "loss": 1.3441,
+      "step": 26129
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.763590391908976e-06,
+      "loss": 2.638,
+      "step": 26130
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7635429835651077e-06,
+      "loss": 2.8244,
+      "step": 26131
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.763495575221239e-06,
+      "loss": 1.8215,
+      "step": 26132
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7634481668773702e-06,
+      "loss": 3.0631,
+      "step": 26133
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.763400758533502e-06,
+      "loss": 1.8773,
+      "step": 26134
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7633533501896334e-06,
+      "loss": 2.3339,
+      "step": 26135
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7633059418457647e-06,
+      "loss": 1.7527,
+      "step": 26136
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7632585335018964e-06,
+      "loss": 2.6577,
+      "step": 26137
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.763211125158028e-06,
+      "loss": 3.0281,
+      "step": 26138
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7631637168141592e-06,
+      "loss": 2.2013,
+      "step": 26139
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7631163084702909e-06,
+      "loss": 2.0642,
+      "step": 26140
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7630689001264224e-06,
+      "loss": 3.2039,
+      "step": 26141
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7630214917825537e-06,
+      "loss": 2.2016,
+      "step": 26142
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7629740834386854e-06,
+      "loss": 1.8969,
+      "step": 26143
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7629266750948168e-06,
+      "loss": 2.0079,
+      "step": 26144
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7628792667509481e-06,
+      "loss": 3.708,
+      "step": 26145
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7628318584070798e-06,
+      "loss": 1.5442,
+      "step": 26146
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7627844500632111e-06,
+      "loss": 3.874,
+      "step": 26147
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7627370417193426e-06,
+      "loss": 3.621,
+      "step": 26148
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7626896333754743e-06,
+      "loss": 1.4241,
+      "step": 26149
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7626422250316056e-06,
+      "loss": 1.8009,
+      "step": 26150
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.762594816687737e-06,
+      "loss": 2.4574,
+      "step": 26151
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7625474083438684e-06,
+      "loss": 2.4365,
+      "step": 26152
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7625e-06,
+      "loss": 1.888,
+      "step": 26153
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7624525916561316e-06,
+      "loss": 1.9304,
+      "step": 26154
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7624051833122628e-06,
+      "loss": 3.4925,
+      "step": 26155
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7623577749683945e-06,
+      "loss": 2.7827,
+      "step": 26156
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.762310366624526e-06,
+      "loss": 2.2035,
+      "step": 26157
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7622629582806573e-06,
+      "loss": 2.6352,
+      "step": 26158
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.762215549936789e-06,
+      "loss": 2.6709,
+      "step": 26159
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7621681415929203e-06,
+      "loss": 3.4869,
+      "step": 26160
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7621207332490518e-06,
+      "loss": 1.9302,
+      "step": 26161
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7620733249051835e-06,
+      "loss": 2.418,
+      "step": 26162
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7620259165613148e-06,
+      "loss": 2.9332,
+      "step": 26163
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7619785082174463e-06,
+      "loss": 3.5085,
+      "step": 26164
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.761931099873578e-06,
+      "loss": 2.3907,
+      "step": 26165
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7618836915297093e-06,
+      "loss": 3.4435,
+      "step": 26166
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7618362831858407e-06,
+      "loss": 2.4821,
+      "step": 26167
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7617888748419724e-06,
+      "loss": 2.8734,
+      "step": 26168
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7617414664981037e-06,
+      "loss": 1.8513,
+      "step": 26169
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7616940581542352e-06,
+      "loss": 2.459,
+      "step": 26170
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.761646649810367e-06,
+      "loss": 2.0773,
+      "step": 26171
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7615992414664982e-06,
+      "loss": 1.3643,
+      "step": 26172
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7615518331226295e-06,
+      "loss": 2.7156,
+      "step": 26173
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.761504424778761e-06,
+      "loss": 2.4106,
+      "step": 26174
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7614570164348927e-06,
+      "loss": 3.382,
+      "step": 26175
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.761409608091024e-06,
+      "loss": 3.5072,
+      "step": 26176
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7613621997471555e-06,
+      "loss": 2.4993,
+      "step": 26177
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7613147914032872e-06,
+      "loss": 2.2445,
+      "step": 26178
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7612673830594184e-06,
+      "loss": 2.2891,
+      "step": 26179
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.76121997471555e-06,
+      "loss": 3.0954,
+      "step": 26180
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7611725663716816e-06,
+      "loss": 1.6964,
+      "step": 26181
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.761125158027813e-06,
+      "loss": 1.8618,
+      "step": 26182
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7610777496839444e-06,
+      "loss": 3.0289,
+      "step": 26183
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7610303413400761e-06,
+      "loss": 3.7137,
+      "step": 26184
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7609829329962074e-06,
+      "loss": 2.6752,
+      "step": 26185
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.7609355246523387e-06,
+      "loss": 2.9612,
+      "step": 26186
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7608881163084704e-06,
+      "loss": 2.1085,
+      "step": 26187
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7608407079646019e-06,
+      "loss": 2.6791,
+      "step": 26188
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7607932996207331e-06,
+      "loss": 3.0057,
+      "step": 26189
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7607458912768649e-06,
+      "loss": 2.019,
+      "step": 26190
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7606984829329963e-06,
+      "loss": 2.8462,
+      "step": 26191
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7606510745891276e-06,
+      "loss": 3.5267,
+      "step": 26192
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7606036662452593e-06,
+      "loss": 1.9005,
+      "step": 26193
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7605562579013908e-06,
+      "loss": 2.3557,
+      "step": 26194
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.760508849557522e-06,
+      "loss": 2.9464,
+      "step": 26195
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7604614412136536e-06,
+      "loss": 2.6833,
+      "step": 26196
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7604140328697853e-06,
+      "loss": 2.6895,
+      "step": 26197
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7603666245259166e-06,
+      "loss": 3.3057,
+      "step": 26198
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7603192161820479e-06,
+      "loss": 2.3869,
+      "step": 26199
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7602718078381796e-06,
+      "loss": 1.9825,
+      "step": 26200
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.760224399494311e-06,
+      "loss": 2.1626,
+      "step": 26201
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7601769911504423e-06,
+      "loss": 3.1035,
+      "step": 26202
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.760129582806574e-06,
+      "loss": 2.538,
+      "step": 26203
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7600821744627055e-06,
+      "loss": 2.0194,
+      "step": 26204
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7600347661188368e-06,
+      "loss": 2.6832,
+      "step": 26205
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7599873577749685e-06,
+      "loss": 2.7176,
+      "step": 26206
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7599399494311e-06,
+      "loss": 2.1324,
+      "step": 26207
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7598925410872313e-06,
+      "loss": 2.3124,
+      "step": 26208
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.759845132743363e-06,
+      "loss": 2.8027,
+      "step": 26209
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7597977243994945e-06,
+      "loss": 2.2251,
+      "step": 26210
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7597503160556258e-06,
+      "loss": 2.1668,
+      "step": 26211
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7597029077117575e-06,
+      "loss": 2.7661,
+      "step": 26212
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7596554993678887e-06,
+      "loss": 2.4961,
+      "step": 26213
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7596080910240202e-06,
+      "loss": 2.4041,
+      "step": 26214
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.759560682680152e-06,
+      "loss": 2.3879,
+      "step": 26215
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7595132743362832e-06,
+      "loss": 4.2136,
+      "step": 26216
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7594658659924147e-06,
+      "loss": 1.9141,
+      "step": 26217
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.759418457648546e-06,
+      "loss": 2.9554,
+      "step": 26218
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7593710493046777e-06,
+      "loss": 1.7715,
+      "step": 26219
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7593236409608092e-06,
+      "loss": 1.7632,
+      "step": 26220
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7592762326169405e-06,
+      "loss": 2.9292,
+      "step": 26221
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7592288242730722e-06,
+      "loss": 1.7312,
+      "step": 26222
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7591814159292035e-06,
+      "loss": 3.4707,
+      "step": 26223
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.759134007585335e-06,
+      "loss": 2.1604,
+      "step": 26224
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7590865992414667e-06,
+      "loss": 1.715,
+      "step": 26225
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.759039190897598e-06,
+      "loss": 1.7797,
+      "step": 26226
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7589917825537294e-06,
+      "loss": 2.6985,
+      "step": 26227
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7589443742098611e-06,
+      "loss": 2.8607,
+      "step": 26228
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7588969658659924e-06,
+      "loss": 3.0943,
+      "step": 26229
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.758849557522124e-06,
+      "loss": 1.9234,
+      "step": 26230
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7588021491782556e-06,
+      "loss": 4.3682,
+      "step": 26231
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7587547408343869e-06,
+      "loss": 1.4857,
+      "step": 26232
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7587073324905184e-06,
+      "loss": 2.22,
+      "step": 26233
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.75865992414665e-06,
+      "loss": 2.4711,
+      "step": 26234
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7586125158027814e-06,
+      "loss": 2.9437,
+      "step": 26235
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7585651074589126e-06,
+      "loss": 2.65,
+      "step": 26236
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7585176991150443e-06,
+      "loss": 3.6032,
+      "step": 26237
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7584702907711758e-06,
+      "loss": 1.2601,
+      "step": 26238
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7584228824273071e-06,
+      "loss": 2.6438,
+      "step": 26239
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7583754740834386e-06,
+      "loss": 3.1678,
+      "step": 26240
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7583280657395703e-06,
+      "loss": 3.9123,
+      "step": 26241
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7582806573957016e-06,
+      "loss": 3.5419,
+      "step": 26242
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.758233249051833e-06,
+      "loss": 2.4292,
+      "step": 26243
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7581858407079648e-06,
+      "loss": 2.4223,
+      "step": 26244
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.758138432364096e-06,
+      "loss": 2.8231,
+      "step": 26245
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7580910240202276e-06,
+      "loss": 2.8021,
+      "step": 26246
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7580436156763593e-06,
+      "loss": 3.5874,
+      "step": 26247
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7579962073324906e-06,
+      "loss": 3.1216,
+      "step": 26248
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7579487989886218e-06,
+      "loss": 1.7408,
+      "step": 26249
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.7579013906447535e-06,
+      "loss": 2.8046,
+      "step": 26250
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.757853982300885e-06,
+      "loss": 2.2094,
+      "step": 26251
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7578065739570163e-06,
+      "loss": 1.9646,
+      "step": 26252
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.757759165613148e-06,
+      "loss": 1.5128,
+      "step": 26253
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7577117572692795e-06,
+      "loss": 3.0551,
+      "step": 26254
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7576643489254108e-06,
+      "loss": 1.9365,
+      "step": 26255
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7576169405815425e-06,
+      "loss": 2.5163,
+      "step": 26256
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.757569532237674e-06,
+      "loss": 1.9377,
+      "step": 26257
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7575221238938053e-06,
+      "loss": 2.9285,
+      "step": 26258
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.757474715549937e-06,
+      "loss": 2.2777,
+      "step": 26259
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7574273072060685e-06,
+      "loss": 1.6529,
+      "step": 26260
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7573798988621997e-06,
+      "loss": 1.8589,
+      "step": 26261
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.757332490518331e-06,
+      "loss": 2.9055,
+      "step": 26262
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7572850821744627e-06,
+      "loss": 3.1961,
+      "step": 26263
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7572376738305942e-06,
+      "loss": 3.3961,
+      "step": 26264
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7571902654867255e-06,
+      "loss": 2.4773,
+      "step": 26265
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7571428571428572e-06,
+      "loss": 2.6023,
+      "step": 26266
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7570954487989887e-06,
+      "loss": 1.3759,
+      "step": 26267
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.75704804045512e-06,
+      "loss": 2.8501,
+      "step": 26268
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7570006321112517e-06,
+      "loss": 2.5994,
+      "step": 26269
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7569532237673832e-06,
+      "loss": 3.8812,
+      "step": 26270
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7569058154235144e-06,
+      "loss": 2.543,
+      "step": 26271
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7568584070796462e-06,
+      "loss": 3.7454,
+      "step": 26272
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7568109987357776e-06,
+      "loss": 2.4338,
+      "step": 26273
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.756763590391909e-06,
+      "loss": 1.2984,
+      "step": 26274
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7567161820480406e-06,
+      "loss": 1.5812,
+      "step": 26275
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.756668773704172e-06,
+      "loss": 1.6845,
+      "step": 26276
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7566213653603034e-06,
+      "loss": 1.7792,
+      "step": 26277
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.756573957016435e-06,
+      "loss": 2.4895,
+      "step": 26278
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7565265486725664e-06,
+      "loss": 2.0802,
+      "step": 26279
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7564791403286979e-06,
+      "loss": 2.4804,
+      "step": 26280
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7564317319848296e-06,
+      "loss": 2.9965,
+      "step": 26281
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7563843236409609e-06,
+      "loss": 2.8013,
+      "step": 26282
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7563369152970924e-06,
+      "loss": 2.7619,
+      "step": 26283
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7562895069532236e-06,
+      "loss": 1.8679,
+      "step": 26284
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7562420986093553e-06,
+      "loss": 2.0202,
+      "step": 26285
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7561946902654868e-06,
+      "loss": 2.8492,
+      "step": 26286
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7561472819216181e-06,
+      "loss": 3.5885,
+      "step": 26287
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7560998735777498e-06,
+      "loss": 2.8295,
+      "step": 26288
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.756052465233881e-06,
+      "loss": 3.1707,
+      "step": 26289
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7560050568900126e-06,
+      "loss": 2.0285,
+      "step": 26290
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7559576485461443e-06,
+      "loss": 2.5522,
+      "step": 26291
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7559102402022756e-06,
+      "loss": 2.4072,
+      "step": 26292
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.755862831858407e-06,
+      "loss": 3.2154,
+      "step": 26293
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7558154235145388e-06,
+      "loss": 2.2444,
+      "step": 26294
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.75576801517067e-06,
+      "loss": 2.1841,
+      "step": 26295
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7557206068268015e-06,
+      "loss": 2.7198,
+      "step": 26296
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7556731984829332e-06,
+      "loss": 0.925,
+      "step": 26297
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7556257901390645e-06,
+      "loss": 1.3126,
+      "step": 26298
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.755578381795196e-06,
+      "loss": 1.8587,
+      "step": 26299
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7555309734513277e-06,
+      "loss": 2.8315,
+      "step": 26300
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.755483565107459e-06,
+      "loss": 3.2729,
+      "step": 26301
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7554361567635903e-06,
+      "loss": 3.2135,
+      "step": 26302
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.755388748419722e-06,
+      "loss": 3.3983,
+      "step": 26303
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7553413400758535e-06,
+      "loss": 1.5769,
+      "step": 26304
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7552939317319848e-06,
+      "loss": 2.874,
+      "step": 26305
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7552465233881163e-06,
+      "loss": 1.8971,
+      "step": 26306
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.755199115044248e-06,
+      "loss": 2.7071,
+      "step": 26307
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7551517067003792e-06,
+      "loss": 1.9841,
+      "step": 26308
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7551042983565107e-06,
+      "loss": 2.1783,
+      "step": 26309
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7550568900126424e-06,
+      "loss": 3.2572,
+      "step": 26310
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7550094816687737e-06,
+      "loss": 2.6091,
+      "step": 26311
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.7549620733249052e-06,
+      "loss": 1.051,
+      "step": 26312
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.754914664981037e-06,
+      "loss": 3.0655,
+      "step": 26313
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7548672566371682e-06,
+      "loss": 3.0583,
+      "step": 26314
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7548198482932995e-06,
+      "loss": 2.1275,
+      "step": 26315
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7547724399494312e-06,
+      "loss": 3.0773,
+      "step": 26316
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7547250316055627e-06,
+      "loss": 3.6731,
+      "step": 26317
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.754677623261694e-06,
+      "loss": 1.6012,
+      "step": 26318
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7546302149178256e-06,
+      "loss": 1.4586,
+      "step": 26319
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7545828065739571e-06,
+      "loss": 2.8563,
+      "step": 26320
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7545353982300884e-06,
+      "loss": 0.9516,
+      "step": 26321
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7544879898862201e-06,
+      "loss": 2.1563,
+      "step": 26322
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7544405815423516e-06,
+      "loss": 1.9137,
+      "step": 26323
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.754393173198483e-06,
+      "loss": 2.1555,
+      "step": 26324
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7543457648546146e-06,
+      "loss": 2.2605,
+      "step": 26325
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.754298356510746e-06,
+      "loss": 2.7014,
+      "step": 26326
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7542509481668774e-06,
+      "loss": 2.2333,
+      "step": 26327
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7542035398230087e-06,
+      "loss": 2.5984,
+      "step": 26328
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7541561314791404e-06,
+      "loss": 1.6595,
+      "step": 26329
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7541087231352719e-06,
+      "loss": 3.4644,
+      "step": 26330
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7540613147914031e-06,
+      "loss": 3.7842,
+      "step": 26331
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7540139064475348e-06,
+      "loss": 3.1673,
+      "step": 26332
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7539664981036663e-06,
+      "loss": 1.8216,
+      "step": 26333
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7539190897597976e-06,
+      "loss": 4.0646,
+      "step": 26334
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7538716814159293e-06,
+      "loss": 2.4454,
+      "step": 26335
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7538242730720608e-06,
+      "loss": 1.5736,
+      "step": 26336
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.753776864728192e-06,
+      "loss": 0.9923,
+      "step": 26337
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7537294563843238e-06,
+      "loss": 2.6453,
+      "step": 26338
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7536820480404553e-06,
+      "loss": 2.1161,
+      "step": 26339
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7536346396965866e-06,
+      "loss": 2.6369,
+      "step": 26340
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7535872313527183e-06,
+      "loss": 3.2885,
+      "step": 26341
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7535398230088495e-06,
+      "loss": 2.995,
+      "step": 26342
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.753492414664981e-06,
+      "loss": 2.883,
+      "step": 26343
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7534450063211127e-06,
+      "loss": 2.5016,
+      "step": 26344
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.753397597977244e-06,
+      "loss": 3.0666,
+      "step": 26345
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7533501896333755e-06,
+      "loss": 1.8258,
+      "step": 26346
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7533027812895072e-06,
+      "loss": 2.4994,
+      "step": 26347
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7532553729456385e-06,
+      "loss": 1.8553,
+      "step": 26348
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.75320796460177e-06,
+      "loss": 2.8129,
+      "step": 26349
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7531605562579013e-06,
+      "loss": 3.2548,
+      "step": 26350
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.753113147914033e-06,
+      "loss": 1.9576,
+      "step": 26351
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7530657395701643e-06,
+      "loss": 2.3802,
+      "step": 26352
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7530183312262957e-06,
+      "loss": 2.136,
+      "step": 26353
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7529709228824275e-06,
+      "loss": 1.3913,
+      "step": 26354
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7529235145385587e-06,
+      "loss": 2.8047,
+      "step": 26355
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7528761061946902e-06,
+      "loss": 2.6394,
+      "step": 26356
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.752828697850822e-06,
+      "loss": 1.3725,
+      "step": 26357
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7527812895069532e-06,
+      "loss": 1.7719,
+      "step": 26358
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7527338811630847e-06,
+      "loss": 1.8493,
+      "step": 26359
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7526864728192164e-06,
+      "loss": 3.6967,
+      "step": 26360
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7526390644753477e-06,
+      "loss": 2.7702,
+      "step": 26361
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7525916561314792e-06,
+      "loss": 3.2429,
+      "step": 26362
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7525442477876109e-06,
+      "loss": 1.0565,
+      "step": 26363
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7524968394437422e-06,
+      "loss": 2.774,
+      "step": 26364
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7524494310998734e-06,
+      "loss": 0.9537,
+      "step": 26365
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7524020227560051e-06,
+      "loss": 1.514,
+      "step": 26366
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7523546144121366e-06,
+      "loss": 2.1035,
+      "step": 26367
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.752307206068268e-06,
+      "loss": 1.858,
+      "step": 26368
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7522597977243996e-06,
+      "loss": 3.0157,
+      "step": 26369
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7522123893805311e-06,
+      "loss": 3.1209,
+      "step": 26370
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7521649810366624e-06,
+      "loss": 3.7016,
+      "step": 26371
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7521175726927939e-06,
+      "loss": 2.9093,
+      "step": 26372
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7520701643489256e-06,
+      "loss": 1.5143,
+      "step": 26373
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7520227560050569e-06,
+      "loss": 3.1645,
+      "step": 26374
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.7519753476611884e-06,
+      "loss": 2.1187,
+      "step": 26375
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.75192793931732e-06,
+      "loss": 3.1863,
+      "step": 26376
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7518805309734513e-06,
+      "loss": 1.9418,
+      "step": 26377
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7518331226295826e-06,
+      "loss": 2.2026,
+      "step": 26378
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7517857142857143e-06,
+      "loss": 2.2827,
+      "step": 26379
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7517383059418458e-06,
+      "loss": 2.7002,
+      "step": 26380
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.751690897597977e-06,
+      "loss": 2.3773,
+      "step": 26381
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7516434892541088e-06,
+      "loss": 1.5731,
+      "step": 26382
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7515960809102403e-06,
+      "loss": 2.5889,
+      "step": 26383
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7515486725663716e-06,
+      "loss": 2.5172,
+      "step": 26384
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7515012642225033e-06,
+      "loss": 2.0185,
+      "step": 26385
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7514538558786348e-06,
+      "loss": 2.8734,
+      "step": 26386
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.751406447534766e-06,
+      "loss": 2.4722,
+      "step": 26387
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7513590391908978e-06,
+      "loss": 2.4808,
+      "step": 26388
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7513116308470293e-06,
+      "loss": 2.1107,
+      "step": 26389
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7512642225031605e-06,
+      "loss": 2.8524,
+      "step": 26390
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7512168141592922e-06,
+      "loss": 0.9912,
+      "step": 26391
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7511694058154235e-06,
+      "loss": 1.3251,
+      "step": 26392
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.751121997471555e-06,
+      "loss": 2.089,
+      "step": 26393
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7510745891276863e-06,
+      "loss": 2.5919,
+      "step": 26394
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.751027180783818e-06,
+      "loss": 2.0133,
+      "step": 26395
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7509797724399495e-06,
+      "loss": 3.2165,
+      "step": 26396
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7509323640960808e-06,
+      "loss": 2.728,
+      "step": 26397
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7508849557522125e-06,
+      "loss": 2.9916,
+      "step": 26398
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.750837547408344e-06,
+      "loss": 1.9398,
+      "step": 26399
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7507901390644752e-06,
+      "loss": 2.0356,
+      "step": 26400
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.750742730720607e-06,
+      "loss": 2.561,
+      "step": 26401
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7506953223767384e-06,
+      "loss": 3.6057,
+      "step": 26402
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7506479140328697e-06,
+      "loss": 2.4593,
+      "step": 26403
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7506005056890014e-06,
+      "loss": 2.4765,
+      "step": 26404
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7505530973451327e-06,
+      "loss": 3.383,
+      "step": 26405
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7505056890012642e-06,
+      "loss": 3.2473,
+      "step": 26406
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.750458280657396e-06,
+      "loss": 1.096,
+      "step": 26407
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7504108723135272e-06,
+      "loss": 1.4384,
+      "step": 26408
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7503634639696587e-06,
+      "loss": 1.5301,
+      "step": 26409
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7503160556257904e-06,
+      "loss": 2.3307,
+      "step": 26410
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7502686472819217e-06,
+      "loss": 2.4035,
+      "step": 26411
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7502212389380532e-06,
+      "loss": 1.951,
+      "step": 26412
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7501738305941849e-06,
+      "loss": 1.0575,
+      "step": 26413
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7501264222503161e-06,
+      "loss": 1.9779,
+      "step": 26414
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7500790139064476e-06,
+      "loss": 1.3393,
+      "step": 26415
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.750031605562579e-06,
+      "loss": 1.8338,
+      "step": 26416
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7499841972187106e-06,
+      "loss": 1.1132,
+      "step": 26417
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7499367888748419e-06,
+      "loss": 3.22,
+      "step": 26418
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7498893805309734e-06,
+      "loss": 2.3895,
+      "step": 26419
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.749841972187105e-06,
+      "loss": 1.3449,
+      "step": 26420
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7497945638432364e-06,
+      "loss": 1.3583,
+      "step": 26421
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7497471554993679e-06,
+      "loss": 2.0188,
+      "step": 26422
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7496997471554996e-06,
+      "loss": 2.1594,
+      "step": 26423
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7496523388116308e-06,
+      "loss": 1.6391,
+      "step": 26424
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7496049304677623e-06,
+      "loss": 1.3472,
+      "step": 26425
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.749557522123894e-06,
+      "loss": 1.2903,
+      "step": 26426
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7495101137800253e-06,
+      "loss": 2.8525,
+      "step": 26427
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7494627054361568e-06,
+      "loss": 3.4747,
+      "step": 26428
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7494152970922885e-06,
+      "loss": 2.2345,
+      "step": 26429
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7493678887484198e-06,
+      "loss": 0.7046,
+      "step": 26430
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.749320480404551e-06,
+      "loss": 2.1547,
+      "step": 26431
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7492730720606828e-06,
+      "loss": 2.0078,
+      "step": 26432
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7492256637168143e-06,
+      "loss": 2.2274,
+      "step": 26433
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7491782553729456e-06,
+      "loss": 3.3352,
+      "step": 26434
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.749130847029077e-06,
+      "loss": 3.3316,
+      "step": 26435
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7490834386852088e-06,
+      "loss": 1.9753,
+      "step": 26436
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.74903603034134e-06,
+      "loss": 2.5088,
+      "step": 26437
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7489886219974715e-06,
+      "loss": 2.8493,
+      "step": 26438
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7489412136536032e-06,
+      "loss": 2.9214,
+      "step": 26439
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.7488938053097345e-06,
+      "loss": 2.6032,
+      "step": 26440
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.748846396965866e-06,
+      "loss": 3.6429,
+      "step": 26441
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7487989886219977e-06,
+      "loss": 2.7505,
+      "step": 26442
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.748751580278129e-06,
+      "loss": 1.6214,
+      "step": 26443
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7487041719342603e-06,
+      "loss": 2.6221,
+      "step": 26444
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.748656763590392e-06,
+      "loss": 2.7885,
+      "step": 26445
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7486093552465235e-06,
+      "loss": 4.0595,
+      "step": 26446
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7485619469026547e-06,
+      "loss": 1.4421,
+      "step": 26447
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7485145385587864e-06,
+      "loss": 2.6799,
+      "step": 26448
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.748467130214918e-06,
+      "loss": 3.1294,
+      "step": 26449
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7484197218710492e-06,
+      "loss": 2.7909,
+      "step": 26450
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.748372313527181e-06,
+      "loss": 2.9893,
+      "step": 26451
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7483249051833124e-06,
+      "loss": 1.6048,
+      "step": 26452
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7482774968394437e-06,
+      "loss": 2.1544,
+      "step": 26453
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7482300884955754e-06,
+      "loss": 2.2483,
+      "step": 26454
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7481826801517069e-06,
+      "loss": 2.3025,
+      "step": 26455
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7481352718078382e-06,
+      "loss": 2.3691,
+      "step": 26456
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7480878634639695e-06,
+      "loss": 1.7086,
+      "step": 26457
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7480404551201012e-06,
+      "loss": 2.0857,
+      "step": 26458
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7479930467762326e-06,
+      "loss": 1.3771,
+      "step": 26459
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.747945638432364e-06,
+      "loss": 2.3912,
+      "step": 26460
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7478982300884956e-06,
+      "loss": 2.8462,
+      "step": 26461
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7478508217446271e-06,
+      "loss": 1.2278,
+      "step": 26462
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7478034134007584e-06,
+      "loss": 3.0221,
+      "step": 26463
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7477560050568901e-06,
+      "loss": 3.1169,
+      "step": 26464
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7477085967130216e-06,
+      "loss": 2.7087,
+      "step": 26465
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7476611883691529e-06,
+      "loss": 0.9629,
+      "step": 26466
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7476137800252846e-06,
+      "loss": 1.3795,
+      "step": 26467
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.747566371681416e-06,
+      "loss": 3.3792,
+      "step": 26468
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7475189633375474e-06,
+      "loss": 1.2974,
+      "step": 26469
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.747471554993679e-06,
+      "loss": 2.0604,
+      "step": 26470
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7474241466498103e-06,
+      "loss": 2.3519,
+      "step": 26471
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7473767383059418e-06,
+      "loss": 3.2594,
+      "step": 26472
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7473293299620735e-06,
+      "loss": 3.0614,
+      "step": 26473
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7472819216182048e-06,
+      "loss": 2.8129,
+      "step": 26474
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7472345132743363e-06,
+      "loss": 3.1839,
+      "step": 26475
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.747187104930468e-06,
+      "loss": 2.4056,
+      "step": 26476
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7471396965865993e-06,
+      "loss": 2.5137,
+      "step": 26477
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7470922882427308e-06,
+      "loss": 2.8877,
+      "step": 26478
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.747044879898862e-06,
+      "loss": 1.4972,
+      "step": 26479
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7469974715549938e-06,
+      "loss": 2.5324,
+      "step": 26480
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7469500632111253e-06,
+      "loss": 3.2751,
+      "step": 26481
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7469026548672565e-06,
+      "loss": 2.4684,
+      "step": 26482
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7468552465233882e-06,
+      "loss": 3.1717,
+      "step": 26483
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7468078381795195e-06,
+      "loss": 2.0231,
+      "step": 26484
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.746760429835651e-06,
+      "loss": 2.3664,
+      "step": 26485
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7467130214917827e-06,
+      "loss": 2.3524,
+      "step": 26486
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.746665613147914e-06,
+      "loss": 3.074,
+      "step": 26487
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7466182048040455e-06,
+      "loss": 2.0795,
+      "step": 26488
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7465707964601772e-06,
+      "loss": 3.5703,
+      "step": 26489
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7465233881163085e-06,
+      "loss": 2.4981,
+      "step": 26490
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.74647597977244e-06,
+      "loss": 3.4094,
+      "step": 26491
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7464285714285717e-06,
+      "loss": 1.8154,
+      "step": 26492
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.746381163084703e-06,
+      "loss": 1.5176,
+      "step": 26493
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7463337547408342e-06,
+      "loss": 2.0184,
+      "step": 26494
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7462863463969662e-06,
+      "loss": 2.4961,
+      "step": 26495
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7462389380530974e-06,
+      "loss": 2.4838,
+      "step": 26496
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7461915297092287e-06,
+      "loss": 3.1275,
+      "step": 26497
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7461441213653604e-06,
+      "loss": 3.3039,
+      "step": 26498
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.746096713021492e-06,
+      "loss": 3.4966,
+      "step": 26499
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7460493046776232e-06,
+      "loss": 2.2673,
+      "step": 26500
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7460018963337547e-06,
+      "loss": 3.6336,
+      "step": 26501
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7459544879898864e-06,
+      "loss": 1.4543,
+      "step": 26502
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.7459070796460177e-06,
+      "loss": 2.3754,
+      "step": 26503
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7458596713021492e-06,
+      "loss": 2.1565,
+      "step": 26504
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7458122629582809e-06,
+      "loss": 0.4876,
+      "step": 26505
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7457648546144121e-06,
+      "loss": 3.3081,
+      "step": 26506
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7457174462705434e-06,
+      "loss": 1.6065,
+      "step": 26507
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7456700379266751e-06,
+      "loss": 2.0772,
+      "step": 26508
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7456226295828066e-06,
+      "loss": 2.51,
+      "step": 26509
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.745575221238938e-06,
+      "loss": 2.7352,
+      "step": 26510
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7455278128950696e-06,
+      "loss": 0.9007,
+      "step": 26511
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.745480404551201e-06,
+      "loss": 2.9735,
+      "step": 26512
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7454329962073324e-06,
+      "loss": 3.4404,
+      "step": 26513
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.745385587863464e-06,
+      "loss": 2.4547,
+      "step": 26514
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7453381795195956e-06,
+      "loss": 2.035,
+      "step": 26515
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7452907711757269e-06,
+      "loss": 2.9821,
+      "step": 26516
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7452433628318586e-06,
+      "loss": 2.9392,
+      "step": 26517
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.74519595448799e-06,
+      "loss": 1.7952,
+      "step": 26518
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7451485461441213e-06,
+      "loss": 3.7399,
+      "step": 26519
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.745101137800253e-06,
+      "loss": 1.8339,
+      "step": 26520
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7450537294563843e-06,
+      "loss": 3.4704,
+      "step": 26521
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7450063211125158e-06,
+      "loss": 3.223,
+      "step": 26522
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.744958912768647e-06,
+      "loss": 2.9727,
+      "step": 26523
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7449115044247788e-06,
+      "loss": 2.9853,
+      "step": 26524
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7448640960809103e-06,
+      "loss": 2.4575,
+      "step": 26525
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7448166877370416e-06,
+      "loss": 2.6585,
+      "step": 26526
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7447692793931733e-06,
+      "loss": 1.6881,
+      "step": 26527
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7447218710493048e-06,
+      "loss": 2.3788,
+      "step": 26528
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.744674462705436e-06,
+      "loss": 2.3238,
+      "step": 26529
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7446270543615677e-06,
+      "loss": 2.9906,
+      "step": 26530
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7445796460176992e-06,
+      "loss": 2.2162,
+      "step": 26531
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7445322376738305e-06,
+      "loss": 2.831,
+      "step": 26532
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7444848293299622e-06,
+      "loss": 1.9169,
+      "step": 26533
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7444374209860935e-06,
+      "loss": 1.3304,
+      "step": 26534
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.744390012642225e-06,
+      "loss": 2.176,
+      "step": 26535
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7443426042983567e-06,
+      "loss": 2.5233,
+      "step": 26536
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.744295195954488e-06,
+      "loss": 1.1413,
+      "step": 26537
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7442477876106195e-06,
+      "loss": 3.6571,
+      "step": 26538
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7442003792667512e-06,
+      "loss": 3.4522,
+      "step": 26539
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7441529709228825e-06,
+      "loss": 2.3167,
+      "step": 26540
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.744105562579014e-06,
+      "loss": 2.923,
+      "step": 26541
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7440581542351457e-06,
+      "loss": 2.5485,
+      "step": 26542
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.744010745891277e-06,
+      "loss": 3.5374,
+      "step": 26543
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7439633375474084e-06,
+      "loss": 1.8912,
+      "step": 26544
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7439159292035397e-06,
+      "loss": 2.2147,
+      "step": 26545
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7438685208596714e-06,
+      "loss": 1.6105,
+      "step": 26546
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7438211125158027e-06,
+      "loss": 2.1661,
+      "step": 26547
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7437737041719342e-06,
+      "loss": 2.6284,
+      "step": 26548
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7437262958280659e-06,
+      "loss": 1.7641,
+      "step": 26549
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7436788874841972e-06,
+      "loss": 3.9591,
+      "step": 26550
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7436314791403287e-06,
+      "loss": 3.3624,
+      "step": 26551
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7435840707964604e-06,
+      "loss": 3.7174,
+      "step": 26552
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7435366624525916e-06,
+      "loss": 3.8908,
+      "step": 26553
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7434892541087231e-06,
+      "loss": 2.3919,
+      "step": 26554
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7434418457648548e-06,
+      "loss": 3.0235,
+      "step": 26555
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7433944374209861e-06,
+      "loss": 3.2635,
+      "step": 26556
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7433470290771176e-06,
+      "loss": 1.6808,
+      "step": 26557
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7432996207332493e-06,
+      "loss": 3.6306,
+      "step": 26558
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7432522123893806e-06,
+      "loss": 2.0889,
+      "step": 26559
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7432048040455119e-06,
+      "loss": 1.9152,
+      "step": 26560
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7431573957016436e-06,
+      "loss": 3.0294,
+      "step": 26561
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.743109987357775e-06,
+      "loss": 2.7593,
+      "step": 26562
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7430625790139064e-06,
+      "loss": 3.8865,
+      "step": 26563
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.743015170670038e-06,
+      "loss": 2.2971,
+      "step": 26564
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7429677623261695e-06,
+      "loss": 2.4731,
+      "step": 26565
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.7429203539823008e-06,
+      "loss": 1.9145,
+      "step": 26566
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7428729456384323e-06,
+      "loss": 2.8824,
+      "step": 26567
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.742825537294564e-06,
+      "loss": 2.6248,
+      "step": 26568
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7427781289506953e-06,
+      "loss": 2.3995,
+      "step": 26569
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7427307206068268e-06,
+      "loss": 1.8109,
+      "step": 26570
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7426833122629585e-06,
+      "loss": 1.2255,
+      "step": 26571
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7426359039190898e-06,
+      "loss": 2.6665,
+      "step": 26572
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.742588495575221e-06,
+      "loss": 2.5435,
+      "step": 26573
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7425410872313528e-06,
+      "loss": 0.8114,
+      "step": 26574
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7424936788874843e-06,
+      "loss": 3.0627,
+      "step": 26575
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7424462705436155e-06,
+      "loss": 1.7058,
+      "step": 26576
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7423988621997472e-06,
+      "loss": 4.612,
+      "step": 26577
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7423514538558787e-06,
+      "loss": 1.8559,
+      "step": 26578
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.74230404551201e-06,
+      "loss": 2.446,
+      "step": 26579
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7422566371681417e-06,
+      "loss": 2.9791,
+      "step": 26580
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7422092288242732e-06,
+      "loss": 2.1543,
+      "step": 26581
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7421618204804045e-06,
+      "loss": 2.6103,
+      "step": 26582
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7421144121365362e-06,
+      "loss": 2.3305,
+      "step": 26583
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7420670037926677e-06,
+      "loss": 2.7029,
+      "step": 26584
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.742019595448799e-06,
+      "loss": 2.6355,
+      "step": 26585
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7419721871049307e-06,
+      "loss": 2.4883,
+      "step": 26586
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.741924778761062e-06,
+      "loss": 1.9551,
+      "step": 26587
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7418773704171934e-06,
+      "loss": 3.2803,
+      "step": 26588
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7418299620733247e-06,
+      "loss": 2.285,
+      "step": 26589
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7417825537294564e-06,
+      "loss": 3.2308,
+      "step": 26590
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.741735145385588e-06,
+      "loss": 2.5556,
+      "step": 26591
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7416877370417192e-06,
+      "loss": 2.6604,
+      "step": 26592
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.741640328697851e-06,
+      "loss": 2.9602,
+      "step": 26593
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7415929203539824e-06,
+      "loss": 0.816,
+      "step": 26594
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7415455120101137e-06,
+      "loss": 3.1536,
+      "step": 26595
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7414981036662454e-06,
+      "loss": 2.6079,
+      "step": 26596
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7414506953223769e-06,
+      "loss": 3.0067,
+      "step": 26597
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7414032869785082e-06,
+      "loss": 2.0178,
+      "step": 26598
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7413558786346399e-06,
+      "loss": 2.4099,
+      "step": 26599
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7413084702907711e-06,
+      "loss": 3.1511,
+      "step": 26600
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7412610619469026e-06,
+      "loss": 3.0903,
+      "step": 26601
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7412136536030343e-06,
+      "loss": 2.1019,
+      "step": 26602
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7411662452591656e-06,
+      "loss": 2.6767,
+      "step": 26603
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7411188369152971e-06,
+      "loss": 0.6974,
+      "step": 26604
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7410714285714288e-06,
+      "loss": 1.4872,
+      "step": 26605
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.74102402022756e-06,
+      "loss": 2.2103,
+      "step": 26606
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7409766118836916e-06,
+      "loss": 2.1327,
+      "step": 26607
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7409292035398233e-06,
+      "loss": 2.5207,
+      "step": 26608
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7408817951959546e-06,
+      "loss": 2.1182,
+      "step": 26609
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.740834386852086e-06,
+      "loss": 2.239,
+      "step": 26610
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7407869785082173e-06,
+      "loss": 1.4184,
+      "step": 26611
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.740739570164349e-06,
+      "loss": 2.5128,
+      "step": 26612
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7406921618204803e-06,
+      "loss": 2.651,
+      "step": 26613
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7406447534766118e-06,
+      "loss": 1.6599,
+      "step": 26614
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7405973451327435e-06,
+      "loss": 3.8865,
+      "step": 26615
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7405499367888748e-06,
+      "loss": 2.5399,
+      "step": 26616
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7405025284450063e-06,
+      "loss": 3.4638,
+      "step": 26617
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.740455120101138e-06,
+      "loss": 2.0928,
+      "step": 26618
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7404077117572693e-06,
+      "loss": 2.8594,
+      "step": 26619
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7403603034134008e-06,
+      "loss": 1.6148,
+      "step": 26620
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7403128950695325e-06,
+      "loss": 4.2366,
+      "step": 26621
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7402654867256638e-06,
+      "loss": 2.4219,
+      "step": 26622
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.740218078381795e-06,
+      "loss": 3.2386,
+      "step": 26623
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.740170670037927e-06,
+      "loss": 2.2499,
+      "step": 26624
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7401232616940582e-06,
+      "loss": 2.654,
+      "step": 26625
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7400758533501895e-06,
+      "loss": 2.8351,
+      "step": 26626
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7400284450063212e-06,
+      "loss": 2.6459,
+      "step": 26627
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7399810366624527e-06,
+      "loss": 0.9627,
+      "step": 26628
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.739933628318584e-06,
+      "loss": 3.0043,
+      "step": 26629
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.7398862199747157e-06,
+      "loss": 1.8482,
+      "step": 26630
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7398388116308472e-06,
+      "loss": 2.0852,
+      "step": 26631
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7397914032869785e-06,
+      "loss": 3.3373,
+      "step": 26632
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.73974399494311e-06,
+      "loss": 1.6025,
+      "step": 26633
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7396965865992417e-06,
+      "loss": 0.9275,
+      "step": 26634
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.739649178255373e-06,
+      "loss": 2.6138,
+      "step": 26635
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7396017699115042e-06,
+      "loss": 2.6995,
+      "step": 26636
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.739554361567636e-06,
+      "loss": 2.2405,
+      "step": 26637
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7395069532237674e-06,
+      "loss": 3.1888,
+      "step": 26638
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7394595448798987e-06,
+      "loss": 1.9149,
+      "step": 26639
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7394121365360304e-06,
+      "loss": 3.1991,
+      "step": 26640
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.739364728192162e-06,
+      "loss": 3.3225,
+      "step": 26641
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7393173198482932e-06,
+      "loss": 1.3953,
+      "step": 26642
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7392699115044249e-06,
+      "loss": 2.1754,
+      "step": 26643
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7392225031605564e-06,
+      "loss": 3.2162,
+      "step": 26644
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7391750948166877e-06,
+      "loss": 2.8647,
+      "step": 26645
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7391276864728194e-06,
+      "loss": 2.3135,
+      "step": 26646
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7390802781289508e-06,
+      "loss": 2.9469,
+      "step": 26647
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7390328697850821e-06,
+      "loss": 3.5459,
+      "step": 26648
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7389854614412138e-06,
+      "loss": 0.897,
+      "step": 26649
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7389380530973451e-06,
+      "loss": 2.162,
+      "step": 26650
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7388906447534766e-06,
+      "loss": 2.6016,
+      "step": 26651
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7388432364096083e-06,
+      "loss": 2.4888,
+      "step": 26652
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7387958280657396e-06,
+      "loss": 1.8208,
+      "step": 26653
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.738748419721871e-06,
+      "loss": 1.9409,
+      "step": 26654
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7387010113780024e-06,
+      "loss": 2.2919,
+      "step": 26655
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.738653603034134e-06,
+      "loss": 2.1727,
+      "step": 26656
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7386061946902656e-06,
+      "loss": 1.2617,
+      "step": 26657
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7385587863463968e-06,
+      "loss": 3.2804,
+      "step": 26658
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7385113780025285e-06,
+      "loss": 3.6971,
+      "step": 26659
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.73846396965866e-06,
+      "loss": 1.713,
+      "step": 26660
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7384165613147913e-06,
+      "loss": 2.4796,
+      "step": 26661
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.738369152970923e-06,
+      "loss": 2.7776,
+      "step": 26662
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7383217446270543e-06,
+      "loss": 1.4731,
+      "step": 26663
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7382743362831858e-06,
+      "loss": 1.5801,
+      "step": 26664
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7382269279393175e-06,
+      "loss": 2.5465,
+      "step": 26665
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7381795195954488e-06,
+      "loss": 2.3337,
+      "step": 26666
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7381321112515803e-06,
+      "loss": 2.3386,
+      "step": 26667
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.738084702907712e-06,
+      "loss": 1.2688,
+      "step": 26668
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7380372945638433e-06,
+      "loss": 3.1725,
+      "step": 26669
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7379898862199747e-06,
+      "loss": 2.0169,
+      "step": 26670
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7379424778761064e-06,
+      "loss": 1.4373,
+      "step": 26671
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7378950695322377e-06,
+      "loss": 2.1883,
+      "step": 26672
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7378476611883692e-06,
+      "loss": 2.3309,
+      "step": 26673
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.737800252844501e-06,
+      "loss": 2.9474,
+      "step": 26674
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7377528445006322e-06,
+      "loss": 3.5208,
+      "step": 26675
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7377054361567635e-06,
+      "loss": 2.3187,
+      "step": 26676
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.737658027812895e-06,
+      "loss": 2.9411,
+      "step": 26677
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7376106194690267e-06,
+      "loss": 2.9536,
+      "step": 26678
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.737563211125158e-06,
+      "loss": 3.594,
+      "step": 26679
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7375158027812895e-06,
+      "loss": 2.1603,
+      "step": 26680
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7374683944374212e-06,
+      "loss": 3.3336,
+      "step": 26681
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7374209860935524e-06,
+      "loss": 2.3057,
+      "step": 26682
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.737373577749684e-06,
+      "loss": 1.171,
+      "step": 26683
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7373261694058156e-06,
+      "loss": 3.4027,
+      "step": 26684
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.737278761061947e-06,
+      "loss": 3.9532,
+      "step": 26685
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7372313527180784e-06,
+      "loss": 3.8868,
+      "step": 26686
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7371839443742101e-06,
+      "loss": 2.2129,
+      "step": 26687
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7371365360303414e-06,
+      "loss": 1.9557,
+      "step": 26688
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7370891276864727e-06,
+      "loss": 2.502,
+      "step": 26689
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7370417193426044e-06,
+      "loss": 2.4773,
+      "step": 26690
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7369943109987359e-06,
+      "loss": 3.3008,
+      "step": 26691
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7369469026548671e-06,
+      "loss": 2.6698,
+      "step": 26692
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.7368994943109989e-06,
+      "loss": 2.8463,
+      "step": 26693
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7368520859671303e-06,
+      "loss": 2.8425,
+      "step": 26694
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7368046776232616e-06,
+      "loss": 3.1447,
+      "step": 26695
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7367572692793933e-06,
+      "loss": 3.3911,
+      "step": 26696
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7367098609355248e-06,
+      "loss": 3.4109,
+      "step": 26697
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.736662452591656e-06,
+      "loss": 2.0847,
+      "step": 26698
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7366150442477876e-06,
+      "loss": 2.1498,
+      "step": 26699
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7365676359039193e-06,
+      "loss": 3.386,
+      "step": 26700
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7365202275600506e-06,
+      "loss": 2.1068,
+      "step": 26701
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7364728192161819e-06,
+      "loss": 1.5786,
+      "step": 26702
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7364254108723136e-06,
+      "loss": 1.2895,
+      "step": 26703
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.736378002528445e-06,
+      "loss": 3.096,
+      "step": 26704
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7363305941845763e-06,
+      "loss": 2.2934,
+      "step": 26705
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.736283185840708e-06,
+      "loss": 3.303,
+      "step": 26706
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7362357774968395e-06,
+      "loss": 2.24,
+      "step": 26707
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7361883691529708e-06,
+      "loss": 2.1332,
+      "step": 26708
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7361409608091025e-06,
+      "loss": 3.4598,
+      "step": 26709
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.736093552465234e-06,
+      "loss": 3.0172,
+      "step": 26710
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7360461441213653e-06,
+      "loss": 1.8105,
+      "step": 26711
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.735998735777497e-06,
+      "loss": 3.2466,
+      "step": 26712
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7359513274336285e-06,
+      "loss": 3.2367,
+      "step": 26713
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7359039190897598e-06,
+      "loss": 3.4323,
+      "step": 26714
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7358565107458915e-06,
+      "loss": 2.7513,
+      "step": 26715
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7358091024020227e-06,
+      "loss": 1.5465,
+      "step": 26716
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7357616940581542e-06,
+      "loss": 2.9413,
+      "step": 26717
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.735714285714286e-06,
+      "loss": 2.8298,
+      "step": 26718
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7356668773704172e-06,
+      "loss": 3.085,
+      "step": 26719
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7356194690265487e-06,
+      "loss": 2.1009,
+      "step": 26720
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.73557206068268e-06,
+      "loss": 2.4129,
+      "step": 26721
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7355246523388117e-06,
+      "loss": 2.1239,
+      "step": 26722
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7354772439949432e-06,
+      "loss": 2.169,
+      "step": 26723
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7354298356510745e-06,
+      "loss": 1.5164,
+      "step": 26724
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7353824273072062e-06,
+      "loss": 4.1057,
+      "step": 26725
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7353350189633377e-06,
+      "loss": 2.9466,
+      "step": 26726
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.735287610619469e-06,
+      "loss": 2.1057,
+      "step": 26727
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7352402022756007e-06,
+      "loss": 3.0846,
+      "step": 26728
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.735192793931732e-06,
+      "loss": 2.6149,
+      "step": 26729
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7351453855878634e-06,
+      "loss": 2.6642,
+      "step": 26730
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7350979772439951e-06,
+      "loss": 2.2749,
+      "step": 26731
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7350505689001264e-06,
+      "loss": 2.0709,
+      "step": 26732
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.735003160556258e-06,
+      "loss": 2.0379,
+      "step": 26733
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7349557522123896e-06,
+      "loss": 1.7357,
+      "step": 26734
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7349083438685209e-06,
+      "loss": 2.4159,
+      "step": 26735
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7348609355246524e-06,
+      "loss": 1.699,
+      "step": 26736
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.734813527180784e-06,
+      "loss": 3.0579,
+      "step": 26737
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7347661188369154e-06,
+      "loss": 3.1181,
+      "step": 26738
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7347187104930469e-06,
+      "loss": 1.3228,
+      "step": 26739
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7346713021491786e-06,
+      "loss": 1.7861,
+      "step": 26740
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7346238938053098e-06,
+      "loss": 2.1326,
+      "step": 26741
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7345764854614411e-06,
+      "loss": 2.27,
+      "step": 26742
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7345290771175726e-06,
+      "loss": 2.2441,
+      "step": 26743
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7344816687737043e-06,
+      "loss": 2.5402,
+      "step": 26744
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7344342604298356e-06,
+      "loss": 3.0685,
+      "step": 26745
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.734386852085967e-06,
+      "loss": 2.1231,
+      "step": 26746
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7343394437420988e-06,
+      "loss": 1.5734,
+      "step": 26747
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.73429203539823e-06,
+      "loss": 3.2823,
+      "step": 26748
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7342446270543616e-06,
+      "loss": 2.3517,
+      "step": 26749
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7341972187104933e-06,
+      "loss": 2.3247,
+      "step": 26750
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7341498103666246e-06,
+      "loss": 1.1247,
+      "step": 26751
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.734102402022756e-06,
+      "loss": 2.5405,
+      "step": 26752
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7340549936788877e-06,
+      "loss": 3.507,
+      "step": 26753
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.734007585335019e-06,
+      "loss": 3.4814,
+      "step": 26754
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.7339601769911503e-06,
+      "loss": 2.9938,
+      "step": 26755
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.733912768647282e-06,
+      "loss": 2.8022,
+      "step": 26756
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7338653603034135e-06,
+      "loss": 1.7552,
+      "step": 26757
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7338179519595448e-06,
+      "loss": 2.1148,
+      "step": 26758
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7337705436156765e-06,
+      "loss": 1.9041,
+      "step": 26759
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.733723135271808e-06,
+      "loss": 1.7598,
+      "step": 26760
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7336757269279393e-06,
+      "loss": 2.3832,
+      "step": 26761
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.733628318584071e-06,
+      "loss": 4.2498,
+      "step": 26762
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7335809102402025e-06,
+      "loss": 2.6543,
+      "step": 26763
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7335335018963337e-06,
+      "loss": 2.2644,
+      "step": 26764
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.733486093552465e-06,
+      "loss": 1.7344,
+      "step": 26765
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.733438685208597e-06,
+      "loss": 2.9146,
+      "step": 26766
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7333912768647282e-06,
+      "loss": 2.2489,
+      "step": 26767
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7333438685208595e-06,
+      "loss": 1.6475,
+      "step": 26768
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7332964601769912e-06,
+      "loss": 1.9008,
+      "step": 26769
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7332490518331227e-06,
+      "loss": 2.3391,
+      "step": 26770
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.733201643489254e-06,
+      "loss": 1.3693,
+      "step": 26771
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7331542351453857e-06,
+      "loss": 3.1047,
+      "step": 26772
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7331068268015172e-06,
+      "loss": 1.0781,
+      "step": 26773
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7330594184576484e-06,
+      "loss": 3.0239,
+      "step": 26774
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7330120101137802e-06,
+      "loss": 2.322,
+      "step": 26775
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7329646017699116e-06,
+      "loss": 1.7808,
+      "step": 26776
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.732917193426043e-06,
+      "loss": 3.6409,
+      "step": 26777
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7328697850821746e-06,
+      "loss": 2.5449,
+      "step": 26778
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.732822376738306e-06,
+      "loss": 2.9666,
+      "step": 26779
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7327749683944374e-06,
+      "loss": 1.6171,
+      "step": 26780
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7327275600505691e-06,
+      "loss": 2.3808,
+      "step": 26781
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7326801517067004e-06,
+      "loss": 2.0136,
+      "step": 26782
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7326327433628319e-06,
+      "loss": 2.6643,
+      "step": 26783
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7325853350189636e-06,
+      "loss": 2.4447,
+      "step": 26784
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7325379266750949e-06,
+      "loss": 3.472,
+      "step": 26785
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7324905183312264e-06,
+      "loss": 3.5053,
+      "step": 26786
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7324431099873576e-06,
+      "loss": 2.9907,
+      "step": 26787
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7323957016434893e-06,
+      "loss": 3.0936,
+      "step": 26788
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7323482932996208e-06,
+      "loss": 3.4445,
+      "step": 26789
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7323008849557521e-06,
+      "loss": 2.5542,
+      "step": 26790
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7322534766118838e-06,
+      "loss": 2.7354,
+      "step": 26791
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.732206068268015e-06,
+      "loss": 2.7903,
+      "step": 26792
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7321586599241466e-06,
+      "loss": 2.5583,
+      "step": 26793
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7321112515802783e-06,
+      "loss": 1.6105,
+      "step": 26794
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7320638432364096e-06,
+      "loss": 3.7471,
+      "step": 26795
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.732016434892541e-06,
+      "loss": 2.2376,
+      "step": 26796
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7319690265486728e-06,
+      "loss": 2.8268,
+      "step": 26797
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.731921618204804e-06,
+      "loss": 2.3187,
+      "step": 26798
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7318742098609355e-06,
+      "loss": 2.1341,
+      "step": 26799
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7318268015170672e-06,
+      "loss": 2.1679,
+      "step": 26800
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7317793931731985e-06,
+      "loss": 2.1088,
+      "step": 26801
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.73173198482933e-06,
+      "loss": 1.4602,
+      "step": 26802
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7316845764854617e-06,
+      "loss": 2.8595,
+      "step": 26803
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.731637168141593e-06,
+      "loss": 2.3739,
+      "step": 26804
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7315897597977243e-06,
+      "loss": 2.6155,
+      "step": 26805
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.731542351453856e-06,
+      "loss": 2.2175,
+      "step": 26806
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7314949431099875e-06,
+      "loss": 2.8086,
+      "step": 26807
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7314475347661188e-06,
+      "loss": 3.9925,
+      "step": 26808
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7314001264222503e-06,
+      "loss": 2.3206,
+      "step": 26809
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.731352718078382e-06,
+      "loss": 1.2619,
+      "step": 26810
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7313053097345132e-06,
+      "loss": 2.0576,
+      "step": 26811
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7312579013906447e-06,
+      "loss": 2.761,
+      "step": 26812
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7312104930467764e-06,
+      "loss": 2.6525,
+      "step": 26813
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7311630847029077e-06,
+      "loss": 2.6394,
+      "step": 26814
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7311156763590392e-06,
+      "loss": 2.8126,
+      "step": 26815
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.731068268015171e-06,
+      "loss": 2.5432,
+      "step": 26816
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7310208596713022e-06,
+      "loss": 2.8229,
+      "step": 26817
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7309734513274335e-06,
+      "loss": 1.9677,
+      "step": 26818
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7309260429835652e-06,
+      "loss": 1.4554,
+      "step": 26819
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.7308786346396967e-06,
+      "loss": 3.7913,
+      "step": 26820
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.730831226295828e-06,
+      "loss": 2.9341,
+      "step": 26821
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7307838179519596e-06,
+      "loss": 1.5073,
+      "step": 26822
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7307364096080911e-06,
+      "loss": 2.9838,
+      "step": 26823
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7306890012642224e-06,
+      "loss": 4.0248,
+      "step": 26824
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7306415929203541e-06,
+      "loss": 2.8185,
+      "step": 26825
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7305941845764856e-06,
+      "loss": 3.0283,
+      "step": 26826
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.730546776232617e-06,
+      "loss": 1.7263,
+      "step": 26827
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7304993678887486e-06,
+      "loss": 2.4117,
+      "step": 26828
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.73045195954488e-06,
+      "loss": 1.838,
+      "step": 26829
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7304045512010114e-06,
+      "loss": 1.9789,
+      "step": 26830
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7303571428571427e-06,
+      "loss": 2.9647,
+      "step": 26831
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7303097345132744e-06,
+      "loss": 2.54,
+      "step": 26832
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7302623261694059e-06,
+      "loss": 4.5471,
+      "step": 26833
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7302149178255371e-06,
+      "loss": 1.9501,
+      "step": 26834
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7301675094816688e-06,
+      "loss": 0.8372,
+      "step": 26835
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7301201011378003e-06,
+      "loss": 2.042,
+      "step": 26836
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7300726927939316e-06,
+      "loss": 1.9042,
+      "step": 26837
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7300252844500633e-06,
+      "loss": 3.7285,
+      "step": 26838
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7299778761061948e-06,
+      "loss": 3.3647,
+      "step": 26839
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.729930467762326e-06,
+      "loss": 2.5376,
+      "step": 26840
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7298830594184578e-06,
+      "loss": 2.6477,
+      "step": 26841
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7298356510745893e-06,
+      "loss": 3.9189,
+      "step": 26842
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7297882427307206e-06,
+      "loss": 2.5387,
+      "step": 26843
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7297408343868523e-06,
+      "loss": 2.1919,
+      "step": 26844
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7296934260429835e-06,
+      "loss": 2.4223,
+      "step": 26845
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.729646017699115e-06,
+      "loss": 2.2694,
+      "step": 26846
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7295986093552467e-06,
+      "loss": 2.4559,
+      "step": 26847
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.729551201011378e-06,
+      "loss": 3.0802,
+      "step": 26848
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7295037926675095e-06,
+      "loss": 2.7047,
+      "step": 26849
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7294563843236412e-06,
+      "loss": 1.9644,
+      "step": 26850
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7294089759797725e-06,
+      "loss": 3.5331,
+      "step": 26851
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.729361567635904e-06,
+      "loss": 3.0077,
+      "step": 26852
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7293141592920353e-06,
+      "loss": 3.5734,
+      "step": 26853
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.729266750948167e-06,
+      "loss": 2.3807,
+      "step": 26854
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7292193426042985e-06,
+      "loss": 1.9478,
+      "step": 26855
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7291719342604297e-06,
+      "loss": 1.7897,
+      "step": 26856
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7291245259165615e-06,
+      "loss": 2.7336,
+      "step": 26857
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7290771175726927e-06,
+      "loss": 2.3495,
+      "step": 26858
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7290297092288242e-06,
+      "loss": 2.9884,
+      "step": 26859
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.728982300884956e-06,
+      "loss": 2.7981,
+      "step": 26860
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7289348925410872e-06,
+      "loss": 2.055,
+      "step": 26861
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7288874841972187e-06,
+      "loss": 1.8944,
+      "step": 26862
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7288400758533504e-06,
+      "loss": 3.5424,
+      "step": 26863
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7287926675094817e-06,
+      "loss": 3.4541,
+      "step": 26864
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7287452591656132e-06,
+      "loss": 2.1461,
+      "step": 26865
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7286978508217449e-06,
+      "loss": 2.5557,
+      "step": 26866
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7286504424778762e-06,
+      "loss": 3.0203,
+      "step": 26867
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7286030341340077e-06,
+      "loss": 1.5573,
+      "step": 26868
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7285556257901394e-06,
+      "loss": 1.4828,
+      "step": 26869
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7285082174462706e-06,
+      "loss": 3.2499,
+      "step": 26870
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.728460809102402e-06,
+      "loss": 2.1975,
+      "step": 26871
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7284134007585334e-06,
+      "loss": 2.7403,
+      "step": 26872
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7283659924146651e-06,
+      "loss": 2.9493,
+      "step": 26873
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7283185840707964e-06,
+      "loss": 1.6508,
+      "step": 26874
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7282711757269279e-06,
+      "loss": 3.8507,
+      "step": 26875
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7282237673830596e-06,
+      "loss": 2.099,
+      "step": 26876
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7281763590391909e-06,
+      "loss": 3.2321,
+      "step": 26877
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7281289506953224e-06,
+      "loss": 1.6827,
+      "step": 26878
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.728081542351454e-06,
+      "loss": 2.3121,
+      "step": 26879
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7280341340075853e-06,
+      "loss": 4.0188,
+      "step": 26880
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7279867256637168e-06,
+      "loss": 2.379,
+      "step": 26881
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7279393173198485e-06,
+      "loss": 2.1309,
+      "step": 26882
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.7278919089759798e-06,
+      "loss": 3.0549,
+      "step": 26883
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7278445006321111e-06,
+      "loss": 2.866,
+      "step": 26884
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7277970922882428e-06,
+      "loss": 1.8341,
+      "step": 26885
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7277496839443743e-06,
+      "loss": 3.0585,
+      "step": 26886
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7277022756005056e-06,
+      "loss": 1.8188,
+      "step": 26887
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7276548672566373e-06,
+      "loss": 2.5066,
+      "step": 26888
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7276074589127688e-06,
+      "loss": 3.7711,
+      "step": 26889
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7275600505689e-06,
+      "loss": 2.7525,
+      "step": 26890
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7275126422250318e-06,
+      "loss": 1.5115,
+      "step": 26891
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7274652338811633e-06,
+      "loss": 2.5718,
+      "step": 26892
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7274178255372945e-06,
+      "loss": 2.5966,
+      "step": 26893
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7273704171934258e-06,
+      "loss": 2.6479,
+      "step": 26894
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7273230088495577e-06,
+      "loss": 3.0213,
+      "step": 26895
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.727275600505689e-06,
+      "loss": 2.3218,
+      "step": 26896
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7272281921618203e-06,
+      "loss": 1.6791,
+      "step": 26897
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.727180783817952e-06,
+      "loss": 3.2307,
+      "step": 26898
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7271333754740835e-06,
+      "loss": 2.9399,
+      "step": 26899
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7270859671302148e-06,
+      "loss": 2.5225,
+      "step": 26900
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7270385587863465e-06,
+      "loss": 2.5596,
+      "step": 26901
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.726991150442478e-06,
+      "loss": 3.3101,
+      "step": 26902
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7269437420986092e-06,
+      "loss": 2.0788,
+      "step": 26903
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.726896333754741e-06,
+      "loss": 1.8645,
+      "step": 26904
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7268489254108724e-06,
+      "loss": 2.6046,
+      "step": 26905
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7268015170670037e-06,
+      "loss": 1.8767,
+      "step": 26906
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7267541087231354e-06,
+      "loss": 3.1306,
+      "step": 26907
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7267067003792667e-06,
+      "loss": 2.5033,
+      "step": 26908
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7266592920353982e-06,
+      "loss": 3.1259,
+      "step": 26909
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.72661188369153e-06,
+      "loss": 1.6254,
+      "step": 26910
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7265644753476612e-06,
+      "loss": 2.3957,
+      "step": 26911
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7265170670037927e-06,
+      "loss": 2.6887,
+      "step": 26912
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7264696586599244e-06,
+      "loss": 1.8245,
+      "step": 26913
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7264222503160557e-06,
+      "loss": 3.24,
+      "step": 26914
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7263748419721872e-06,
+      "loss": 3.0408,
+      "step": 26915
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7263274336283184e-06,
+      "loss": 1.8738,
+      "step": 26916
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7262800252844501e-06,
+      "loss": 2.0699,
+      "step": 26917
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7262326169405816e-06,
+      "loss": 1.6177,
+      "step": 26918
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.726185208596713e-06,
+      "loss": 1.3371,
+      "step": 26919
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7261378002528446e-06,
+      "loss": 1.5347,
+      "step": 26920
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7260903919089759e-06,
+      "loss": 2.0736,
+      "step": 26921
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7260429835651074e-06,
+      "loss": 1.774,
+      "step": 26922
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.725995575221239e-06,
+      "loss": 3.0241,
+      "step": 26923
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7259481668773704e-06,
+      "loss": 1.7412,
+      "step": 26924
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7259007585335019e-06,
+      "loss": 2.6694,
+      "step": 26925
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7258533501896336e-06,
+      "loss": 2.8371,
+      "step": 26926
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7258059418457648e-06,
+      "loss": 2.0505,
+      "step": 26927
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7257585335018963e-06,
+      "loss": 2.9041,
+      "step": 26928
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.725711125158028e-06,
+      "loss": 2.2294,
+      "step": 26929
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7256637168141593e-06,
+      "loss": 2.2926,
+      "step": 26930
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7256163084702908e-06,
+      "loss": 4.947,
+      "step": 26931
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7255689001264225e-06,
+      "loss": 1.7698,
+      "step": 26932
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7255214917825538e-06,
+      "loss": 3.6766,
+      "step": 26933
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.725474083438685e-06,
+      "loss": 1.4265,
+      "step": 26934
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7254266750948168e-06,
+      "loss": 2.5257,
+      "step": 26935
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7253792667509483e-06,
+      "loss": 2.8672,
+      "step": 26936
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7253318584070796e-06,
+      "loss": 2.6386,
+      "step": 26937
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.725284450063211e-06,
+      "loss": 3.7321,
+      "step": 26938
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7252370417193428e-06,
+      "loss": 2.4632,
+      "step": 26939
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.725189633375474e-06,
+      "loss": 1.7174,
+      "step": 26940
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7251422250316055e-06,
+      "loss": 3.3585,
+      "step": 26941
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7250948166877372e-06,
+      "loss": 3.5708,
+      "step": 26942
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7250474083438685e-06,
+      "loss": 2.8238,
+      "step": 26943
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.725e-06,
+      "loss": 3.5802,
+      "step": 26944
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.7249525916561317e-06,
+      "loss": 3.404,
+      "step": 26945
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.724905183312263e-06,
+      "loss": 2.5501,
+      "step": 26946
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7248577749683943e-06,
+      "loss": 3.8168,
+      "step": 26947
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.724810366624526e-06,
+      "loss": 2.3478,
+      "step": 26948
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7247629582806575e-06,
+      "loss": 1.5653,
+      "step": 26949
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7247155499367887e-06,
+      "loss": 2.8133,
+      "step": 26950
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7246681415929204e-06,
+      "loss": 3.4727,
+      "step": 26951
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.724620733249052e-06,
+      "loss": 1.461,
+      "step": 26952
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7245733249051832e-06,
+      "loss": 2.7462,
+      "step": 26953
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.724525916561315e-06,
+      "loss": 2.25,
+      "step": 26954
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7244785082174464e-06,
+      "loss": 2.9924,
+      "step": 26955
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7244310998735777e-06,
+      "loss": 1.9333,
+      "step": 26956
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7243836915297094e-06,
+      "loss": 2.9464,
+      "step": 26957
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7243362831858409e-06,
+      "loss": 3.0584,
+      "step": 26958
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7242888748419722e-06,
+      "loss": 2.4301,
+      "step": 26959
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7242414664981035e-06,
+      "loss": 2.7487,
+      "step": 26960
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7241940581542352e-06,
+      "loss": 2.2358,
+      "step": 26961
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7241466498103666e-06,
+      "loss": 2.9984,
+      "step": 26962
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.724099241466498e-06,
+      "loss": 1.7643,
+      "step": 26963
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7240518331226296e-06,
+      "loss": 1.962,
+      "step": 26964
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7240044247787611e-06,
+      "loss": 3.8057,
+      "step": 26965
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7239570164348924e-06,
+      "loss": 3.9348,
+      "step": 26966
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7239096080910241e-06,
+      "loss": 2.472,
+      "step": 26967
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7238621997471556e-06,
+      "loss": 3.0231,
+      "step": 26968
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7238147914032869e-06,
+      "loss": 0.8463,
+      "step": 26969
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7237673830594186e-06,
+      "loss": 2.1461,
+      "step": 26970
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.72371997471555e-06,
+      "loss": 2.466,
+      "step": 26971
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7236725663716814e-06,
+      "loss": 1.6984,
+      "step": 26972
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.723625158027813e-06,
+      "loss": 1.3151,
+      "step": 26973
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7235777496839443e-06,
+      "loss": 2.1562,
+      "step": 26974
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7235303413400758e-06,
+      "loss": 3.3523,
+      "step": 26975
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7234829329962075e-06,
+      "loss": 1.9742,
+      "step": 26976
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7234355246523388e-06,
+      "loss": 1.9873,
+      "step": 26977
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7233881163084703e-06,
+      "loss": 2.1094,
+      "step": 26978
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.723340707964602e-06,
+      "loss": 1.4105,
+      "step": 26979
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7232932996207333e-06,
+      "loss": 2.9446,
+      "step": 26980
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7232458912768648e-06,
+      "loss": 4.8814,
+      "step": 26981
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.723198482932996e-06,
+      "loss": 1.8029,
+      "step": 26982
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7231510745891278e-06,
+      "loss": 2.9625,
+      "step": 26983
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7231036662452593e-06,
+      "loss": 3.4909,
+      "step": 26984
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7230562579013905e-06,
+      "loss": 2.7496,
+      "step": 26985
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7230088495575222e-06,
+      "loss": 3.4183,
+      "step": 26986
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7229614412136535e-06,
+      "loss": 2.3925,
+      "step": 26987
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.722914032869785e-06,
+      "loss": 2.5548,
+      "step": 26988
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7228666245259167e-06,
+      "loss": 2.8256,
+      "step": 26989
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.722819216182048e-06,
+      "loss": 2.1566,
+      "step": 26990
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7227718078381795e-06,
+      "loss": 1.7571,
+      "step": 26991
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7227243994943112e-06,
+      "loss": 3.7234,
+      "step": 26992
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7226769911504425e-06,
+      "loss": 2.6217,
+      "step": 26993
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.722629582806574e-06,
+      "loss": 1.9921,
+      "step": 26994
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7225821744627057e-06,
+      "loss": 2.1758,
+      "step": 26995
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.722534766118837e-06,
+      "loss": 3.2657,
+      "step": 26996
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7224873577749685e-06,
+      "loss": 2.5831,
+      "step": 26997
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7224399494311002e-06,
+      "loss": 3.4484,
+      "step": 26998
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7223925410872314e-06,
+      "loss": 2.6732,
+      "step": 26999
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7223451327433627e-06,
+      "loss": 2.4452,
+      "step": 27000
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7222977243994944e-06,
+      "loss": 1.4453,
+      "step": 27001
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.722250316055626e-06,
+      "loss": 2.8742,
+      "step": 27002
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7222029077117572e-06,
+      "loss": 2.6547,
+      "step": 27003
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7221554993678887e-06,
+      "loss": 3.2402,
+      "step": 27004
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7221080910240204e-06,
+      "loss": 2.4282,
+      "step": 27005
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7220606826801517e-06,
+      "loss": 2.7799,
+      "step": 27006
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7220132743362832e-06,
+      "loss": 2.0969,
+      "step": 27007
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7219658659924149e-06,
+      "loss": 2.3106,
+      "step": 27008
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7219184576485461e-06,
+      "loss": 2.7415,
+      "step": 27009
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.7218710493046776e-06,
+      "loss": 1.5777,
+      "step": 27010
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7218236409608093e-06,
+      "loss": 1.6069,
+      "step": 27011
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7217762326169406e-06,
+      "loss": 2.4437,
+      "step": 27012
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.721728824273072e-06,
+      "loss": 2.9151,
+      "step": 27013
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7216814159292036e-06,
+      "loss": 1.6472,
+      "step": 27014
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.721634007585335e-06,
+      "loss": 2.3469,
+      "step": 27015
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7215865992414664e-06,
+      "loss": 1.229,
+      "step": 27016
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.721539190897598e-06,
+      "loss": 2.4345,
+      "step": 27017
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7214917825537296e-06,
+      "loss": 2.774,
+      "step": 27018
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7214443742098609e-06,
+      "loss": 2.7049,
+      "step": 27019
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7213969658659926e-06,
+      "loss": 2.2194,
+      "step": 27020
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.721349557522124e-06,
+      "loss": 4.3191,
+      "step": 27021
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7213021491782553e-06,
+      "loss": 2.029,
+      "step": 27022
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.721254740834387e-06,
+      "loss": 2.0877,
+      "step": 27023
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7212073324905185e-06,
+      "loss": 2.7738,
+      "step": 27024
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7211599241466498e-06,
+      "loss": 1.2203,
+      "step": 27025
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.721112515802781e-06,
+      "loss": 2.9937,
+      "step": 27026
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7210651074589128e-06,
+      "loss": 2.4087,
+      "step": 27027
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7210176991150443e-06,
+      "loss": 3.5709,
+      "step": 27028
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7209702907711756e-06,
+      "loss": 1.4622,
+      "step": 27029
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7209228824273073e-06,
+      "loss": 3.2261,
+      "step": 27030
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7208754740834388e-06,
+      "loss": 3.4763,
+      "step": 27031
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.72082806573957e-06,
+      "loss": 2.5466,
+      "step": 27032
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7207806573957017e-06,
+      "loss": 2.628,
+      "step": 27033
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7207332490518332e-06,
+      "loss": 1.6122,
+      "step": 27034
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7206858407079645e-06,
+      "loss": 3.5647,
+      "step": 27035
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7206384323640962e-06,
+      "loss": 2.3903,
+      "step": 27036
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7205910240202277e-06,
+      "loss": 3.0772,
+      "step": 27037
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.720543615676359e-06,
+      "loss": 3.2477,
+      "step": 27038
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7204962073324907e-06,
+      "loss": 1.0272,
+      "step": 27039
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.720448798988622e-06,
+      "loss": 3.0505,
+      "step": 27040
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7204013906447535e-06,
+      "loss": 2.8696,
+      "step": 27041
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7203539823008852e-06,
+      "loss": 2.2519,
+      "step": 27042
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7203065739570165e-06,
+      "loss": 1.6316,
+      "step": 27043
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.720259165613148e-06,
+      "loss": 2.6681,
+      "step": 27044
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7202117572692797e-06,
+      "loss": 2.0055,
+      "step": 27045
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.720164348925411e-06,
+      "loss": 0.973,
+      "step": 27046
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7201169405815424e-06,
+      "loss": 3.6345,
+      "step": 27047
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7200695322376737e-06,
+      "loss": 3.1656,
+      "step": 27048
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7200221238938054e-06,
+      "loss": 1.908,
+      "step": 27049
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7199747155499367e-06,
+      "loss": 1.7384,
+      "step": 27050
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7199273072060682e-06,
+      "loss": 1.8161,
+      "step": 27051
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7198798988621999e-06,
+      "loss": 2.0191,
+      "step": 27052
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7198324905183312e-06,
+      "loss": 1.9789,
+      "step": 27053
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7197850821744627e-06,
+      "loss": 3.7337,
+      "step": 27054
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7197376738305944e-06,
+      "loss": 2.6417,
+      "step": 27055
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7196902654867256e-06,
+      "loss": 1.8982,
+      "step": 27056
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7196428571428571e-06,
+      "loss": 2.97,
+      "step": 27057
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7195954487989888e-06,
+      "loss": 2.2896,
+      "step": 27058
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7195480404551201e-06,
+      "loss": 1.9361,
+      "step": 27059
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7195006321112516e-06,
+      "loss": 1.6137,
+      "step": 27060
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7194532237673833e-06,
+      "loss": 1.0894,
+      "step": 27061
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7194058154235146e-06,
+      "loss": 2.4974,
+      "step": 27062
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7193584070796459e-06,
+      "loss": 1.8163,
+      "step": 27063
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7193109987357776e-06,
+      "loss": 2.435,
+      "step": 27064
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.719263590391909e-06,
+      "loss": 1.2959,
+      "step": 27065
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7192161820480404e-06,
+      "loss": 3.5785,
+      "step": 27066
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.719168773704172e-06,
+      "loss": 1.7624,
+      "step": 27067
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7191213653603035e-06,
+      "loss": 1.9941,
+      "step": 27068
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7190739570164348e-06,
+      "loss": 1.4793,
+      "step": 27069
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7190265486725663e-06,
+      "loss": 3.2457,
+      "step": 27070
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.718979140328698e-06,
+      "loss": 2.4581,
+      "step": 27071
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7189317319848293e-06,
+      "loss": 2.4918,
+      "step": 27072
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.7188843236409608e-06,
+      "loss": 3.5393,
+      "step": 27073
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7188369152970925e-06,
+      "loss": 2.4992,
+      "step": 27074
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7187895069532238e-06,
+      "loss": 2.6163,
+      "step": 27075
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.718742098609355e-06,
+      "loss": 4.0352,
+      "step": 27076
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7186946902654868e-06,
+      "loss": 2.6002,
+      "step": 27077
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7186472819216183e-06,
+      "loss": 2.1057,
+      "step": 27078
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7185998735777495e-06,
+      "loss": 2.5945,
+      "step": 27079
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7185524652338812e-06,
+      "loss": 3.1214,
+      "step": 27080
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7185050568900127e-06,
+      "loss": 2.7902,
+      "step": 27081
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.718457648546144e-06,
+      "loss": 1.4779,
+      "step": 27082
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7184102402022757e-06,
+      "loss": 2.5042,
+      "step": 27083
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7183628318584072e-06,
+      "loss": 1.64,
+      "step": 27084
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7183154235145385e-06,
+      "loss": 3.0953,
+      "step": 27085
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7182680151706702e-06,
+      "loss": 2.1382,
+      "step": 27086
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7182206068268017e-06,
+      "loss": 0.894,
+      "step": 27087
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.718173198482933e-06,
+      "loss": 2.4205,
+      "step": 27088
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7181257901390647e-06,
+      "loss": 2.8788,
+      "step": 27089
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.718078381795196e-06,
+      "loss": 3.2619,
+      "step": 27090
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7180309734513274e-06,
+      "loss": 2.2981,
+      "step": 27091
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7179835651074587e-06,
+      "loss": 1.9532,
+      "step": 27092
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7179361567635904e-06,
+      "loss": 2.6317,
+      "step": 27093
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.717888748419722e-06,
+      "loss": 1.7492,
+      "step": 27094
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7178413400758532e-06,
+      "loss": 3.6552,
+      "step": 27095
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.717793931731985e-06,
+      "loss": 3.0652,
+      "step": 27096
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7177465233881164e-06,
+      "loss": 2.6365,
+      "step": 27097
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7176991150442477e-06,
+      "loss": 3.1657,
+      "step": 27098
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7176517067003794e-06,
+      "loss": 3.2687,
+      "step": 27099
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7176042983565109e-06,
+      "loss": 2.4186,
+      "step": 27100
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7175568900126422e-06,
+      "loss": 2.5436,
+      "step": 27101
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7175094816687739e-06,
+      "loss": 2.3321,
+      "step": 27102
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7174620733249051e-06,
+      "loss": 2.1474,
+      "step": 27103
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7174146649810366e-06,
+      "loss": 2.7934,
+      "step": 27104
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7173672566371683e-06,
+      "loss": 2.0008,
+      "step": 27105
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7173198482932996e-06,
+      "loss": 2.9703,
+      "step": 27106
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7172724399494311e-06,
+      "loss": 4.2817,
+      "step": 27107
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7172250316055628e-06,
+      "loss": 2.6161,
+      "step": 27108
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.717177623261694e-06,
+      "loss": 2.3022,
+      "step": 27109
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7171302149178256e-06,
+      "loss": 1.9043,
+      "step": 27110
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7170828065739573e-06,
+      "loss": 3.0837,
+      "step": 27111
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7170353982300886e-06,
+      "loss": 1.9526,
+      "step": 27112
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.71698798988622e-06,
+      "loss": 1.8924,
+      "step": 27113
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7169405815423513e-06,
+      "loss": 1.8058,
+      "step": 27114
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.716893173198483e-06,
+      "loss": 3.7653,
+      "step": 27115
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7168457648546143e-06,
+      "loss": 2.1269,
+      "step": 27116
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7167983565107458e-06,
+      "loss": 3.3127,
+      "step": 27117
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7167509481668775e-06,
+      "loss": 2.5204,
+      "step": 27118
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7167035398230088e-06,
+      "loss": 2.0278,
+      "step": 27119
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7166561314791403e-06,
+      "loss": 2.0827,
+      "step": 27120
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.716608723135272e-06,
+      "loss": 1.8175,
+      "step": 27121
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7165613147914033e-06,
+      "loss": 2.074,
+      "step": 27122
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7165139064475348e-06,
+      "loss": 3.1866,
+      "step": 27123
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7164664981036665e-06,
+      "loss": 2.3957,
+      "step": 27124
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7164190897597978e-06,
+      "loss": 1.4907,
+      "step": 27125
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7163716814159292e-06,
+      "loss": 2.5098,
+      "step": 27126
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.716324273072061e-06,
+      "loss": 2.2753,
+      "step": 27127
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7162768647281922e-06,
+      "loss": 1.5128,
+      "step": 27128
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7162294563843235e-06,
+      "loss": 1.7548,
+      "step": 27129
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7161820480404552e-06,
+      "loss": 1.903,
+      "step": 27130
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7161346396965867e-06,
+      "loss": 2.0907,
+      "step": 27131
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.716087231352718e-06,
+      "loss": 2.1056,
+      "step": 27132
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7160398230088497e-06,
+      "loss": 2.7298,
+      "step": 27133
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7159924146649812e-06,
+      "loss": 2.0828,
+      "step": 27134
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.7159450063211125e-06,
+      "loss": 2.3774,
+      "step": 27135
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.715897597977244e-06,
+      "loss": 2.8101,
+      "step": 27136
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7158501896333757e-06,
+      "loss": 2.7089,
+      "step": 27137
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.715802781289507e-06,
+      "loss": 3.0039,
+      "step": 27138
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7157553729456384e-06,
+      "loss": 2.4988,
+      "step": 27139
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7157079646017701e-06,
+      "loss": 2.025,
+      "step": 27140
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7156605562579014e-06,
+      "loss": 2.2003,
+      "step": 27141
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7156131479140327e-06,
+      "loss": 3.0128,
+      "step": 27142
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7155657395701644e-06,
+      "loss": 1.3657,
+      "step": 27143
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.715518331226296e-06,
+      "loss": 3.4189,
+      "step": 27144
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7154709228824272e-06,
+      "loss": 2.7261,
+      "step": 27145
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7154235145385589e-06,
+      "loss": 1.5002,
+      "step": 27146
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7153761061946904e-06,
+      "loss": 1.6544,
+      "step": 27147
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7153286978508217e-06,
+      "loss": 1.4465,
+      "step": 27148
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7152812895069534e-06,
+      "loss": 3.1101,
+      "step": 27149
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7152338811630848e-06,
+      "loss": 3.6915,
+      "step": 27150
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7151864728192161e-06,
+      "loss": 2.0488,
+      "step": 27151
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7151390644753478e-06,
+      "loss": 2.6498,
+      "step": 27152
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7150916561314793e-06,
+      "loss": 1.0409,
+      "step": 27153
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7150442477876106e-06,
+      "loss": 2.2736,
+      "step": 27154
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7149968394437423e-06,
+      "loss": 2.1304,
+      "step": 27155
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7149494310998736e-06,
+      "loss": 0.9898,
+      "step": 27156
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.714902022756005e-06,
+      "loss": 2.9209,
+      "step": 27157
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7148546144121364e-06,
+      "loss": 2.0562,
+      "step": 27158
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.714807206068268e-06,
+      "loss": 2.2288,
+      "step": 27159
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7147597977243996e-06,
+      "loss": 2.1787,
+      "step": 27160
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7147123893805308e-06,
+      "loss": 2.4422,
+      "step": 27161
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7146649810366625e-06,
+      "loss": 2.4861,
+      "step": 27162
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.714617572692794e-06,
+      "loss": 2.2886,
+      "step": 27163
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7145701643489253e-06,
+      "loss": 1.0179,
+      "step": 27164
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.714522756005057e-06,
+      "loss": 2.0396,
+      "step": 27165
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7144753476611885e-06,
+      "loss": 1.673,
+      "step": 27166
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7144279393173198e-06,
+      "loss": 5.2825,
+      "step": 27167
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7143805309734515e-06,
+      "loss": 2.2834,
+      "step": 27168
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7143331226295828e-06,
+      "loss": 3.6248,
+      "step": 27169
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7142857142857143e-06,
+      "loss": 2.6689,
+      "step": 27170
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.714238305941846e-06,
+      "loss": 1.6661,
+      "step": 27171
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7141908975979773e-06,
+      "loss": 2.0494,
+      "step": 27172
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7141434892541087e-06,
+      "loss": 2.9742,
+      "step": 27173
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7140960809102404e-06,
+      "loss": 1.9974,
+      "step": 27174
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7140486725663717e-06,
+      "loss": 3.3382,
+      "step": 27175
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7140012642225032e-06,
+      "loss": 2.9117,
+      "step": 27176
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.713953855878635e-06,
+      "loss": 2.7487,
+      "step": 27177
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7139064475347662e-06,
+      "loss": 2.9184,
+      "step": 27178
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7138590391908975e-06,
+      "loss": 2.745,
+      "step": 27179
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.713811630847029e-06,
+      "loss": 3.4463,
+      "step": 27180
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7137642225031607e-06,
+      "loss": 2.4118,
+      "step": 27181
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.713716814159292e-06,
+      "loss": 1.3444,
+      "step": 27182
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7136694058154235e-06,
+      "loss": 1.9447,
+      "step": 27183
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7136219974715552e-06,
+      "loss": 3.2043,
+      "step": 27184
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7135745891276864e-06,
+      "loss": 2.139,
+      "step": 27185
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.713527180783818e-06,
+      "loss": 1.8074,
+      "step": 27186
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7134797724399496e-06,
+      "loss": 3.0634,
+      "step": 27187
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.713432364096081e-06,
+      "loss": 1.8592,
+      "step": 27188
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7133849557522124e-06,
+      "loss": 2.9306,
+      "step": 27189
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7133375474083441e-06,
+      "loss": 2.0107,
+      "step": 27190
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7132901390644754e-06,
+      "loss": 2.998,
+      "step": 27191
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7132427307206067e-06,
+      "loss": 2.1985,
+      "step": 27192
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7131953223767384e-06,
+      "loss": 3.2776,
+      "step": 27193
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7131479140328699e-06,
+      "loss": 3.895,
+      "step": 27194
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7131005056890011e-06,
+      "loss": 1.8037,
+      "step": 27195
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7130530973451329e-06,
+      "loss": 1.1035,
+      "step": 27196
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7130056890012643e-06,
+      "loss": 2.3844,
+      "step": 27197
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7129582806573956e-06,
+      "loss": 3.1211,
+      "step": 27198
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7129108723135273e-06,
+      "loss": 2.0367,
+      "step": 27199
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.7128634639696588e-06,
+      "loss": 1.7204,
+      "step": 27200
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.71281605562579e-06,
+      "loss": 1.8543,
+      "step": 27201
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7127686472819216e-06,
+      "loss": 1.4796,
+      "step": 27202
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7127212389380533e-06,
+      "loss": 1.548,
+      "step": 27203
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7126738305941846e-06,
+      "loss": 1.5853,
+      "step": 27204
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7126264222503159e-06,
+      "loss": 1.5429,
+      "step": 27205
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7125790139064476e-06,
+      "loss": 2.7776,
+      "step": 27206
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.712531605562579e-06,
+      "loss": 3.4997,
+      "step": 27207
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7124841972187103e-06,
+      "loss": 1.9747,
+      "step": 27208
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.712436788874842e-06,
+      "loss": 1.6579,
+      "step": 27209
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7123893805309735e-06,
+      "loss": 2.1428,
+      "step": 27210
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7123419721871048e-06,
+      "loss": 2.5466,
+      "step": 27211
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7122945638432365e-06,
+      "loss": 2.5955,
+      "step": 27212
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.712247155499368e-06,
+      "loss": 2.5794,
+      "step": 27213
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7121997471554993e-06,
+      "loss": 3.0523,
+      "step": 27214
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.712152338811631e-06,
+      "loss": 2.6611,
+      "step": 27215
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7121049304677625e-06,
+      "loss": 1.7069,
+      "step": 27216
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7120575221238938e-06,
+      "loss": 2.8594,
+      "step": 27217
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7120101137800255e-06,
+      "loss": 1.7784,
+      "step": 27218
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7119627054361568e-06,
+      "loss": 1.8474,
+      "step": 27219
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7119152970922882e-06,
+      "loss": 1.9701,
+      "step": 27220
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.71186788874842e-06,
+      "loss": 2.6376,
+      "step": 27221
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7118204804045512e-06,
+      "loss": 2.2836,
+      "step": 27222
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7117730720606827e-06,
+      "loss": 1.8502,
+      "step": 27223
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.711725663716814e-06,
+      "loss": 1.5996,
+      "step": 27224
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7116782553729457e-06,
+      "loss": 1.8112,
+      "step": 27225
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7116308470290772e-06,
+      "loss": 3.4877,
+      "step": 27226
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7115834386852085e-06,
+      "loss": 1.8211,
+      "step": 27227
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7115360303413402e-06,
+      "loss": 3.7261,
+      "step": 27228
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7114886219974717e-06,
+      "loss": 1.4107,
+      "step": 27229
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.711441213653603e-06,
+      "loss": 1.5063,
+      "step": 27230
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7113938053097347e-06,
+      "loss": 3.0404,
+      "step": 27231
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.711346396965866e-06,
+      "loss": 2.3932,
+      "step": 27232
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7112989886219974e-06,
+      "loss": 2.3043,
+      "step": 27233
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7112515802781291e-06,
+      "loss": 2.7863,
+      "step": 27234
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7112041719342604e-06,
+      "loss": 2.2251,
+      "step": 27235
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.711156763590392e-06,
+      "loss": 2.5104,
+      "step": 27236
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7111093552465236e-06,
+      "loss": 1.4252,
+      "step": 27237
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7110619469026549e-06,
+      "loss": 2.9717,
+      "step": 27238
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7110145385587864e-06,
+      "loss": 3.4436,
+      "step": 27239
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.710967130214918e-06,
+      "loss": 2.8267,
+      "step": 27240
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7109197218710494e-06,
+      "loss": 2.6184,
+      "step": 27241
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7108723135271809e-06,
+      "loss": 2.2175,
+      "step": 27242
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7108249051833126e-06,
+      "loss": 2.6605,
+      "step": 27243
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7107774968394438e-06,
+      "loss": 2.5104,
+      "step": 27244
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7107300884955751e-06,
+      "loss": 2.3207,
+      "step": 27245
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7106826801517066e-06,
+      "loss": 3.2202,
+      "step": 27246
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7106352718078383e-06,
+      "loss": 2.8871,
+      "step": 27247
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7105878634639696e-06,
+      "loss": 3.3469,
+      "step": 27248
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.710540455120101e-06,
+      "loss": 1.4881,
+      "step": 27249
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7104930467762328e-06,
+      "loss": 2.2761,
+      "step": 27250
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.710445638432364e-06,
+      "loss": 2.7025,
+      "step": 27251
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7103982300884956e-06,
+      "loss": 3.5665,
+      "step": 27252
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7103508217446273e-06,
+      "loss": 1.1482,
+      "step": 27253
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7103034134007586e-06,
+      "loss": 1.897,
+      "step": 27254
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.71025600505689e-06,
+      "loss": 3.2161,
+      "step": 27255
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7102085967130217e-06,
+      "loss": 3.8386,
+      "step": 27256
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.710161188369153e-06,
+      "loss": 3.7491,
+      "step": 27257
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7101137800252843e-06,
+      "loss": 2.5754,
+      "step": 27258
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.710066371681416e-06,
+      "loss": 2.4635,
+      "step": 27259
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7100189633375475e-06,
+      "loss": 3.5304,
+      "step": 27260
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7099715549936788e-06,
+      "loss": 2.3488,
+      "step": 27261
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.7099241466498105e-06,
+      "loss": 1.2468,
+      "step": 27262
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.709876738305942e-06,
+      "loss": 2.8094,
+      "step": 27263
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7098293299620733e-06,
+      "loss": 4.1129,
+      "step": 27264
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.709781921618205e-06,
+      "loss": 2.0332,
+      "step": 27265
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7097345132743365e-06,
+      "loss": 1.2753,
+      "step": 27266
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7096871049304677e-06,
+      "loss": 3.0885,
+      "step": 27267
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7096396965865992e-06,
+      "loss": 3.1591,
+      "step": 27268
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.709592288242731e-06,
+      "loss": 2.2133,
+      "step": 27269
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7095448798988622e-06,
+      "loss": 1.8911,
+      "step": 27270
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7094974715549935e-06,
+      "loss": 1.9166,
+      "step": 27271
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7094500632111252e-06,
+      "loss": 2.4994,
+      "step": 27272
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7094026548672567e-06,
+      "loss": 3.1945,
+      "step": 27273
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.709355246523388e-06,
+      "loss": 2.2301,
+      "step": 27274
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7093078381795197e-06,
+      "loss": 3.1303,
+      "step": 27275
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7092604298356512e-06,
+      "loss": 2.6271,
+      "step": 27276
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7092130214917824e-06,
+      "loss": 2.5063,
+      "step": 27277
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7091656131479142e-06,
+      "loss": 2.6832,
+      "step": 27278
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7091182048040456e-06,
+      "loss": 2.9011,
+      "step": 27279
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.709070796460177e-06,
+      "loss": 2.3819,
+      "step": 27280
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7090233881163086e-06,
+      "loss": 2.3572,
+      "step": 27281
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7089759797724401e-06,
+      "loss": 2.5096,
+      "step": 27282
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7089285714285714e-06,
+      "loss": 2.8538,
+      "step": 27283
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7088811630847031e-06,
+      "loss": 2.1399,
+      "step": 27284
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7088337547408344e-06,
+      "loss": 3.9591,
+      "step": 27285
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7087863463969659e-06,
+      "loss": 3.0178,
+      "step": 27286
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7087389380530976e-06,
+      "loss": 1.913,
+      "step": 27287
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7086915297092289e-06,
+      "loss": 2.2175,
+      "step": 27288
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7086441213653604e-06,
+      "loss": 1.9924,
+      "step": 27289
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7085967130214916e-06,
+      "loss": 1.5581,
+      "step": 27290
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7085493046776233e-06,
+      "loss": 2.3418,
+      "step": 27291
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7085018963337548e-06,
+      "loss": 1.6507,
+      "step": 27292
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7084544879898861e-06,
+      "loss": 1.8293,
+      "step": 27293
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7084070796460178e-06,
+      "loss": 4.1017,
+      "step": 27294
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7083596713021493e-06,
+      "loss": 2.0259,
+      "step": 27295
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7083122629582806e-06,
+      "loss": 3.4614,
+      "step": 27296
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7082648546144123e-06,
+      "loss": 1.5472,
+      "step": 27297
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7082174462705436e-06,
+      "loss": 1.3991,
+      "step": 27298
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.708170037926675e-06,
+      "loss": 3.4246,
+      "step": 27299
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7081226295828068e-06,
+      "loss": 1.8394,
+      "step": 27300
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.708075221238938e-06,
+      "loss": 2.3236,
+      "step": 27301
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7080278128950695e-06,
+      "loss": 1.4831,
+      "step": 27302
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7079804045512012e-06,
+      "loss": 3.5064,
+      "step": 27303
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7079329962073325e-06,
+      "loss": 2.7484,
+      "step": 27304
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.707885587863464e-06,
+      "loss": 2.5266,
+      "step": 27305
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7078381795195957e-06,
+      "loss": 1.6168,
+      "step": 27306
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.707790771175727e-06,
+      "loss": 2.9142,
+      "step": 27307
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7077433628318585e-06,
+      "loss": 2.5565,
+      "step": 27308
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7076959544879898e-06,
+      "loss": 2.3986,
+      "step": 27309
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7076485461441215e-06,
+      "loss": 2.0336,
+      "step": 27310
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7076011378002528e-06,
+      "loss": 2.6965,
+      "step": 27311
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7075537294563843e-06,
+      "loss": 1.9306,
+      "step": 27312
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.707506321112516e-06,
+      "loss": 2.1418,
+      "step": 27313
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7074589127686472e-06,
+      "loss": 1.6558,
+      "step": 27314
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7074115044247787e-06,
+      "loss": 2.2064,
+      "step": 27315
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7073640960809104e-06,
+      "loss": 1.9102,
+      "step": 27316
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7073166877370417e-06,
+      "loss": 2.2254,
+      "step": 27317
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7072692793931732e-06,
+      "loss": 2.9445,
+      "step": 27318
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.707221871049305e-06,
+      "loss": 2.6235,
+      "step": 27319
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7071744627054362e-06,
+      "loss": 1.0028,
+      "step": 27320
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7071270543615675e-06,
+      "loss": 2.0776,
+      "step": 27321
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7070796460176994e-06,
+      "loss": 3.1828,
+      "step": 27322
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7070322376738307e-06,
+      "loss": 2.9405,
+      "step": 27323
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.706984829329962e-06,
+      "loss": 2.2466,
+      "step": 27324
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7069374209860937e-06,
+      "loss": 2.4931,
+      "step": 27325
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.7068900126422251e-06,
+      "loss": 2.4962,
+      "step": 27326
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7068426042983564e-06,
+      "loss": 2.0963,
+      "step": 27327
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7067951959544881e-06,
+      "loss": 1.5058,
+      "step": 27328
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7067477876106196e-06,
+      "loss": 2.6948,
+      "step": 27329
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.706700379266751e-06,
+      "loss": 1.7463,
+      "step": 27330
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7066529709228824e-06,
+      "loss": 3.5501,
+      "step": 27331
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.706605562579014e-06,
+      "loss": 4.2758,
+      "step": 27332
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7065581542351454e-06,
+      "loss": 1.4217,
+      "step": 27333
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7065107458912767e-06,
+      "loss": 2.6418,
+      "step": 27334
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7064633375474084e-06,
+      "loss": 2.2338,
+      "step": 27335
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7064159292035399e-06,
+      "loss": 3.1024,
+      "step": 27336
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7063685208596711e-06,
+      "loss": 1.9138,
+      "step": 27337
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7063211125158028e-06,
+      "loss": 1.649,
+      "step": 27338
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7062737041719343e-06,
+      "loss": 1.6798,
+      "step": 27339
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7062262958280656e-06,
+      "loss": 2.9888,
+      "step": 27340
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7061788874841973e-06,
+      "loss": 1.9883,
+      "step": 27341
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7061314791403288e-06,
+      "loss": 2.1021,
+      "step": 27342
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.70608407079646e-06,
+      "loss": 1.737,
+      "step": 27343
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7060366624525918e-06,
+      "loss": 2.7551,
+      "step": 27344
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7059892541087233e-06,
+      "loss": 3.0407,
+      "step": 27345
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7059418457648546e-06,
+      "loss": 2.5255,
+      "step": 27346
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7058944374209863e-06,
+      "loss": 2.6072,
+      "step": 27347
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7058470290771175e-06,
+      "loss": 2.8954,
+      "step": 27348
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.705799620733249e-06,
+      "loss": 1.5274,
+      "step": 27349
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7057522123893807e-06,
+      "loss": 3.7943,
+      "step": 27350
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.705704804045512e-06,
+      "loss": 3.4013,
+      "step": 27351
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7056573957016435e-06,
+      "loss": 3.5334,
+      "step": 27352
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7056099873577748e-06,
+      "loss": 3.4452,
+      "step": 27353
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7055625790139065e-06,
+      "loss": 1.5928,
+      "step": 27354
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.705515170670038e-06,
+      "loss": 1.358,
+      "step": 27355
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7054677623261693e-06,
+      "loss": 2.8481,
+      "step": 27356
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.705420353982301e-06,
+      "loss": 2.7089,
+      "step": 27357
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7053729456384325e-06,
+      "loss": 1.5607,
+      "step": 27358
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7053255372945637e-06,
+      "loss": 1.9602,
+      "step": 27359
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7052781289506955e-06,
+      "loss": 1.5857,
+      "step": 27360
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7052307206068267e-06,
+      "loss": 2.0309,
+      "step": 27361
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7051833122629582e-06,
+      "loss": 3.253,
+      "step": 27362
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.70513590391909e-06,
+      "loss": 1.9089,
+      "step": 27363
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7050884955752212e-06,
+      "loss": 1.9297,
+      "step": 27364
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7050410872313527e-06,
+      "loss": 2.328,
+      "step": 27365
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7049936788874844e-06,
+      "loss": 2.0832,
+      "step": 27366
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7049462705436157e-06,
+      "loss": 2.5669,
+      "step": 27367
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7048988621997472e-06,
+      "loss": 2.4282,
+      "step": 27368
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7048514538558789e-06,
+      "loss": 1.3461,
+      "step": 27369
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7048040455120102e-06,
+      "loss": 2.4632,
+      "step": 27370
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7047566371681417e-06,
+      "loss": 1.7333,
+      "step": 27371
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7047092288242734e-06,
+      "loss": 2.1342,
+      "step": 27372
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7046618204804046e-06,
+      "loss": 1.598,
+      "step": 27373
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.704614412136536e-06,
+      "loss": 3.6587,
+      "step": 27374
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7045670037926674e-06,
+      "loss": 3.0236,
+      "step": 27375
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7045195954487991e-06,
+      "loss": 3.1711,
+      "step": 27376
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7044721871049304e-06,
+      "loss": 2.1495,
+      "step": 27377
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7044247787610619e-06,
+      "loss": 2.7453,
+      "step": 27378
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7043773704171936e-06,
+      "loss": 2.8884,
+      "step": 27379
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7043299620733249e-06,
+      "loss": 1.9181,
+      "step": 27380
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7042825537294564e-06,
+      "loss": 2.7002,
+      "step": 27381
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.704235145385588e-06,
+      "loss": 1.0417,
+      "step": 27382
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7041877370417193e-06,
+      "loss": 2.1629,
+      "step": 27383
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7041403286978508e-06,
+      "loss": 3.1757,
+      "step": 27384
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7040929203539825e-06,
+      "loss": 3.0298,
+      "step": 27385
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7040455120101138e-06,
+      "loss": 0.9012,
+      "step": 27386
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7039981036662451e-06,
+      "loss": 2.4217,
+      "step": 27387
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7039506953223768e-06,
+      "loss": 2.0706,
+      "step": 27388
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7039032869785083e-06,
+      "loss": 2.2085,
+      "step": 27389
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.7038558786346396e-06,
+      "loss": 2.0992,
+      "step": 27390
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7038084702907713e-06,
+      "loss": 3.5787,
+      "step": 27391
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7037610619469028e-06,
+      "loss": 2.3304,
+      "step": 27392
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.703713653603034e-06,
+      "loss": 2.9631,
+      "step": 27393
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7036662452591658e-06,
+      "loss": 2.2924,
+      "step": 27394
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7036188369152973e-06,
+      "loss": 3.8092,
+      "step": 27395
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7035714285714285e-06,
+      "loss": 3.1443,
+      "step": 27396
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.70352402022756e-06,
+      "loss": 2.4761,
+      "step": 27397
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7034766118836917e-06,
+      "loss": 4.5194,
+      "step": 27398
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.703429203539823e-06,
+      "loss": 1.9874,
+      "step": 27399
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7033817951959543e-06,
+      "loss": 2.8982,
+      "step": 27400
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.703334386852086e-06,
+      "loss": 2.7081,
+      "step": 27401
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7032869785082175e-06,
+      "loss": 0.758,
+      "step": 27402
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7032395701643488e-06,
+      "loss": 2.6746,
+      "step": 27403
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7031921618204805e-06,
+      "loss": 1.6079,
+      "step": 27404
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.703144753476612e-06,
+      "loss": 1.1122,
+      "step": 27405
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7030973451327432e-06,
+      "loss": 2.158,
+      "step": 27406
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.703049936788875e-06,
+      "loss": 1.6284,
+      "step": 27407
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7030025284450064e-06,
+      "loss": 2.8409,
+      "step": 27408
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7029551201011377e-06,
+      "loss": 3.3326,
+      "step": 27409
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7029077117572694e-06,
+      "loss": 3.5429,
+      "step": 27410
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.702860303413401e-06,
+      "loss": 2.5836,
+      "step": 27411
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7028128950695322e-06,
+      "loss": 2.283,
+      "step": 27412
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.702765486725664e-06,
+      "loss": 1.8136,
+      "step": 27413
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7027180783817952e-06,
+      "loss": 2.7953,
+      "step": 27414
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7026706700379267e-06,
+      "loss": 1.9296,
+      "step": 27415
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7026232616940584e-06,
+      "loss": 2.8634,
+      "step": 27416
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7025758533501897e-06,
+      "loss": 2.7232,
+      "step": 27417
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7025284450063212e-06,
+      "loss": 2.7296,
+      "step": 27418
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7024810366624524e-06,
+      "loss": 2.2051,
+      "step": 27419
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7024336283185841e-06,
+      "loss": 2.0635,
+      "step": 27420
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7023862199747156e-06,
+      "loss": 2.8257,
+      "step": 27421
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.702338811630847e-06,
+      "loss": 3.1357,
+      "step": 27422
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7022914032869786e-06,
+      "loss": 2.31,
+      "step": 27423
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.70224399494311e-06,
+      "loss": 2.4775,
+      "step": 27424
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7021965865992414e-06,
+      "loss": 3.0416,
+      "step": 27425
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.702149178255373e-06,
+      "loss": 2.1523,
+      "step": 27426
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7021017699115044e-06,
+      "loss": 1.136,
+      "step": 27427
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7020543615676359e-06,
+      "loss": 1.7673,
+      "step": 27428
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7020069532237676e-06,
+      "loss": 2.7342,
+      "step": 27429
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7019595448798988e-06,
+      "loss": 2.1271,
+      "step": 27430
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7019121365360303e-06,
+      "loss": 2.1707,
+      "step": 27431
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.701864728192162e-06,
+      "loss": 3.1771,
+      "step": 27432
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7018173198482933e-06,
+      "loss": 2.0587,
+      "step": 27433
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7017699115044248e-06,
+      "loss": 2.5379,
+      "step": 27434
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7017225031605565e-06,
+      "loss": 4.2172,
+      "step": 27435
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7016750948166878e-06,
+      "loss": 2.9273,
+      "step": 27436
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7016276864728193e-06,
+      "loss": 2.2987,
+      "step": 27437
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.701580278128951e-06,
+      "loss": 1.5521,
+      "step": 27438
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7015328697850823e-06,
+      "loss": 1.3489,
+      "step": 27439
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7014854614412136e-06,
+      "loss": 2.0161,
+      "step": 27440
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.701438053097345e-06,
+      "loss": 1.5246,
+      "step": 27441
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7013906447534768e-06,
+      "loss": 2.2146,
+      "step": 27442
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.701343236409608e-06,
+      "loss": 2.0842,
+      "step": 27443
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7012958280657395e-06,
+      "loss": 0.8962,
+      "step": 27444
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7012484197218712e-06,
+      "loss": 3.3989,
+      "step": 27445
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7012010113780025e-06,
+      "loss": 1.2911,
+      "step": 27446
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.701153603034134e-06,
+      "loss": 2.8935,
+      "step": 27447
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7011061946902657e-06,
+      "loss": 2.4957,
+      "step": 27448
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.701058786346397e-06,
+      "loss": 2.452,
+      "step": 27449
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7010113780025283e-06,
+      "loss": 2.3865,
+      "step": 27450
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7009639696586602e-06,
+      "loss": 1.4945,
+      "step": 27451
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7009165613147915e-06,
+      "loss": 2.2501,
+      "step": 27452
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.7008691529709227e-06,
+      "loss": 2.6614,
+      "step": 27453
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7008217446270544e-06,
+      "loss": 2.6864,
+      "step": 27454
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.700774336283186e-06,
+      "loss": 1.414,
+      "step": 27455
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7007269279393172e-06,
+      "loss": 2.5358,
+      "step": 27456
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.700679519595449e-06,
+      "loss": 2.3694,
+      "step": 27457
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7006321112515804e-06,
+      "loss": 2.3583,
+      "step": 27458
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7005847029077117e-06,
+      "loss": 1.674,
+      "step": 27459
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7005372945638434e-06,
+      "loss": 2.014,
+      "step": 27460
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7004898862199749e-06,
+      "loss": 1.8893,
+      "step": 27461
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7004424778761062e-06,
+      "loss": 2.2563,
+      "step": 27462
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7003950695322375e-06,
+      "loss": 2.5946,
+      "step": 27463
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7003476611883692e-06,
+      "loss": 3.3698,
+      "step": 27464
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7003002528445006e-06,
+      "loss": 2.0772,
+      "step": 27465
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.700252844500632e-06,
+      "loss": 2.029,
+      "step": 27466
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7002054361567636e-06,
+      "loss": 2.3852,
+      "step": 27467
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7001580278128951e-06,
+      "loss": 2.5481,
+      "step": 27468
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7001106194690264e-06,
+      "loss": 2.7793,
+      "step": 27469
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7000632111251581e-06,
+      "loss": 2.2739,
+      "step": 27470
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.7000158027812896e-06,
+      "loss": 2.6671,
+      "step": 27471
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6999683944374209e-06,
+      "loss": 3.2648,
+      "step": 27472
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6999209860935526e-06,
+      "loss": 3.5877,
+      "step": 27473
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.699873577749684e-06,
+      "loss": 3.4805,
+      "step": 27474
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6998261694058154e-06,
+      "loss": 3.9441,
+      "step": 27475
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.699778761061947e-06,
+      "loss": 2.7204,
+      "step": 27476
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6997313527180783e-06,
+      "loss": 2.9061,
+      "step": 27477
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6996839443742098e-06,
+      "loss": 1.7965,
+      "step": 27478
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6996365360303415e-06,
+      "loss": 3.2658,
+      "step": 27479
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6995891276864728e-06,
+      "loss": 2.8596,
+      "step": 27480
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6995417193426043e-06,
+      "loss": 3.2215,
+      "step": 27481
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.699494310998736e-06,
+      "loss": 4.1793,
+      "step": 27482
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6994469026548673e-06,
+      "loss": 2.967,
+      "step": 27483
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6993994943109988e-06,
+      "loss": 2.6101,
+      "step": 27484
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.69935208596713e-06,
+      "loss": 2.2467,
+      "step": 27485
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6993046776232618e-06,
+      "loss": 2.6153,
+      "step": 27486
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6992572692793933e-06,
+      "loss": 1.3052,
+      "step": 27487
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6992098609355245e-06,
+      "loss": 1.7784,
+      "step": 27488
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6991624525916562e-06,
+      "loss": 1.4455,
+      "step": 27489
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6991150442477875e-06,
+      "loss": 1.6985,
+      "step": 27490
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.699067635903919e-06,
+      "loss": 1.6268,
+      "step": 27491
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6990202275600507e-06,
+      "loss": 1.7561,
+      "step": 27492
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.698972819216182e-06,
+      "loss": 1.7915,
+      "step": 27493
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6989254108723135e-06,
+      "loss": 4.0873,
+      "step": 27494
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6988780025284452e-06,
+      "loss": 1.5851,
+      "step": 27495
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6988305941845765e-06,
+      "loss": 2.1346,
+      "step": 27496
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.698783185840708e-06,
+      "loss": 1.796,
+      "step": 27497
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6987357774968397e-06,
+      "loss": 2.6649,
+      "step": 27498
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.698688369152971e-06,
+      "loss": 1.986,
+      "step": 27499
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6986409608091025e-06,
+      "loss": 2.4683,
+      "step": 27500
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6985935524652342e-06,
+      "loss": 2.5772,
+      "step": 27501
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6985461441213654e-06,
+      "loss": 2.0515,
+      "step": 27502
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6984987357774967e-06,
+      "loss": 3.0068,
+      "step": 27503
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6984513274336284e-06,
+      "loss": 2.6955,
+      "step": 27504
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.69840391908976e-06,
+      "loss": 2.2633,
+      "step": 27505
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6983565107458912e-06,
+      "loss": 3.694,
+      "step": 27506
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6983091024020227e-06,
+      "loss": 3.1695,
+      "step": 27507
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6982616940581544e-06,
+      "loss": 1.9772,
+      "step": 27508
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6982142857142857e-06,
+      "loss": 1.1763,
+      "step": 27509
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6981668773704172e-06,
+      "loss": 1.6055,
+      "step": 27510
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6981194690265489e-06,
+      "loss": 3.829,
+      "step": 27511
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6980720606826801e-06,
+      "loss": 2.4186,
+      "step": 27512
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6980246523388116e-06,
+      "loss": 3.0502,
+      "step": 27513
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6979772439949433e-06,
+      "loss": 2.0768,
+      "step": 27514
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.6979298356510746e-06,
+      "loss": 2.2603,
+      "step": 27515
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.697882427307206e-06,
+      "loss": 3.0427,
+      "step": 27516
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6978350189633376e-06,
+      "loss": 2.1657,
+      "step": 27517
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.697787610619469e-06,
+      "loss": 2.1892,
+      "step": 27518
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6977402022756004e-06,
+      "loss": 3.5787,
+      "step": 27519
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.697692793931732e-06,
+      "loss": 1.9153,
+      "step": 27520
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6976453855878636e-06,
+      "loss": 1.5301,
+      "step": 27521
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6975979772439949e-06,
+      "loss": 2.6371,
+      "step": 27522
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6975505689001266e-06,
+      "loss": 2.4886,
+      "step": 27523
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.697503160556258e-06,
+      "loss": 2.3304,
+      "step": 27524
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6974557522123893e-06,
+      "loss": 2.1275,
+      "step": 27525
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.697408343868521e-06,
+      "loss": 3.1248,
+      "step": 27526
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6973609355246525e-06,
+      "loss": 2.4312,
+      "step": 27527
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6973135271807838e-06,
+      "loss": 1.7948,
+      "step": 27528
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.697266118836915e-06,
+      "loss": 2.5236,
+      "step": 27529
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6972187104930468e-06,
+      "loss": 1.8239,
+      "step": 27530
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6971713021491783e-06,
+      "loss": 2.0909,
+      "step": 27531
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6971238938053096e-06,
+      "loss": 1.569,
+      "step": 27532
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6970764854614413e-06,
+      "loss": 1.5395,
+      "step": 27533
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6970290771175728e-06,
+      "loss": 1.9155,
+      "step": 27534
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.696981668773704e-06,
+      "loss": 2.0948,
+      "step": 27535
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6969342604298357e-06,
+      "loss": 3.0545,
+      "step": 27536
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6968868520859672e-06,
+      "loss": 3.4024,
+      "step": 27537
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6968394437420985e-06,
+      "loss": 1.5359,
+      "step": 27538
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6967920353982302e-06,
+      "loss": 3.395,
+      "step": 27539
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6967446270543617e-06,
+      "loss": 2.1053,
+      "step": 27540
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.696697218710493e-06,
+      "loss": 2.6466,
+      "step": 27541
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6966498103666247e-06,
+      "loss": 1.9993,
+      "step": 27542
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.696602402022756e-06,
+      "loss": 2.81,
+      "step": 27543
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6965549936788875e-06,
+      "loss": 1.0832,
+      "step": 27544
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6965075853350192e-06,
+      "loss": 1.106,
+      "step": 27545
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6964601769911505e-06,
+      "loss": 2.6317,
+      "step": 27546
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.696412768647282e-06,
+      "loss": 2.2594,
+      "step": 27547
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6963653603034137e-06,
+      "loss": 1.833,
+      "step": 27548
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.696317951959545e-06,
+      "loss": 2.3739,
+      "step": 27549
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6962705436156764e-06,
+      "loss": 1.8199,
+      "step": 27550
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6962231352718077e-06,
+      "loss": 2.9762,
+      "step": 27551
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6961757269279394e-06,
+      "loss": 1.7496,
+      "step": 27552
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.696128318584071e-06,
+      "loss": 2.684,
+      "step": 27553
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6960809102402022e-06,
+      "loss": 1.6054,
+      "step": 27554
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6960335018963339e-06,
+      "loss": 1.1107,
+      "step": 27555
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6959860935524652e-06,
+      "loss": 0.966,
+      "step": 27556
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6959386852085967e-06,
+      "loss": 2.1005,
+      "step": 27557
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6958912768647284e-06,
+      "loss": 2.3091,
+      "step": 27558
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6958438685208596e-06,
+      "loss": 3.1498,
+      "step": 27559
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6957964601769911e-06,
+      "loss": 2.5473,
+      "step": 27560
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6957490518331228e-06,
+      "loss": 3.5544,
+      "step": 27561
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6957016434892541e-06,
+      "loss": 3.3069,
+      "step": 27562
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6956542351453856e-06,
+      "loss": 3.0568,
+      "step": 27563
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6956068268015173e-06,
+      "loss": 2.8018,
+      "step": 27564
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6955594184576486e-06,
+      "loss": 3.7649,
+      "step": 27565
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.69551201011378e-06,
+      "loss": 3.566,
+      "step": 27566
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6954646017699118e-06,
+      "loss": 2.5283,
+      "step": 27567
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.695417193426043e-06,
+      "loss": 1.9851,
+      "step": 27568
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6953697850821744e-06,
+      "loss": 3.6528,
+      "step": 27569
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.695322376738306e-06,
+      "loss": 2.0695,
+      "step": 27570
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6952749683944375e-06,
+      "loss": 3.9103,
+      "step": 27571
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6952275600505688e-06,
+      "loss": 2.4334,
+      "step": 27572
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6951801517067003e-06,
+      "loss": 3.434,
+      "step": 27573
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.695132743362832e-06,
+      "loss": 2.053,
+      "step": 27574
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6950853350189633e-06,
+      "loss": 2.3354,
+      "step": 27575
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6950379266750948e-06,
+      "loss": 2.3515,
+      "step": 27576
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6949905183312265e-06,
+      "loss": 2.6625,
+      "step": 27577
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6949431099873578e-06,
+      "loss": 1.9914,
+      "step": 27578
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.6948957016434893e-06,
+      "loss": 2.1586,
+      "step": 27579
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.694848293299621e-06,
+      "loss": 2.0829,
+      "step": 27580
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6948008849557523e-06,
+      "loss": 2.2622,
+      "step": 27581
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6947534766118835e-06,
+      "loss": 1.5323,
+      "step": 27582
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6947060682680152e-06,
+      "loss": 2.6448,
+      "step": 27583
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6946586599241467e-06,
+      "loss": 3.5842,
+      "step": 27584
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.694611251580278e-06,
+      "loss": 2.5618,
+      "step": 27585
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6945638432364097e-06,
+      "loss": 2.4319,
+      "step": 27586
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6945164348925412e-06,
+      "loss": 0.636,
+      "step": 27587
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6944690265486725e-06,
+      "loss": 1.3885,
+      "step": 27588
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6944216182048042e-06,
+      "loss": 2.42,
+      "step": 27589
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6943742098609357e-06,
+      "loss": 1.8527,
+      "step": 27590
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.694326801517067e-06,
+      "loss": 0.9825,
+      "step": 27591
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6942793931731987e-06,
+      "loss": 3.2445,
+      "step": 27592
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6942319848293302e-06,
+      "loss": 2.0539,
+      "step": 27593
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6941845764854614e-06,
+      "loss": 2.4353,
+      "step": 27594
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6941371681415927e-06,
+      "loss": 1.2527,
+      "step": 27595
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6940897597977244e-06,
+      "loss": 3.8527,
+      "step": 27596
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.694042351453856e-06,
+      "loss": 2.1091,
+      "step": 27597
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6939949431099872e-06,
+      "loss": 3.5624,
+      "step": 27598
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.693947534766119e-06,
+      "loss": 2.6385,
+      "step": 27599
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6939001264222504e-06,
+      "loss": 2.8473,
+      "step": 27600
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6938527180783817e-06,
+      "loss": 2.768,
+      "step": 27601
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6938053097345134e-06,
+      "loss": 1.7254,
+      "step": 27602
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6937579013906449e-06,
+      "loss": 2.6759,
+      "step": 27603
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6937104930467762e-06,
+      "loss": 3.147,
+      "step": 27604
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6936630847029079e-06,
+      "loss": 1.3957,
+      "step": 27605
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6936156763590391e-06,
+      "loss": 2.5206,
+      "step": 27606
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6935682680151706e-06,
+      "loss": 3.2026,
+      "step": 27607
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6935208596713023e-06,
+      "loss": 2.9937,
+      "step": 27608
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6934734513274336e-06,
+      "loss": 1.5904,
+      "step": 27609
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6934260429835651e-06,
+      "loss": 2.0481,
+      "step": 27610
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6933786346396968e-06,
+      "loss": 2.462,
+      "step": 27611
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.693331226295828e-06,
+      "loss": 3.4404,
+      "step": 27612
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6932838179519596e-06,
+      "loss": 2.966,
+      "step": 27613
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6932364096080913e-06,
+      "loss": 3.3895,
+      "step": 27614
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6931890012642226e-06,
+      "loss": 2.3488,
+      "step": 27615
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.693141592920354e-06,
+      "loss": 1.8316,
+      "step": 27616
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6930941845764853e-06,
+      "loss": 2.3413,
+      "step": 27617
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.693046776232617e-06,
+      "loss": 2.4969,
+      "step": 27618
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6929993678887483e-06,
+      "loss": 2.2882,
+      "step": 27619
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6929519595448798e-06,
+      "loss": 2.3114,
+      "step": 27620
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6929045512010115e-06,
+      "loss": 1.8444,
+      "step": 27621
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6928571428571428e-06,
+      "loss": 2.9994,
+      "step": 27622
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6928097345132743e-06,
+      "loss": 3.0297,
+      "step": 27623
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.692762326169406e-06,
+      "loss": 1.6879,
+      "step": 27624
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6927149178255373e-06,
+      "loss": 0.999,
+      "step": 27625
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6926675094816688e-06,
+      "loss": 3.5534,
+      "step": 27626
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6926201011378005e-06,
+      "loss": 1.7744,
+      "step": 27627
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6925726927939318e-06,
+      "loss": 2.7612,
+      "step": 27628
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6925252844500632e-06,
+      "loss": 2.2244,
+      "step": 27629
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.692477876106195e-06,
+      "loss": 2.4186,
+      "step": 27630
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6924304677623262e-06,
+      "loss": 0.5212,
+      "step": 27631
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6923830594184575e-06,
+      "loss": 2.2609,
+      "step": 27632
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6923356510745892e-06,
+      "loss": 2.171,
+      "step": 27633
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6922882427307207e-06,
+      "loss": 1.9628,
+      "step": 27634
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.692240834386852e-06,
+      "loss": 3.0397,
+      "step": 27635
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6921934260429837e-06,
+      "loss": 1.5118,
+      "step": 27636
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6921460176991152e-06,
+      "loss": 2.1146,
+      "step": 27637
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6920986093552465e-06,
+      "loss": 2.3939,
+      "step": 27638
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.692051201011378e-06,
+      "loss": 1.7167,
+      "step": 27639
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6920037926675097e-06,
+      "loss": 3.9978,
+      "step": 27640
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.691956384323641e-06,
+      "loss": 3.2268,
+      "step": 27641
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6919089759797724e-06,
+      "loss": 2.8164,
+      "step": 27642
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.6918615676359041e-06,
+      "loss": 1.97,
+      "step": 27643
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6918141592920354e-06,
+      "loss": 2.8004,
+      "step": 27644
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6917667509481667e-06,
+      "loss": 3.8145,
+      "step": 27645
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6917193426042984e-06,
+      "loss": 2.1724,
+      "step": 27646
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.69167193426043e-06,
+      "loss": 3.1017,
+      "step": 27647
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6916245259165612e-06,
+      "loss": 1.992,
+      "step": 27648
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6915771175726929e-06,
+      "loss": 1.6667,
+      "step": 27649
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6915297092288244e-06,
+      "loss": 2.6121,
+      "step": 27650
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6914823008849557e-06,
+      "loss": 1.4704,
+      "step": 27651
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6914348925410874e-06,
+      "loss": 3.4843,
+      "step": 27652
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6913874841972188e-06,
+      "loss": 3.2222,
+      "step": 27653
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6913400758533501e-06,
+      "loss": 2.8564,
+      "step": 27654
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6912926675094818e-06,
+      "loss": 2.3646,
+      "step": 27655
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6912452591656133e-06,
+      "loss": 2.3262,
+      "step": 27656
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6911978508217446e-06,
+      "loss": 2.7077,
+      "step": 27657
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6911504424778763e-06,
+      "loss": 2.0548,
+      "step": 27658
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6911030341340076e-06,
+      "loss": 2.8493,
+      "step": 27659
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.691055625790139e-06,
+      "loss": 2.0169,
+      "step": 27660
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6910082174462704e-06,
+      "loss": 1.9683,
+      "step": 27661
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.690960809102402e-06,
+      "loss": 3.5646,
+      "step": 27662
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6909134007585336e-06,
+      "loss": 2.3782,
+      "step": 27663
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6908659924146648e-06,
+      "loss": 3.0912,
+      "step": 27664
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6908185840707965e-06,
+      "loss": 3.3712,
+      "step": 27665
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.690771175726928e-06,
+      "loss": 1.8716,
+      "step": 27666
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6907237673830593e-06,
+      "loss": 3.0224,
+      "step": 27667
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.690676359039191e-06,
+      "loss": 1.6482,
+      "step": 27668
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6906289506953225e-06,
+      "loss": 2.6756,
+      "step": 27669
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6905815423514538e-06,
+      "loss": 3.6508,
+      "step": 27670
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6905341340075855e-06,
+      "loss": 1.3021,
+      "step": 27671
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6904867256637168e-06,
+      "loss": 3.0802,
+      "step": 27672
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6904393173198483e-06,
+      "loss": 2.6591,
+      "step": 27673
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.69039190897598e-06,
+      "loss": 2.9093,
+      "step": 27674
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6903445006321113e-06,
+      "loss": 1.8275,
+      "step": 27675
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6902970922882427e-06,
+      "loss": 3.5013,
+      "step": 27676
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6902496839443744e-06,
+      "loss": 2.398,
+      "step": 27677
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6902022756005057e-06,
+      "loss": 3.5908,
+      "step": 27678
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6901548672566372e-06,
+      "loss": 1.569,
+      "step": 27679
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.690107458912769e-06,
+      "loss": 2.3524,
+      "step": 27680
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6900600505689002e-06,
+      "loss": 3.3946,
+      "step": 27681
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6900126422250317e-06,
+      "loss": 2.7589,
+      "step": 27682
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.689965233881163e-06,
+      "loss": 3.0472,
+      "step": 27683
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6899178255372947e-06,
+      "loss": 1.7646,
+      "step": 27684
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.689870417193426e-06,
+      "loss": 2.2168,
+      "step": 27685
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6898230088495575e-06,
+      "loss": 2.0789,
+      "step": 27686
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6897756005056892e-06,
+      "loss": 2.3404,
+      "step": 27687
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6897281921618204e-06,
+      "loss": 1.9947,
+      "step": 27688
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.689680783817952e-06,
+      "loss": 2.0593,
+      "step": 27689
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6896333754740836e-06,
+      "loss": 4.0036,
+      "step": 27690
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.689585967130215e-06,
+      "loss": 3.1962,
+      "step": 27691
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6895385587863464e-06,
+      "loss": 1.8275,
+      "step": 27692
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6894911504424781e-06,
+      "loss": 1.0664,
+      "step": 27693
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6894437420986094e-06,
+      "loss": 3.2618,
+      "step": 27694
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6893963337547409e-06,
+      "loss": 2.913,
+      "step": 27695
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6893489254108726e-06,
+      "loss": 2.4494,
+      "step": 27696
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6893015170670039e-06,
+      "loss": 1.2589,
+      "step": 27697
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6892541087231351e-06,
+      "loss": 2.5725,
+      "step": 27698
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6892067003792669e-06,
+      "loss": 2.5753,
+      "step": 27699
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6891592920353983e-06,
+      "loss": 2.1894,
+      "step": 27700
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6891118836915296e-06,
+      "loss": 2.5366,
+      "step": 27701
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6890644753476613e-06,
+      "loss": 2.7077,
+      "step": 27702
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6890170670037928e-06,
+      "loss": 2.4187,
+      "step": 27703
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.688969658659924e-06,
+      "loss": 3.9514,
+      "step": 27704
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6889222503160556e-06,
+      "loss": 1.8142,
+      "step": 27705
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.6888748419721873e-06,
+      "loss": 2.6715,
+      "step": 27706
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6888274336283186e-06,
+      "loss": 2.786,
+      "step": 27707
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.68878002528445e-06,
+      "loss": 2.5086,
+      "step": 27708
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6887326169405818e-06,
+      "loss": 3.2691,
+      "step": 27709
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.688685208596713e-06,
+      "loss": 1.613,
+      "step": 27710
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6886378002528443e-06,
+      "loss": 2.7582,
+      "step": 27711
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.688590391908976e-06,
+      "loss": 2.6504,
+      "step": 27712
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6885429835651075e-06,
+      "loss": 2.2853,
+      "step": 27713
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6884955752212388e-06,
+      "loss": 2.2908,
+      "step": 27714
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6884481668773705e-06,
+      "loss": 2.1337,
+      "step": 27715
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.688400758533502e-06,
+      "loss": 2.3854,
+      "step": 27716
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6883533501896333e-06,
+      "loss": 3.0159,
+      "step": 27717
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.688305941845765e-06,
+      "loss": 2.2066,
+      "step": 27718
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6882585335018965e-06,
+      "loss": 3.6976,
+      "step": 27719
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6882111251580278e-06,
+      "loss": 1.264,
+      "step": 27720
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6881637168141595e-06,
+      "loss": 2.5151,
+      "step": 27721
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.688116308470291e-06,
+      "loss": 1.7252,
+      "step": 27722
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6880689001264222e-06,
+      "loss": 1.6911,
+      "step": 27723
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.688021491782554e-06,
+      "loss": 2.1932,
+      "step": 27724
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6879740834386852e-06,
+      "loss": 2.8508,
+      "step": 27725
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6879266750948167e-06,
+      "loss": 2.161,
+      "step": 27726
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.687879266750948e-06,
+      "loss": 2.3518,
+      "step": 27727
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6878318584070797e-06,
+      "loss": 2.5361,
+      "step": 27728
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6877844500632112e-06,
+      "loss": 3.148,
+      "step": 27729
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6877370417193425e-06,
+      "loss": 2.2443,
+      "step": 27730
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6876896333754742e-06,
+      "loss": 2.6335,
+      "step": 27731
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6876422250316057e-06,
+      "loss": 3.4603,
+      "step": 27732
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.687594816687737e-06,
+      "loss": 0.6715,
+      "step": 27733
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6875474083438687e-06,
+      "loss": 2.3908,
+      "step": 27734
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6875e-06,
+      "loss": 2.6574,
+      "step": 27735
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6874525916561314e-06,
+      "loss": 3.2074,
+      "step": 27736
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6874051833122631e-06,
+      "loss": 3.0576,
+      "step": 27737
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6873577749683944e-06,
+      "loss": 1.4773,
+      "step": 27738
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.687310366624526e-06,
+      "loss": 1.9048,
+      "step": 27739
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6872629582806576e-06,
+      "loss": 2.2236,
+      "step": 27740
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6872155499367889e-06,
+      "loss": 3.3403,
+      "step": 27741
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6871681415929204e-06,
+      "loss": 3.0146,
+      "step": 27742
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.687120733249052e-06,
+      "loss": 2.5201,
+      "step": 27743
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6870733249051834e-06,
+      "loss": 2.6267,
+      "step": 27744
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6870259165613149e-06,
+      "loss": 1.7209,
+      "step": 27745
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6869785082174461e-06,
+      "loss": 2.6145,
+      "step": 27746
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6869310998735778e-06,
+      "loss": 1.4743,
+      "step": 27747
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6868836915297091e-06,
+      "loss": 3.2547,
+      "step": 27748
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6868362831858406e-06,
+      "loss": 2.9477,
+      "step": 27749
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6867888748419723e-06,
+      "loss": 3.0631,
+      "step": 27750
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6867414664981036e-06,
+      "loss": 1.5461,
+      "step": 27751
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.686694058154235e-06,
+      "loss": 2.2431,
+      "step": 27752
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6866466498103668e-06,
+      "loss": 3.4717,
+      "step": 27753
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.686599241466498e-06,
+      "loss": 2.825,
+      "step": 27754
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6865518331226296e-06,
+      "loss": 1.3171,
+      "step": 27755
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6865044247787613e-06,
+      "loss": 1.8664,
+      "step": 27756
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6864570164348926e-06,
+      "loss": 2.4613,
+      "step": 27757
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.686409608091024e-06,
+      "loss": 2.0191,
+      "step": 27758
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6863621997471557e-06,
+      "loss": 2.6572,
+      "step": 27759
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.686314791403287e-06,
+      "loss": 1.1541,
+      "step": 27760
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6862673830594183e-06,
+      "loss": 2.9089,
+      "step": 27761
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.68621997471555e-06,
+      "loss": 2.9455,
+      "step": 27762
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6861725663716815e-06,
+      "loss": 3.0797,
+      "step": 27763
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6861251580278128e-06,
+      "loss": 1.8429,
+      "step": 27764
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6860777496839445e-06,
+      "loss": 1.9802,
+      "step": 27765
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.686030341340076e-06,
+      "loss": 3.6413,
+      "step": 27766
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6859829329962073e-06,
+      "loss": 2.2546,
+      "step": 27767
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6859355246523388e-06,
+      "loss": 2.4448,
+      "step": 27768
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6858881163084705e-06,
+      "loss": 2.6595,
+      "step": 27769
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.6858407079646017e-06,
+      "loss": 3.3481,
+      "step": 27770
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6857932996207332e-06,
+      "loss": 2.7263,
+      "step": 27771
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.685745891276865e-06,
+      "loss": 2.4823,
+      "step": 27772
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6856984829329962e-06,
+      "loss": 2.1523,
+      "step": 27773
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6856510745891275e-06,
+      "loss": 1.2089,
+      "step": 27774
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6856036662452592e-06,
+      "loss": 1.6733,
+      "step": 27775
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6855562579013907e-06,
+      "loss": 1.7371,
+      "step": 27776
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.685508849557522e-06,
+      "loss": 2.5192,
+      "step": 27777
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6854614412136537e-06,
+      "loss": 1.4435,
+      "step": 27778
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6854140328697852e-06,
+      "loss": 2.7909,
+      "step": 27779
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6853666245259164e-06,
+      "loss": 2.6791,
+      "step": 27780
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6853192161820482e-06,
+      "loss": 2.3815,
+      "step": 27781
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6852718078381796e-06,
+      "loss": 2.9911,
+      "step": 27782
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.685224399494311e-06,
+      "loss": 2.6122,
+      "step": 27783
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6851769911504426e-06,
+      "loss": 2.9239,
+      "step": 27784
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6851295828065741e-06,
+      "loss": 2.2266,
+      "step": 27785
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6850821744627054e-06,
+      "loss": 2.985,
+      "step": 27786
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6850347661188371e-06,
+      "loss": 3.26,
+      "step": 27787
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6849873577749684e-06,
+      "loss": 1.2209,
+      "step": 27788
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6849399494310999e-06,
+      "loss": 2.4934,
+      "step": 27789
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6848925410872312e-06,
+      "loss": 3.0477,
+      "step": 27790
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6848451327433629e-06,
+      "loss": 1.9154,
+      "step": 27791
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6847977243994944e-06,
+      "loss": 1.3793,
+      "step": 27792
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6847503160556256e-06,
+      "loss": 2.2375,
+      "step": 27793
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6847029077117573e-06,
+      "loss": 2.4585,
+      "step": 27794
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6846554993678888e-06,
+      "loss": 2.6142,
+      "step": 27795
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6846080910240201e-06,
+      "loss": 1.5649,
+      "step": 27796
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6845606826801518e-06,
+      "loss": 3.2442,
+      "step": 27797
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6845132743362833e-06,
+      "loss": 2.7626,
+      "step": 27798
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6844658659924146e-06,
+      "loss": 2.3644,
+      "step": 27799
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6844184576485463e-06,
+      "loss": 1.8294,
+      "step": 27800
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6843710493046776e-06,
+      "loss": 2.0261,
+      "step": 27801
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.684323640960809e-06,
+      "loss": 1.9404,
+      "step": 27802
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6842762326169408e-06,
+      "loss": 2.5996,
+      "step": 27803
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.684228824273072e-06,
+      "loss": 2.2983,
+      "step": 27804
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6841814159292035e-06,
+      "loss": 1.0228,
+      "step": 27805
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6841340075853352e-06,
+      "loss": 3.8225,
+      "step": 27806
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6840865992414665e-06,
+      "loss": 2.6953,
+      "step": 27807
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.684039190897598e-06,
+      "loss": 1.4227,
+      "step": 27808
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6839917825537297e-06,
+      "loss": 1.3766,
+      "step": 27809
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.683944374209861e-06,
+      "loss": 2.7423,
+      "step": 27810
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6838969658659925e-06,
+      "loss": 2.7462,
+      "step": 27811
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6838495575221238e-06,
+      "loss": 2.6491,
+      "step": 27812
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6838021491782555e-06,
+      "loss": 3.7712,
+      "step": 27813
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6837547408343868e-06,
+      "loss": 1.5413,
+      "step": 27814
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6837073324905183e-06,
+      "loss": 2.1763,
+      "step": 27815
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.68365992414665e-06,
+      "loss": 2.5233,
+      "step": 27816
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6836125158027812e-06,
+      "loss": 3.5282,
+      "step": 27817
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6835651074589127e-06,
+      "loss": 2.7639,
+      "step": 27818
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6835176991150444e-06,
+      "loss": 3.0049,
+      "step": 27819
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6834702907711757e-06,
+      "loss": 2.6008,
+      "step": 27820
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6834228824273072e-06,
+      "loss": 1.7796,
+      "step": 27821
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.683375474083439e-06,
+      "loss": 3.4022,
+      "step": 27822
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6833280657395702e-06,
+      "loss": 2.5391,
+      "step": 27823
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6832806573957017e-06,
+      "loss": 2.9736,
+      "step": 27824
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6832332490518334e-06,
+      "loss": 3.4105,
+      "step": 27825
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6831858407079647e-06,
+      "loss": 2.4221,
+      "step": 27826
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.683138432364096e-06,
+      "loss": 3.2555,
+      "step": 27827
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6830910240202277e-06,
+      "loss": 2.0339,
+      "step": 27828
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6830436156763591e-06,
+      "loss": 2.8999,
+      "step": 27829
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6829962073324904e-06,
+      "loss": 1.957,
+      "step": 27830
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6829487989886221e-06,
+      "loss": 1.7277,
+      "step": 27831
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.6829013906447536e-06,
+      "loss": 4.015,
+      "step": 27832
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.682853982300885e-06,
+      "loss": 2.3675,
+      "step": 27833
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6828065739570164e-06,
+      "loss": 1.9315,
+      "step": 27834
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.682759165613148e-06,
+      "loss": 2.2141,
+      "step": 27835
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6827117572692794e-06,
+      "loss": 2.6063,
+      "step": 27836
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6826643489254109e-06,
+      "loss": 1.7859,
+      "step": 27837
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6826169405815426e-06,
+      "loss": 3.5035,
+      "step": 27838
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6825695322376739e-06,
+      "loss": 2.6055,
+      "step": 27839
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6825221238938051e-06,
+      "loss": 1.6575,
+      "step": 27840
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6824747155499368e-06,
+      "loss": 1.8311,
+      "step": 27841
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6824273072060683e-06,
+      "loss": 0.7494,
+      "step": 27842
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6823798988621996e-06,
+      "loss": 2.9757,
+      "step": 27843
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6823324905183313e-06,
+      "loss": 3.286,
+      "step": 27844
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6822850821744628e-06,
+      "loss": 1.6349,
+      "step": 27845
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.682237673830594e-06,
+      "loss": 3.4373,
+      "step": 27846
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6821902654867258e-06,
+      "loss": 1.8478,
+      "step": 27847
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6821428571428573e-06,
+      "loss": 1.0191,
+      "step": 27848
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6820954487989886e-06,
+      "loss": 2.1637,
+      "step": 27849
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6820480404551203e-06,
+      "loss": 2.5142,
+      "step": 27850
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6820006321112518e-06,
+      "loss": 3.1489,
+      "step": 27851
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.681953223767383e-06,
+      "loss": 2.778,
+      "step": 27852
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6819058154235147e-06,
+      "loss": 2.7667,
+      "step": 27853
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.681858407079646e-06,
+      "loss": 2.4179,
+      "step": 27854
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6818109987357775e-06,
+      "loss": 2.6789,
+      "step": 27855
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6817635903919088e-06,
+      "loss": 1.8645,
+      "step": 27856
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6817161820480405e-06,
+      "loss": 3.4661,
+      "step": 27857
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.681668773704172e-06,
+      "loss": 2.3131,
+      "step": 27858
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6816213653603033e-06,
+      "loss": 2.5678,
+      "step": 27859
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.681573957016435e-06,
+      "loss": 2.3933,
+      "step": 27860
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6815265486725665e-06,
+      "loss": 3.2584,
+      "step": 27861
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6814791403286977e-06,
+      "loss": 2.2739,
+      "step": 27862
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6814317319848295e-06,
+      "loss": 1.1531,
+      "step": 27863
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.681384323640961e-06,
+      "loss": 1.509,
+      "step": 27864
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6813369152970922e-06,
+      "loss": 2.2945,
+      "step": 27865
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.681289506953224e-06,
+      "loss": 1.4017,
+      "step": 27866
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6812420986093552e-06,
+      "loss": 2.5364,
+      "step": 27867
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6811946902654867e-06,
+      "loss": 2.7979,
+      "step": 27868
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6811472819216184e-06,
+      "loss": 2.466,
+      "step": 27869
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6810998735777497e-06,
+      "loss": 2.442,
+      "step": 27870
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6810524652338812e-06,
+      "loss": 2.3209,
+      "step": 27871
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6810050568900129e-06,
+      "loss": 2.4132,
+      "step": 27872
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6809576485461442e-06,
+      "loss": 3.5574,
+      "step": 27873
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6809102402022757e-06,
+      "loss": 2.2857,
+      "step": 27874
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6808628318584074e-06,
+      "loss": 1.8178,
+      "step": 27875
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6808154235145386e-06,
+      "loss": 0.8989,
+      "step": 27876
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.68076801517067e-06,
+      "loss": 2.5059,
+      "step": 27877
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6807206068268014e-06,
+      "loss": 1.1955,
+      "step": 27878
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6806731984829331e-06,
+      "loss": 2.5017,
+      "step": 27879
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6806257901390644e-06,
+      "loss": 2.0236,
+      "step": 27880
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6805783817951959e-06,
+      "loss": 2.3759,
+      "step": 27881
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6805309734513276e-06,
+      "loss": 1.8424,
+      "step": 27882
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6804835651074589e-06,
+      "loss": 1.8124,
+      "step": 27883
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6804361567635904e-06,
+      "loss": 2.7375,
+      "step": 27884
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.680388748419722e-06,
+      "loss": 3.3512,
+      "step": 27885
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6803413400758533e-06,
+      "loss": 3.2927,
+      "step": 27886
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6802939317319848e-06,
+      "loss": 1.4929,
+      "step": 27887
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6802465233881165e-06,
+      "loss": 2.6129,
+      "step": 27888
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6801991150442478e-06,
+      "loss": 3.3597,
+      "step": 27889
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6801517067003791e-06,
+      "loss": 2.1772,
+      "step": 27890
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6801042983565108e-06,
+      "loss": 2.4794,
+      "step": 27891
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6800568900126423e-06,
+      "loss": 1.6193,
+      "step": 27892
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6800094816687736e-06,
+      "loss": 1.6643,
+      "step": 27893
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6799620733249053e-06,
+      "loss": 1.1965,
+      "step": 27894
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.6799146649810368e-06,
+      "loss": 1.8811,
+      "step": 27895
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.679867256637168e-06,
+      "loss": 2.1697,
+      "step": 27896
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6798198482932998e-06,
+      "loss": 1.3086,
+      "step": 27897
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6797724399494313e-06,
+      "loss": 1.0019,
+      "step": 27898
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6797250316055625e-06,
+      "loss": 2.7238,
+      "step": 27899
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.679677623261694e-06,
+      "loss": 2.498,
+      "step": 27900
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6796302149178257e-06,
+      "loss": 2.9313,
+      "step": 27901
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.679582806573957e-06,
+      "loss": 2.0842,
+      "step": 27902
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6795353982300883e-06,
+      "loss": 2.6453,
+      "step": 27903
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.67948798988622e-06,
+      "loss": 2.6345,
+      "step": 27904
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6794405815423515e-06,
+      "loss": 3.147,
+      "step": 27905
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6793931731984828e-06,
+      "loss": 2.6156,
+      "step": 27906
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6793457648546145e-06,
+      "loss": 2.18,
+      "step": 27907
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.679298356510746e-06,
+      "loss": 3.2991,
+      "step": 27908
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6792509481668772e-06,
+      "loss": 1.5124,
+      "step": 27909
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.679203539823009e-06,
+      "loss": 2.8711,
+      "step": 27910
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6791561314791404e-06,
+      "loss": 3.3958,
+      "step": 27911
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6791087231352717e-06,
+      "loss": 1.5833,
+      "step": 27912
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6790613147914034e-06,
+      "loss": 2.1892,
+      "step": 27913
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.679013906447535e-06,
+      "loss": 2.0594,
+      "step": 27914
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6789664981036662e-06,
+      "loss": 2.692,
+      "step": 27915
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.678919089759798e-06,
+      "loss": 2.2421,
+      "step": 27916
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6788716814159292e-06,
+      "loss": 2.3389,
+      "step": 27917
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6788242730720607e-06,
+      "loss": 2.5109,
+      "step": 27918
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6787768647281924e-06,
+      "loss": 2.7789,
+      "step": 27919
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6787294563843237e-06,
+      "loss": 3.1293,
+      "step": 27920
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6786820480404552e-06,
+      "loss": 3.615,
+      "step": 27921
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6786346396965864e-06,
+      "loss": 1.3962,
+      "step": 27922
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6785872313527181e-06,
+      "loss": 1.8698,
+      "step": 27923
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6785398230088496e-06,
+      "loss": 2.5945,
+      "step": 27924
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.678492414664981e-06,
+      "loss": 1.2342,
+      "step": 27925
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6784450063211126e-06,
+      "loss": 3.5151,
+      "step": 27926
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.678397597977244e-06,
+      "loss": 2.5904,
+      "step": 27927
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6783501896333754e-06,
+      "loss": 1.3696,
+      "step": 27928
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.678302781289507e-06,
+      "loss": 2.9926,
+      "step": 27929
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6782553729456384e-06,
+      "loss": 2.1771,
+      "step": 27930
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6782079646017699e-06,
+      "loss": 2.4608,
+      "step": 27931
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6781605562579016e-06,
+      "loss": 3.0503,
+      "step": 27932
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6781131479140328e-06,
+      "loss": 3.6258,
+      "step": 27933
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6780657395701643e-06,
+      "loss": 3.5424,
+      "step": 27934
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.678018331226296e-06,
+      "loss": 2.8633,
+      "step": 27935
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6779709228824273e-06,
+      "loss": 1.3222,
+      "step": 27936
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6779235145385588e-06,
+      "loss": 1.9184,
+      "step": 27937
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6778761061946905e-06,
+      "loss": 2.7311,
+      "step": 27938
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6778286978508218e-06,
+      "loss": 1.4013,
+      "step": 27939
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6777812895069533e-06,
+      "loss": 2.6091,
+      "step": 27940
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.677733881163085e-06,
+      "loss": 1.3326,
+      "step": 27941
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6776864728192163e-06,
+      "loss": 2.9239,
+      "step": 27942
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6776390644753476e-06,
+      "loss": 2.7774,
+      "step": 27943
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.677591656131479e-06,
+      "loss": 1.4843,
+      "step": 27944
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6775442477876108e-06,
+      "loss": 0.7532,
+      "step": 27945
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.677496839443742e-06,
+      "loss": 2.9285,
+      "step": 27946
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6774494310998735e-06,
+      "loss": 3.0162,
+      "step": 27947
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6774020227560052e-06,
+      "loss": 2.9384,
+      "step": 27948
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6773546144121365e-06,
+      "loss": 2.7324,
+      "step": 27949
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.677307206068268e-06,
+      "loss": 2.1516,
+      "step": 27950
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6772597977243997e-06,
+      "loss": 0.3709,
+      "step": 27951
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.677212389380531e-06,
+      "loss": 1.5446,
+      "step": 27952
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6771649810366625e-06,
+      "loss": 2.3192,
+      "step": 27953
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6771175726927942e-06,
+      "loss": 3.743,
+      "step": 27954
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6770701643489255e-06,
+      "loss": 3.806,
+      "step": 27955
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6770227560050567e-06,
+      "loss": 3.1393,
+      "step": 27956
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6769753476611884e-06,
+      "loss": 2.1417,
+      "step": 27957
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.67692793931732e-06,
+      "loss": 0.8877,
+      "step": 27958
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.6768805309734512e-06,
+      "loss": 3.0331,
+      "step": 27959
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.676833122629583e-06,
+      "loss": 2.5661,
+      "step": 27960
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6767857142857144e-06,
+      "loss": 2.5449,
+      "step": 27961
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6767383059418457e-06,
+      "loss": 1.4828,
+      "step": 27962
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6766908975979774e-06,
+      "loss": 1.4063,
+      "step": 27963
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6766434892541089e-06,
+      "loss": 3.9762,
+      "step": 27964
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6765960809102402e-06,
+      "loss": 3.5722,
+      "step": 27965
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6765486725663717e-06,
+      "loss": 1.8724,
+      "step": 27966
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6765012642225034e-06,
+      "loss": 3.9761,
+      "step": 27967
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6764538558786346e-06,
+      "loss": 1.9699,
+      "step": 27968
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.676406447534766e-06,
+      "loss": 3.3347,
+      "step": 27969
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6763590391908976e-06,
+      "loss": 1.7735,
+      "step": 27970
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6763116308470291e-06,
+      "loss": 2.2842,
+      "step": 27971
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6762642225031604e-06,
+      "loss": 2.7089,
+      "step": 27972
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6762168141592921e-06,
+      "loss": 1.1985,
+      "step": 27973
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6761694058154236e-06,
+      "loss": 4.1151,
+      "step": 27974
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6761219974715549e-06,
+      "loss": 3.801,
+      "step": 27975
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6760745891276866e-06,
+      "loss": 1.2081,
+      "step": 27976
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.676027180783818e-06,
+      "loss": 1.9599,
+      "step": 27977
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6759797724399494e-06,
+      "loss": 2.4362,
+      "step": 27978
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.675932364096081e-06,
+      "loss": 2.3641,
+      "step": 27979
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6758849557522126e-06,
+      "loss": 1.7741,
+      "step": 27980
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6758375474083438e-06,
+      "loss": 2.729,
+      "step": 27981
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6757901390644755e-06,
+      "loss": 1.2229,
+      "step": 27982
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6757427307206068e-06,
+      "loss": 2.3284,
+      "step": 27983
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6756953223767383e-06,
+      "loss": 2.3048,
+      "step": 27984
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.67564791403287e-06,
+      "loss": 1.4288,
+      "step": 27985
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6756005056890013e-06,
+      "loss": 2.7291,
+      "step": 27986
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6755530973451328e-06,
+      "loss": 1.9484,
+      "step": 27987
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.675505689001264e-06,
+      "loss": 3.3167,
+      "step": 27988
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6754582806573958e-06,
+      "loss": 2.0948,
+      "step": 27989
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6754108723135273e-06,
+      "loss": 1.2467,
+      "step": 27990
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6753634639696585e-06,
+      "loss": 0.9646,
+      "step": 27991
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6753160556257902e-06,
+      "loss": 2.0221,
+      "step": 27992
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6752686472819217e-06,
+      "loss": 2.8228,
+      "step": 27993
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.675221238938053e-06,
+      "loss": 2.6448,
+      "step": 27994
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6751738305941847e-06,
+      "loss": 2.4713,
+      "step": 27995
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.675126422250316e-06,
+      "loss": 2.7641,
+      "step": 27996
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6750790139064475e-06,
+      "loss": 3.669,
+      "step": 27997
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6750316055625792e-06,
+      "loss": 0.7666,
+      "step": 27998
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6749841972187105e-06,
+      "loss": 2.6356,
+      "step": 27999
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.674936788874842e-06,
+      "loss": 2.5106,
+      "step": 28000
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6748893805309737e-06,
+      "loss": 1.5638,
+      "step": 28001
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.674841972187105e-06,
+      "loss": 2.371,
+      "step": 28002
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6747945638432365e-06,
+      "loss": 3.1829,
+      "step": 28003
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6747471554993682e-06,
+      "loss": 3.0465,
+      "step": 28004
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6746997471554994e-06,
+      "loss": 3.0394,
+      "step": 28005
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.674652338811631e-06,
+      "loss": 3.3787,
+      "step": 28006
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6746049304677626e-06,
+      "loss": 2.8665,
+      "step": 28007
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.674557522123894e-06,
+      "loss": 1.7839,
+      "step": 28008
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6745101137800252e-06,
+      "loss": 2.1064,
+      "step": 28009
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6744627054361567e-06,
+      "loss": 1.6635,
+      "step": 28010
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6744152970922884e-06,
+      "loss": 1.2164,
+      "step": 28011
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6743678887484197e-06,
+      "loss": 2.9208,
+      "step": 28012
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6743204804045512e-06,
+      "loss": 2.878,
+      "step": 28013
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6742730720606829e-06,
+      "loss": 3.5752,
+      "step": 28014
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6742256637168141e-06,
+      "loss": 3.8937,
+      "step": 28015
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6741782553729456e-06,
+      "loss": 2.4731,
+      "step": 28016
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6741308470290773e-06,
+      "loss": 1.6989,
+      "step": 28017
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6740834386852086e-06,
+      "loss": 2.0877,
+      "step": 28018
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.67403603034134e-06,
+      "loss": 2.8049,
+      "step": 28019
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6739886219974718e-06,
+      "loss": 2.9388,
+      "step": 28020
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.673941213653603e-06,
+      "loss": 2.1724,
+      "step": 28021
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.6738938053097344e-06,
+      "loss": 2.7099,
+      "step": 28022
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.673846396965866e-06,
+      "loss": 2.7963,
+      "step": 28023
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6737989886219976e-06,
+      "loss": 1.3699,
+      "step": 28024
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6737515802781289e-06,
+      "loss": 2.7492,
+      "step": 28025
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6737041719342606e-06,
+      "loss": 3.4402,
+      "step": 28026
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.673656763590392e-06,
+      "loss": 3.0536,
+      "step": 28027
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6736093552465233e-06,
+      "loss": 2.9537,
+      "step": 28028
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.673561946902655e-06,
+      "loss": 2.2057,
+      "step": 28029
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6735145385587865e-06,
+      "loss": 2.7708,
+      "step": 28030
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6734671302149178e-06,
+      "loss": 2.1322,
+      "step": 28031
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.673419721871049e-06,
+      "loss": 2.4409,
+      "step": 28032
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6733723135271808e-06,
+      "loss": 2.3386,
+      "step": 28033
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6733249051833123e-06,
+      "loss": 2.3376,
+      "step": 28034
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6732774968394436e-06,
+      "loss": 1.9845,
+      "step": 28035
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6732300884955753e-06,
+      "loss": 1.4559,
+      "step": 28036
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6731826801517068e-06,
+      "loss": 2.16,
+      "step": 28037
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.673135271807838e-06,
+      "loss": 1.872,
+      "step": 28038
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6730878634639697e-06,
+      "loss": 2.7612,
+      "step": 28039
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6730404551201012e-06,
+      "loss": 2.0639,
+      "step": 28040
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6729930467762325e-06,
+      "loss": 3.9673,
+      "step": 28041
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6729456384323642e-06,
+      "loss": 1.8206,
+      "step": 28042
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6728982300884957e-06,
+      "loss": 1.503,
+      "step": 28043
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.672850821744627e-06,
+      "loss": 2.0021,
+      "step": 28044
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6728034134007587e-06,
+      "loss": 2.6575,
+      "step": 28045
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.67275600505689e-06,
+      "loss": 2.9993,
+      "step": 28046
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6727085967130215e-06,
+      "loss": 3.0032,
+      "step": 28047
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6726611883691532e-06,
+      "loss": 2.6793,
+      "step": 28048
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6726137800252845e-06,
+      "loss": 2.2433,
+      "step": 28049
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.672566371681416e-06,
+      "loss": 2.539,
+      "step": 28050
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6725189633375477e-06,
+      "loss": 1.7527,
+      "step": 28051
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.672471554993679e-06,
+      "loss": 3.2208,
+      "step": 28052
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6724241466498104e-06,
+      "loss": 2.193,
+      "step": 28053
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6723767383059417e-06,
+      "loss": 2.1597,
+      "step": 28054
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6723293299620734e-06,
+      "loss": 2.321,
+      "step": 28055
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.672281921618205e-06,
+      "loss": 1.4021,
+      "step": 28056
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6722345132743362e-06,
+      "loss": 2.084,
+      "step": 28057
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6721871049304679e-06,
+      "loss": 2.617,
+      "step": 28058
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6721396965865992e-06,
+      "loss": 1.156,
+      "step": 28059
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6720922882427307e-06,
+      "loss": 2.4644,
+      "step": 28060
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6720448798988624e-06,
+      "loss": 3.5782,
+      "step": 28061
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6719974715549936e-06,
+      "loss": 1.1611,
+      "step": 28062
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6719500632111251e-06,
+      "loss": 1.8769,
+      "step": 28063
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6719026548672568e-06,
+      "loss": 1.7455,
+      "step": 28064
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6718552465233881e-06,
+      "loss": 2.2023,
+      "step": 28065
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6718078381795196e-06,
+      "loss": 3.0521,
+      "step": 28066
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6717604298356513e-06,
+      "loss": 1.6207,
+      "step": 28067
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6717130214917826e-06,
+      "loss": 2.5849,
+      "step": 28068
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.671665613147914e-06,
+      "loss": 2.5381,
+      "step": 28069
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6716182048040458e-06,
+      "loss": 2.4895,
+      "step": 28070
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.671570796460177e-06,
+      "loss": 1.9841,
+      "step": 28071
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6715233881163084e-06,
+      "loss": 2.4927,
+      "step": 28072
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.67147597977244e-06,
+      "loss": 3.4746,
+      "step": 28073
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6714285714285715e-06,
+      "loss": 2.8098,
+      "step": 28074
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6713811630847028e-06,
+      "loss": 1.9269,
+      "step": 28075
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6713337547408343e-06,
+      "loss": 2.7085,
+      "step": 28076
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.671286346396966e-06,
+      "loss": 2.4113,
+      "step": 28077
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6712389380530973e-06,
+      "loss": 2.8838,
+      "step": 28078
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6711915297092288e-06,
+      "loss": 3.0495,
+      "step": 28079
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6711441213653605e-06,
+      "loss": 3.0793,
+      "step": 28080
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6710967130214918e-06,
+      "loss": 2.2605,
+      "step": 28081
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6710493046776233e-06,
+      "loss": 1.9141,
+      "step": 28082
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.671001896333755e-06,
+      "loss": 1.7205,
+      "step": 28083
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6709544879898863e-06,
+      "loss": 2.8699,
+      "step": 28084
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6709070796460175e-06,
+      "loss": 3.1135,
+      "step": 28085
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.6708596713021492e-06,
+      "loss": 2.0593,
+      "step": 28086
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6708122629582807e-06,
+      "loss": 2.1322,
+      "step": 28087
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.670764854614412e-06,
+      "loss": 3.2356,
+      "step": 28088
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6707174462705437e-06,
+      "loss": 1.0589,
+      "step": 28089
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6706700379266752e-06,
+      "loss": 2.3429,
+      "step": 28090
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6706226295828065e-06,
+      "loss": 1.04,
+      "step": 28091
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6705752212389382e-06,
+      "loss": 2.1813,
+      "step": 28092
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6705278128950697e-06,
+      "loss": 2.9558,
+      "step": 28093
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.670480404551201e-06,
+      "loss": 1.5855,
+      "step": 28094
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6704329962073327e-06,
+      "loss": 3.0335,
+      "step": 28095
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6703855878634642e-06,
+      "loss": 3.8052,
+      "step": 28096
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6703381795195954e-06,
+      "loss": 2.7282,
+      "step": 28097
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6702907711757267e-06,
+      "loss": 3.5351,
+      "step": 28098
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6702433628318584e-06,
+      "loss": 1.6314,
+      "step": 28099
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.67019595448799e-06,
+      "loss": 1.4138,
+      "step": 28100
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6701485461441212e-06,
+      "loss": 0.9178,
+      "step": 28101
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.670101137800253e-06,
+      "loss": 1.4286,
+      "step": 28102
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6700537294563844e-06,
+      "loss": 3.5217,
+      "step": 28103
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6700063211125157e-06,
+      "loss": 1.5768,
+      "step": 28104
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6699589127686474e-06,
+      "loss": 1.9959,
+      "step": 28105
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6699115044247789e-06,
+      "loss": 2.0344,
+      "step": 28106
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6698640960809102e-06,
+      "loss": 3.9264,
+      "step": 28107
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6698166877370419e-06,
+      "loss": 2.9806,
+      "step": 28108
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6697692793931734e-06,
+      "loss": 2.379,
+      "step": 28109
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6697218710493046e-06,
+      "loss": 2.4858,
+      "step": 28110
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6696744627054363e-06,
+      "loss": 1.9339,
+      "step": 28111
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6696270543615676e-06,
+      "loss": 2.6569,
+      "step": 28112
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6695796460176991e-06,
+      "loss": 1.6472,
+      "step": 28113
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6695322376738308e-06,
+      "loss": 1.0381,
+      "step": 28114
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.669484829329962e-06,
+      "loss": 3.0866,
+      "step": 28115
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6694374209860936e-06,
+      "loss": 1.4373,
+      "step": 28116
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6693900126422253e-06,
+      "loss": 1.8766,
+      "step": 28117
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6693426042983566e-06,
+      "loss": 2.6796,
+      "step": 28118
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.669295195954488e-06,
+      "loss": 3.52,
+      "step": 28119
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6692477876106193e-06,
+      "loss": 2.6527,
+      "step": 28120
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.669200379266751e-06,
+      "loss": 2.5502,
+      "step": 28121
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6691529709228825e-06,
+      "loss": 2.6069,
+      "step": 28122
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6691055625790138e-06,
+      "loss": 2.711,
+      "step": 28123
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6690581542351455e-06,
+      "loss": 2.6836,
+      "step": 28124
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6690107458912768e-06,
+      "loss": 2.6533,
+      "step": 28125
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6689633375474083e-06,
+      "loss": 1.4474,
+      "step": 28126
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.66891592920354e-06,
+      "loss": 2.4135,
+      "step": 28127
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6688685208596713e-06,
+      "loss": 1.4777,
+      "step": 28128
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6688211125158028e-06,
+      "loss": 2.5349,
+      "step": 28129
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6687737041719345e-06,
+      "loss": 1.244,
+      "step": 28130
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6687262958280658e-06,
+      "loss": 3.1433,
+      "step": 28131
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6686788874841972e-06,
+      "loss": 2.9229,
+      "step": 28132
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.668631479140329e-06,
+      "loss": 1.6241,
+      "step": 28133
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6685840707964602e-06,
+      "loss": 2.3122,
+      "step": 28134
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6685366624525917e-06,
+      "loss": 2.2873,
+      "step": 28135
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6684892541087234e-06,
+      "loss": 3.0175,
+      "step": 28136
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6684418457648547e-06,
+      "loss": 1.7248,
+      "step": 28137
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.668394437420986e-06,
+      "loss": 2.5063,
+      "step": 28138
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6683470290771177e-06,
+      "loss": 1.5283,
+      "step": 28139
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6682996207332492e-06,
+      "loss": 2.3327,
+      "step": 28140
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6682522123893805e-06,
+      "loss": 2.2255,
+      "step": 28141
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.668204804045512e-06,
+      "loss": 3.0313,
+      "step": 28142
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6681573957016437e-06,
+      "loss": 3.1842,
+      "step": 28143
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.668109987357775e-06,
+      "loss": 1.352,
+      "step": 28144
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6680625790139064e-06,
+      "loss": 2.698,
+      "step": 28145
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6680151706700381e-06,
+      "loss": 2.2244,
+      "step": 28146
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6679677623261694e-06,
+      "loss": 2.4844,
+      "step": 28147
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6679203539823007e-06,
+      "loss": 3.0088,
+      "step": 28148
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.6678729456384326e-06,
+      "loss": 3.247,
+      "step": 28149
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.667825537294564e-06,
+      "loss": 1.7997,
+      "step": 28150
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6677781289506952e-06,
+      "loss": 2.6288,
+      "step": 28151
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6677307206068269e-06,
+      "loss": 2.4694,
+      "step": 28152
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6676833122629584e-06,
+      "loss": 2.5043,
+      "step": 28153
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6676359039190897e-06,
+      "loss": 2.7173,
+      "step": 28154
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6675884955752214e-06,
+      "loss": 2.3264,
+      "step": 28155
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6675410872313528e-06,
+      "loss": 3.222,
+      "step": 28156
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6674936788874841e-06,
+      "loss": 3.1959,
+      "step": 28157
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6674462705436158e-06,
+      "loss": 3.8355,
+      "step": 28158
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6673988621997473e-06,
+      "loss": 2.3916,
+      "step": 28159
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6673514538558786e-06,
+      "loss": 1.9795,
+      "step": 28160
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6673040455120103e-06,
+      "loss": 2.6107,
+      "step": 28161
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6672566371681416e-06,
+      "loss": 3.0893,
+      "step": 28162
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.667209228824273e-06,
+      "loss": 2.1773,
+      "step": 28163
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6671618204804044e-06,
+      "loss": 1.4098,
+      "step": 28164
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.667114412136536e-06,
+      "loss": 2.5901,
+      "step": 28165
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6670670037926676e-06,
+      "loss": 3.1324,
+      "step": 28166
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6670195954487988e-06,
+      "loss": 3.1635,
+      "step": 28167
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6669721871049305e-06,
+      "loss": 1.6726,
+      "step": 28168
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.666924778761062e-06,
+      "loss": 1.3952,
+      "step": 28169
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6668773704171933e-06,
+      "loss": 2.2864,
+      "step": 28170
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.666829962073325e-06,
+      "loss": 2.2693,
+      "step": 28171
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6667825537294565e-06,
+      "loss": 2.5123,
+      "step": 28172
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6667351453855878e-06,
+      "loss": 3.634,
+      "step": 28173
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6666877370417195e-06,
+      "loss": 2.0582,
+      "step": 28174
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6666403286978508e-06,
+      "loss": 2.5527,
+      "step": 28175
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6665929203539823e-06,
+      "loss": 1.9447,
+      "step": 28176
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.666545512010114e-06,
+      "loss": 2.7388,
+      "step": 28177
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6664981036662453e-06,
+      "loss": 3.1654,
+      "step": 28178
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6664506953223767e-06,
+      "loss": 3.1958,
+      "step": 28179
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6664032869785084e-06,
+      "loss": 2.3006,
+      "step": 28180
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6663558786346397e-06,
+      "loss": 1.5939,
+      "step": 28181
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6663084702907712e-06,
+      "loss": 3.9728,
+      "step": 28182
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6662610619469025e-06,
+      "loss": 2.1298,
+      "step": 28183
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6662136536030342e-06,
+      "loss": 1.4184,
+      "step": 28184
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6661662452591657e-06,
+      "loss": 3.1271,
+      "step": 28185
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.666118836915297e-06,
+      "loss": 2.1211,
+      "step": 28186
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6660714285714287e-06,
+      "loss": 1.8502,
+      "step": 28187
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.66602402022756e-06,
+      "loss": 3.5811,
+      "step": 28188
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6659766118836915e-06,
+      "loss": 2.1063,
+      "step": 28189
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6659292035398232e-06,
+      "loss": 2.858,
+      "step": 28190
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6658817951959544e-06,
+      "loss": 1.7883,
+      "step": 28191
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.665834386852086e-06,
+      "loss": 2.2384,
+      "step": 28192
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6657869785082176e-06,
+      "loss": 2.0948,
+      "step": 28193
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.665739570164349e-06,
+      "loss": 2.6045,
+      "step": 28194
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6656921618204804e-06,
+      "loss": 3.1181,
+      "step": 28195
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6656447534766121e-06,
+      "loss": 2.1198,
+      "step": 28196
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6655973451327434e-06,
+      "loss": 2.8698,
+      "step": 28197
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6655499367888749e-06,
+      "loss": 2.2248,
+      "step": 28198
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6655025284450066e-06,
+      "loss": 2.3869,
+      "step": 28199
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6654551201011379e-06,
+      "loss": 2.563,
+      "step": 28200
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6654077117572692e-06,
+      "loss": 2.5302,
+      "step": 28201
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6653603034134009e-06,
+      "loss": 2.1456,
+      "step": 28202
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6653128950695323e-06,
+      "loss": 1.4395,
+      "step": 28203
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6652654867256636e-06,
+      "loss": 1.8153,
+      "step": 28204
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6652180783817951e-06,
+      "loss": 3.7055,
+      "step": 28205
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6651706700379268e-06,
+      "loss": 2.7598,
+      "step": 28206
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.665123261694058e-06,
+      "loss": 2.3762,
+      "step": 28207
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6650758533501896e-06,
+      "loss": 1.5639,
+      "step": 28208
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6650284450063213e-06,
+      "loss": 2.7096,
+      "step": 28209
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6649810366624526e-06,
+      "loss": 2.2568,
+      "step": 28210
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.664933628318584e-06,
+      "loss": 2.1391,
+      "step": 28211
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.6648862199747158e-06,
+      "loss": 2.331,
+      "step": 28212
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.664838811630847e-06,
+      "loss": 2.5553,
+      "step": 28213
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6647914032869783e-06,
+      "loss": 2.2618,
+      "step": 28214
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.66474399494311e-06,
+      "loss": 1.9967,
+      "step": 28215
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6646965865992415e-06,
+      "loss": 3.637,
+      "step": 28216
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6646491782553728e-06,
+      "loss": 2.7483,
+      "step": 28217
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6646017699115045e-06,
+      "loss": 2.8586,
+      "step": 28218
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.664554361567636e-06,
+      "loss": 2.1155,
+      "step": 28219
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6645069532237673e-06,
+      "loss": 2.8538,
+      "step": 28220
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.664459544879899e-06,
+      "loss": 2.6831,
+      "step": 28221
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6644121365360305e-06,
+      "loss": 2.8634,
+      "step": 28222
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6643647281921618e-06,
+      "loss": 1.8438,
+      "step": 28223
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6643173198482935e-06,
+      "loss": 2.3557,
+      "step": 28224
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.664269911504425e-06,
+      "loss": 2.0934,
+      "step": 28225
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6642225031605562e-06,
+      "loss": 2.7162,
+      "step": 28226
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6641750948166875e-06,
+      "loss": 2.5099,
+      "step": 28227
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6641276864728192e-06,
+      "loss": 3.1486,
+      "step": 28228
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6640802781289507e-06,
+      "loss": 2.8709,
+      "step": 28229
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.664032869785082e-06,
+      "loss": 3.0739,
+      "step": 28230
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6639854614412137e-06,
+      "loss": 1.2557,
+      "step": 28231
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6639380530973452e-06,
+      "loss": 0.6656,
+      "step": 28232
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6638906447534765e-06,
+      "loss": 1.964,
+      "step": 28233
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6638432364096082e-06,
+      "loss": 2.4857,
+      "step": 28234
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6637958280657397e-06,
+      "loss": 2.1876,
+      "step": 28235
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.663748419721871e-06,
+      "loss": 3.1048,
+      "step": 28236
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6637010113780027e-06,
+      "loss": 1.419,
+      "step": 28237
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6636536030341341e-06,
+      "loss": 2.7473,
+      "step": 28238
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6636061946902654e-06,
+      "loss": 3.0876,
+      "step": 28239
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6635587863463971e-06,
+      "loss": 1.7,
+      "step": 28240
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6635113780025284e-06,
+      "loss": 2.4966,
+      "step": 28241
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.66346396965866e-06,
+      "loss": 1.6036,
+      "step": 28242
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6634165613147916e-06,
+      "loss": 2.0802,
+      "step": 28243
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6633691529709229e-06,
+      "loss": 1.5759,
+      "step": 28244
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6633217446270544e-06,
+      "loss": 3.0345,
+      "step": 28245
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.663274336283186e-06,
+      "loss": 3.4504,
+      "step": 28246
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6632269279393174e-06,
+      "loss": 2.1722,
+      "step": 28247
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6631795195954489e-06,
+      "loss": 3.1166,
+      "step": 28248
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6631321112515801e-06,
+      "loss": 2.413,
+      "step": 28249
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6630847029077118e-06,
+      "loss": 3.1119,
+      "step": 28250
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6630372945638433e-06,
+      "loss": 2.7097,
+      "step": 28251
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6629898862199746e-06,
+      "loss": 2.7897,
+      "step": 28252
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6629424778761063e-06,
+      "loss": 0.5082,
+      "step": 28253
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6628950695322376e-06,
+      "loss": 2.5255,
+      "step": 28254
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.662847661188369e-06,
+      "loss": 2.2685,
+      "step": 28255
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6628002528445008e-06,
+      "loss": 2.7929,
+      "step": 28256
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.662752844500632e-06,
+      "loss": 2.2532,
+      "step": 28257
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6627054361567636e-06,
+      "loss": 2.6981,
+      "step": 28258
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6626580278128953e-06,
+      "loss": 2.6335,
+      "step": 28259
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6626106194690266e-06,
+      "loss": 3.2128,
+      "step": 28260
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.662563211125158e-06,
+      "loss": 0.8513,
+      "step": 28261
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6625158027812897e-06,
+      "loss": 3.8107,
+      "step": 28262
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.662468394437421e-06,
+      "loss": 1.8356,
+      "step": 28263
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6624209860935525e-06,
+      "loss": 1.5892,
+      "step": 28264
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6623735777496842e-06,
+      "loss": 2.3024,
+      "step": 28265
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6623261694058155e-06,
+      "loss": 2.0996,
+      "step": 28266
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6622787610619468e-06,
+      "loss": 2.873,
+      "step": 28267
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6622313527180785e-06,
+      "loss": 3.217,
+      "step": 28268
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.66218394437421e-06,
+      "loss": 1.6863,
+      "step": 28269
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6621365360303413e-06,
+      "loss": 2.8893,
+      "step": 28270
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6620891276864728e-06,
+      "loss": 2.1703,
+      "step": 28271
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6620417193426045e-06,
+      "loss": 1.3107,
+      "step": 28272
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6619943109987357e-06,
+      "loss": 3.4332,
+      "step": 28273
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6619469026548672e-06,
+      "loss": 3.2756,
+      "step": 28274
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.661899494310999e-06,
+      "loss": 1.4043,
+      "step": 28275
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.6618520859671302e-06,
+      "loss": 1.5687,
+      "step": 28276
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6618046776232617e-06,
+      "loss": 3.3447,
+      "step": 28277
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6617572692793934e-06,
+      "loss": 2.4607,
+      "step": 28278
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6617098609355247e-06,
+      "loss": 2.9036,
+      "step": 28279
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.661662452591656e-06,
+      "loss": 2.3832,
+      "step": 28280
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6616150442477877e-06,
+      "loss": 4.1265,
+      "step": 28281
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6615676359039192e-06,
+      "loss": 1.5353,
+      "step": 28282
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6615202275600504e-06,
+      "loss": 2.3197,
+      "step": 28283
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6614728192161822e-06,
+      "loss": 2.1927,
+      "step": 28284
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6614254108723136e-06,
+      "loss": 1.6493,
+      "step": 28285
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.661378002528445e-06,
+      "loss": 1.5443,
+      "step": 28286
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6613305941845766e-06,
+      "loss": 3.321,
+      "step": 28287
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6612831858407081e-06,
+      "loss": 2.9974,
+      "step": 28288
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6612357774968394e-06,
+      "loss": 2.4728,
+      "step": 28289
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6611883691529711e-06,
+      "loss": 3.8881,
+      "step": 28290
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6611409608091026e-06,
+      "loss": 2.165,
+      "step": 28291
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6610935524652339e-06,
+      "loss": 2.4156,
+      "step": 28292
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6610461441213652e-06,
+      "loss": 2.1383,
+      "step": 28293
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6609987357774969e-06,
+      "loss": 2.3809,
+      "step": 28294
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6609513274336284e-06,
+      "loss": 3.5849,
+      "step": 28295
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6609039190897596e-06,
+      "loss": 3.1378,
+      "step": 28296
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6608565107458913e-06,
+      "loss": 2.4317,
+      "step": 28297
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6608091024020228e-06,
+      "loss": 2.8445,
+      "step": 28298
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6607616940581541e-06,
+      "loss": 2.3182,
+      "step": 28299
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6607142857142858e-06,
+      "loss": 3.547,
+      "step": 28300
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6606668773704173e-06,
+      "loss": 3.0151,
+      "step": 28301
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6606194690265486e-06,
+      "loss": 2.9091,
+      "step": 28302
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6605720606826803e-06,
+      "loss": 1.1774,
+      "step": 28303
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6605246523388116e-06,
+      "loss": 4.1994,
+      "step": 28304
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.660477243994943e-06,
+      "loss": 1.8619,
+      "step": 28305
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6604298356510748e-06,
+      "loss": 1.8564,
+      "step": 28306
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.660382427307206e-06,
+      "loss": 1.1058,
+      "step": 28307
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6603350189633375e-06,
+      "loss": 1.8312,
+      "step": 28308
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6602876106194692e-06,
+      "loss": 1.1569,
+      "step": 28309
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6602402022756005e-06,
+      "loss": 2.306,
+      "step": 28310
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.660192793931732e-06,
+      "loss": 2.7855,
+      "step": 28311
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6601453855878637e-06,
+      "loss": 2.2,
+      "step": 28312
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.660097977243995e-06,
+      "loss": 0.7903,
+      "step": 28313
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6600505689001265e-06,
+      "loss": 2.3028,
+      "step": 28314
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6600031605562578e-06,
+      "loss": 2.4177,
+      "step": 28315
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6599557522123895e-06,
+      "loss": 2.9733,
+      "step": 28316
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6599083438685208e-06,
+      "loss": 2.732,
+      "step": 28317
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6598609355246523e-06,
+      "loss": 3.2754,
+      "step": 28318
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.659813527180784e-06,
+      "loss": 2.512,
+      "step": 28319
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6597661188369152e-06,
+      "loss": 2.0235,
+      "step": 28320
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6597187104930467e-06,
+      "loss": 2.6968,
+      "step": 28321
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6596713021491784e-06,
+      "loss": 2.3296,
+      "step": 28322
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6596238938053097e-06,
+      "loss": 3.4336,
+      "step": 28323
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6595764854614412e-06,
+      "loss": 2.0241,
+      "step": 28324
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.659529077117573e-06,
+      "loss": 1.7223,
+      "step": 28325
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6594816687737042e-06,
+      "loss": 2.5854,
+      "step": 28326
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6594342604298357e-06,
+      "loss": 3.5999,
+      "step": 28327
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6593868520859674e-06,
+      "loss": 2.5105,
+      "step": 28328
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6593394437420987e-06,
+      "loss": 2.365,
+      "step": 28329
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.65929203539823e-06,
+      "loss": 1.7983,
+      "step": 28330
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6592446270543617e-06,
+      "loss": 3.0532,
+      "step": 28331
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6591972187104931e-06,
+      "loss": 2.9614,
+      "step": 28332
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6591498103666244e-06,
+      "loss": 2.5008,
+      "step": 28333
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6591024020227561e-06,
+      "loss": 1.5953,
+      "step": 28334
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6590549936788876e-06,
+      "loss": 2.8052,
+      "step": 28335
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.659007585335019e-06,
+      "loss": 3.5556,
+      "step": 28336
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6589601769911504e-06,
+      "loss": 1.0778,
+      "step": 28337
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.658912768647282e-06,
+      "loss": 1.9776,
+      "step": 28338
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6588653603034134e-06,
+      "loss": 3.4996,
+      "step": 28339
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.6588179519595449e-06,
+      "loss": 1.9326,
+      "step": 28340
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6587705436156766e-06,
+      "loss": 3.4123,
+      "step": 28341
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6587231352718079e-06,
+      "loss": 2.3284,
+      "step": 28342
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6586757269279391e-06,
+      "loss": 2.5839,
+      "step": 28343
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6586283185840708e-06,
+      "loss": 2.7129,
+      "step": 28344
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6585809102402023e-06,
+      "loss": 1.7218,
+      "step": 28345
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6585335018963336e-06,
+      "loss": 2.1392,
+      "step": 28346
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6584860935524653e-06,
+      "loss": 1.1655,
+      "step": 28347
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6584386852085968e-06,
+      "loss": 1.4036,
+      "step": 28348
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.658391276864728e-06,
+      "loss": 2.4444,
+      "step": 28349
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6583438685208598e-06,
+      "loss": 2.4105,
+      "step": 28350
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6582964601769913e-06,
+      "loss": 2.2906,
+      "step": 28351
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6582490518331226e-06,
+      "loss": 3.4466,
+      "step": 28352
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6582016434892543e-06,
+      "loss": 2.3374,
+      "step": 28353
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6581542351453858e-06,
+      "loss": 2.0161,
+      "step": 28354
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.658106826801517e-06,
+      "loss": 2.4179,
+      "step": 28355
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6580594184576487e-06,
+      "loss": 1.883,
+      "step": 28356
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.65801201011378e-06,
+      "loss": 0.8845,
+      "step": 28357
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6579646017699115e-06,
+      "loss": 2.5501,
+      "step": 28358
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6579171934260428e-06,
+      "loss": 3.0865,
+      "step": 28359
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6578697850821745e-06,
+      "loss": 1.3091,
+      "step": 28360
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.657822376738306e-06,
+      "loss": 1.9957,
+      "step": 28361
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6577749683944373e-06,
+      "loss": 1.0797,
+      "step": 28362
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.657727560050569e-06,
+      "loss": 3.2298,
+      "step": 28363
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6576801517067005e-06,
+      "loss": 2.863,
+      "step": 28364
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6576327433628317e-06,
+      "loss": 1.3714,
+      "step": 28365
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6575853350189635e-06,
+      "loss": 3.4669,
+      "step": 28366
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.657537926675095e-06,
+      "loss": 2.4978,
+      "step": 28367
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6574905183312262e-06,
+      "loss": 2.0367,
+      "step": 28368
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.657443109987358e-06,
+      "loss": 2.5366,
+      "step": 28369
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6573957016434892e-06,
+      "loss": 2.4761,
+      "step": 28370
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6573482932996207e-06,
+      "loss": 0.9695,
+      "step": 28371
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6573008849557524e-06,
+      "loss": 1.7138,
+      "step": 28372
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6572534766118837e-06,
+      "loss": 2.2121,
+      "step": 28373
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6572060682680152e-06,
+      "loss": 3.1497,
+      "step": 28374
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6571586599241469e-06,
+      "loss": 2.7149,
+      "step": 28375
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6571112515802782e-06,
+      "loss": 1.1726,
+      "step": 28376
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6570638432364097e-06,
+      "loss": 3.1929,
+      "step": 28377
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6570164348925414e-06,
+      "loss": 3.3792,
+      "step": 28378
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6569690265486726e-06,
+      "loss": 2.505,
+      "step": 28379
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6569216182048041e-06,
+      "loss": 2.2611,
+      "step": 28380
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6568742098609354e-06,
+      "loss": 2.5079,
+      "step": 28381
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6568268015170671e-06,
+      "loss": 1.7996,
+      "step": 28382
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6567793931731984e-06,
+      "loss": 1.6351,
+      "step": 28383
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6567319848293299e-06,
+      "loss": 2.0183,
+      "step": 28384
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6566845764854616e-06,
+      "loss": 2.752,
+      "step": 28385
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6566371681415929e-06,
+      "loss": 2.2696,
+      "step": 28386
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6565897597977244e-06,
+      "loss": 3.3403,
+      "step": 28387
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.656542351453856e-06,
+      "loss": 2.825,
+      "step": 28388
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6564949431099874e-06,
+      "loss": 3.08,
+      "step": 28389
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6564475347661188e-06,
+      "loss": 1.5994,
+      "step": 28390
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6564001264222505e-06,
+      "loss": 2.6924,
+      "step": 28391
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6563527180783818e-06,
+      "loss": 2.5016,
+      "step": 28392
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6563053097345133e-06,
+      "loss": 2.0408,
+      "step": 28393
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.656257901390645e-06,
+      "loss": 2.0298,
+      "step": 28394
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6562104930467763e-06,
+      "loss": 0.5487,
+      "step": 28395
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6561630847029076e-06,
+      "loss": 1.9302,
+      "step": 28396
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6561156763590393e-06,
+      "loss": 2.2932,
+      "step": 28397
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6560682680151708e-06,
+      "loss": 3.8086,
+      "step": 28398
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.656020859671302e-06,
+      "loss": 2.3792,
+      "step": 28399
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6559734513274338e-06,
+      "loss": 2.6499,
+      "step": 28400
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6559260429835653e-06,
+      "loss": 2.3329,
+      "step": 28401
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.6558786346396965e-06,
+      "loss": 2.6634,
+      "step": 28402
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.655831226295828e-06,
+      "loss": 1.9834,
+      "step": 28403
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6557838179519597e-06,
+      "loss": 4.1903,
+      "step": 28404
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.655736409608091e-06,
+      "loss": 0.8038,
+      "step": 28405
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6556890012642225e-06,
+      "loss": 2.1732,
+      "step": 28406
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6556415929203542e-06,
+      "loss": 2.3135,
+      "step": 28407
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6555941845764855e-06,
+      "loss": 2.9161,
+      "step": 28408
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6555467762326168e-06,
+      "loss": 1.2603,
+      "step": 28409
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6554993678887485e-06,
+      "loss": 1.4246,
+      "step": 28410
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.65545195954488e-06,
+      "loss": 2.6596,
+      "step": 28411
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6554045512010112e-06,
+      "loss": 3.4541,
+      "step": 28412
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.655357142857143e-06,
+      "loss": 4.4006,
+      "step": 28413
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6553097345132744e-06,
+      "loss": 4.0001,
+      "step": 28414
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6552623261694057e-06,
+      "loss": 2.9295,
+      "step": 28415
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6552149178255374e-06,
+      "loss": 2.1972,
+      "step": 28416
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.655167509481669e-06,
+      "loss": 2.0398,
+      "step": 28417
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6551201011378002e-06,
+      "loss": 1.5542,
+      "step": 28418
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.655072692793932e-06,
+      "loss": 1.8143,
+      "step": 28419
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6550252844500634e-06,
+      "loss": 2.1996,
+      "step": 28420
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6549778761061947e-06,
+      "loss": 2.6803,
+      "step": 28421
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6549304677623264e-06,
+      "loss": 2.0513,
+      "step": 28422
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6548830594184577e-06,
+      "loss": 2.5775,
+      "step": 28423
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6548356510745892e-06,
+      "loss": 1.9206,
+      "step": 28424
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6547882427307204e-06,
+      "loss": 2.8477,
+      "step": 28425
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6547408343868521e-06,
+      "loss": 2.1548,
+      "step": 28426
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6546934260429836e-06,
+      "loss": 1.7809,
+      "step": 28427
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.654646017699115e-06,
+      "loss": 4.0101,
+      "step": 28428
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6545986093552466e-06,
+      "loss": 2.2072,
+      "step": 28429
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.654551201011378e-06,
+      "loss": 1.9414,
+      "step": 28430
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6545037926675094e-06,
+      "loss": 3.2491,
+      "step": 28431
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.654456384323641e-06,
+      "loss": 2.4639,
+      "step": 28432
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6544089759797724e-06,
+      "loss": 2.4792,
+      "step": 28433
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6543615676359039e-06,
+      "loss": 3.8978,
+      "step": 28434
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6543141592920356e-06,
+      "loss": 2.0256,
+      "step": 28435
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6542667509481668e-06,
+      "loss": 1.3827,
+      "step": 28436
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6542193426042983e-06,
+      "loss": 2.1827,
+      "step": 28437
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.65417193426043e-06,
+      "loss": 2.9807,
+      "step": 28438
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6541245259165613e-06,
+      "loss": 2.8976,
+      "step": 28439
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6540771175726928e-06,
+      "loss": 1.8674,
+      "step": 28440
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6540297092288245e-06,
+      "loss": 1.5648,
+      "step": 28441
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6539823008849558e-06,
+      "loss": 3.2783,
+      "step": 28442
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6539348925410873e-06,
+      "loss": 2.3303,
+      "step": 28443
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.653887484197219e-06,
+      "loss": 1.7793,
+      "step": 28444
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6538400758533503e-06,
+      "loss": 1.6789,
+      "step": 28445
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6537926675094816e-06,
+      "loss": 2.4626,
+      "step": 28446
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.653745259165613e-06,
+      "loss": 2.9189,
+      "step": 28447
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6536978508217448e-06,
+      "loss": 2.3175,
+      "step": 28448
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.653650442477876e-06,
+      "loss": 3.1402,
+      "step": 28449
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6536030341340075e-06,
+      "loss": 2.5156,
+      "step": 28450
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6535556257901392e-06,
+      "loss": 3.4752,
+      "step": 28451
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6535082174462705e-06,
+      "loss": 3.3585,
+      "step": 28452
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.653460809102402e-06,
+      "loss": 2.4712,
+      "step": 28453
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6534134007585337e-06,
+      "loss": 2.6851,
+      "step": 28454
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.653365992414665e-06,
+      "loss": 1.9891,
+      "step": 28455
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6533185840707965e-06,
+      "loss": 2.379,
+      "step": 28456
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6532711757269282e-06,
+      "loss": 1.6816,
+      "step": 28457
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6532237673830595e-06,
+      "loss": 2.8026,
+      "step": 28458
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6531763590391907e-06,
+      "loss": 2.6526,
+      "step": 28459
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6531289506953224e-06,
+      "loss": 2.2723,
+      "step": 28460
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.653081542351454e-06,
+      "loss": 0.5,
+      "step": 28461
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6530341340075852e-06,
+      "loss": 2.4761,
+      "step": 28462
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.652986725663717e-06,
+      "loss": 3.406,
+      "step": 28463
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6529393173198484e-06,
+      "loss": 2.51,
+      "step": 28464
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6528919089759797e-06,
+      "loss": 1.3283,
+      "step": 28465
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.6528445006321114e-06,
+      "loss": 2.5909,
+      "step": 28466
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6527970922882429e-06,
+      "loss": 2.8585,
+      "step": 28467
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6527496839443742e-06,
+      "loss": 2.7767,
+      "step": 28468
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6527022756005057e-06,
+      "loss": 3.1957,
+      "step": 28469
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6526548672566374e-06,
+      "loss": 3.8896,
+      "step": 28470
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6526074589127686e-06,
+      "loss": 2.7886,
+      "step": 28471
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6525600505689e-06,
+      "loss": 2.429,
+      "step": 28472
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6525126422250316e-06,
+      "loss": 1.2039,
+      "step": 28473
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6524652338811631e-06,
+      "loss": 2.1692,
+      "step": 28474
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6524178255372944e-06,
+      "loss": 3.7515,
+      "step": 28475
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6523704171934261e-06,
+      "loss": 2.0214,
+      "step": 28476
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6523230088495576e-06,
+      "loss": 4.1502,
+      "step": 28477
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6522756005056889e-06,
+      "loss": 2.1274,
+      "step": 28478
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6522281921618206e-06,
+      "loss": 3.0339,
+      "step": 28479
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.652180783817952e-06,
+      "loss": 2.4143,
+      "step": 28480
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6521333754740834e-06,
+      "loss": 1.8707,
+      "step": 28481
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.652085967130215e-06,
+      "loss": 3.2579,
+      "step": 28482
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6520385587863466e-06,
+      "loss": 2.5535,
+      "step": 28483
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6519911504424778e-06,
+      "loss": 3.4439,
+      "step": 28484
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6519437420986095e-06,
+      "loss": 1.5448,
+      "step": 28485
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6518963337547408e-06,
+      "loss": 1.469,
+      "step": 28486
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6518489254108723e-06,
+      "loss": 3.3983,
+      "step": 28487
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.651801517067004e-06,
+      "loss": 2.2315,
+      "step": 28488
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6517541087231353e-06,
+      "loss": 1.5609,
+      "step": 28489
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6517067003792668e-06,
+      "loss": 2.4048,
+      "step": 28490
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.651659292035398e-06,
+      "loss": 1.9026,
+      "step": 28491
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6516118836915298e-06,
+      "loss": 2.6593,
+      "step": 28492
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6515644753476613e-06,
+      "loss": 2.5936,
+      "step": 28493
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6515170670037925e-06,
+      "loss": 1.633,
+      "step": 28494
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6514696586599243e-06,
+      "loss": 2.6571,
+      "step": 28495
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6514222503160557e-06,
+      "loss": 1.8812,
+      "step": 28496
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.651374841972187e-06,
+      "loss": 2.3015,
+      "step": 28497
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6513274336283187e-06,
+      "loss": 3.7173,
+      "step": 28498
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.65128002528445e-06,
+      "loss": 2.632,
+      "step": 28499
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6512326169405815e-06,
+      "loss": 2.2757,
+      "step": 28500
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6511852085967132e-06,
+      "loss": 1.7868,
+      "step": 28501
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6511378002528445e-06,
+      "loss": 1.0413,
+      "step": 28502
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.651090391908976e-06,
+      "loss": 2.2412,
+      "step": 28503
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6510429835651077e-06,
+      "loss": 4.0475,
+      "step": 28504
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.650995575221239e-06,
+      "loss": 2.5028,
+      "step": 28505
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6509481668773705e-06,
+      "loss": 1.4101,
+      "step": 28506
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6509007585335022e-06,
+      "loss": 2.8743,
+      "step": 28507
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6508533501896334e-06,
+      "loss": 3.4419,
+      "step": 28508
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.650805941845765e-06,
+      "loss": 2.9194,
+      "step": 28509
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6507585335018966e-06,
+      "loss": 2.3436,
+      "step": 28510
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.650711125158028e-06,
+      "loss": 1.075,
+      "step": 28511
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6506637168141592e-06,
+      "loss": 1.9747,
+      "step": 28512
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6506163084702907e-06,
+      "loss": 2.6168,
+      "step": 28513
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6505689001264224e-06,
+      "loss": 2.693,
+      "step": 28514
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6505214917825537e-06,
+      "loss": 2.2649,
+      "step": 28515
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6504740834386852e-06,
+      "loss": 3.3461,
+      "step": 28516
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6504266750948169e-06,
+      "loss": 2.4228,
+      "step": 28517
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6503792667509481e-06,
+      "loss": 1.3987,
+      "step": 28518
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6503318584070796e-06,
+      "loss": 1.9323,
+      "step": 28519
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6502844500632113e-06,
+      "loss": 1.843,
+      "step": 28520
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6502370417193426e-06,
+      "loss": 2.3742,
+      "step": 28521
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6501896333754741e-06,
+      "loss": 2.064,
+      "step": 28522
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6501422250316058e-06,
+      "loss": 2.5052,
+      "step": 28523
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.650094816687737e-06,
+      "loss": 3.3505,
+      "step": 28524
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6500474083438684e-06,
+      "loss": 2.0899,
+      "step": 28525
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.65e-06,
+      "loss": 3.7028,
+      "step": 28526
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6499525916561316e-06,
+      "loss": 2.4897,
+      "step": 28527
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6499051833122629e-06,
+      "loss": 1.844,
+      "step": 28528
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.6498577749683946e-06,
+      "loss": 2.5169,
+      "step": 28529
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.649810366624526e-06,
+      "loss": 1.5797,
+      "step": 28530
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6497629582806573e-06,
+      "loss": 1.8613,
+      "step": 28531
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.649715549936789e-06,
+      "loss": 1.9632,
+      "step": 28532
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6496681415929205e-06,
+      "loss": 3.424,
+      "step": 28533
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6496207332490518e-06,
+      "loss": 2.4657,
+      "step": 28534
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6495733249051833e-06,
+      "loss": 2.5074,
+      "step": 28535
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.649525916561315e-06,
+      "loss": 3.1022,
+      "step": 28536
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6494785082174463e-06,
+      "loss": 2.6677,
+      "step": 28537
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6494310998735776e-06,
+      "loss": 2.3006,
+      "step": 28538
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6493836915297093e-06,
+      "loss": 3.8077,
+      "step": 28539
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6493362831858408e-06,
+      "loss": 1.3777,
+      "step": 28540
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.649288874841972e-06,
+      "loss": 1.6539,
+      "step": 28541
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6492414664981037e-06,
+      "loss": 2.4762,
+      "step": 28542
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6491940581542352e-06,
+      "loss": 2.4029,
+      "step": 28543
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6491466498103665e-06,
+      "loss": 1.7882,
+      "step": 28544
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6490992414664982e-06,
+      "loss": 3.7415,
+      "step": 28545
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6490518331226297e-06,
+      "loss": 1.9467,
+      "step": 28546
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.649004424778761e-06,
+      "loss": 2.5004,
+      "step": 28547
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6489570164348927e-06,
+      "loss": 1.1916,
+      "step": 28548
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6489096080910242e-06,
+      "loss": 1.3536,
+      "step": 28549
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6488621997471555e-06,
+      "loss": 1.8414,
+      "step": 28550
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6488147914032872e-06,
+      "loss": 1.5145,
+      "step": 28551
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6487673830594185e-06,
+      "loss": 3.3117,
+      "step": 28552
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.64871997471555e-06,
+      "loss": 1.3059,
+      "step": 28553
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6486725663716817e-06,
+      "loss": 3.7023,
+      "step": 28554
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.648625158027813e-06,
+      "loss": 1.5686,
+      "step": 28555
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6485777496839444e-06,
+      "loss": 2.3353,
+      "step": 28556
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6485303413400757e-06,
+      "loss": 1.3888,
+      "step": 28557
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6484829329962074e-06,
+      "loss": 2.5768,
+      "step": 28558
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.648435524652339e-06,
+      "loss": 2.5634,
+      "step": 28559
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6483881163084702e-06,
+      "loss": 3.0227,
+      "step": 28560
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6483407079646019e-06,
+      "loss": 2.4533,
+      "step": 28561
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6482932996207334e-06,
+      "loss": 2.7819,
+      "step": 28562
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6482458912768647e-06,
+      "loss": 1.8635,
+      "step": 28563
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6481984829329964e-06,
+      "loss": 2.324,
+      "step": 28564
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6481510745891276e-06,
+      "loss": 2.3981,
+      "step": 28565
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6481036662452591e-06,
+      "loss": 3.046,
+      "step": 28566
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6480562579013908e-06,
+      "loss": 1.0588,
+      "step": 28567
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6480088495575221e-06,
+      "loss": 1.8218,
+      "step": 28568
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6479614412136536e-06,
+      "loss": 1.8896,
+      "step": 28569
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6479140328697853e-06,
+      "loss": 3.2698,
+      "step": 28570
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6478666245259166e-06,
+      "loss": 1.7362,
+      "step": 28571
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.647819216182048e-06,
+      "loss": 3.3985,
+      "step": 28572
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6477718078381798e-06,
+      "loss": 2.2602,
+      "step": 28573
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.647724399494311e-06,
+      "loss": 2.7179,
+      "step": 28574
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6476769911504424e-06,
+      "loss": 1.6619,
+      "step": 28575
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6476295828065743e-06,
+      "loss": 2.2002,
+      "step": 28576
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6475821744627055e-06,
+      "loss": 3.9338,
+      "step": 28577
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6475347661188368e-06,
+      "loss": 2.3496,
+      "step": 28578
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6474873577749683e-06,
+      "loss": 1.7977,
+      "step": 28579
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6474399494311e-06,
+      "loss": 4.1721,
+      "step": 28580
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6473925410872313e-06,
+      "loss": 3.1771,
+      "step": 28581
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6473451327433628e-06,
+      "loss": 2.4386,
+      "step": 28582
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6472977243994945e-06,
+      "loss": 1.9057,
+      "step": 28583
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6472503160556258e-06,
+      "loss": 4.3462,
+      "step": 28584
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6472029077117573e-06,
+      "loss": 2.6286,
+      "step": 28585
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.647155499367889e-06,
+      "loss": 4.1073,
+      "step": 28586
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6471080910240203e-06,
+      "loss": 2.4815,
+      "step": 28587
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6470606826801515e-06,
+      "loss": 2.2425,
+      "step": 28588
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6470132743362832e-06,
+      "loss": 2.7241,
+      "step": 28589
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6469658659924147e-06,
+      "loss": 2.5676,
+      "step": 28590
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.646918457648546e-06,
+      "loss": 2.3008,
+      "step": 28591
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6468710493046777e-06,
+      "loss": 1.634,
+      "step": 28592
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.6468236409608092e-06,
+      "loss": 2.112,
+      "step": 28593
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6467762326169405e-06,
+      "loss": 2.7082,
+      "step": 28594
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6467288242730722e-06,
+      "loss": 2.1817,
+      "step": 28595
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6466814159292037e-06,
+      "loss": 2.9161,
+      "step": 28596
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.646634007585335e-06,
+      "loss": 2.2374,
+      "step": 28597
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6465865992414667e-06,
+      "loss": 1.245,
+      "step": 28598
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6465391908975982e-06,
+      "loss": 3.1005,
+      "step": 28599
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6464917825537294e-06,
+      "loss": 2.6587,
+      "step": 28600
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6464443742098607e-06,
+      "loss": 3.6709,
+      "step": 28601
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6463969658659924e-06,
+      "loss": 1.1368,
+      "step": 28602
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.646349557522124e-06,
+      "loss": 1.3661,
+      "step": 28603
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6463021491782552e-06,
+      "loss": 2.6023,
+      "step": 28604
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.646254740834387e-06,
+      "loss": 2.5153,
+      "step": 28605
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6462073324905184e-06,
+      "loss": 2.0884,
+      "step": 28606
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6461599241466497e-06,
+      "loss": 2.4114,
+      "step": 28607
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6461125158027814e-06,
+      "loss": 1.8392,
+      "step": 28608
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6460651074589129e-06,
+      "loss": 2.1368,
+      "step": 28609
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6460176991150442e-06,
+      "loss": 2.6167,
+      "step": 28610
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6459702907711759e-06,
+      "loss": 1.8994,
+      "step": 28611
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6459228824273074e-06,
+      "loss": 2.1576,
+      "step": 28612
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6458754740834386e-06,
+      "loss": 1.1517,
+      "step": 28613
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6458280657395703e-06,
+      "loss": 1.934,
+      "step": 28614
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6457806573957016e-06,
+      "loss": 1.9873,
+      "step": 28615
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6457332490518331e-06,
+      "loss": 1.4501,
+      "step": 28616
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6456858407079648e-06,
+      "loss": 2.0158,
+      "step": 28617
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.645638432364096e-06,
+      "loss": 2.279,
+      "step": 28618
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6455910240202276e-06,
+      "loss": 2.3697,
+      "step": 28619
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6455436156763589e-06,
+      "loss": 1.9564,
+      "step": 28620
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6454962073324906e-06,
+      "loss": 2.7788,
+      "step": 28621
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.645448798988622e-06,
+      "loss": 2.9558,
+      "step": 28622
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6454013906447533e-06,
+      "loss": 2.1103,
+      "step": 28623
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.645353982300885e-06,
+      "loss": 1.8528,
+      "step": 28624
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6453065739570165e-06,
+      "loss": 2.4763,
+      "step": 28625
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6452591656131478e-06,
+      "loss": 2.7672,
+      "step": 28626
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6452117572692795e-06,
+      "loss": 2.6303,
+      "step": 28627
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6451643489254108e-06,
+      "loss": 1.7686,
+      "step": 28628
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6451169405815423e-06,
+      "loss": 3.5181,
+      "step": 28629
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.645069532237674e-06,
+      "loss": 4.1276,
+      "step": 28630
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6450221238938053e-06,
+      "loss": 3.6769,
+      "step": 28631
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6449747155499368e-06,
+      "loss": 3.2598,
+      "step": 28632
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6449273072060685e-06,
+      "loss": 1.6082,
+      "step": 28633
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6448798988621998e-06,
+      "loss": 1.989,
+      "step": 28634
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6448324905183312e-06,
+      "loss": 2.7339,
+      "step": 28635
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.644785082174463e-06,
+      "loss": 3.4668,
+      "step": 28636
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6447376738305942e-06,
+      "loss": 1.2266,
+      "step": 28637
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6446902654867257e-06,
+      "loss": 1.5254,
+      "step": 28638
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6446428571428574e-06,
+      "loss": 2.1451,
+      "step": 28639
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6445954487989887e-06,
+      "loss": 2.5031,
+      "step": 28640
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.64454804045512e-06,
+      "loss": 2.5929,
+      "step": 28641
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6445006321112515e-06,
+      "loss": 1.5085,
+      "step": 28642
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6444532237673832e-06,
+      "loss": 1.5422,
+      "step": 28643
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6444058154235145e-06,
+      "loss": 1.9271,
+      "step": 28644
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.644358407079646e-06,
+      "loss": 1.251,
+      "step": 28645
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6443109987357777e-06,
+      "loss": 1.8901,
+      "step": 28646
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.644263590391909e-06,
+      "loss": 2.0765,
+      "step": 28647
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6442161820480404e-06,
+      "loss": 2.4747,
+      "step": 28648
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6441687737041721e-06,
+      "loss": 2.2272,
+      "step": 28649
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6441213653603034e-06,
+      "loss": 4.0757,
+      "step": 28650
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.644073957016435e-06,
+      "loss": 3.2503,
+      "step": 28651
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6440265486725666e-06,
+      "loss": 1.9923,
+      "step": 28652
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.643979140328698e-06,
+      "loss": 2.501,
+      "step": 28653
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6439317319848292e-06,
+      "loss": 2.8237,
+      "step": 28654
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6438843236409609e-06,
+      "loss": 2.2206,
+      "step": 28655
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.6438369152970924e-06,
+      "loss": 2.4908,
+      "step": 28656
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6437895069532237e-06,
+      "loss": 2.4378,
+      "step": 28657
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6437420986093554e-06,
+      "loss": 1.5154,
+      "step": 28658
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6436946902654868e-06,
+      "loss": 1.1462,
+      "step": 28659
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6436472819216181e-06,
+      "loss": 1.2424,
+      "step": 28660
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6435998735777498e-06,
+      "loss": 2.9292,
+      "step": 28661
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6435524652338813e-06,
+      "loss": 2.358,
+      "step": 28662
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6435050568900126e-06,
+      "loss": 2.9379,
+      "step": 28663
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.643457648546144e-06,
+      "loss": 2.5789,
+      "step": 28664
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6434102402022758e-06,
+      "loss": 2.8177,
+      "step": 28665
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.643362831858407e-06,
+      "loss": 3.2046,
+      "step": 28666
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6433154235145384e-06,
+      "loss": 2.0523,
+      "step": 28667
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.64326801517067e-06,
+      "loss": 3.0921,
+      "step": 28668
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6432206068268016e-06,
+      "loss": 1.404,
+      "step": 28669
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6431731984829328e-06,
+      "loss": 1.9823,
+      "step": 28670
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6431257901390645e-06,
+      "loss": 2.2104,
+      "step": 28671
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.643078381795196e-06,
+      "loss": 2.7252,
+      "step": 28672
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6430309734513273e-06,
+      "loss": 2.832,
+      "step": 28673
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.642983565107459e-06,
+      "loss": 2.6091,
+      "step": 28674
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6429361567635905e-06,
+      "loss": 2.9243,
+      "step": 28675
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6428887484197218e-06,
+      "loss": 1.5793,
+      "step": 28676
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6428413400758535e-06,
+      "loss": 1.3225,
+      "step": 28677
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.642793931731985e-06,
+      "loss": 2.7022,
+      "step": 28678
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6427465233881163e-06,
+      "loss": 1.9421,
+      "step": 28679
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.642699115044248e-06,
+      "loss": 2.2746,
+      "step": 28680
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6426517067003793e-06,
+      "loss": 2.1063,
+      "step": 28681
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6426042983565107e-06,
+      "loss": 2.249,
+      "step": 28682
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6425568900126425e-06,
+      "loss": 2.4319,
+      "step": 28683
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6425094816687737e-06,
+      "loss": 2.8111,
+      "step": 28684
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6424620733249052e-06,
+      "loss": 1.3087,
+      "step": 28685
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6424146649810365e-06,
+      "loss": 2.945,
+      "step": 28686
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6423672566371682e-06,
+      "loss": 1.2954,
+      "step": 28687
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6423198482932997e-06,
+      "loss": 3.6031,
+      "step": 28688
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.642272439949431e-06,
+      "loss": 2.9098,
+      "step": 28689
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6422250316055627e-06,
+      "loss": 3.694,
+      "step": 28690
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6421776232616942e-06,
+      "loss": 2.1957,
+      "step": 28691
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6421302149178255e-06,
+      "loss": 2.1451,
+      "step": 28692
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6420828065739572e-06,
+      "loss": 2.5672,
+      "step": 28693
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6420353982300884e-06,
+      "loss": 2.1592,
+      "step": 28694
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.64198798988622e-06,
+      "loss": 2.5444,
+      "step": 28695
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6419405815423516e-06,
+      "loss": 2.9091,
+      "step": 28696
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.641893173198483e-06,
+      "loss": 2.6472,
+      "step": 28697
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6418457648546144e-06,
+      "loss": 3.6939,
+      "step": 28698
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6417983565107461e-06,
+      "loss": 2.4757,
+      "step": 28699
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6417509481668774e-06,
+      "loss": 1.3436,
+      "step": 28700
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6417035398230089e-06,
+      "loss": 3.0763,
+      "step": 28701
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6416561314791406e-06,
+      "loss": 2.5406,
+      "step": 28702
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6416087231352719e-06,
+      "loss": 1.8648,
+      "step": 28703
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6415613147914032e-06,
+      "loss": 1.965,
+      "step": 28704
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.641513906447535e-06,
+      "loss": 2.3597,
+      "step": 28705
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6414664981036663e-06,
+      "loss": 1.7613,
+      "step": 28706
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6414190897597976e-06,
+      "loss": 2.1096,
+      "step": 28707
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6413716814159291e-06,
+      "loss": 2.5618,
+      "step": 28708
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6413242730720608e-06,
+      "loss": 3.2233,
+      "step": 28709
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.641276864728192e-06,
+      "loss": 2.3965,
+      "step": 28710
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6412294563843236e-06,
+      "loss": 2.7172,
+      "step": 28711
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6411820480404553e-06,
+      "loss": 1.8166,
+      "step": 28712
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6411346396965866e-06,
+      "loss": 2.4579,
+      "step": 28713
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.641087231352718e-06,
+      "loss": 0.7549,
+      "step": 28714
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6410398230088498e-06,
+      "loss": 3.1791,
+      "step": 28715
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.640992414664981e-06,
+      "loss": 1.8255,
+      "step": 28716
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6409450063211123e-06,
+      "loss": 1.7581,
+      "step": 28717
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.640897597977244e-06,
+      "loss": 3.4123,
+      "step": 28718
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6408501896333755e-06,
+      "loss": 1.8497,
+      "step": 28719
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.6408027812895068e-06,
+      "loss": 2.6753,
+      "step": 28720
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6407553729456385e-06,
+      "loss": 2.7937,
+      "step": 28721
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.64070796460177e-06,
+      "loss": 2.7686,
+      "step": 28722
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6406605562579013e-06,
+      "loss": 2.1729,
+      "step": 28723
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.640613147914033e-06,
+      "loss": 2.1852,
+      "step": 28724
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6405657395701645e-06,
+      "loss": 2.4149,
+      "step": 28725
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6405183312262958e-06,
+      "loss": 1.847,
+      "step": 28726
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6404709228824275e-06,
+      "loss": 2.441,
+      "step": 28727
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.640423514538559e-06,
+      "loss": 0.836,
+      "step": 28728
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6403761061946902e-06,
+      "loss": 2.2835,
+      "step": 28729
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6403286978508215e-06,
+      "loss": 2.3669,
+      "step": 28730
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6402812895069532e-06,
+      "loss": 0.7757,
+      "step": 28731
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6402338811630847e-06,
+      "loss": 1.597,
+      "step": 28732
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.640186472819216e-06,
+      "loss": 1.1782,
+      "step": 28733
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6401390644753477e-06,
+      "loss": 3.5071,
+      "step": 28734
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6400916561314792e-06,
+      "loss": 3.9903,
+      "step": 28735
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6400442477876105e-06,
+      "loss": 1.461,
+      "step": 28736
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6399968394437422e-06,
+      "loss": 2.762,
+      "step": 28737
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6399494310998737e-06,
+      "loss": 2.4582,
+      "step": 28738
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.639902022756005e-06,
+      "loss": 2.2576,
+      "step": 28739
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6398546144121367e-06,
+      "loss": 3.0464,
+      "step": 28740
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6398072060682681e-06,
+      "loss": 1.8273,
+      "step": 28741
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6397597977243994e-06,
+      "loss": 1.8374,
+      "step": 28742
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6397123893805311e-06,
+      "loss": 1.6235,
+      "step": 28743
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6396649810366624e-06,
+      "loss": 1.7243,
+      "step": 28744
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.639617572692794e-06,
+      "loss": 2.9071,
+      "step": 28745
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6395701643489256e-06,
+      "loss": 2.2057,
+      "step": 28746
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6395227560050569e-06,
+      "loss": 2.841,
+      "step": 28747
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6394753476611884e-06,
+      "loss": 2.772,
+      "step": 28748
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.63942793931732e-06,
+      "loss": 1.7368,
+      "step": 28749
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6393805309734514e-06,
+      "loss": 3.0029,
+      "step": 28750
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6393331226295829e-06,
+      "loss": 2.1381,
+      "step": 28751
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6392857142857141e-06,
+      "loss": 2.1076,
+      "step": 28752
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6392383059418458e-06,
+      "loss": 2.4866,
+      "step": 28753
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6391908975979773e-06,
+      "loss": 2.6681,
+      "step": 28754
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6391434892541086e-06,
+      "loss": 2.2069,
+      "step": 28755
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6390960809102403e-06,
+      "loss": 1.973,
+      "step": 28756
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6390486725663716e-06,
+      "loss": 1.4155,
+      "step": 28757
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.639001264222503e-06,
+      "loss": 2.0967,
+      "step": 28758
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6389538558786348e-06,
+      "loss": 2.0557,
+      "step": 28759
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.638906447534766e-06,
+      "loss": 2.0041,
+      "step": 28760
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6388590391908976e-06,
+      "loss": 1.9306,
+      "step": 28761
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6388116308470293e-06,
+      "loss": 3.3694,
+      "step": 28762
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6387642225031606e-06,
+      "loss": 2.7123,
+      "step": 28763
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.638716814159292e-06,
+      "loss": 1.8907,
+      "step": 28764
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6386694058154237e-06,
+      "loss": 2.299,
+      "step": 28765
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.638621997471555e-06,
+      "loss": 2.7937,
+      "step": 28766
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6385745891276865e-06,
+      "loss": 1.2272,
+      "step": 28767
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6385271807838182e-06,
+      "loss": 2.6641,
+      "step": 28768
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6384797724399495e-06,
+      "loss": 3.7185,
+      "step": 28769
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6384323640960808e-06,
+      "loss": 2.6713,
+      "step": 28770
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6383849557522125e-06,
+      "loss": 2.2016,
+      "step": 28771
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.638337547408344e-06,
+      "loss": 3.1817,
+      "step": 28772
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6382901390644753e-06,
+      "loss": 1.7961,
+      "step": 28773
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6382427307206068e-06,
+      "loss": 1.2451,
+      "step": 28774
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6381953223767385e-06,
+      "loss": 0.4573,
+      "step": 28775
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6381479140328697e-06,
+      "loss": 1.8146,
+      "step": 28776
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6381005056890012e-06,
+      "loss": 2.1877,
+      "step": 28777
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.638053097345133e-06,
+      "loss": 2.1495,
+      "step": 28778
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6380056890012642e-06,
+      "loss": 2.2592,
+      "step": 28779
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6379582806573957e-06,
+      "loss": 2.7729,
+      "step": 28780
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6379108723135274e-06,
+      "loss": 2.8873,
+      "step": 28781
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.6378634639696587e-06,
+      "loss": 1.4372,
+      "step": 28782
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.63781605562579e-06,
+      "loss": 1.667,
+      "step": 28783
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6377686472819217e-06,
+      "loss": 2.374,
+      "step": 28784
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6377212389380532e-06,
+      "loss": 2.7591,
+      "step": 28785
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6376738305941845e-06,
+      "loss": 1.6765,
+      "step": 28786
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6376264222503162e-06,
+      "loss": 1.8535,
+      "step": 28787
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6375790139064476e-06,
+      "loss": 3.1739,
+      "step": 28788
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.637531605562579e-06,
+      "loss": 2.5117,
+      "step": 28789
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6374841972187106e-06,
+      "loss": 2.0394,
+      "step": 28790
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6374367888748421e-06,
+      "loss": 2.2071,
+      "step": 28791
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6373893805309734e-06,
+      "loss": 2.2351,
+      "step": 28792
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6373419721871051e-06,
+      "loss": 1.5156,
+      "step": 28793
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6372945638432366e-06,
+      "loss": 1.6779,
+      "step": 28794
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6372471554993679e-06,
+      "loss": 3.0223,
+      "step": 28795
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6371997471554992e-06,
+      "loss": 1.8532,
+      "step": 28796
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6371523388116309e-06,
+      "loss": 2.7567,
+      "step": 28797
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6371049304677624e-06,
+      "loss": 2.2595,
+      "step": 28798
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6370575221238936e-06,
+      "loss": 3.0688,
+      "step": 28799
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6370101137800253e-06,
+      "loss": 2.0686,
+      "step": 28800
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6369627054361568e-06,
+      "loss": 3.444,
+      "step": 28801
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6369152970922881e-06,
+      "loss": 2.8718,
+      "step": 28802
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6368678887484198e-06,
+      "loss": 1.9235,
+      "step": 28803
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6368204804045513e-06,
+      "loss": 2.4087,
+      "step": 28804
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6367730720606826e-06,
+      "loss": 1.8099,
+      "step": 28805
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6367256637168143e-06,
+      "loss": 2.0388,
+      "step": 28806
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6366782553729458e-06,
+      "loss": 2.0257,
+      "step": 28807
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.636630847029077e-06,
+      "loss": 3.5047,
+      "step": 28808
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6365834386852088e-06,
+      "loss": 3.054,
+      "step": 28809
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.63653603034134e-06,
+      "loss": 2.0527,
+      "step": 28810
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6364886219974715e-06,
+      "loss": 2.5337,
+      "step": 28811
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6364412136536032e-06,
+      "loss": 2.9999,
+      "step": 28812
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6363938053097345e-06,
+      "loss": 2.1438,
+      "step": 28813
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.636346396965866e-06,
+      "loss": 1.7263,
+      "step": 28814
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6362989886219977e-06,
+      "loss": 1.7699,
+      "step": 28815
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.636251580278129e-06,
+      "loss": 2.4992,
+      "step": 28816
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6362041719342605e-06,
+      "loss": 2.1109,
+      "step": 28817
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6361567635903918e-06,
+      "loss": 1.7048,
+      "step": 28818
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6361093552465235e-06,
+      "loss": 2.0256,
+      "step": 28819
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.636061946902655e-06,
+      "loss": 2.9663,
+      "step": 28820
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6360145385587863e-06,
+      "loss": 2.019,
+      "step": 28821
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.635967130214918e-06,
+      "loss": 2.4476,
+      "step": 28822
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6359197218710492e-06,
+      "loss": 0.9793,
+      "step": 28823
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6358723135271807e-06,
+      "loss": 2.083,
+      "step": 28824
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6358249051833124e-06,
+      "loss": 2.044,
+      "step": 28825
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6357774968394437e-06,
+      "loss": 3.2983,
+      "step": 28826
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6357300884955752e-06,
+      "loss": 2.2248,
+      "step": 28827
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.635682680151707e-06,
+      "loss": 2.1953,
+      "step": 28828
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6356352718078382e-06,
+      "loss": 2.6808,
+      "step": 28829
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6355878634639697e-06,
+      "loss": 2.3174,
+      "step": 28830
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6355404551201014e-06,
+      "loss": 3.6786,
+      "step": 28831
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6354930467762327e-06,
+      "loss": 3.0121,
+      "step": 28832
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6354456384323642e-06,
+      "loss": 3.0341,
+      "step": 28833
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6353982300884959e-06,
+      "loss": 2.1928,
+      "step": 28834
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6353508217446271e-06,
+      "loss": 2.5127,
+      "step": 28835
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6353034134007584e-06,
+      "loss": 2.9201,
+      "step": 28836
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6352560050568901e-06,
+      "loss": 3.501,
+      "step": 28837
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6352085967130216e-06,
+      "loss": 3.1941,
+      "step": 28838
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.635161188369153e-06,
+      "loss": 2.3646,
+      "step": 28839
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6351137800252844e-06,
+      "loss": 3.2911,
+      "step": 28840
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.635066371681416e-06,
+      "loss": 2.421,
+      "step": 28841
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6350189633375474e-06,
+      "loss": 1.0821,
+      "step": 28842
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6349715549936789e-06,
+      "loss": 2.1592,
+      "step": 28843
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6349241466498106e-06,
+      "loss": 1.7952,
+      "step": 28844
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6348767383059419e-06,
+      "loss": 1.6272,
+      "step": 28845
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.6348293299620731e-06,
+      "loss": 2.6302,
+      "step": 28846
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.634781921618205e-06,
+      "loss": 1.217,
+      "step": 28847
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6347345132743363e-06,
+      "loss": 1.0816,
+      "step": 28848
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6346871049304676e-06,
+      "loss": 1.1974,
+      "step": 28849
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6346396965865993e-06,
+      "loss": 2.4295,
+      "step": 28850
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6345922882427308e-06,
+      "loss": 2.5466,
+      "step": 28851
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.634544879898862e-06,
+      "loss": 2.7307,
+      "step": 28852
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6344974715549938e-06,
+      "loss": 3.2028,
+      "step": 28853
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6344500632111253e-06,
+      "loss": 2.938,
+      "step": 28854
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6344026548672566e-06,
+      "loss": 2.5002,
+      "step": 28855
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6343552465233883e-06,
+      "loss": 1.4595,
+      "step": 28856
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6343078381795198e-06,
+      "loss": 2.6148,
+      "step": 28857
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.634260429835651e-06,
+      "loss": 2.8556,
+      "step": 28858
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6342130214917827e-06,
+      "loss": 1.9147,
+      "step": 28859
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.634165613147914e-06,
+      "loss": 3.1056,
+      "step": 28860
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6341182048040455e-06,
+      "loss": 2.1086,
+      "step": 28861
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6340707964601768e-06,
+      "loss": 1.6085,
+      "step": 28862
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6340233881163085e-06,
+      "loss": 2.7551,
+      "step": 28863
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.63397597977244e-06,
+      "loss": 2.3466,
+      "step": 28864
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6339285714285713e-06,
+      "loss": 2.5242,
+      "step": 28865
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.633881163084703e-06,
+      "loss": 1.5606,
+      "step": 28866
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6338337547408345e-06,
+      "loss": 3.6299,
+      "step": 28867
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6337863463969657e-06,
+      "loss": 3.7398,
+      "step": 28868
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6337389380530975e-06,
+      "loss": 1.8228,
+      "step": 28869
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.633691529709229e-06,
+      "loss": 3.5295,
+      "step": 28870
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6336441213653602e-06,
+      "loss": 1.9429,
+      "step": 28871
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.633596713021492e-06,
+      "loss": 2.7163,
+      "step": 28872
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6335493046776232e-06,
+      "loss": 1.8094,
+      "step": 28873
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6335018963337547e-06,
+      "loss": 2.5321,
+      "step": 28874
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6334544879898864e-06,
+      "loss": 2.4289,
+      "step": 28875
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6334070796460177e-06,
+      "loss": 1.8755,
+      "step": 28876
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6333596713021492e-06,
+      "loss": 3.2691,
+      "step": 28877
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6333122629582809e-06,
+      "loss": 1.6201,
+      "step": 28878
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6332648546144122e-06,
+      "loss": 2.0644,
+      "step": 28879
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6332174462705437e-06,
+      "loss": 1.3151,
+      "step": 28880
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6331700379266754e-06,
+      "loss": 1.7201,
+      "step": 28881
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6331226295828066e-06,
+      "loss": 1.7046,
+      "step": 28882
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6330752212389381e-06,
+      "loss": 1.7637,
+      "step": 28883
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6330278128950694e-06,
+      "loss": 2.4868,
+      "step": 28884
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6329804045512011e-06,
+      "loss": 2.1855,
+      "step": 28885
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6329329962073324e-06,
+      "loss": 0.8276,
+      "step": 28886
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6328855878634639e-06,
+      "loss": 2.1427,
+      "step": 28887
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6328381795195956e-06,
+      "loss": 1.9294,
+      "step": 28888
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6327907711757269e-06,
+      "loss": 1.9323,
+      "step": 28889
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6327433628318584e-06,
+      "loss": 1.8279,
+      "step": 28890
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.63269595448799e-06,
+      "loss": 1.7098,
+      "step": 28891
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6326485461441214e-06,
+      "loss": 3.762,
+      "step": 28892
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6326011378002528e-06,
+      "loss": 2.5581,
+      "step": 28893
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6325537294563845e-06,
+      "loss": 2.3331,
+      "step": 28894
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6325063211125158e-06,
+      "loss": 1.9349,
+      "step": 28895
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6324589127686473e-06,
+      "loss": 1.7816,
+      "step": 28896
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.632411504424779e-06,
+      "loss": 2.3969,
+      "step": 28897
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6323640960809103e-06,
+      "loss": 2.3666,
+      "step": 28898
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6323166877370416e-06,
+      "loss": 2.0538,
+      "step": 28899
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6322692793931733e-06,
+      "loss": 2.7814,
+      "step": 28900
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6322218710493048e-06,
+      "loss": 1.2394,
+      "step": 28901
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.632174462705436e-06,
+      "loss": 3.3079,
+      "step": 28902
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6321270543615678e-06,
+      "loss": 3.4774,
+      "step": 28903
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6320796460176993e-06,
+      "loss": 1.9414,
+      "step": 28904
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6320322376738305e-06,
+      "loss": 2.1763,
+      "step": 28905
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.631984829329962e-06,
+      "loss": 2.9487,
+      "step": 28906
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6319374209860937e-06,
+      "loss": 2.5253,
+      "step": 28907
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.631890012642225e-06,
+      "loss": 2.9354,
+      "step": 28908
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6318426042983565e-06,
+      "loss": 2.6445,
+      "step": 28909
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.6317951959544882e-06,
+      "loss": 0.8855,
+      "step": 28910
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6317477876106195e-06,
+      "loss": 1.4774,
+      "step": 28911
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6317003792667508e-06,
+      "loss": 1.965,
+      "step": 28912
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6316529709228825e-06,
+      "loss": 4.018,
+      "step": 28913
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.631605562579014e-06,
+      "loss": 3.2376,
+      "step": 28914
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6315581542351452e-06,
+      "loss": 1.7008,
+      "step": 28915
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.631510745891277e-06,
+      "loss": 3.4087,
+      "step": 28916
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6314633375474084e-06,
+      "loss": 2.494,
+      "step": 28917
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6314159292035397e-06,
+      "loss": 4.5363,
+      "step": 28918
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6313685208596714e-06,
+      "loss": 3.7943,
+      "step": 28919
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.631321112515803e-06,
+      "loss": 3.0968,
+      "step": 28920
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6312737041719342e-06,
+      "loss": 2.4932,
+      "step": 28921
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.631226295828066e-06,
+      "loss": 1.2208,
+      "step": 28922
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6311788874841974e-06,
+      "loss": 2.7159,
+      "step": 28923
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6311314791403287e-06,
+      "loss": 2.1701,
+      "step": 28924
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6310840707964604e-06,
+      "loss": 3.1618,
+      "step": 28925
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6310366624525917e-06,
+      "loss": 2.8203,
+      "step": 28926
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6309892541087232e-06,
+      "loss": 2.2893,
+      "step": 28927
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6309418457648544e-06,
+      "loss": 2.2227,
+      "step": 28928
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6308944374209861e-06,
+      "loss": 1.0318,
+      "step": 28929
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6308470290771176e-06,
+      "loss": 3.7534,
+      "step": 28930
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.630799620733249e-06,
+      "loss": 2.7053,
+      "step": 28931
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6307522123893806e-06,
+      "loss": 2.7456,
+      "step": 28932
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6307048040455121e-06,
+      "loss": 2.2621,
+      "step": 28933
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6306573957016434e-06,
+      "loss": 2.7256,
+      "step": 28934
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.630609987357775e-06,
+      "loss": 2.1268,
+      "step": 28935
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6305625790139066e-06,
+      "loss": 2.8023,
+      "step": 28936
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6305151706700379e-06,
+      "loss": 1.197,
+      "step": 28937
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6304677623261696e-06,
+      "loss": 2.8761,
+      "step": 28938
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6304203539823008e-06,
+      "loss": 3.6856,
+      "step": 28939
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6303729456384323e-06,
+      "loss": 3.0856,
+      "step": 28940
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.630325537294564e-06,
+      "loss": 3.2572,
+      "step": 28941
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6302781289506953e-06,
+      "loss": 2.4638,
+      "step": 28942
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6302307206068268e-06,
+      "loss": 2.6495,
+      "step": 28943
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6301833122629585e-06,
+      "loss": 3.5173,
+      "step": 28944
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6301359039190898e-06,
+      "loss": 3.195,
+      "step": 28945
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6300884955752213e-06,
+      "loss": 3.6956,
+      "step": 28946
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.630041087231353e-06,
+      "loss": 2.9768,
+      "step": 28947
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6299936788874843e-06,
+      "loss": 1.3144,
+      "step": 28948
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6299462705436158e-06,
+      "loss": 2.1579,
+      "step": 28949
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.629898862199747e-06,
+      "loss": 2.7531,
+      "step": 28950
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6298514538558788e-06,
+      "loss": 3.5789,
+      "step": 28951
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.62980404551201e-06,
+      "loss": 1.9556,
+      "step": 28952
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6297566371681415e-06,
+      "loss": 2.3002,
+      "step": 28953
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6297092288242732e-06,
+      "loss": 3.3026,
+      "step": 28954
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6296618204804045e-06,
+      "loss": 3.8247,
+      "step": 28955
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.629614412136536e-06,
+      "loss": 1.1261,
+      "step": 28956
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6295670037926677e-06,
+      "loss": 1.9236,
+      "step": 28957
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.629519595448799e-06,
+      "loss": 2.036,
+      "step": 28958
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6294721871049305e-06,
+      "loss": 2.8652,
+      "step": 28959
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6294247787610622e-06,
+      "loss": 2.4814,
+      "step": 28960
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6293773704171935e-06,
+      "loss": 2.0098,
+      "step": 28961
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.629329962073325e-06,
+      "loss": 2.2414,
+      "step": 28962
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6292825537294567e-06,
+      "loss": 1.3922,
+      "step": 28963
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.629235145385588e-06,
+      "loss": 2.4514,
+      "step": 28964
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6291877370417192e-06,
+      "loss": 1.7492,
+      "step": 28965
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.629140328697851e-06,
+      "loss": 3.0275,
+      "step": 28966
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6290929203539824e-06,
+      "loss": 2.0738,
+      "step": 28967
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6290455120101137e-06,
+      "loss": 1.4583,
+      "step": 28968
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6289981036662454e-06,
+      "loss": 2.7476,
+      "step": 28969
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6289506953223769e-06,
+      "loss": 1.8982,
+      "step": 28970
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6289032869785082e-06,
+      "loss": 2.6923,
+      "step": 28971
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6288558786346397e-06,
+      "loss": 3.4016,
+      "step": 28972
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.6288084702907714e-06,
+      "loss": 2.1136,
+      "step": 28973
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6287610619469027e-06,
+      "loss": 2.5378,
+      "step": 28974
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.628713653603034e-06,
+      "loss": 0.952,
+      "step": 28975
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6286662452591658e-06,
+      "loss": 1.2721,
+      "step": 28976
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6286188369152971e-06,
+      "loss": 1.8952,
+      "step": 28977
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6285714285714284e-06,
+      "loss": 2.4088,
+      "step": 28978
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6285240202275601e-06,
+      "loss": 3.386,
+      "step": 28979
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6284766118836916e-06,
+      "loss": 1.7271,
+      "step": 28980
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6284292035398229e-06,
+      "loss": 3.0978,
+      "step": 28981
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6283817951959546e-06,
+      "loss": 1.988,
+      "step": 28982
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.628334386852086e-06,
+      "loss": 2.5356,
+      "step": 28983
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6282869785082174e-06,
+      "loss": 2.1374,
+      "step": 28984
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.628239570164349e-06,
+      "loss": 2.3052,
+      "step": 28985
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6281921618204806e-06,
+      "loss": 4.4742,
+      "step": 28986
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6281447534766118e-06,
+      "loss": 2.0427,
+      "step": 28987
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6280973451327435e-06,
+      "loss": 2.519,
+      "step": 28988
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6280499367888748e-06,
+      "loss": 1.9027,
+      "step": 28989
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6280025284450063e-06,
+      "loss": 3.2825,
+      "step": 28990
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.627955120101138e-06,
+      "loss": 1.8677,
+      "step": 28991
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6279077117572693e-06,
+      "loss": 1.2614,
+      "step": 28992
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6278603034134008e-06,
+      "loss": 3.1109,
+      "step": 28993
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.627812895069532e-06,
+      "loss": 2.3729,
+      "step": 28994
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6277654867256638e-06,
+      "loss": 1.7902,
+      "step": 28995
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6277180783817953e-06,
+      "loss": 3.1021,
+      "step": 28996
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6276706700379265e-06,
+      "loss": 2.3814,
+      "step": 28997
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6276232616940583e-06,
+      "loss": 2.9513,
+      "step": 28998
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6275758533501897e-06,
+      "loss": 2.3487,
+      "step": 28999
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.627528445006321e-06,
+      "loss": 2.1941,
+      "step": 29000
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6274810366624527e-06,
+      "loss": 3.0258,
+      "step": 29001
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.627433628318584e-06,
+      "loss": 1.5395,
+      "step": 29002
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6273862199747155e-06,
+      "loss": 2.4806,
+      "step": 29003
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6273388116308472e-06,
+      "loss": 2.1372,
+      "step": 29004
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6272914032869785e-06,
+      "loss": 0.9658,
+      "step": 29005
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.62724399494311e-06,
+      "loss": 2.1294,
+      "step": 29006
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6271965865992417e-06,
+      "loss": 3.6133,
+      "step": 29007
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.627149178255373e-06,
+      "loss": 3.2844,
+      "step": 29008
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6271017699115045e-06,
+      "loss": 2.3859,
+      "step": 29009
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6270543615676362e-06,
+      "loss": 2.6239,
+      "step": 29010
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6270069532237674e-06,
+      "loss": 2.3794,
+      "step": 29011
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.626959544879899e-06,
+      "loss": 1.2518,
+      "step": 29012
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6269121365360306e-06,
+      "loss": 2.5574,
+      "step": 29013
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.626864728192162e-06,
+      "loss": 2.8835,
+      "step": 29014
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6268173198482932e-06,
+      "loss": 2.0994,
+      "step": 29015
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6267699115044247e-06,
+      "loss": 2.2499,
+      "step": 29016
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6267225031605564e-06,
+      "loss": 0.6711,
+      "step": 29017
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6266750948166877e-06,
+      "loss": 0.8136,
+      "step": 29018
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6266276864728192e-06,
+      "loss": 2.3461,
+      "step": 29019
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6265802781289509e-06,
+      "loss": 3.4152,
+      "step": 29020
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6265328697850821e-06,
+      "loss": 3.3702,
+      "step": 29021
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6264854614412136e-06,
+      "loss": 0.8312,
+      "step": 29022
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6264380530973453e-06,
+      "loss": 1.8799,
+      "step": 29023
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6263906447534766e-06,
+      "loss": 2.3607,
+      "step": 29024
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6263432364096081e-06,
+      "loss": 3.1827,
+      "step": 29025
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6262958280657398e-06,
+      "loss": 1.787,
+      "step": 29026
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.626248419721871e-06,
+      "loss": 3.412,
+      "step": 29027
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6262010113780024e-06,
+      "loss": 3.5302,
+      "step": 29028
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.626153603034134e-06,
+      "loss": 2.9629,
+      "step": 29029
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6261061946902656e-06,
+      "loss": 1.5329,
+      "step": 29030
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6260587863463969e-06,
+      "loss": 2.2787,
+      "step": 29031
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6260113780025286e-06,
+      "loss": 2.9095,
+      "step": 29032
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.62596396965866e-06,
+      "loss": 3.4289,
+      "step": 29033
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6259165613147913e-06,
+      "loss": 3.1133,
+      "step": 29034
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.625869152970923e-06,
+      "loss": 1.0722,
+      "step": 29035
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.6258217446270545e-06,
+      "loss": 2.6847,
+      "step": 29036
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6257743362831858e-06,
+      "loss": 1.8549,
+      "step": 29037
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6257269279393173e-06,
+      "loss": 2.6437,
+      "step": 29038
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.625679519595449e-06,
+      "loss": 3.1467,
+      "step": 29039
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6256321112515803e-06,
+      "loss": 1.6783,
+      "step": 29040
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6255847029077116e-06,
+      "loss": 3.4243,
+      "step": 29041
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6255372945638433e-06,
+      "loss": 1.7021,
+      "step": 29042
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6254898862199748e-06,
+      "loss": 2.6767,
+      "step": 29043
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.625442477876106e-06,
+      "loss": 2.3276,
+      "step": 29044
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6253950695322377e-06,
+      "loss": 1.7971,
+      "step": 29045
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6253476611883692e-06,
+      "loss": 2.6179,
+      "step": 29046
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6253002528445005e-06,
+      "loss": 2.1797,
+      "step": 29047
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6252528445006322e-06,
+      "loss": 3.317,
+      "step": 29048
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6252054361567637e-06,
+      "loss": 2.0697,
+      "step": 29049
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.625158027812895e-06,
+      "loss": 1.9572,
+      "step": 29050
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6251106194690267e-06,
+      "loss": 2.1728,
+      "step": 29051
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6250632111251582e-06,
+      "loss": 4.1225,
+      "step": 29052
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6250158027812895e-06,
+      "loss": 3.1796,
+      "step": 29053
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6249683944374212e-06,
+      "loss": 4.1002,
+      "step": 29054
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6249209860935525e-06,
+      "loss": 1.8301,
+      "step": 29055
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.624873577749684e-06,
+      "loss": 2.3642,
+      "step": 29056
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6248261694058152e-06,
+      "loss": 2.758,
+      "step": 29057
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.624778761061947e-06,
+      "loss": 2.1613,
+      "step": 29058
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6247313527180784e-06,
+      "loss": 2.2925,
+      "step": 29059
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6246839443742097e-06,
+      "loss": 1.9199,
+      "step": 29060
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6246365360303414e-06,
+      "loss": 2.4857,
+      "step": 29061
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.624589127686473e-06,
+      "loss": 2.3541,
+      "step": 29062
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6245417193426042e-06,
+      "loss": 2.443,
+      "step": 29063
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6244943109987359e-06,
+      "loss": 2.8348,
+      "step": 29064
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6244469026548674e-06,
+      "loss": 2.6704,
+      "step": 29065
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6243994943109987e-06,
+      "loss": 3.4677,
+      "step": 29066
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6243520859671304e-06,
+      "loss": 1.7837,
+      "step": 29067
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6243046776232616e-06,
+      "loss": 3.4217,
+      "step": 29068
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6242572692793931e-06,
+      "loss": 1.0154,
+      "step": 29069
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6242098609355248e-06,
+      "loss": 2.0742,
+      "step": 29070
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6241624525916561e-06,
+      "loss": 2.2536,
+      "step": 29071
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6241150442477876e-06,
+      "loss": 3.7008,
+      "step": 29072
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6240676359039193e-06,
+      "loss": 2.3838,
+      "step": 29073
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6240202275600506e-06,
+      "loss": 1.539,
+      "step": 29074
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.623972819216182e-06,
+      "loss": 2.3854,
+      "step": 29075
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6239254108723138e-06,
+      "loss": 2.8966,
+      "step": 29076
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.623878002528445e-06,
+      "loss": 1.8769,
+      "step": 29077
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6238305941845766e-06,
+      "loss": 1.5835,
+      "step": 29078
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6237831858407078e-06,
+      "loss": 1.9057,
+      "step": 29079
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6237357774968396e-06,
+      "loss": 3.7545,
+      "step": 29080
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6236883691529708e-06,
+      "loss": 2.4233,
+      "step": 29081
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6236409608091023e-06,
+      "loss": 1.511,
+      "step": 29082
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.623593552465234e-06,
+      "loss": 1.3436,
+      "step": 29083
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6235461441213653e-06,
+      "loss": 0.8453,
+      "step": 29084
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6234987357774968e-06,
+      "loss": 2.5474,
+      "step": 29085
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6234513274336285e-06,
+      "loss": 2.4352,
+      "step": 29086
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6234039190897598e-06,
+      "loss": 2.2461,
+      "step": 29087
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6233565107458913e-06,
+      "loss": 3.2608,
+      "step": 29088
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.623309102402023e-06,
+      "loss": 2.7463,
+      "step": 29089
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6232616940581543e-06,
+      "loss": 3.7739,
+      "step": 29090
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6232142857142858e-06,
+      "loss": 2.1568,
+      "step": 29091
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6231668773704175e-06,
+      "loss": 2.5326,
+      "step": 29092
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6231194690265487e-06,
+      "loss": 1.9835,
+      "step": 29093
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.62307206068268e-06,
+      "loss": 2.5472,
+      "step": 29094
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6230246523388117e-06,
+      "loss": 1.9171,
+      "step": 29095
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6229772439949432e-06,
+      "loss": 2.252,
+      "step": 29096
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6229298356510745e-06,
+      "loss": 1.5312,
+      "step": 29097
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6228824273072062e-06,
+      "loss": 3.6581,
+      "step": 29098
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6228350189633377e-06,
+      "loss": 2.393,
+      "step": 29099
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.622787610619469e-06,
+      "loss": 1.9075,
+      "step": 29100
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6227402022756005e-06,
+      "loss": 2.3473,
+      "step": 29101
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6226927939317322e-06,
+      "loss": 1.3192,
+      "step": 29102
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6226453855878634e-06,
+      "loss": 3.2633,
+      "step": 29103
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.622597977243995e-06,
+      "loss": 2.2862,
+      "step": 29104
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6225505689001266e-06,
+      "loss": 2.4552,
+      "step": 29105
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.622503160556258e-06,
+      "loss": 2.5763,
+      "step": 29106
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6224557522123892e-06,
+      "loss": 1.9172,
+      "step": 29107
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.622408343868521e-06,
+      "loss": 1.2755,
+      "step": 29108
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6223609355246524e-06,
+      "loss": 2.405,
+      "step": 29109
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6223135271807837e-06,
+      "loss": 1.5856,
+      "step": 29110
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6222661188369154e-06,
+      "loss": 1.8338,
+      "step": 29111
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6222187104930469e-06,
+      "loss": 2.0127,
+      "step": 29112
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6221713021491782e-06,
+      "loss": 3.3735,
+      "step": 29113
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6221238938053099e-06,
+      "loss": 1.4464,
+      "step": 29114
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6220764854614414e-06,
+      "loss": 1.761,
+      "step": 29115
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6220290771175726e-06,
+      "loss": 2.6825,
+      "step": 29116
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6219816687737043e-06,
+      "loss": 2.2995,
+      "step": 29117
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6219342604298358e-06,
+      "loss": 0.8282,
+      "step": 29118
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6218868520859671e-06,
+      "loss": 2.71,
+      "step": 29119
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6218394437420988e-06,
+      "loss": 3.0195,
+      "step": 29120
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.62179203539823e-06,
+      "loss": 2.8296,
+      "step": 29121
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6217446270543616e-06,
+      "loss": 1.8515,
+      "step": 29122
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6216972187104929e-06,
+      "loss": 2.1672,
+      "step": 29123
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6216498103666246e-06,
+      "loss": 1.8969,
+      "step": 29124
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.621602402022756e-06,
+      "loss": 3.1393,
+      "step": 29125
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6215549936788873e-06,
+      "loss": 1.4201,
+      "step": 29126
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.621507585335019e-06,
+      "loss": 1.9235,
+      "step": 29127
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6214601769911505e-06,
+      "loss": 2.7637,
+      "step": 29128
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6214127686472818e-06,
+      "loss": 1.8422,
+      "step": 29129
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6213653603034135e-06,
+      "loss": 2.6445,
+      "step": 29130
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6213179519595448e-06,
+      "loss": 1.1115,
+      "step": 29131
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6212705436156763e-06,
+      "loss": 2.8592,
+      "step": 29132
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.621223135271808e-06,
+      "loss": 2.2347,
+      "step": 29133
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6211757269279393e-06,
+      "loss": 2.1148,
+      "step": 29134
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6211283185840708e-06,
+      "loss": 3.434,
+      "step": 29135
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6210809102402025e-06,
+      "loss": 1.7545,
+      "step": 29136
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6210335018963338e-06,
+      "loss": 3.6138,
+      "step": 29137
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6209860935524652e-06,
+      "loss": 2.9861,
+      "step": 29138
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.620938685208597e-06,
+      "loss": 2.2315,
+      "step": 29139
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6208912768647282e-06,
+      "loss": 3.5714,
+      "step": 29140
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6208438685208597e-06,
+      "loss": 2.1862,
+      "step": 29141
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6207964601769914e-06,
+      "loss": 2.5958,
+      "step": 29142
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6207490518331227e-06,
+      "loss": 3.2247,
+      "step": 29143
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.620701643489254e-06,
+      "loss": 1.3506,
+      "step": 29144
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6206542351453855e-06,
+      "loss": 2.8048,
+      "step": 29145
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6206068268015172e-06,
+      "loss": 1.8154,
+      "step": 29146
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6205594184576485e-06,
+      "loss": 1.7889,
+      "step": 29147
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.62051201011378e-06,
+      "loss": 2.3092,
+      "step": 29148
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6204646017699117e-06,
+      "loss": 1.8864,
+      "step": 29149
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.620417193426043e-06,
+      "loss": 2.7795,
+      "step": 29150
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6203697850821744e-06,
+      "loss": 2.6405,
+      "step": 29151
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6203223767383061e-06,
+      "loss": 2.3694,
+      "step": 29152
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6202749683944374e-06,
+      "loss": 2.5022,
+      "step": 29153
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.620227560050569e-06,
+      "loss": 1.7663,
+      "step": 29154
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6201801517067006e-06,
+      "loss": 3.5317,
+      "step": 29155
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.620132743362832e-06,
+      "loss": 1.4936,
+      "step": 29156
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6200853350189632e-06,
+      "loss": 2.2674,
+      "step": 29157
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6200379266750949e-06,
+      "loss": 2.124,
+      "step": 29158
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6199905183312264e-06,
+      "loss": 2.2332,
+      "step": 29159
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6199431099873577e-06,
+      "loss": 3.076,
+      "step": 29160
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6198957016434894e-06,
+      "loss": 2.8081,
+      "step": 29161
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6198482932996208e-06,
+      "loss": 2.4487,
+      "step": 29162
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6198008849557521e-06,
+      "loss": 2.2979,
+      "step": 29163
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6197534766118838e-06,
+      "loss": 2.2837,
+      "step": 29164
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6197060682680153e-06,
+      "loss": 2.0428,
+      "step": 29165
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6196586599241466e-06,
+      "loss": 1.9721,
+      "step": 29166
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.619611251580278e-06,
+      "loss": 1.8214,
+      "step": 29167
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6195638432364098e-06,
+      "loss": 1.1392,
+      "step": 29168
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.619516434892541e-06,
+      "loss": 2.2395,
+      "step": 29169
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6194690265486724e-06,
+      "loss": 1.9947,
+      "step": 29170
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.619421618204804e-06,
+      "loss": 3.5528,
+      "step": 29171
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6193742098609356e-06,
+      "loss": 2.2292,
+      "step": 29172
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6193268015170668e-06,
+      "loss": 1.8384,
+      "step": 29173
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6192793931731985e-06,
+      "loss": 4.3022,
+      "step": 29174
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.61923198482933e-06,
+      "loss": 3.4728,
+      "step": 29175
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6191845764854613e-06,
+      "loss": 3.3288,
+      "step": 29176
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.619137168141593e-06,
+      "loss": 3.3926,
+      "step": 29177
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6190897597977245e-06,
+      "loss": 2.7354,
+      "step": 29178
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6190423514538558e-06,
+      "loss": 2.9788,
+      "step": 29179
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6189949431099875e-06,
+      "loss": 1.5553,
+      "step": 29180
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.618947534766119e-06,
+      "loss": 2.4492,
+      "step": 29181
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6189001264222503e-06,
+      "loss": 2.9556,
+      "step": 29182
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.618852718078382e-06,
+      "loss": 3.4572,
+      "step": 29183
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6188053097345133e-06,
+      "loss": 2.7463,
+      "step": 29184
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6187579013906447e-06,
+      "loss": 2.5275,
+      "step": 29185
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6187104930467765e-06,
+      "loss": 2.3531,
+      "step": 29186
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6186630847029077e-06,
+      "loss": 1.1306,
+      "step": 29187
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6186156763590392e-06,
+      "loss": 1.7835,
+      "step": 29188
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6185682680151705e-06,
+      "loss": 2.9856,
+      "step": 29189
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6185208596713022e-06,
+      "loss": 2.2578,
+      "step": 29190
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6184734513274337e-06,
+      "loss": 1.7467,
+      "step": 29191
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.618426042983565e-06,
+      "loss": 4.309,
+      "step": 29192
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6183786346396967e-06,
+      "loss": 3.6939,
+      "step": 29193
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6183312262958282e-06,
+      "loss": 1.5145,
+      "step": 29194
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6182838179519595e-06,
+      "loss": 2.4758,
+      "step": 29195
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6182364096080912e-06,
+      "loss": 2.3908,
+      "step": 29196
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6181890012642224e-06,
+      "loss": 2.6207,
+      "step": 29197
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.618141592920354e-06,
+      "loss": 1.266,
+      "step": 29198
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6180941845764856e-06,
+      "loss": 2.6916,
+      "step": 29199
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.618046776232617e-06,
+      "loss": 2.7624,
+      "step": 29200
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6179993678887484e-06,
+      "loss": 1.2282,
+      "step": 29201
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6179519595448801e-06,
+      "loss": 1.9459,
+      "step": 29202
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6179045512010114e-06,
+      "loss": 3.0079,
+      "step": 29203
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6178571428571429e-06,
+      "loss": 3.143,
+      "step": 29204
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6178097345132746e-06,
+      "loss": 2.6737,
+      "step": 29205
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6177623261694059e-06,
+      "loss": 1.9287,
+      "step": 29206
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6177149178255374e-06,
+      "loss": 2.0028,
+      "step": 29207
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.617667509481669e-06,
+      "loss": 1.3917,
+      "step": 29208
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6176201011378003e-06,
+      "loss": 3.2841,
+      "step": 29209
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6175726927939316e-06,
+      "loss": 2.6223,
+      "step": 29210
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6175252844500631e-06,
+      "loss": 3.5175,
+      "step": 29211
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6174778761061948e-06,
+      "loss": 3.489,
+      "step": 29212
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.617430467762326e-06,
+      "loss": 2.0895,
+      "step": 29213
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6173830594184576e-06,
+      "loss": 3.4592,
+      "step": 29214
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6173356510745893e-06,
+      "loss": 2.3274,
+      "step": 29215
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6172882427307206e-06,
+      "loss": 2.0021,
+      "step": 29216
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.617240834386852e-06,
+      "loss": 3.2861,
+      "step": 29217
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6171934260429838e-06,
+      "loss": 3.4084,
+      "step": 29218
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.617146017699115e-06,
+      "loss": 3.6321,
+      "step": 29219
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6170986093552465e-06,
+      "loss": 1.2464,
+      "step": 29220
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6170512010113783e-06,
+      "loss": 4.2059,
+      "step": 29221
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6170037926675095e-06,
+      "loss": 2.2493,
+      "step": 29222
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6169563843236408e-06,
+      "loss": 2.8914,
+      "step": 29223
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6169089759797725e-06,
+      "loss": 2.512,
+      "step": 29224
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.616861567635904e-06,
+      "loss": 1.6716,
+      "step": 29225
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.6168141592920353e-06,
+      "loss": 2.7388,
+      "step": 29226
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.616766750948167e-06,
+      "loss": 2.5234,
+      "step": 29227
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6167193426042985e-06,
+      "loss": 1.7235,
+      "step": 29228
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6166719342604298e-06,
+      "loss": 2.7053,
+      "step": 29229
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6166245259165615e-06,
+      "loss": 2.1206,
+      "step": 29230
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.616577117572693e-06,
+      "loss": 2.3242,
+      "step": 29231
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6165297092288242e-06,
+      "loss": 3.9414,
+      "step": 29232
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6164823008849557e-06,
+      "loss": 1.7012,
+      "step": 29233
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6164348925410874e-06,
+      "loss": 2.5304,
+      "step": 29234
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6163874841972187e-06,
+      "loss": 1.7547,
+      "step": 29235
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.61634007585335e-06,
+      "loss": 2.6677,
+      "step": 29236
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6162926675094817e-06,
+      "loss": 2.8154,
+      "step": 29237
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6162452591656132e-06,
+      "loss": 3.2181,
+      "step": 29238
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6161978508217445e-06,
+      "loss": 1.775,
+      "step": 29239
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6161504424778762e-06,
+      "loss": 2.2299,
+      "step": 29240
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6161030341340077e-06,
+      "loss": 2.6853,
+      "step": 29241
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.616055625790139e-06,
+      "loss": 1.8125,
+      "step": 29242
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6160082174462707e-06,
+      "loss": 3.699,
+      "step": 29243
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6159608091024021e-06,
+      "loss": 2.2734,
+      "step": 29244
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6159134007585334e-06,
+      "loss": 2.6812,
+      "step": 29245
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6158659924146651e-06,
+      "loss": 1.4041,
+      "step": 29246
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6158185840707966e-06,
+      "loss": 1.8636,
+      "step": 29247
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.615771175726928e-06,
+      "loss": 2.0564,
+      "step": 29248
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6157237673830596e-06,
+      "loss": 3.0926,
+      "step": 29249
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6156763590391909e-06,
+      "loss": 2.9621,
+      "step": 29250
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6156289506953224e-06,
+      "loss": 2.5703,
+      "step": 29251
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.615581542351454e-06,
+      "loss": 3.2239,
+      "step": 29252
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6155341340075854e-06,
+      "loss": 2.3936,
+      "step": 29253
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6154867256637169e-06,
+      "loss": 2.2102,
+      "step": 29254
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6154393173198481e-06,
+      "loss": 1.5725,
+      "step": 29255
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6153919089759798e-06,
+      "loss": 1.6036,
+      "step": 29256
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6153445006321113e-06,
+      "loss": 1.3475,
+      "step": 29257
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6152970922882426e-06,
+      "loss": 2.3606,
+      "step": 29258
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6152496839443743e-06,
+      "loss": 1.5008,
+      "step": 29259
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6152022756005056e-06,
+      "loss": 3.6019,
+      "step": 29260
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.615154867256637e-06,
+      "loss": 2.173,
+      "step": 29261
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6151074589127688e-06,
+      "loss": 2.1375,
+      "step": 29262
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6150600505689e-06,
+      "loss": 0.8153,
+      "step": 29263
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6150126422250316e-06,
+      "loss": 2.9348,
+      "step": 29264
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6149652338811633e-06,
+      "loss": 2.7366,
+      "step": 29265
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6149178255372946e-06,
+      "loss": 2.3684,
+      "step": 29266
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.614870417193426e-06,
+      "loss": 3.155,
+      "step": 29267
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6148230088495578e-06,
+      "loss": 0.7136,
+      "step": 29268
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.614775600505689e-06,
+      "loss": 2.9186,
+      "step": 29269
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6147281921618205e-06,
+      "loss": 2.0603,
+      "step": 29270
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6146807838179522e-06,
+      "loss": 3.9102,
+      "step": 29271
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6146333754740835e-06,
+      "loss": 1.3229,
+      "step": 29272
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6145859671302148e-06,
+      "loss": 2.5306,
+      "step": 29273
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6145385587863465e-06,
+      "loss": 2.392,
+      "step": 29274
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.614491150442478e-06,
+      "loss": 1.91,
+      "step": 29275
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6144437420986093e-06,
+      "loss": 2.8722,
+      "step": 29276
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6143963337547408e-06,
+      "loss": 3.2629,
+      "step": 29277
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6143489254108725e-06,
+      "loss": 1.2472,
+      "step": 29278
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6143015170670037e-06,
+      "loss": 2.1329,
+      "step": 29279
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6142541087231352e-06,
+      "loss": 3.1914,
+      "step": 29280
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.614206700379267e-06,
+      "loss": 2.6026,
+      "step": 29281
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6141592920353982e-06,
+      "loss": 2.4148,
+      "step": 29282
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6141118836915297e-06,
+      "loss": 2.0054,
+      "step": 29283
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6140644753476614e-06,
+      "loss": 2.1853,
+      "step": 29284
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6140170670037927e-06,
+      "loss": 2.1252,
+      "step": 29285
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.613969658659924e-06,
+      "loss": 2.4813,
+      "step": 29286
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6139222503160557e-06,
+      "loss": 1.9208,
+      "step": 29287
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6138748419721872e-06,
+      "loss": 2.4556,
+      "step": 29288
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6138274336283185e-06,
+      "loss": 1.6628,
+      "step": 29289
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.6137800252844502e-06,
+      "loss": 3.9095,
+      "step": 29290
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6137326169405816e-06,
+      "loss": 3.3295,
+      "step": 29291
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.613685208596713e-06,
+      "loss": 3.7313,
+      "step": 29292
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6136378002528446e-06,
+      "loss": 1.9485,
+      "step": 29293
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6135903919089761e-06,
+      "loss": 2.3149,
+      "step": 29294
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6135429835651074e-06,
+      "loss": 1.6553,
+      "step": 29295
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6134955752212391e-06,
+      "loss": 2.6167,
+      "step": 29296
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6134481668773706e-06,
+      "loss": 2.4379,
+      "step": 29297
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6134007585335019e-06,
+      "loss": 2.6499,
+      "step": 29298
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6133533501896332e-06,
+      "loss": 1.1303,
+      "step": 29299
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6133059418457649e-06,
+      "loss": 2.6367,
+      "step": 29300
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6132585335018964e-06,
+      "loss": 3.3146,
+      "step": 29301
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6132111251580276e-06,
+      "loss": 2.0788,
+      "step": 29302
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6131637168141593e-06,
+      "loss": 4.0199,
+      "step": 29303
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6131163084702908e-06,
+      "loss": 3.2093,
+      "step": 29304
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6130689001264221e-06,
+      "loss": 2.3112,
+      "step": 29305
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6130214917825538e-06,
+      "loss": 1.4912,
+      "step": 29306
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6129740834386853e-06,
+      "loss": 2.4018,
+      "step": 29307
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6129266750948166e-06,
+      "loss": 1.9646,
+      "step": 29308
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6128792667509483e-06,
+      "loss": 2.1453,
+      "step": 29309
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6128318584070798e-06,
+      "loss": 2.7848,
+      "step": 29310
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.612784450063211e-06,
+      "loss": 2.3675,
+      "step": 29311
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6127370417193428e-06,
+      "loss": 1.9652,
+      "step": 29312
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.612689633375474e-06,
+      "loss": 3.4713,
+      "step": 29313
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6126422250316055e-06,
+      "loss": 2.6522,
+      "step": 29314
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6125948166877372e-06,
+      "loss": 3.1807,
+      "step": 29315
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6125474083438685e-06,
+      "loss": 3.1491,
+      "step": 29316
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6125e-06,
+      "loss": 1.6082,
+      "step": 29317
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6124525916561317e-06,
+      "loss": 1.2567,
+      "step": 29318
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.612405183312263e-06,
+      "loss": 2.0466,
+      "step": 29319
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6123577749683945e-06,
+      "loss": 3.0569,
+      "step": 29320
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6123103666245258e-06,
+      "loss": 2.6807,
+      "step": 29321
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6122629582806575e-06,
+      "loss": 2.0377,
+      "step": 29322
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.612215549936789e-06,
+      "loss": 2.1672,
+      "step": 29323
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6121681415929203e-06,
+      "loss": 2.2914,
+      "step": 29324
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.612120733249052e-06,
+      "loss": 1.9193,
+      "step": 29325
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6120733249051832e-06,
+      "loss": 3.7023,
+      "step": 29326
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6120259165613147e-06,
+      "loss": 2.2257,
+      "step": 29327
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6119785082174464e-06,
+      "loss": 2.7039,
+      "step": 29328
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6119310998735777e-06,
+      "loss": 1.9313,
+      "step": 29329
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6118836915297092e-06,
+      "loss": 1.8483,
+      "step": 29330
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.611836283185841e-06,
+      "loss": 3.9281,
+      "step": 29331
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6117888748419722e-06,
+      "loss": 3.4683,
+      "step": 29332
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6117414664981037e-06,
+      "loss": 2.135,
+      "step": 29333
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6116940581542354e-06,
+      "loss": 2.6698,
+      "step": 29334
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6116466498103667e-06,
+      "loss": 2.4828,
+      "step": 29335
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6115992414664982e-06,
+      "loss": 2.6166,
+      "step": 29336
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6115518331226299e-06,
+      "loss": 2.5685,
+      "step": 29337
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6115044247787611e-06,
+      "loss": 2.541,
+      "step": 29338
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6114570164348924e-06,
+      "loss": 1.8772,
+      "step": 29339
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6114096080910241e-06,
+      "loss": 2.8183,
+      "step": 29340
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6113621997471556e-06,
+      "loss": 0.7379,
+      "step": 29341
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.611314791403287e-06,
+      "loss": 2.4042,
+      "step": 29342
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6112673830594184e-06,
+      "loss": 1.8582,
+      "step": 29343
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.61121997471555e-06,
+      "loss": 2.5719,
+      "step": 29344
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6111725663716814e-06,
+      "loss": 2.0004,
+      "step": 29345
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6111251580278129e-06,
+      "loss": 2.0094,
+      "step": 29346
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6110777496839446e-06,
+      "loss": 1.6301,
+      "step": 29347
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6110303413400759e-06,
+      "loss": 1.7334,
+      "step": 29348
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6109829329962073e-06,
+      "loss": 1.6181,
+      "step": 29349
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.610935524652339e-06,
+      "loss": 3.2356,
+      "step": 29350
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6108881163084703e-06,
+      "loss": 3.2819,
+      "step": 29351
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6108407079646016e-06,
+      "loss": 3.1116,
+      "step": 29352
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.6107932996207333e-06,
+      "loss": 1.4818,
+      "step": 29353
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6107458912768648e-06,
+      "loss": 2.4344,
+      "step": 29354
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.610698482932996e-06,
+      "loss": 2.2257,
+      "step": 29355
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6106510745891278e-06,
+      "loss": 3.3644,
+      "step": 29356
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6106036662452593e-06,
+      "loss": 1.258,
+      "step": 29357
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6105562579013906e-06,
+      "loss": 2.4745,
+      "step": 29358
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6105088495575223e-06,
+      "loss": 3.0138,
+      "step": 29359
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6104614412136538e-06,
+      "loss": 1.1554,
+      "step": 29360
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.610414032869785e-06,
+      "loss": 3.5385,
+      "step": 29361
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6103666245259167e-06,
+      "loss": 1.9676,
+      "step": 29362
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6103192161820482e-06,
+      "loss": 2.1544,
+      "step": 29363
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6102718078381795e-06,
+      "loss": 0.4006,
+      "step": 29364
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6102243994943108e-06,
+      "loss": 2.2999,
+      "step": 29365
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6101769911504425e-06,
+      "loss": 1.4149,
+      "step": 29366
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.610129582806574e-06,
+      "loss": 3.0904,
+      "step": 29367
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6100821744627053e-06,
+      "loss": 2.9345,
+      "step": 29368
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.610034766118837e-06,
+      "loss": 1.7224,
+      "step": 29369
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6099873577749685e-06,
+      "loss": 1.4369,
+      "step": 29370
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6099399494310998e-06,
+      "loss": 2.4588,
+      "step": 29371
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6098925410872315e-06,
+      "loss": 2.4409,
+      "step": 29372
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.609845132743363e-06,
+      "loss": 2.7374,
+      "step": 29373
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6097977243994942e-06,
+      "loss": 2.5691,
+      "step": 29374
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.609750316055626e-06,
+      "loss": 3.4457,
+      "step": 29375
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6097029077117574e-06,
+      "loss": 3.1253,
+      "step": 29376
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6096554993678887e-06,
+      "loss": 1.8618,
+      "step": 29377
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6096080910240204e-06,
+      "loss": 2.4694,
+      "step": 29378
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6095606826801517e-06,
+      "loss": 2.4148,
+      "step": 29379
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6095132743362832e-06,
+      "loss": 2.4281,
+      "step": 29380
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6094658659924149e-06,
+      "loss": 2.6711,
+      "step": 29381
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6094184576485462e-06,
+      "loss": 1.9768,
+      "step": 29382
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6093710493046777e-06,
+      "loss": 1.0457,
+      "step": 29383
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6093236409608094e-06,
+      "loss": 2.1012,
+      "step": 29384
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6092762326169406e-06,
+      "loss": 2.1648,
+      "step": 29385
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6092288242730721e-06,
+      "loss": 1.5893,
+      "step": 29386
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6091814159292034e-06,
+      "loss": 4.467,
+      "step": 29387
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6091340075853351e-06,
+      "loss": 3.5052,
+      "step": 29388
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6090865992414666e-06,
+      "loss": 2.4404,
+      "step": 29389
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6090391908975979e-06,
+      "loss": 2.7671,
+      "step": 29390
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6089917825537296e-06,
+      "loss": 2.9216,
+      "step": 29391
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6089443742098609e-06,
+      "loss": 2.6561,
+      "step": 29392
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6088969658659924e-06,
+      "loss": 2.8282,
+      "step": 29393
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.608849557522124e-06,
+      "loss": 2.8787,
+      "step": 29394
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6088021491782554e-06,
+      "loss": 1.2685,
+      "step": 29395
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6087547408343868e-06,
+      "loss": 2.6562,
+      "step": 29396
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6087073324905185e-06,
+      "loss": 2.7997,
+      "step": 29397
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6086599241466498e-06,
+      "loss": 1.7426,
+      "step": 29398
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6086125158027813e-06,
+      "loss": 1.8992,
+      "step": 29399
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.608565107458913e-06,
+      "loss": 2.7431,
+      "step": 29400
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6085176991150443e-06,
+      "loss": 3.5381,
+      "step": 29401
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6084702907711756e-06,
+      "loss": 2.0689,
+      "step": 29402
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6084228824273075e-06,
+      "loss": 2.2501,
+      "step": 29403
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6083754740834388e-06,
+      "loss": 3.165,
+      "step": 29404
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.60832806573957e-06,
+      "loss": 3.2031,
+      "step": 29405
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6082806573957018e-06,
+      "loss": 3.6159,
+      "step": 29406
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6082332490518333e-06,
+      "loss": 3.3747,
+      "step": 29407
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6081858407079645e-06,
+      "loss": 2.4727,
+      "step": 29408
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.608138432364096e-06,
+      "loss": 2.9543,
+      "step": 29409
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6080910240202277e-06,
+      "loss": 1.7857,
+      "step": 29410
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.608043615676359e-06,
+      "loss": 2.6412,
+      "step": 29411
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6079962073324905e-06,
+      "loss": 2.8415,
+      "step": 29412
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6079487989886222e-06,
+      "loss": 1.8895,
+      "step": 29413
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6079013906447535e-06,
+      "loss": 2.6936,
+      "step": 29414
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6078539823008848e-06,
+      "loss": 2.4177,
+      "step": 29415
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.6078065739570165e-06,
+      "loss": 3.1451,
+      "step": 29416
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.607759165613148e-06,
+      "loss": 2.7151,
+      "step": 29417
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6077117572692792e-06,
+      "loss": 1.2439,
+      "step": 29418
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.607664348925411e-06,
+      "loss": 2.4127,
+      "step": 29419
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6076169405815424e-06,
+      "loss": 3.1656,
+      "step": 29420
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6075695322376737e-06,
+      "loss": 3.0144,
+      "step": 29421
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6075221238938054e-06,
+      "loss": 1.5708,
+      "step": 29422
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.607474715549937e-06,
+      "loss": 3.1498,
+      "step": 29423
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6074273072060682e-06,
+      "loss": 3.3157,
+      "step": 29424
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6073798988622e-06,
+      "loss": 2.8018,
+      "step": 29425
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6073324905183314e-06,
+      "loss": 2.5618,
+      "step": 29426
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6072850821744627e-06,
+      "loss": 3.0869,
+      "step": 29427
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6072376738305944e-06,
+      "loss": 3.2148,
+      "step": 29428
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6071902654867257e-06,
+      "loss": 2.9712,
+      "step": 29429
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6071428571428572e-06,
+      "loss": 2.1922,
+      "step": 29430
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6070954487989884e-06,
+      "loss": 2.2805,
+      "step": 29431
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6070480404551201e-06,
+      "loss": 2.8709,
+      "step": 29432
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6070006321112516e-06,
+      "loss": 3.1488,
+      "step": 29433
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.606953223767383e-06,
+      "loss": 2.832,
+      "step": 29434
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6069058154235146e-06,
+      "loss": 2.1822,
+      "step": 29435
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6068584070796461e-06,
+      "loss": 2.4119,
+      "step": 29436
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6068109987357774e-06,
+      "loss": 3.5687,
+      "step": 29437
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.606763590391909e-06,
+      "loss": 2.4573,
+      "step": 29438
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6067161820480406e-06,
+      "loss": 2.5522,
+      "step": 29439
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6066687737041719e-06,
+      "loss": 3.1898,
+      "step": 29440
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6066213653603036e-06,
+      "loss": 2.3433,
+      "step": 29441
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6065739570164348e-06,
+      "loss": 2.2941,
+      "step": 29442
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6065265486725663e-06,
+      "loss": 2.4472,
+      "step": 29443
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.606479140328698e-06,
+      "loss": 1.7987,
+      "step": 29444
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6064317319848293e-06,
+      "loss": 2.866,
+      "step": 29445
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6063843236409608e-06,
+      "loss": 2.2062,
+      "step": 29446
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6063369152970925e-06,
+      "loss": 2.1836,
+      "step": 29447
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6062895069532238e-06,
+      "loss": 2.3852,
+      "step": 29448
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6062420986093553e-06,
+      "loss": 2.5818,
+      "step": 29449
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.606194690265487e-06,
+      "loss": 3.4088,
+      "step": 29450
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6061472819216183e-06,
+      "loss": 2.1202,
+      "step": 29451
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6060998735777498e-06,
+      "loss": 2.6106,
+      "step": 29452
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.606052465233881e-06,
+      "loss": 3.1239,
+      "step": 29453
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6060050568900128e-06,
+      "loss": 2.6962,
+      "step": 29454
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.605957648546144e-06,
+      "loss": 1.7765,
+      "step": 29455
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6059102402022755e-06,
+      "loss": 2.6467,
+      "step": 29456
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6058628318584072e-06,
+      "loss": 2.2586,
+      "step": 29457
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6058154235145385e-06,
+      "loss": 3.2008,
+      "step": 29458
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.60576801517067e-06,
+      "loss": 1.7445,
+      "step": 29459
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6057206068268017e-06,
+      "loss": 3.3284,
+      "step": 29460
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.605673198482933e-06,
+      "loss": 2.8516,
+      "step": 29461
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6056257901390645e-06,
+      "loss": 2.7529,
+      "step": 29462
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6055783817951962e-06,
+      "loss": 1.6511,
+      "step": 29463
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6055309734513275e-06,
+      "loss": 0.7092,
+      "step": 29464
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.605483565107459e-06,
+      "loss": 3.2004,
+      "step": 29465
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6054361567635907e-06,
+      "loss": 3.2356,
+      "step": 29466
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.605388748419722e-06,
+      "loss": 2.7076,
+      "step": 29467
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6053413400758532e-06,
+      "loss": 1.2703,
+      "step": 29468
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.605293931731985e-06,
+      "loss": 2.244,
+      "step": 29469
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6052465233881164e-06,
+      "loss": 0.4379,
+      "step": 29470
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6051991150442477e-06,
+      "loss": 2.1385,
+      "step": 29471
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6051517067003794e-06,
+      "loss": 1.8847,
+      "step": 29472
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6051042983565109e-06,
+      "loss": 2.0376,
+      "step": 29473
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6050568900126422e-06,
+      "loss": 1.7584,
+      "step": 29474
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6050094816687737e-06,
+      "loss": 3.0159,
+      "step": 29475
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6049620733249054e-06,
+      "loss": 2.1274,
+      "step": 29476
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6049146649810367e-06,
+      "loss": 2.525,
+      "step": 29477
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6048672566371681e-06,
+      "loss": 2.5777,
+      "step": 29478
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6048198482932998e-06,
+      "loss": 1.9586,
+      "step": 29479
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.6047724399494311e-06,
+      "loss": 1.8003,
+      "step": 29480
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6047250316055624e-06,
+      "loss": 1.4787,
+      "step": 29481
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6046776232616941e-06,
+      "loss": 3.1871,
+      "step": 29482
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6046302149178256e-06,
+      "loss": 2.8212,
+      "step": 29483
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6045828065739569e-06,
+      "loss": 1.2001,
+      "step": 29484
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6045353982300886e-06,
+      "loss": 3.1842,
+      "step": 29485
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.60448798988622e-06,
+      "loss": 2.8842,
+      "step": 29486
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6044405815423514e-06,
+      "loss": 3.8229,
+      "step": 29487
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.604393173198483e-06,
+      "loss": 1.6986,
+      "step": 29488
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6043457648546146e-06,
+      "loss": 3.3076,
+      "step": 29489
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6042983565107458e-06,
+      "loss": 2.7899,
+      "step": 29490
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6042509481668775e-06,
+      "loss": 3.1141,
+      "step": 29491
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.604203539823009e-06,
+      "loss": 0.896,
+      "step": 29492
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6041561314791403e-06,
+      "loss": 2.9487,
+      "step": 29493
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6041087231352716e-06,
+      "loss": 1.0449,
+      "step": 29494
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6040613147914033e-06,
+      "loss": 2.4682,
+      "step": 29495
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6040139064475348e-06,
+      "loss": 1.15,
+      "step": 29496
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.603966498103666e-06,
+      "loss": 3.278,
+      "step": 29497
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6039190897597978e-06,
+      "loss": 1.4535,
+      "step": 29498
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6038716814159293e-06,
+      "loss": 1.5579,
+      "step": 29499
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6038242730720605e-06,
+      "loss": 2.7466,
+      "step": 29500
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6037768647281923e-06,
+      "loss": 1.743,
+      "step": 29501
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6037294563843237e-06,
+      "loss": 1.5714,
+      "step": 29502
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.603682048040455e-06,
+      "loss": 2.11,
+      "step": 29503
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6036346396965867e-06,
+      "loss": 2.0203,
+      "step": 29504
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6035872313527182e-06,
+      "loss": 2.8255,
+      "step": 29505
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6035398230088495e-06,
+      "loss": 2.6772,
+      "step": 29506
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6034924146649812e-06,
+      "loss": 2.6937,
+      "step": 29507
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6034450063211125e-06,
+      "loss": 3.0143,
+      "step": 29508
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.603397597977244e-06,
+      "loss": 1.6999,
+      "step": 29509
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6033501896333757e-06,
+      "loss": 2.8443,
+      "step": 29510
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.603302781289507e-06,
+      "loss": 2.7684,
+      "step": 29511
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6032553729456385e-06,
+      "loss": 2.1232,
+      "step": 29512
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6032079646017702e-06,
+      "loss": 2.2293,
+      "step": 29513
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6031605562579014e-06,
+      "loss": 2.8084,
+      "step": 29514
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.603113147914033e-06,
+      "loss": 1.4455,
+      "step": 29515
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6030657395701642e-06,
+      "loss": 2.5356,
+      "step": 29516
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.603018331226296e-06,
+      "loss": 1.4534,
+      "step": 29517
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6029709228824274e-06,
+      "loss": 2.6908,
+      "step": 29518
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6029235145385587e-06,
+      "loss": 2.7819,
+      "step": 29519
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6028761061946904e-06,
+      "loss": 2.2357,
+      "step": 29520
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6028286978508217e-06,
+      "loss": 1.7307,
+      "step": 29521
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6027812895069532e-06,
+      "loss": 2.2478,
+      "step": 29522
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6027338811630849e-06,
+      "loss": 4.1164,
+      "step": 29523
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6026864728192161e-06,
+      "loss": 2.0394,
+      "step": 29524
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6026390644753476e-06,
+      "loss": 2.2973,
+      "step": 29525
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6025916561314793e-06,
+      "loss": 1.5736,
+      "step": 29526
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6025442477876106e-06,
+      "loss": 1.8187,
+      "step": 29527
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6024968394437421e-06,
+      "loss": 1.6885,
+      "step": 29528
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6024494310998738e-06,
+      "loss": 2.9244,
+      "step": 29529
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.602402022756005e-06,
+      "loss": 1.4893,
+      "step": 29530
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6023546144121366e-06,
+      "loss": 3.655,
+      "step": 29531
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6023072060682683e-06,
+      "loss": 2.6474,
+      "step": 29532
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6022597977243996e-06,
+      "loss": 2.3402,
+      "step": 29533
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6022123893805309e-06,
+      "loss": 2.6492,
+      "step": 29534
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6021649810366626e-06,
+      "loss": 3.368,
+      "step": 29535
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.602117572692794e-06,
+      "loss": 2.6913,
+      "step": 29536
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6020701643489253e-06,
+      "loss": 2.4704,
+      "step": 29537
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6020227560050568e-06,
+      "loss": 2.0404,
+      "step": 29538
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6019753476611885e-06,
+      "loss": 2.2052,
+      "step": 29539
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6019279393173198e-06,
+      "loss": 2.4167,
+      "step": 29540
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6018805309734513e-06,
+      "loss": 1.7888,
+      "step": 29541
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.601833122629583e-06,
+      "loss": 1.3382,
+      "step": 29542
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.6017857142857143e-06,
+      "loss": 3.4146,
+      "step": 29543
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6017383059418456e-06,
+      "loss": 3.1735,
+      "step": 29544
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6016908975979775e-06,
+      "loss": 3.0524,
+      "step": 29545
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6016434892541088e-06,
+      "loss": 1.6229,
+      "step": 29546
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.60159608091024e-06,
+      "loss": 2.7799,
+      "step": 29547
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6015486725663717e-06,
+      "loss": 2.6218,
+      "step": 29548
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6015012642225032e-06,
+      "loss": 3.1739,
+      "step": 29549
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6014538558786345e-06,
+      "loss": 1.6279,
+      "step": 29550
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6014064475347662e-06,
+      "loss": 1.656,
+      "step": 29551
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6013590391908977e-06,
+      "loss": 3.8936,
+      "step": 29552
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.601311630847029e-06,
+      "loss": 3.2118,
+      "step": 29553
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6012642225031607e-06,
+      "loss": 2.7115,
+      "step": 29554
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6012168141592922e-06,
+      "loss": 3.0625,
+      "step": 29555
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6011694058154235e-06,
+      "loss": 1.2252,
+      "step": 29556
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6011219974715552e-06,
+      "loss": 2.7046,
+      "step": 29557
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6010745891276865e-06,
+      "loss": 2.2735,
+      "step": 29558
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.601027180783818e-06,
+      "loss": 3.7917,
+      "step": 29559
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6009797724399492e-06,
+      "loss": 2.565,
+      "step": 29560
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.600932364096081e-06,
+      "loss": 2.4111,
+      "step": 29561
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6008849557522124e-06,
+      "loss": 1.776,
+      "step": 29562
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6008375474083437e-06,
+      "loss": 1.8639,
+      "step": 29563
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6007901390644754e-06,
+      "loss": 1.9304,
+      "step": 29564
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.600742730720607e-06,
+      "loss": 1.7932,
+      "step": 29565
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6006953223767382e-06,
+      "loss": 1.7004,
+      "step": 29566
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6006479140328699e-06,
+      "loss": 3.1787,
+      "step": 29567
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6006005056890014e-06,
+      "loss": 2.1182,
+      "step": 29568
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6005530973451327e-06,
+      "loss": 2.1948,
+      "step": 29569
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6005056890012644e-06,
+      "loss": 2.2567,
+      "step": 29570
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6004582806573956e-06,
+      "loss": 2.6627,
+      "step": 29571
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6004108723135271e-06,
+      "loss": 1.7328,
+      "step": 29572
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6003634639696588e-06,
+      "loss": 0.8045,
+      "step": 29573
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6003160556257901e-06,
+      "loss": 3.295,
+      "step": 29574
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6002686472819216e-06,
+      "loss": 2.2531,
+      "step": 29575
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6002212389380533e-06,
+      "loss": 2.6184,
+      "step": 29576
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6001738305941846e-06,
+      "loss": 1.9465,
+      "step": 29577
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.600126422250316e-06,
+      "loss": 2.4881,
+      "step": 29578
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.6000790139064478e-06,
+      "loss": 2.3472,
+      "step": 29579
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.600031605562579e-06,
+      "loss": 1.7432,
+      "step": 29580
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5999841972187106e-06,
+      "loss": 1.3103,
+      "step": 29581
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5999367888748418e-06,
+      "loss": 2.8424,
+      "step": 29582
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5998893805309736e-06,
+      "loss": 3.2718,
+      "step": 29583
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5998419721871048e-06,
+      "loss": 2.7187,
+      "step": 29584
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5997945638432363e-06,
+      "loss": 2.6297,
+      "step": 29585
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.599747155499368e-06,
+      "loss": 2.033,
+      "step": 29586
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5996997471554993e-06,
+      "loss": 2.8468,
+      "step": 29587
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5996523388116308e-06,
+      "loss": 2.7576,
+      "step": 29588
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5996049304677625e-06,
+      "loss": 2.1895,
+      "step": 29589
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5995575221238938e-06,
+      "loss": 2.7862,
+      "step": 29590
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5995101137800253e-06,
+      "loss": 1.7532,
+      "step": 29591
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.599462705436157e-06,
+      "loss": 1.8367,
+      "step": 29592
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5994152970922883e-06,
+      "loss": 3.3146,
+      "step": 29593
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5993678887484198e-06,
+      "loss": 3.3019,
+      "step": 29594
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5993204804045515e-06,
+      "loss": 2.8409,
+      "step": 29595
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5992730720606827e-06,
+      "loss": 1.3839,
+      "step": 29596
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.599225663716814e-06,
+      "loss": 2.9978,
+      "step": 29597
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5991782553729457e-06,
+      "loss": 1.5607,
+      "step": 29598
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5991308470290772e-06,
+      "loss": 3.6131,
+      "step": 29599
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5990834386852085e-06,
+      "loss": 2.3691,
+      "step": 29600
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5990360303413402e-06,
+      "loss": 2.7374,
+      "step": 29601
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5989886219974717e-06,
+      "loss": 1.7374,
+      "step": 29602
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.598941213653603e-06,
+      "loss": 2.4377,
+      "step": 29603
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5988938053097345e-06,
+      "loss": 1.5398,
+      "step": 29604
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5988463969658662e-06,
+      "loss": 2.4946,
+      "step": 29605
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.5987989886219974e-06,
+      "loss": 1.7069,
+      "step": 29606
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.598751580278129e-06,
+      "loss": 2.9417,
+      "step": 29607
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5987041719342606e-06,
+      "loss": 1.9986,
+      "step": 29608
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.598656763590392e-06,
+      "loss": 1.4634,
+      "step": 29609
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5986093552465232e-06,
+      "loss": 2.906,
+      "step": 29610
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.598561946902655e-06,
+      "loss": 2.3062,
+      "step": 29611
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5985145385587864e-06,
+      "loss": 1.5749,
+      "step": 29612
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5984671302149177e-06,
+      "loss": 2.7679,
+      "step": 29613
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5984197218710494e-06,
+      "loss": 2.8345,
+      "step": 29614
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5983723135271809e-06,
+      "loss": 3.3052,
+      "step": 29615
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5983249051833122e-06,
+      "loss": 1.5402,
+      "step": 29616
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5982774968394439e-06,
+      "loss": 3.0985,
+      "step": 29617
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5982300884955754e-06,
+      "loss": 2.281,
+      "step": 29618
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5981826801517066e-06,
+      "loss": 2.3573,
+      "step": 29619
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5981352718078383e-06,
+      "loss": 2.0685,
+      "step": 29620
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5980878634639698e-06,
+      "loss": 1.4778,
+      "step": 29621
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5980404551201011e-06,
+      "loss": 1.7838,
+      "step": 29622
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5979930467762328e-06,
+      "loss": 2.2392,
+      "step": 29623
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.597945638432364e-06,
+      "loss": 1.9019,
+      "step": 29624
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5978982300884956e-06,
+      "loss": 2.4103,
+      "step": 29625
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5978508217446269e-06,
+      "loss": 1.3841,
+      "step": 29626
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5978034134007586e-06,
+      "loss": 3.853,
+      "step": 29627
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.59775600505689e-06,
+      "loss": 1.8669,
+      "step": 29628
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5977085967130213e-06,
+      "loss": 2.2987,
+      "step": 29629
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.597661188369153e-06,
+      "loss": 2.6505,
+      "step": 29630
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5976137800252845e-06,
+      "loss": 2.5292,
+      "step": 29631
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5975663716814158e-06,
+      "loss": 3.8836,
+      "step": 29632
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5975189633375475e-06,
+      "loss": 2.2078,
+      "step": 29633
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.597471554993679e-06,
+      "loss": 2.1744,
+      "step": 29634
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5974241466498103e-06,
+      "loss": 2.936,
+      "step": 29635
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.597376738305942e-06,
+      "loss": 2.6331,
+      "step": 29636
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5973293299620733e-06,
+      "loss": 4.2621,
+      "step": 29637
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5972819216182048e-06,
+      "loss": 2.5723,
+      "step": 29638
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5972345132743365e-06,
+      "loss": 2.6206,
+      "step": 29639
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5971871049304678e-06,
+      "loss": 1.5089,
+      "step": 29640
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5971396965865992e-06,
+      "loss": 0.7356,
+      "step": 29641
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.597092288242731e-06,
+      "loss": 1.6296,
+      "step": 29642
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5970448798988622e-06,
+      "loss": 1.9532,
+      "step": 29643
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5969974715549937e-06,
+      "loss": 2.8642,
+      "step": 29644
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5969500632111254e-06,
+      "loss": 2.6343,
+      "step": 29645
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5969026548672567e-06,
+      "loss": 2.1155,
+      "step": 29646
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5968552465233882e-06,
+      "loss": 1.8179,
+      "step": 29647
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5968078381795195e-06,
+      "loss": 2.1103,
+      "step": 29648
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5967604298356512e-06,
+      "loss": 2.4573,
+      "step": 29649
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5967130214917825e-06,
+      "loss": 1.7265,
+      "step": 29650
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.596665613147914e-06,
+      "loss": 2.6396,
+      "step": 29651
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5966182048040457e-06,
+      "loss": 1.9668,
+      "step": 29652
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.596570796460177e-06,
+      "loss": 1.7727,
+      "step": 29653
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5965233881163084e-06,
+      "loss": 2.3775,
+      "step": 29654
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5964759797724401e-06,
+      "loss": 1.9794,
+      "step": 29655
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5964285714285714e-06,
+      "loss": 2.5159,
+      "step": 29656
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.596381163084703e-06,
+      "loss": 2.509,
+      "step": 29657
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5963337547408346e-06,
+      "loss": 3.0118,
+      "step": 29658
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.596286346396966e-06,
+      "loss": 2.1308,
+      "step": 29659
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5962389380530974e-06,
+      "loss": 1.134,
+      "step": 29660
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.596191529709229e-06,
+      "loss": 2.2359,
+      "step": 29661
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5961441213653604e-06,
+      "loss": 1.942,
+      "step": 29662
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5960967130214917e-06,
+      "loss": 2.6259,
+      "step": 29663
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5960493046776234e-06,
+      "loss": 2.0589,
+      "step": 29664
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5960018963337549e-06,
+      "loss": 1.8641,
+      "step": 29665
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5959544879898861e-06,
+      "loss": 2.6426,
+      "step": 29666
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5959070796460178e-06,
+      "loss": 0.9935,
+      "step": 29667
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5958596713021493e-06,
+      "loss": 1.5718,
+      "step": 29668
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.5958122629582806e-06,
+      "loss": 2.0678,
+      "step": 29669
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.595764854614412e-06,
+      "loss": 1.469,
+      "step": 29670
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5957174462705438e-06,
+      "loss": 2.7416,
+      "step": 29671
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.595670037926675e-06,
+      "loss": 1.037,
+      "step": 29672
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5956226295828064e-06,
+      "loss": 2.9489,
+      "step": 29673
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5955752212389383e-06,
+      "loss": 1.5576,
+      "step": 29674
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5955278128950696e-06,
+      "loss": 1.8477,
+      "step": 29675
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5954804045512008e-06,
+      "loss": 3.0146,
+      "step": 29676
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5954329962073325e-06,
+      "loss": 2.6967,
+      "step": 29677
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.595385587863464e-06,
+      "loss": 2.6241,
+      "step": 29678
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5953381795195953e-06,
+      "loss": 2.6076,
+      "step": 29679
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.595290771175727e-06,
+      "loss": 2.0867,
+      "step": 29680
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5952433628318585e-06,
+      "loss": 2.3929,
+      "step": 29681
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5951959544879898e-06,
+      "loss": 2.5043,
+      "step": 29682
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5951485461441215e-06,
+      "loss": 1.6495,
+      "step": 29683
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.595101137800253e-06,
+      "loss": 1.7509,
+      "step": 29684
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5950537294563843e-06,
+      "loss": 2.1286,
+      "step": 29685
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.595006321112516e-06,
+      "loss": 1.0867,
+      "step": 29686
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5949589127686473e-06,
+      "loss": 2.3628,
+      "step": 29687
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5949115044247787e-06,
+      "loss": 1.9705,
+      "step": 29688
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5948640960809105e-06,
+      "loss": 1.5251,
+      "step": 29689
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5948166877370417e-06,
+      "loss": 2.7045,
+      "step": 29690
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5947692793931732e-06,
+      "loss": 1.8404,
+      "step": 29691
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5947218710493045e-06,
+      "loss": 1.6417,
+      "step": 29692
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5946744627054362e-06,
+      "loss": 1.6659,
+      "step": 29693
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5946270543615677e-06,
+      "loss": 3.0067,
+      "step": 29694
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.594579646017699e-06,
+      "loss": 1.9494,
+      "step": 29695
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5945322376738307e-06,
+      "loss": 4.5241,
+      "step": 29696
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5944848293299622e-06,
+      "loss": 2.8273,
+      "step": 29697
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5944374209860935e-06,
+      "loss": 2.4833,
+      "step": 29698
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5943900126422252e-06,
+      "loss": 2.4688,
+      "step": 29699
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5943426042983564e-06,
+      "loss": 2.1051,
+      "step": 29700
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.594295195954488e-06,
+      "loss": 2.7609,
+      "step": 29701
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5942477876106196e-06,
+      "loss": 1.0192,
+      "step": 29702
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.594200379266751e-06,
+      "loss": 2.0042,
+      "step": 29703
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5941529709228824e-06,
+      "loss": 3.7825,
+      "step": 29704
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5941055625790141e-06,
+      "loss": 2.1606,
+      "step": 29705
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5940581542351454e-06,
+      "loss": 2.9485,
+      "step": 29706
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5940107458912769e-06,
+      "loss": 2.9636,
+      "step": 29707
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5939633375474086e-06,
+      "loss": 2.6645,
+      "step": 29708
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5939159292035399e-06,
+      "loss": 1.1723,
+      "step": 29709
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5938685208596714e-06,
+      "loss": 2.4923,
+      "step": 29710
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.593821112515803e-06,
+      "loss": 2.5207,
+      "step": 29711
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5937737041719343e-06,
+      "loss": 2.1993,
+      "step": 29712
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5937262958280656e-06,
+      "loss": 3.1023,
+      "step": 29713
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5936788874841971e-06,
+      "loss": 2.7632,
+      "step": 29714
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5936314791403288e-06,
+      "loss": 2.4583,
+      "step": 29715
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.59358407079646e-06,
+      "loss": 3.6163,
+      "step": 29716
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5935366624525916e-06,
+      "loss": 2.0613,
+      "step": 29717
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5934892541087233e-06,
+      "loss": 2.4764,
+      "step": 29718
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5934418457648546e-06,
+      "loss": 1.6817,
+      "step": 29719
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.593394437420986e-06,
+      "loss": 2.77,
+      "step": 29720
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5933470290771178e-06,
+      "loss": 2.5284,
+      "step": 29721
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.593299620733249e-06,
+      "loss": 1.923,
+      "step": 29722
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5932522123893805e-06,
+      "loss": 2.1031,
+      "step": 29723
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5932048040455123e-06,
+      "loss": 3.9069,
+      "step": 29724
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5931573957016435e-06,
+      "loss": 2.5839,
+      "step": 29725
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5931099873577748e-06,
+      "loss": 2.9231,
+      "step": 29726
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5930625790139065e-06,
+      "loss": 2.4751,
+      "step": 29727
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.593015170670038e-06,
+      "loss": 1.7566,
+      "step": 29728
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5929677623261693e-06,
+      "loss": 1.9946,
+      "step": 29729
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.592920353982301e-06,
+      "loss": 3.0217,
+      "step": 29730
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5928729456384325e-06,
+      "loss": 2.8219,
+      "step": 29731
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5928255372945638e-06,
+      "loss": 2.1663,
+      "step": 29732
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.5927781289506955e-06,
+      "loss": 1.4011,
+      "step": 29733
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.592730720606827e-06,
+      "loss": 2.8142,
+      "step": 29734
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5926833122629582e-06,
+      "loss": 2.229,
+      "step": 29735
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5926359039190897e-06,
+      "loss": 2.7804,
+      "step": 29736
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5925884955752214e-06,
+      "loss": 1.7598,
+      "step": 29737
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5925410872313527e-06,
+      "loss": 3.2151,
+      "step": 29738
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.592493678887484e-06,
+      "loss": 1.448,
+      "step": 29739
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5924462705436157e-06,
+      "loss": 2.5694,
+      "step": 29740
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5923988621997472e-06,
+      "loss": 1.4928,
+      "step": 29741
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5923514538558785e-06,
+      "loss": 2.0199,
+      "step": 29742
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5923040455120102e-06,
+      "loss": 2.6442,
+      "step": 29743
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5922566371681417e-06,
+      "loss": 1.2622,
+      "step": 29744
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.592209228824273e-06,
+      "loss": 1.7282,
+      "step": 29745
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5921618204804047e-06,
+      "loss": 2.9757,
+      "step": 29746
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5921144121365361e-06,
+      "loss": 2.4615,
+      "step": 29747
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5920670037926674e-06,
+      "loss": 2.0376,
+      "step": 29748
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5920195954487991e-06,
+      "loss": 2.6989,
+      "step": 29749
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5919721871049306e-06,
+      "loss": 3.6954,
+      "step": 29750
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.591924778761062e-06,
+      "loss": 2.2874,
+      "step": 29751
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5918773704171936e-06,
+      "loss": 3.2286,
+      "step": 29752
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5918299620733249e-06,
+      "loss": 2.4865,
+      "step": 29753
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5917825537294564e-06,
+      "loss": 2.8143,
+      "step": 29754
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.591735145385588e-06,
+      "loss": 1.2999,
+      "step": 29755
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5916877370417194e-06,
+      "loss": 3.3952,
+      "step": 29756
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5916403286978509e-06,
+      "loss": 1.9017,
+      "step": 29757
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5915929203539821e-06,
+      "loss": 0.8024,
+      "step": 29758
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5915455120101138e-06,
+      "loss": 2.3352,
+      "step": 29759
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5914981036662453e-06,
+      "loss": 1.4458,
+      "step": 29760
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5914506953223766e-06,
+      "loss": 2.1269,
+      "step": 29761
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5914032869785083e-06,
+      "loss": 2.843,
+      "step": 29762
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5913558786346398e-06,
+      "loss": 3.746,
+      "step": 29763
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.591308470290771e-06,
+      "loss": 1.328,
+      "step": 29764
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5912610619469028e-06,
+      "loss": 2.1623,
+      "step": 29765
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.591213653603034e-06,
+      "loss": 2.213,
+      "step": 29766
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5911662452591656e-06,
+      "loss": 1.1866,
+      "step": 29767
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5911188369152973e-06,
+      "loss": 1.6057,
+      "step": 29768
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5910714285714286e-06,
+      "loss": 2.7288,
+      "step": 29769
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.59102402022756e-06,
+      "loss": 3.1096,
+      "step": 29770
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5909766118836918e-06,
+      "loss": 2.8703,
+      "step": 29771
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.590929203539823e-06,
+      "loss": 2.4145,
+      "step": 29772
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5908817951959545e-06,
+      "loss": 2.8905,
+      "step": 29773
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5908343868520862e-06,
+      "loss": 1.5297,
+      "step": 29774
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5907869785082175e-06,
+      "loss": 2.6775,
+      "step": 29775
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.590739570164349e-06,
+      "loss": 3.0529,
+      "step": 29776
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5906921618204807e-06,
+      "loss": 3.1111,
+      "step": 29777
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.590644753476612e-06,
+      "loss": 3.3138,
+      "step": 29778
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5905973451327433e-06,
+      "loss": 3.4558,
+      "step": 29779
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5905499367888748e-06,
+      "loss": 2.0409,
+      "step": 29780
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5905025284450065e-06,
+      "loss": 2.7705,
+      "step": 29781
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5904551201011377e-06,
+      "loss": 1.5704,
+      "step": 29782
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5904077117572692e-06,
+      "loss": 3.0688,
+      "step": 29783
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.590360303413401e-06,
+      "loss": 1.9506,
+      "step": 29784
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5903128950695322e-06,
+      "loss": 2.4904,
+      "step": 29785
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5902654867256637e-06,
+      "loss": 1.3404,
+      "step": 29786
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5902180783817954e-06,
+      "loss": 3.1692,
+      "step": 29787
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5901706700379267e-06,
+      "loss": 3.0365,
+      "step": 29788
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5901232616940582e-06,
+      "loss": 4.385,
+      "step": 29789
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5900758533501899e-06,
+      "loss": 3.0199,
+      "step": 29790
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5900284450063212e-06,
+      "loss": 1.3028,
+      "step": 29791
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5899810366624525e-06,
+      "loss": 1.1994,
+      "step": 29792
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5899336283185842e-06,
+      "loss": 2.6829,
+      "step": 29793
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5898862199747156e-06,
+      "loss": 1.6193,
+      "step": 29794
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.589838811630847e-06,
+      "loss": 2.0692,
+      "step": 29795
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.5897914032869786e-06,
+      "loss": 1.6451,
+      "step": 29796
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5897439949431101e-06,
+      "loss": 2.8865,
+      "step": 29797
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5896965865992414e-06,
+      "loss": 1.4167,
+      "step": 29798
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5896491782553731e-06,
+      "loss": 2.0388,
+      "step": 29799
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5896017699115046e-06,
+      "loss": 2.1209,
+      "step": 29800
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5895543615676359e-06,
+      "loss": 2.8164,
+      "step": 29801
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5895069532237674e-06,
+      "loss": 2.811,
+      "step": 29802
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.589459544879899e-06,
+      "loss": 1.5887,
+      "step": 29803
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5894121365360304e-06,
+      "loss": 3.7653,
+      "step": 29804
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5893647281921616e-06,
+      "loss": 3.5922,
+      "step": 29805
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5893173198482933e-06,
+      "loss": 2.4206,
+      "step": 29806
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5892699115044248e-06,
+      "loss": 3.3199,
+      "step": 29807
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5892225031605561e-06,
+      "loss": 2.3844,
+      "step": 29808
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5891750948166878e-06,
+      "loss": 2.6046,
+      "step": 29809
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5891276864728193e-06,
+      "loss": 3.0056,
+      "step": 29810
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5890802781289506e-06,
+      "loss": 2.6536,
+      "step": 29811
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5890328697850823e-06,
+      "loss": 2.5329,
+      "step": 29812
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5889854614412138e-06,
+      "loss": 2.5533,
+      "step": 29813
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.588938053097345e-06,
+      "loss": 2.066,
+      "step": 29814
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5888906447534768e-06,
+      "loss": 2.3756,
+      "step": 29815
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5888432364096083e-06,
+      "loss": 2.0669,
+      "step": 29816
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5887958280657395e-06,
+      "loss": 2.1482,
+      "step": 29817
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5887484197218712e-06,
+      "loss": 1.3726,
+      "step": 29818
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5887010113780025e-06,
+      "loss": 1.9633,
+      "step": 29819
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.588653603034134e-06,
+      "loss": 1.6997,
+      "step": 29820
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5886061946902657e-06,
+      "loss": 2.3297,
+      "step": 29821
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.588558786346397e-06,
+      "loss": 3.1197,
+      "step": 29822
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5885113780025285e-06,
+      "loss": 1.974,
+      "step": 29823
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5884639696586598e-06,
+      "loss": 2.1287,
+      "step": 29824
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5884165613147915e-06,
+      "loss": 3.3906,
+      "step": 29825
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.588369152970923e-06,
+      "loss": 2.4525,
+      "step": 29826
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5883217446270543e-06,
+      "loss": 2.8211,
+      "step": 29827
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.588274336283186e-06,
+      "loss": 2.7582,
+      "step": 29828
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5882269279393172e-06,
+      "loss": 3.3824,
+      "step": 29829
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5881795195954487e-06,
+      "loss": 3.0293,
+      "step": 29830
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5881321112515804e-06,
+      "loss": 1.3971,
+      "step": 29831
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5880847029077117e-06,
+      "loss": 2.735,
+      "step": 29832
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5880372945638432e-06,
+      "loss": 0.9865,
+      "step": 29833
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.587989886219975e-06,
+      "loss": 2.7224,
+      "step": 29834
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5879424778761062e-06,
+      "loss": 3.0743,
+      "step": 29835
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5878950695322377e-06,
+      "loss": 2.1794,
+      "step": 29836
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5878476611883694e-06,
+      "loss": 3.2612,
+      "step": 29837
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5878002528445007e-06,
+      "loss": 2.5619,
+      "step": 29838
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5877528445006322e-06,
+      "loss": 1.8458,
+      "step": 29839
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5877054361567639e-06,
+      "loss": 2.8459,
+      "step": 29840
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5876580278128951e-06,
+      "loss": 2.7535,
+      "step": 29841
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5876106194690264e-06,
+      "loss": 2.4475,
+      "step": 29842
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5875632111251581e-06,
+      "loss": 2.2628,
+      "step": 29843
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5875158027812896e-06,
+      "loss": 2.2152,
+      "step": 29844
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.587468394437421e-06,
+      "loss": 2.7025,
+      "step": 29845
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5874209860935524e-06,
+      "loss": 0.7889,
+      "step": 29846
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.587373577749684e-06,
+      "loss": 2.404,
+      "step": 29847
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5873261694058154e-06,
+      "loss": 1.4522,
+      "step": 29848
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5872787610619469e-06,
+      "loss": 2.5887,
+      "step": 29849
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5872313527180786e-06,
+      "loss": 3.2269,
+      "step": 29850
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5871839443742099e-06,
+      "loss": 1.6213,
+      "step": 29851
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5871365360303413e-06,
+      "loss": 3.3079,
+      "step": 29852
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.587089127686473e-06,
+      "loss": 1.9701,
+      "step": 29853
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5870417193426043e-06,
+      "loss": 1.7148,
+      "step": 29854
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5869943109987356e-06,
+      "loss": 2.7486,
+      "step": 29855
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5869469026548673e-06,
+      "loss": 2.3092,
+      "step": 29856
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5868994943109988e-06,
+      "loss": 2.7464,
+      "step": 29857
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.58685208596713e-06,
+      "loss": 3.4111,
+      "step": 29858
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5868046776232618e-06,
+      "loss": 0.6114,
+      "step": 29859
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.5867572692793933e-06,
+      "loss": 3.8753,
+      "step": 29860
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5867098609355246e-06,
+      "loss": 2.3631,
+      "step": 29861
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5866624525916563e-06,
+      "loss": 2.4624,
+      "step": 29862
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5866150442477878e-06,
+      "loss": 2.0503,
+      "step": 29863
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.586567635903919e-06,
+      "loss": 2.5692,
+      "step": 29864
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5865202275600507e-06,
+      "loss": 1.5227,
+      "step": 29865
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5864728192161822e-06,
+      "loss": 1.9214,
+      "step": 29866
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5864254108723135e-06,
+      "loss": 3.0477,
+      "step": 29867
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5863780025284448e-06,
+      "loss": 3.086,
+      "step": 29868
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5863305941845765e-06,
+      "loss": 0.8767,
+      "step": 29869
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.586283185840708e-06,
+      "loss": 1.2098,
+      "step": 29870
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5862357774968393e-06,
+      "loss": 1.8475,
+      "step": 29871
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.586188369152971e-06,
+      "loss": 2.4089,
+      "step": 29872
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5861409608091025e-06,
+      "loss": 3.402,
+      "step": 29873
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5860935524652338e-06,
+      "loss": 2.2461,
+      "step": 29874
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5860461441213655e-06,
+      "loss": 2.7939,
+      "step": 29875
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.585998735777497e-06,
+      "loss": 2.4023,
+      "step": 29876
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5859513274336282e-06,
+      "loss": 2.9088,
+      "step": 29877
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.58590391908976e-06,
+      "loss": 1.9856,
+      "step": 29878
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5858565107458914e-06,
+      "loss": 3.0455,
+      "step": 29879
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5858091024020227e-06,
+      "loss": 1.5279,
+      "step": 29880
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5857616940581544e-06,
+      "loss": 2.5099,
+      "step": 29881
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5857142857142857e-06,
+      "loss": 2.6788,
+      "step": 29882
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5856668773704172e-06,
+      "loss": 3.2498,
+      "step": 29883
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5856194690265489e-06,
+      "loss": 1.9245,
+      "step": 29884
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5855720606826802e-06,
+      "loss": 1.122,
+      "step": 29885
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5855246523388117e-06,
+      "loss": 2.5812,
+      "step": 29886
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5854772439949434e-06,
+      "loss": 2.0017,
+      "step": 29887
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5854298356510746e-06,
+      "loss": 3.0242,
+      "step": 29888
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5853824273072061e-06,
+      "loss": 2.4762,
+      "step": 29889
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5853350189633374e-06,
+      "loss": 2.5121,
+      "step": 29890
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5852876106194691e-06,
+      "loss": 1.6679,
+      "step": 29891
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5852402022756006e-06,
+      "loss": 2.5782,
+      "step": 29892
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5851927939317319e-06,
+      "loss": 2.7219,
+      "step": 29893
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5851453855878636e-06,
+      "loss": 2.88,
+      "step": 29894
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5850979772439949e-06,
+      "loss": 1.736,
+      "step": 29895
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5850505689001264e-06,
+      "loss": 2.7467,
+      "step": 29896
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.585003160556258e-06,
+      "loss": 2.6789,
+      "step": 29897
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5849557522123894e-06,
+      "loss": 1.3736,
+      "step": 29898
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5849083438685208e-06,
+      "loss": 1.1025,
+      "step": 29899
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5848609355246525e-06,
+      "loss": 2.3858,
+      "step": 29900
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5848135271807838e-06,
+      "loss": 3.2605,
+      "step": 29901
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5847661188369153e-06,
+      "loss": 2.0876,
+      "step": 29902
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.584718710493047e-06,
+      "loss": 2.3107,
+      "step": 29903
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5846713021491783e-06,
+      "loss": 2.0872,
+      "step": 29904
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5846238938053098e-06,
+      "loss": 1.896,
+      "step": 29905
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5845764854614415e-06,
+      "loss": 0.838,
+      "step": 29906
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5845290771175728e-06,
+      "loss": 1.9356,
+      "step": 29907
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.584481668773704e-06,
+      "loss": 2.0112,
+      "step": 29908
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5844342604298358e-06,
+      "loss": 1.5574,
+      "step": 29909
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5843868520859673e-06,
+      "loss": 2.3847,
+      "step": 29910
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5843394437420985e-06,
+      "loss": 2.9009,
+      "step": 29911
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.58429203539823e-06,
+      "loss": 2.2896,
+      "step": 29912
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5842446270543617e-06,
+      "loss": 2.7137,
+      "step": 29913
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.584197218710493e-06,
+      "loss": 1.5071,
+      "step": 29914
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5841498103666245e-06,
+      "loss": 2.5073,
+      "step": 29915
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5841024020227562e-06,
+      "loss": 2.1102,
+      "step": 29916
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5840549936788875e-06,
+      "loss": 3.1801,
+      "step": 29917
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.584007585335019e-06,
+      "loss": 2.7841,
+      "step": 29918
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5839601769911507e-06,
+      "loss": 1.7383,
+      "step": 29919
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.583912768647282e-06,
+      "loss": 3.1059,
+      "step": 29920
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5838653603034132e-06,
+      "loss": 2.2562,
+      "step": 29921
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.583817951959545e-06,
+      "loss": 3.2179,
+      "step": 29922
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.5837705436156764e-06,
+      "loss": 1.8224,
+      "step": 29923
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5837231352718077e-06,
+      "loss": 3.2035,
+      "step": 29924
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5836757269279394e-06,
+      "loss": 1.7053,
+      "step": 29925
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.583628318584071e-06,
+      "loss": 2.3531,
+      "step": 29926
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5835809102402022e-06,
+      "loss": 2.2932,
+      "step": 29927
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.583533501896334e-06,
+      "loss": 2.0818,
+      "step": 29928
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5834860935524654e-06,
+      "loss": 2.579,
+      "step": 29929
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5834386852085967e-06,
+      "loss": 1.6532,
+      "step": 29930
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5833912768647284e-06,
+      "loss": 2.908,
+      "step": 29931
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5833438685208599e-06,
+      "loss": 2.672,
+      "step": 29932
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5832964601769912e-06,
+      "loss": 1.7123,
+      "step": 29933
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5832490518331224e-06,
+      "loss": 3.8942,
+      "step": 29934
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5832016434892541e-06,
+      "loss": 3.1439,
+      "step": 29935
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5831542351453856e-06,
+      "loss": 2.7514,
+      "step": 29936
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.583106826801517e-06,
+      "loss": 2.4147,
+      "step": 29937
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5830594184576486e-06,
+      "loss": 2.3042,
+      "step": 29938
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5830120101137801e-06,
+      "loss": 1.3115,
+      "step": 29939
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5829646017699114e-06,
+      "loss": 1.6247,
+      "step": 29940
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.582917193426043e-06,
+      "loss": 3.2619,
+      "step": 29941
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5828697850821746e-06,
+      "loss": 1.5443,
+      "step": 29942
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5828223767383059e-06,
+      "loss": 2.4399,
+      "step": 29943
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5827749683944376e-06,
+      "loss": 2.3483,
+      "step": 29944
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.582727560050569e-06,
+      "loss": 1.7614,
+      "step": 29945
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5826801517067003e-06,
+      "loss": 1.0548,
+      "step": 29946
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.582632743362832e-06,
+      "loss": 1.9709,
+      "step": 29947
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5825853350189633e-06,
+      "loss": 2.3365,
+      "step": 29948
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5825379266750948e-06,
+      "loss": 2.4871,
+      "step": 29949
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5824905183312265e-06,
+      "loss": 2.7387,
+      "step": 29950
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5824431099873578e-06,
+      "loss": 2.5804,
+      "step": 29951
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5823957016434893e-06,
+      "loss": 2.2836,
+      "step": 29952
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5823482932996206e-06,
+      "loss": 1.5671,
+      "step": 29953
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5823008849557523e-06,
+      "loss": 2.188,
+      "step": 29954
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5822534766118838e-06,
+      "loss": 1.6791,
+      "step": 29955
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.582206068268015e-06,
+      "loss": 3.0897,
+      "step": 29956
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5821586599241468e-06,
+      "loss": 2.6682,
+      "step": 29957
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.582111251580278e-06,
+      "loss": 2.3856,
+      "step": 29958
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5820638432364095e-06,
+      "loss": 3.6065,
+      "step": 29959
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5820164348925412e-06,
+      "loss": 3.2033,
+      "step": 29960
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5819690265486725e-06,
+      "loss": 0.764,
+      "step": 29961
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.581921618204804e-06,
+      "loss": 2.9067,
+      "step": 29962
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5818742098609357e-06,
+      "loss": 2.2302,
+      "step": 29963
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.581826801517067e-06,
+      "loss": 1.9817,
+      "step": 29964
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5817793931731985e-06,
+      "loss": 2.8176,
+      "step": 29965
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5817319848293302e-06,
+      "loss": 2.2525,
+      "step": 29966
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5816845764854615e-06,
+      "loss": 1.2194,
+      "step": 29967
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.581637168141593e-06,
+      "loss": 2.5673,
+      "step": 29968
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5815897597977247e-06,
+      "loss": 0.4297,
+      "step": 29969
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.581542351453856e-06,
+      "loss": 2.4414,
+      "step": 29970
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5814949431099872e-06,
+      "loss": 2.933,
+      "step": 29971
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.581447534766119e-06,
+      "loss": 1.8977,
+      "step": 29972
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5814001264222504e-06,
+      "loss": 1.0793,
+      "step": 29973
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5813527180783817e-06,
+      "loss": 3.0524,
+      "step": 29974
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5813053097345132e-06,
+      "loss": 1.5268,
+      "step": 29975
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5812579013906449e-06,
+      "loss": 1.9125,
+      "step": 29976
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5812104930467762e-06,
+      "loss": 2.3792,
+      "step": 29977
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5811630847029077e-06,
+      "loss": 2.0851,
+      "step": 29978
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5811156763590394e-06,
+      "loss": 2.9558,
+      "step": 29979
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5810682680151707e-06,
+      "loss": 2.8093,
+      "step": 29980
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5810208596713021e-06,
+      "loss": 2.9398,
+      "step": 29981
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5809734513274338e-06,
+      "loss": 2.343,
+      "step": 29982
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5809260429835651e-06,
+      "loss": 1.8138,
+      "step": 29983
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5808786346396964e-06,
+      "loss": 1.6713,
+      "step": 29984
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5808312262958281e-06,
+      "loss": 2.1237,
+      "step": 29985
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.5807838179519596e-06,
+      "loss": 2.1352,
+      "step": 29986
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5807364096080909e-06,
+      "loss": 3.2152,
+      "step": 29987
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5806890012642226e-06,
+      "loss": 2.3201,
+      "step": 29988
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.580641592920354e-06,
+      "loss": 2.0826,
+      "step": 29989
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5805941845764854e-06,
+      "loss": 3.3389,
+      "step": 29990
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.580546776232617e-06,
+      "loss": 1.5115,
+      "step": 29991
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5804993678887486e-06,
+      "loss": 3.0588,
+      "step": 29992
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5804519595448798e-06,
+      "loss": 3.0381,
+      "step": 29993
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5804045512010115e-06,
+      "loss": 2.3067,
+      "step": 29994
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.580357142857143e-06,
+      "loss": 1.2569,
+      "step": 29995
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5803097345132743e-06,
+      "loss": 1.8808,
+      "step": 29996
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5802623261694056e-06,
+      "loss": 3.1948,
+      "step": 29997
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5802149178255373e-06,
+      "loss": 3.3971,
+      "step": 29998
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5801675094816688e-06,
+      "loss": 1.6753,
+      "step": 29999
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5801201011378e-06,
+      "loss": 3.0124,
+      "step": 30000
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5800726927939318e-06,
+      "loss": 2.6435,
+      "step": 30001
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5800252844500633e-06,
+      "loss": 3.1126,
+      "step": 30002
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5799778761061945e-06,
+      "loss": 2.2044,
+      "step": 30003
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5799304677623263e-06,
+      "loss": 1.6795,
+      "step": 30004
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5798830594184577e-06,
+      "loss": 2.5103,
+      "step": 30005
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.579835651074589e-06,
+      "loss": 3.2487,
+      "step": 30006
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5797882427307207e-06,
+      "loss": 2.7722,
+      "step": 30007
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5797408343868522e-06,
+      "loss": 1.8546,
+      "step": 30008
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5796934260429835e-06,
+      "loss": 2.2077,
+      "step": 30009
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5796460176991152e-06,
+      "loss": 2.8382,
+      "step": 30010
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5795986093552465e-06,
+      "loss": 1.309,
+      "step": 30011
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.579551201011378e-06,
+      "loss": 0.8472,
+      "step": 30012
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5795037926675097e-06,
+      "loss": 2.1508,
+      "step": 30013
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.579456384323641e-06,
+      "loss": 2.412,
+      "step": 30014
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5794089759797725e-06,
+      "loss": 1.7284,
+      "step": 30015
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5793615676359042e-06,
+      "loss": 2.0882,
+      "step": 30016
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5793141592920354e-06,
+      "loss": 1.03,
+      "step": 30017
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.579266750948167e-06,
+      "loss": 1.2945,
+      "step": 30018
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5792193426042982e-06,
+      "loss": 1.8961,
+      "step": 30019
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.57917193426043e-06,
+      "loss": 2.6147,
+      "step": 30020
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5791245259165614e-06,
+      "loss": 3.5493,
+      "step": 30021
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5790771175726927e-06,
+      "loss": 3.5329,
+      "step": 30022
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5790297092288244e-06,
+      "loss": 3.0718,
+      "step": 30023
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5789823008849557e-06,
+      "loss": 2.1129,
+      "step": 30024
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5789348925410872e-06,
+      "loss": 1.8574,
+      "step": 30025
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5788874841972189e-06,
+      "loss": 3.3549,
+      "step": 30026
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5788400758533501e-06,
+      "loss": 2.0411,
+      "step": 30027
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5787926675094816e-06,
+      "loss": 1.8273,
+      "step": 30028
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5787452591656133e-06,
+      "loss": 1.5189,
+      "step": 30029
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5786978508217446e-06,
+      "loss": 2.0757,
+      "step": 30030
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5786504424778761e-06,
+      "loss": 1.6565,
+      "step": 30031
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5786030341340078e-06,
+      "loss": 1.7337,
+      "step": 30032
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.578555625790139e-06,
+      "loss": 1.652,
+      "step": 30033
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5785082174462706e-06,
+      "loss": 1.1605,
+      "step": 30034
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5784608091024023e-06,
+      "loss": 1.8656,
+      "step": 30035
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5784134007585336e-06,
+      "loss": 3.3841,
+      "step": 30036
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5783659924146649e-06,
+      "loss": 0.9342,
+      "step": 30037
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5783185840707966e-06,
+      "loss": 1.543,
+      "step": 30038
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.578271175726928e-06,
+      "loss": 2.554,
+      "step": 30039
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5782237673830593e-06,
+      "loss": 1.7607,
+      "step": 30040
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5781763590391908e-06,
+      "loss": 2.2103,
+      "step": 30041
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5781289506953225e-06,
+      "loss": 2.1844,
+      "step": 30042
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5780815423514538e-06,
+      "loss": 1.8599,
+      "step": 30043
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5780341340075853e-06,
+      "loss": 1.7376,
+      "step": 30044
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.577986725663717e-06,
+      "loss": 1.2553,
+      "step": 30045
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5779393173198483e-06,
+      "loss": 2.9241,
+      "step": 30046
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5778919089759798e-06,
+      "loss": 3.1249,
+      "step": 30047
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5778445006321115e-06,
+      "loss": 1.6237,
+      "step": 30048
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.5777970922882428e-06,
+      "loss": 2.2868,
+      "step": 30049
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.577749683944374e-06,
+      "loss": 1.2659,
+      "step": 30050
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5777022756005057e-06,
+      "loss": 1.0913,
+      "step": 30051
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5776548672566372e-06,
+      "loss": 2.0614,
+      "step": 30052
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5776074589127685e-06,
+      "loss": 2.1402,
+      "step": 30053
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5775600505689002e-06,
+      "loss": 2.1414,
+      "step": 30054
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5775126422250317e-06,
+      "loss": 2.69,
+      "step": 30055
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.577465233881163e-06,
+      "loss": 2.3986,
+      "step": 30056
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5774178255372947e-06,
+      "loss": 2.1344,
+      "step": 30057
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5773704171934262e-06,
+      "loss": 1.5369,
+      "step": 30058
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5773230088495575e-06,
+      "loss": 3.0699,
+      "step": 30059
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5772756005056892e-06,
+      "loss": 3.8083,
+      "step": 30060
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5772281921618207e-06,
+      "loss": 1.9914,
+      "step": 30061
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.577180783817952e-06,
+      "loss": 1.4291,
+      "step": 30062
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5771333754740832e-06,
+      "loss": 3.4225,
+      "step": 30063
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.577085967130215e-06,
+      "loss": 3.0907,
+      "step": 30064
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5770385587863464e-06,
+      "loss": 2.6317,
+      "step": 30065
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5769911504424777e-06,
+      "loss": 2.7126,
+      "step": 30066
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5769437420986094e-06,
+      "loss": 0.9273,
+      "step": 30067
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.576896333754741e-06,
+      "loss": 1.3025,
+      "step": 30068
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5768489254108722e-06,
+      "loss": 2.8104,
+      "step": 30069
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5768015170670039e-06,
+      "loss": 3.0647,
+      "step": 30070
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5767541087231354e-06,
+      "loss": 3.7127,
+      "step": 30071
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5767067003792667e-06,
+      "loss": 2.5492,
+      "step": 30072
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5766592920353984e-06,
+      "loss": 1.3612,
+      "step": 30073
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5766118836915299e-06,
+      "loss": 3.3928,
+      "step": 30074
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5765644753476611e-06,
+      "loss": 2.5736,
+      "step": 30075
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5765170670037928e-06,
+      "loss": 1.8559,
+      "step": 30076
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5764696586599241e-06,
+      "loss": 1.7168,
+      "step": 30077
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5764222503160556e-06,
+      "loss": 1.5151,
+      "step": 30078
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5763748419721873e-06,
+      "loss": 2.7696,
+      "step": 30079
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5763274336283186e-06,
+      "loss": 2.3515,
+      "step": 30080
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.57628002528445e-06,
+      "loss": 2.9276,
+      "step": 30081
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5762326169405818e-06,
+      "loss": 2.0577,
+      "step": 30082
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.576185208596713e-06,
+      "loss": 3.2618,
+      "step": 30083
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5761378002528446e-06,
+      "loss": 3.1944,
+      "step": 30084
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5760903919089758e-06,
+      "loss": 3.4467,
+      "step": 30085
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5760429835651076e-06,
+      "loss": 2.9213,
+      "step": 30086
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.575995575221239e-06,
+      "loss": 2.7865,
+      "step": 30087
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5759481668773703e-06,
+      "loss": 3.1107,
+      "step": 30088
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.575900758533502e-06,
+      "loss": 2.7315,
+      "step": 30089
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5758533501896333e-06,
+      "loss": 2.2984,
+      "step": 30090
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5758059418457648e-06,
+      "loss": 3.029,
+      "step": 30091
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5757585335018965e-06,
+      "loss": 2.3102,
+      "step": 30092
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5757111251580278e-06,
+      "loss": 3.2348,
+      "step": 30093
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5756637168141593e-06,
+      "loss": 1.5782,
+      "step": 30094
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.575616308470291e-06,
+      "loss": 3.3969,
+      "step": 30095
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5755689001264223e-06,
+      "loss": 2.0164,
+      "step": 30096
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5755214917825538e-06,
+      "loss": 2.2231,
+      "step": 30097
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5754740834386855e-06,
+      "loss": 3.0406,
+      "step": 30098
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5754266750948167e-06,
+      "loss": 1.9798,
+      "step": 30099
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.575379266750948e-06,
+      "loss": 2.1856,
+      "step": 30100
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.57533185840708e-06,
+      "loss": 0.7634,
+      "step": 30101
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5752844500632112e-06,
+      "loss": 2.4279,
+      "step": 30102
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5752370417193425e-06,
+      "loss": 1.8382,
+      "step": 30103
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5751896333754742e-06,
+      "loss": 1.6561,
+      "step": 30104
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5751422250316057e-06,
+      "loss": 2.135,
+      "step": 30105
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.575094816687737e-06,
+      "loss": 3.2765,
+      "step": 30106
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5750474083438685e-06,
+      "loss": 2.9988,
+      "step": 30107
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5750000000000002e-06,
+      "loss": 2.4307,
+      "step": 30108
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5749525916561314e-06,
+      "loss": 2.0323,
+      "step": 30109
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.574905183312263e-06,
+      "loss": 2.4951,
+      "step": 30110
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5748577749683946e-06,
+      "loss": 2.385,
+      "step": 30111
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.574810366624526e-06,
+      "loss": 2.1529,
+      "step": 30112
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.5747629582806572e-06,
+      "loss": 1.9567,
+      "step": 30113
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.574715549936789e-06,
+      "loss": 2.1316,
+      "step": 30114
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5746681415929204e-06,
+      "loss": 3.0631,
+      "step": 30115
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5746207332490517e-06,
+      "loss": 2.1845,
+      "step": 30116
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5745733249051834e-06,
+      "loss": 1.9727,
+      "step": 30117
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5745259165613149e-06,
+      "loss": 4.7571,
+      "step": 30118
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5744785082174462e-06,
+      "loss": 3.0557,
+      "step": 30119
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5744310998735779e-06,
+      "loss": 3.0421,
+      "step": 30120
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5743836915297094e-06,
+      "loss": 2.7511,
+      "step": 30121
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5743362831858406e-06,
+      "loss": 2.699,
+      "step": 30122
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5742888748419723e-06,
+      "loss": 3.2635,
+      "step": 30123
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5742414664981038e-06,
+      "loss": 2.8048,
+      "step": 30124
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5741940581542351e-06,
+      "loss": 2.6919,
+      "step": 30125
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5741466498103668e-06,
+      "loss": 2.0604,
+      "step": 30126
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.574099241466498e-06,
+      "loss": 3.443,
+      "step": 30127
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5740518331226296e-06,
+      "loss": 2.2779,
+      "step": 30128
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5740044247787609e-06,
+      "loss": 2.7055,
+      "step": 30129
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5739570164348926e-06,
+      "loss": 1.35,
+      "step": 30130
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.573909608091024e-06,
+      "loss": 3.0222,
+      "step": 30131
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5738621997471553e-06,
+      "loss": 1.4592,
+      "step": 30132
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.573814791403287e-06,
+      "loss": 1.4042,
+      "step": 30133
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5737673830594185e-06,
+      "loss": 1.6532,
+      "step": 30134
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5737199747155498e-06,
+      "loss": 2.4706,
+      "step": 30135
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5736725663716815e-06,
+      "loss": 2.0562,
+      "step": 30136
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.573625158027813e-06,
+      "loss": 2.1504,
+      "step": 30137
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5735777496839443e-06,
+      "loss": 1.601,
+      "step": 30138
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.573530341340076e-06,
+      "loss": 1.8769,
+      "step": 30139
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5734829329962073e-06,
+      "loss": 1.4013,
+      "step": 30140
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5734355246523388e-06,
+      "loss": 1.5798,
+      "step": 30141
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5733881163084705e-06,
+      "loss": 3.0451,
+      "step": 30142
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5733407079646018e-06,
+      "loss": 2.8302,
+      "step": 30143
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5732932996207333e-06,
+      "loss": 2.8069,
+      "step": 30144
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.573245891276865e-06,
+      "loss": 1.963,
+      "step": 30145
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5731984829329962e-06,
+      "loss": 1.9899,
+      "step": 30146
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5731510745891277e-06,
+      "loss": 3.32,
+      "step": 30147
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5731036662452594e-06,
+      "loss": 1.3881,
+      "step": 30148
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5730562579013907e-06,
+      "loss": 3.0796,
+      "step": 30149
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5730088495575222e-06,
+      "loss": 2.679,
+      "step": 30150
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5729614412136535e-06,
+      "loss": 3.0657,
+      "step": 30151
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5729140328697852e-06,
+      "loss": 2.6836,
+      "step": 30152
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5728666245259165e-06,
+      "loss": 2.2578,
+      "step": 30153
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.572819216182048e-06,
+      "loss": 3.0708,
+      "step": 30154
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5727718078381797e-06,
+      "loss": 2.3987,
+      "step": 30155
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.572724399494311e-06,
+      "loss": 2.0442,
+      "step": 30156
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5726769911504424e-06,
+      "loss": 2.4918,
+      "step": 30157
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5726295828065741e-06,
+      "loss": 2.4944,
+      "step": 30158
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5725821744627054e-06,
+      "loss": 2.2884,
+      "step": 30159
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.572534766118837e-06,
+      "loss": 2.5877,
+      "step": 30160
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5724873577749686e-06,
+      "loss": 2.4716,
+      "step": 30161
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5724399494311e-06,
+      "loss": 2.062,
+      "step": 30162
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5723925410872314e-06,
+      "loss": 3.2816,
+      "step": 30163
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.572345132743363e-06,
+      "loss": 2.9655,
+      "step": 30164
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5722977243994944e-06,
+      "loss": 2.7754,
+      "step": 30165
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5722503160556257e-06,
+      "loss": 2.8396,
+      "step": 30166
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5722029077117574e-06,
+      "loss": 2.619,
+      "step": 30167
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5721554993678889e-06,
+      "loss": 2.1524,
+      "step": 30168
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5721080910240201e-06,
+      "loss": 1.0074,
+      "step": 30169
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5720606826801518e-06,
+      "loss": 2.3751,
+      "step": 30170
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5720132743362833e-06,
+      "loss": 1.4631,
+      "step": 30171
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5719658659924146e-06,
+      "loss": 2.6987,
+      "step": 30172
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.571918457648546e-06,
+      "loss": 2.279,
+      "step": 30173
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5718710493046778e-06,
+      "loss": 2.5924,
+      "step": 30174
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.571823640960809e-06,
+      "loss": 1.6952,
+      "step": 30175
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.5717762326169406e-06,
+      "loss": 2.5048,
+      "step": 30176
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5717288242730723e-06,
+      "loss": 0.6943,
+      "step": 30177
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5716814159292036e-06,
+      "loss": 3.9494,
+      "step": 30178
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5716340075853348e-06,
+      "loss": 1.6096,
+      "step": 30179
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5715865992414665e-06,
+      "loss": 1.7179,
+      "step": 30180
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.571539190897598e-06,
+      "loss": 2.365,
+      "step": 30181
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5714917825537293e-06,
+      "loss": 0.7694,
+      "step": 30182
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.571444374209861e-06,
+      "loss": 4.1048,
+      "step": 30183
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5713969658659925e-06,
+      "loss": 1.5777,
+      "step": 30184
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5713495575221238e-06,
+      "loss": 1.4872,
+      "step": 30185
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5713021491782555e-06,
+      "loss": 2.3003,
+      "step": 30186
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.571254740834387e-06,
+      "loss": 1.6123,
+      "step": 30187
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5712073324905183e-06,
+      "loss": 2.3157,
+      "step": 30188
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.57115992414665e-06,
+      "loss": 2.6137,
+      "step": 30189
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5711125158027815e-06,
+      "loss": 2.5066,
+      "step": 30190
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5710651074589127e-06,
+      "loss": 2.2816,
+      "step": 30191
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5710176991150445e-06,
+      "loss": 2.8793,
+      "step": 30192
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5709702907711757e-06,
+      "loss": 2.2862,
+      "step": 30193
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5709228824273072e-06,
+      "loss": 3.452,
+      "step": 30194
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5708754740834385e-06,
+      "loss": 2.4747,
+      "step": 30195
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5708280657395702e-06,
+      "loss": 1.2237,
+      "step": 30196
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5707806573957017e-06,
+      "loss": 1.8996,
+      "step": 30197
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.570733249051833e-06,
+      "loss": 2.2507,
+      "step": 30198
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5706858407079647e-06,
+      "loss": 2.3325,
+      "step": 30199
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5706384323640962e-06,
+      "loss": 0.6588,
+      "step": 30200
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5705910240202275e-06,
+      "loss": 2.2727,
+      "step": 30201
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5705436156763592e-06,
+      "loss": 3.4876,
+      "step": 30202
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5704962073324907e-06,
+      "loss": 2.7642,
+      "step": 30203
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.570448798988622e-06,
+      "loss": 3.3979,
+      "step": 30204
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5704013906447536e-06,
+      "loss": 2.3587,
+      "step": 30205
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.570353982300885e-06,
+      "loss": 1.4505,
+      "step": 30206
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5703065739570164e-06,
+      "loss": 2.8753,
+      "step": 30207
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5702591656131481e-06,
+      "loss": 1.6848,
+      "step": 30208
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5702117572692794e-06,
+      "loss": 4.0372,
+      "step": 30209
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5701643489254109e-06,
+      "loss": 2.5641,
+      "step": 30210
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5701169405815426e-06,
+      "loss": 2.6406,
+      "step": 30211
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5700695322376739e-06,
+      "loss": 3.7542,
+      "step": 30212
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5700221238938054e-06,
+      "loss": 2.1325,
+      "step": 30213
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.569974715549937e-06,
+      "loss": 1.7644,
+      "step": 30214
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5699273072060683e-06,
+      "loss": 3.9288,
+      "step": 30215
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5698798988621998e-06,
+      "loss": 2.1712,
+      "step": 30216
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5698324905183311e-06,
+      "loss": 2.3971,
+      "step": 30217
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5697850821744628e-06,
+      "loss": 3.8944,
+      "step": 30218
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.569737673830594e-06,
+      "loss": 2.01,
+      "step": 30219
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5696902654867256e-06,
+      "loss": 2.2056,
+      "step": 30220
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5696428571428573e-06,
+      "loss": 3.0303,
+      "step": 30221
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5695954487989886e-06,
+      "loss": 2.0959,
+      "step": 30222
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.56954804045512e-06,
+      "loss": 3.1449,
+      "step": 30223
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5695006321112518e-06,
+      "loss": 1.9641,
+      "step": 30224
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.569453223767383e-06,
+      "loss": 2.1399,
+      "step": 30225
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5694058154235145e-06,
+      "loss": 2.536,
+      "step": 30226
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5693584070796463e-06,
+      "loss": 1.2241,
+      "step": 30227
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5693109987357775e-06,
+      "loss": 1.9085,
+      "step": 30228
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5692635903919088e-06,
+      "loss": 2.202,
+      "step": 30229
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5692161820480407e-06,
+      "loss": 3.0144,
+      "step": 30230
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.569168773704172e-06,
+      "loss": 2.5125,
+      "step": 30231
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5691213653603033e-06,
+      "loss": 2.1415,
+      "step": 30232
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.569073957016435e-06,
+      "loss": 1.5833,
+      "step": 30233
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5690265486725665e-06,
+      "loss": 2.9743,
+      "step": 30234
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5689791403286978e-06,
+      "loss": 3.0115,
+      "step": 30235
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5689317319848295e-06,
+      "loss": 3.0857,
+      "step": 30236
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.568884323640961e-06,
+      "loss": 2.7103,
+      "step": 30237
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5688369152970922e-06,
+      "loss": 1.1086,
+      "step": 30238
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5687895069532237e-06,
+      "loss": 2.8097,
+      "step": 30239
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.5687420986093554e-06,
+      "loss": 2.2429,
+      "step": 30240
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5686946902654867e-06,
+      "loss": 0.9984,
+      "step": 30241
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.568647281921618e-06,
+      "loss": 1.312,
+      "step": 30242
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5685998735777497e-06,
+      "loss": 2.3274,
+      "step": 30243
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5685524652338812e-06,
+      "loss": 2.3255,
+      "step": 30244
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5685050568900125e-06,
+      "loss": 3.0257,
+      "step": 30245
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5684576485461442e-06,
+      "loss": 2.5011,
+      "step": 30246
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5684102402022757e-06,
+      "loss": 2.9811,
+      "step": 30247
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.568362831858407e-06,
+      "loss": 2.2765,
+      "step": 30248
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5683154235145387e-06,
+      "loss": 2.6357,
+      "step": 30249
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5682680151706702e-06,
+      "loss": 2.5339,
+      "step": 30250
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5682206068268014e-06,
+      "loss": 3.0473,
+      "step": 30251
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5681731984829331e-06,
+      "loss": 1.9331,
+      "step": 30252
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5681257901390646e-06,
+      "loss": 2.6968,
+      "step": 30253
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.568078381795196e-06,
+      "loss": 2.6672,
+      "step": 30254
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5680309734513276e-06,
+      "loss": 2.3614,
+      "step": 30255
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5679835651074589e-06,
+      "loss": 3.1795,
+      "step": 30256
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5679361567635904e-06,
+      "loss": 2.6195,
+      "step": 30257
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.567888748419722e-06,
+      "loss": 2.3029,
+      "step": 30258
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5678413400758534e-06,
+      "loss": 3.363,
+      "step": 30259
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5677939317319849e-06,
+      "loss": 2.5755,
+      "step": 30260
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5677465233881161e-06,
+      "loss": 2.7917,
+      "step": 30261
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5676991150442478e-06,
+      "loss": 3.4869,
+      "step": 30262
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5676517067003793e-06,
+      "loss": 1.6207,
+      "step": 30263
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5676042983565106e-06,
+      "loss": 2.7858,
+      "step": 30264
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5675568900126423e-06,
+      "loss": 1.9109,
+      "step": 30265
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5675094816687738e-06,
+      "loss": 1.5966,
+      "step": 30266
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.567462073324905e-06,
+      "loss": 2.2779,
+      "step": 30267
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5674146649810368e-06,
+      "loss": 2.3776,
+      "step": 30268
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.567367256637168e-06,
+      "loss": 1.3696,
+      "step": 30269
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5673198482932996e-06,
+      "loss": 4.4302,
+      "step": 30270
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5672724399494313e-06,
+      "loss": 2.9522,
+      "step": 30271
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5672250316055626e-06,
+      "loss": 3.7659,
+      "step": 30272
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.567177623261694e-06,
+      "loss": 1.6215,
+      "step": 30273
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5671302149178258e-06,
+      "loss": 2.7141,
+      "step": 30274
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.567082806573957e-06,
+      "loss": 1.9635,
+      "step": 30275
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5670353982300885e-06,
+      "loss": 0.9147,
+      "step": 30276
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5669879898862202e-06,
+      "loss": 2.5666,
+      "step": 30277
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5669405815423515e-06,
+      "loss": 3.1389,
+      "step": 30278
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.566893173198483e-06,
+      "loss": 0.9881,
+      "step": 30279
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5668457648546147e-06,
+      "loss": 1.8366,
+      "step": 30280
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.566798356510746e-06,
+      "loss": 3.3817,
+      "step": 30281
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5667509481668773e-06,
+      "loss": 1.6948,
+      "step": 30282
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5667035398230088e-06,
+      "loss": 1.9399,
+      "step": 30283
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5666561314791405e-06,
+      "loss": 2.2969,
+      "step": 30284
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5666087231352717e-06,
+      "loss": 2.6351,
+      "step": 30285
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5665613147914032e-06,
+      "loss": 1.3991,
+      "step": 30286
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.566513906447535e-06,
+      "loss": 3.0014,
+      "step": 30287
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5664664981036662e-06,
+      "loss": 3.0391,
+      "step": 30288
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5664190897597977e-06,
+      "loss": 2.1858,
+      "step": 30289
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5663716814159294e-06,
+      "loss": 2.2272,
+      "step": 30290
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5663242730720607e-06,
+      "loss": 3.0664,
+      "step": 30291
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5662768647281922e-06,
+      "loss": 1.8293,
+      "step": 30292
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5662294563843239e-06,
+      "loss": 1.9203,
+      "step": 30293
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5661820480404552e-06,
+      "loss": 2.6194,
+      "step": 30294
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5661346396965865e-06,
+      "loss": 3.4285,
+      "step": 30295
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5660872313527182e-06,
+      "loss": 2.2087,
+      "step": 30296
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5660398230088496e-06,
+      "loss": 2.7246,
+      "step": 30297
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.565992414664981e-06,
+      "loss": 1.0915,
+      "step": 30298
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5659450063211126e-06,
+      "loss": 2.6914,
+      "step": 30299
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5658975979772441e-06,
+      "loss": 2.067,
+      "step": 30300
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5658501896333754e-06,
+      "loss": 2.9006,
+      "step": 30301
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5658027812895071e-06,
+      "loss": 2.793,
+      "step": 30302
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.5657553729456386e-06,
+      "loss": 3.7144,
+      "step": 30303
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5657079646017699e-06,
+      "loss": 1.8073,
+      "step": 30304
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5656605562579014e-06,
+      "loss": 2.715,
+      "step": 30305
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.565613147914033e-06,
+      "loss": 1.9499,
+      "step": 30306
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5655657395701644e-06,
+      "loss": 2.9026,
+      "step": 30307
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5655183312262956e-06,
+      "loss": 2.5895,
+      "step": 30308
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5654709228824273e-06,
+      "loss": 2.7606,
+      "step": 30309
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5654235145385588e-06,
+      "loss": 2.3408,
+      "step": 30310
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5653761061946901e-06,
+      "loss": 1.7861,
+      "step": 30311
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5653286978508218e-06,
+      "loss": 1.7989,
+      "step": 30312
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5652812895069533e-06,
+      "loss": 2.1057,
+      "step": 30313
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5652338811630846e-06,
+      "loss": 2.8485,
+      "step": 30314
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5651864728192163e-06,
+      "loss": 2.9842,
+      "step": 30315
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5651390644753478e-06,
+      "loss": 2.5314,
+      "step": 30316
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.565091656131479e-06,
+      "loss": 0.4198,
+      "step": 30317
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5650442477876108e-06,
+      "loss": 2.8196,
+      "step": 30318
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5649968394437423e-06,
+      "loss": 2.1548,
+      "step": 30319
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5649494310998735e-06,
+      "loss": 0.9507,
+      "step": 30320
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5649020227560052e-06,
+      "loss": 3.0059,
+      "step": 30321
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5648546144121365e-06,
+      "loss": 2.1062,
+      "step": 30322
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.564807206068268e-06,
+      "loss": 2.9075,
+      "step": 30323
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5647597977243997e-06,
+      "loss": 1.7626,
+      "step": 30324
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.564712389380531e-06,
+      "loss": 3.4225,
+      "step": 30325
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5646649810366625e-06,
+      "loss": 2.436,
+      "step": 30326
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5646175726927938e-06,
+      "loss": 1.5024,
+      "step": 30327
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5645701643489255e-06,
+      "loss": 3.4958,
+      "step": 30328
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.564522756005057e-06,
+      "loss": 2.7024,
+      "step": 30329
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5644753476611883e-06,
+      "loss": 2.5441,
+      "step": 30330
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.56442793931732e-06,
+      "loss": 1.5777,
+      "step": 30331
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5643805309734514e-06,
+      "loss": 3.3029,
+      "step": 30332
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5643331226295827e-06,
+      "loss": 1.293,
+      "step": 30333
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5642857142857144e-06,
+      "loss": 1.9532,
+      "step": 30334
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5642383059418457e-06,
+      "loss": 2.2538,
+      "step": 30335
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5641908975979772e-06,
+      "loss": 2.8682,
+      "step": 30336
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.564143489254109e-06,
+      "loss": 0.9545,
+      "step": 30337
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5640960809102402e-06,
+      "loss": 1.7813,
+      "step": 30338
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5640486725663717e-06,
+      "loss": 2.7388,
+      "step": 30339
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5640012642225034e-06,
+      "loss": 1.9745,
+      "step": 30340
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5639538558786347e-06,
+      "loss": 3.4844,
+      "step": 30341
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5639064475347662e-06,
+      "loss": 2.6335,
+      "step": 30342
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5638590391908979e-06,
+      "loss": 0.9967,
+      "step": 30343
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5638116308470291e-06,
+      "loss": 1.8538,
+      "step": 30344
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5637642225031606e-06,
+      "loss": 1.8947,
+      "step": 30345
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5637168141592923e-06,
+      "loss": 3.3719,
+      "step": 30346
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5636694058154236e-06,
+      "loss": 2.4979,
+      "step": 30347
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.563621997471555e-06,
+      "loss": 3.2277,
+      "step": 30348
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5635745891276864e-06,
+      "loss": 2.4921,
+      "step": 30349
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.563527180783818e-06,
+      "loss": 2.2078,
+      "step": 30350
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5634797724399494e-06,
+      "loss": 1.7122,
+      "step": 30351
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5634323640960809e-06,
+      "loss": 2.0271,
+      "step": 30352
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5633849557522126e-06,
+      "loss": 2.5578,
+      "step": 30353
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5633375474083439e-06,
+      "loss": 4.8395,
+      "step": 30354
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5632901390644753e-06,
+      "loss": 2.6405,
+      "step": 30355
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.563242730720607e-06,
+      "loss": 2.3829,
+      "step": 30356
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5631953223767383e-06,
+      "loss": 1.8921,
+      "step": 30357
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5631479140328698e-06,
+      "loss": 3.1289,
+      "step": 30358
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5631005056890015e-06,
+      "loss": 2.8965,
+      "step": 30359
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5630530973451328e-06,
+      "loss": 1.9287,
+      "step": 30360
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.563005689001264e-06,
+      "loss": 1.7363,
+      "step": 30361
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5629582806573958e-06,
+      "loss": 2.5186,
+      "step": 30362
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5629108723135273e-06,
+      "loss": 3.5379,
+      "step": 30363
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5628634639696586e-06,
+      "loss": 1.475,
+      "step": 30364
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5628160556257903e-06,
+      "loss": 2.5297,
+      "step": 30365
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.5627686472819218e-06,
+      "loss": 1.6147,
+      "step": 30366
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.562721238938053e-06,
+      "loss": 3.4975,
+      "step": 30367
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5626738305941847e-06,
+      "loss": 2.8088,
+      "step": 30368
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5626264222503162e-06,
+      "loss": 2.6535,
+      "step": 30369
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5625790139064475e-06,
+      "loss": 2.5075,
+      "step": 30370
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5625316055625788e-06,
+      "loss": 2.7704,
+      "step": 30371
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5624841972187107e-06,
+      "loss": 1.8808,
+      "step": 30372
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.562436788874842e-06,
+      "loss": 3.9275,
+      "step": 30373
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5623893805309733e-06,
+      "loss": 1.8348,
+      "step": 30374
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.562341972187105e-06,
+      "loss": 2.3666,
+      "step": 30375
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5622945638432365e-06,
+      "loss": 2.6814,
+      "step": 30376
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5622471554993678e-06,
+      "loss": 1.8827,
+      "step": 30377
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5621997471554995e-06,
+      "loss": 1.5418,
+      "step": 30378
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.562152338811631e-06,
+      "loss": 3.4078,
+      "step": 30379
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5621049304677622e-06,
+      "loss": 3.3976,
+      "step": 30380
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.562057522123894e-06,
+      "loss": 1.7944,
+      "step": 30381
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5620101137800254e-06,
+      "loss": 2.4563,
+      "step": 30382
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5619627054361567e-06,
+      "loss": 2.9235,
+      "step": 30383
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5619152970922884e-06,
+      "loss": 1.2589,
+      "step": 30384
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5618678887484197e-06,
+      "loss": 2.2652,
+      "step": 30385
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5618204804045512e-06,
+      "loss": 1.5051,
+      "step": 30386
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5617730720606829e-06,
+      "loss": 2.0402,
+      "step": 30387
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5617256637168142e-06,
+      "loss": 1.8861,
+      "step": 30388
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5616782553729457e-06,
+      "loss": 2.7824,
+      "step": 30389
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.561630847029077e-06,
+      "loss": 2.2637,
+      "step": 30390
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5615834386852086e-06,
+      "loss": 2.8543,
+      "step": 30391
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5615360303413401e-06,
+      "loss": 2.7381,
+      "step": 30392
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5614886219974714e-06,
+      "loss": 3.33,
+      "step": 30393
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5614412136536031e-06,
+      "loss": 2.7734,
+      "step": 30394
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5613938053097346e-06,
+      "loss": 1.9449,
+      "step": 30395
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5613463969658659e-06,
+      "loss": 2.3519,
+      "step": 30396
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5612989886219976e-06,
+      "loss": 2.7714,
+      "step": 30397
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5612515802781289e-06,
+      "loss": 2.576,
+      "step": 30398
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5612041719342604e-06,
+      "loss": 2.2436,
+      "step": 30399
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.561156763590392e-06,
+      "loss": 1.2138,
+      "step": 30400
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5611093552465234e-06,
+      "loss": 2.4095,
+      "step": 30401
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5610619469026548e-06,
+      "loss": 2.4001,
+      "step": 30402
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5610145385587865e-06,
+      "loss": 3.621,
+      "step": 30403
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5609671302149178e-06,
+      "loss": 0.8937,
+      "step": 30404
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5609197218710493e-06,
+      "loss": 2.2578,
+      "step": 30405
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.560872313527181e-06,
+      "loss": 1.618,
+      "step": 30406
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5608249051833123e-06,
+      "loss": 3.1309,
+      "step": 30407
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5607774968394438e-06,
+      "loss": 2.7602,
+      "step": 30408
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5607300884955755e-06,
+      "loss": 1.836,
+      "step": 30409
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5606826801517068e-06,
+      "loss": 1.9179,
+      "step": 30410
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.560635271807838e-06,
+      "loss": 2.2224,
+      "step": 30411
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5605878634639696e-06,
+      "loss": 1.5777,
+      "step": 30412
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5605404551201013e-06,
+      "loss": 2.7354,
+      "step": 30413
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5604930467762325e-06,
+      "loss": 2.6387,
+      "step": 30414
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.560445638432364e-06,
+      "loss": 2.5563,
+      "step": 30415
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5603982300884957e-06,
+      "loss": 1.2563,
+      "step": 30416
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.560350821744627e-06,
+      "loss": 3.0974,
+      "step": 30417
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5603034134007585e-06,
+      "loss": 1.4961,
+      "step": 30418
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5602560050568902e-06,
+      "loss": 2.4926,
+      "step": 30419
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5602085967130215e-06,
+      "loss": 1.6098,
+      "step": 30420
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.560161188369153e-06,
+      "loss": 2.4142,
+      "step": 30421
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5601137800252847e-06,
+      "loss": 2.1044,
+      "step": 30422
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.560066371681416e-06,
+      "loss": 1.493,
+      "step": 30423
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5600189633375472e-06,
+      "loss": 0.8601,
+      "step": 30424
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.559971554993679e-06,
+      "loss": 2.6969,
+      "step": 30425
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5599241466498104e-06,
+      "loss": 2.2284,
+      "step": 30426
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5598767383059417e-06,
+      "loss": 1.49,
+      "step": 30427
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5598293299620734e-06,
+      "loss": 2.8372,
+      "step": 30428
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.559781921618205e-06,
+      "loss": 2.3079,
+      "step": 30429
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.5597345132743362e-06,
+      "loss": 2.3359,
+      "step": 30430
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.559687104930468e-06,
+      "loss": 1.9194,
+      "step": 30431
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5596396965865994e-06,
+      "loss": 3.0395,
+      "step": 30432
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5595922882427307e-06,
+      "loss": 2.8145,
+      "step": 30433
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5595448798988622e-06,
+      "loss": 0.8969,
+      "step": 30434
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5594974715549939e-06,
+      "loss": 1.4187,
+      "step": 30435
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5594500632111252e-06,
+      "loss": 2.0969,
+      "step": 30436
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5594026548672564e-06,
+      "loss": 0.9436,
+      "step": 30437
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5593552465233881e-06,
+      "loss": 2.6175,
+      "step": 30438
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5593078381795196e-06,
+      "loss": 2.6911,
+      "step": 30439
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.559260429835651e-06,
+      "loss": 2.3479,
+      "step": 30440
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5592130214917826e-06,
+      "loss": 1.5702,
+      "step": 30441
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5591656131479141e-06,
+      "loss": 2.1964,
+      "step": 30442
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5591182048040454e-06,
+      "loss": 3.6383,
+      "step": 30443
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.559070796460177e-06,
+      "loss": 2.7012,
+      "step": 30444
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5590233881163086e-06,
+      "loss": 2.6572,
+      "step": 30445
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5589759797724399e-06,
+      "loss": 2.8568,
+      "step": 30446
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5589285714285716e-06,
+      "loss": 2.8942,
+      "step": 30447
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.558881163084703e-06,
+      "loss": 1.9083,
+      "step": 30448
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5588337547408343e-06,
+      "loss": 2.0205,
+      "step": 30449
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.558786346396966e-06,
+      "loss": 1.8876,
+      "step": 30450
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5587389380530973e-06,
+      "loss": 3.112,
+      "step": 30451
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5586915297092288e-06,
+      "loss": 2.5922,
+      "step": 30452
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5586441213653605e-06,
+      "loss": 3.4174,
+      "step": 30453
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5585967130214918e-06,
+      "loss": 3.7976,
+      "step": 30454
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5585493046776233e-06,
+      "loss": 0.7353,
+      "step": 30455
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5585018963337546e-06,
+      "loss": 1.9606,
+      "step": 30456
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5584544879898863e-06,
+      "loss": 2.5218,
+      "step": 30457
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5584070796460178e-06,
+      "loss": 3.8817,
+      "step": 30458
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.558359671302149e-06,
+      "loss": 4.4267,
+      "step": 30459
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5583122629582808e-06,
+      "loss": 2.7278,
+      "step": 30460
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5582648546144122e-06,
+      "loss": 3.0878,
+      "step": 30461
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5582174462705435e-06,
+      "loss": 1.8807,
+      "step": 30462
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5581700379266752e-06,
+      "loss": 1.7689,
+      "step": 30463
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5581226295828065e-06,
+      "loss": 1.7665,
+      "step": 30464
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.558075221238938e-06,
+      "loss": 1.6973,
+      "step": 30465
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5580278128950697e-06,
+      "loss": 2.1053,
+      "step": 30466
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.557980404551201e-06,
+      "loss": 2.7514,
+      "step": 30467
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5579329962073325e-06,
+      "loss": 2.9111,
+      "step": 30468
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5578855878634642e-06,
+      "loss": 2.8227,
+      "step": 30469
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5578381795195955e-06,
+      "loss": 1.767,
+      "step": 30470
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.557790771175727e-06,
+      "loss": 1.0575,
+      "step": 30471
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5577433628318587e-06,
+      "loss": 3.431,
+      "step": 30472
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.55769595448799e-06,
+      "loss": 2.6563,
+      "step": 30473
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5576485461441214e-06,
+      "loss": 1.5891,
+      "step": 30474
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5576011378002531e-06,
+      "loss": 1.4955,
+      "step": 30475
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5575537294563844e-06,
+      "loss": 2.362,
+      "step": 30476
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5575063211125157e-06,
+      "loss": 3.0973,
+      "step": 30477
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5574589127686472e-06,
+      "loss": 3.0037,
+      "step": 30478
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5574115044247789e-06,
+      "loss": 2.0286,
+      "step": 30479
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5573640960809102e-06,
+      "loss": 1.9062,
+      "step": 30480
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5573166877370417e-06,
+      "loss": 2.4564,
+      "step": 30481
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5572692793931734e-06,
+      "loss": 3.7995,
+      "step": 30482
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5572218710493047e-06,
+      "loss": 1.9262,
+      "step": 30483
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5571744627054361e-06,
+      "loss": 2.4499,
+      "step": 30484
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5571270543615678e-06,
+      "loss": 1.3149,
+      "step": 30485
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5570796460176991e-06,
+      "loss": 2.0179,
+      "step": 30486
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5570322376738306e-06,
+      "loss": 3.9917,
+      "step": 30487
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5569848293299623e-06,
+      "loss": 1.7396,
+      "step": 30488
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5569374209860936e-06,
+      "loss": 3.5023,
+      "step": 30489
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5568900126422249e-06,
+      "loss": 0.7482,
+      "step": 30490
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5568426042983566e-06,
+      "loss": 2.1113,
+      "step": 30491
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.556795195954488e-06,
+      "loss": 3.1535,
+      "step": 30492
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.5567477876106194e-06,
+      "loss": 3.252,
+      "step": 30493
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.556700379266751e-06,
+      "loss": 1.7554,
+      "step": 30494
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5566529709228826e-06,
+      "loss": 2.3334,
+      "step": 30495
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5566055625790138e-06,
+      "loss": 2.4728,
+      "step": 30496
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5565581542351455e-06,
+      "loss": 1.8766,
+      "step": 30497
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.556510745891277e-06,
+      "loss": 1.8452,
+      "step": 30498
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5564633375474083e-06,
+      "loss": 1.9606,
+      "step": 30499
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5564159292035396e-06,
+      "loss": 1.7658,
+      "step": 30500
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5563685208596715e-06,
+      "loss": 2.8668,
+      "step": 30501
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5563211125158028e-06,
+      "loss": 3.0991,
+      "step": 30502
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.556273704171934e-06,
+      "loss": 2.2492,
+      "step": 30503
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5562262958280658e-06,
+      "loss": 1.9972,
+      "step": 30504
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5561788874841973e-06,
+      "loss": 3.5045,
+      "step": 30505
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5561314791403285e-06,
+      "loss": 2.51,
+      "step": 30506
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5560840707964603e-06,
+      "loss": 2.3122,
+      "step": 30507
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5560366624525917e-06,
+      "loss": 1.9421,
+      "step": 30508
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.555989254108723e-06,
+      "loss": 1.3233,
+      "step": 30509
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5559418457648547e-06,
+      "loss": 3.743,
+      "step": 30510
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5558944374209862e-06,
+      "loss": 2.5263,
+      "step": 30511
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5558470290771175e-06,
+      "loss": 3.507,
+      "step": 30512
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5557996207332492e-06,
+      "loss": 2.8251,
+      "step": 30513
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5557522123893805e-06,
+      "loss": 1.9676,
+      "step": 30514
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.555704804045512e-06,
+      "loss": 1.9022,
+      "step": 30515
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5556573957016437e-06,
+      "loss": 2.8341,
+      "step": 30516
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.555609987357775e-06,
+      "loss": 0.7411,
+      "step": 30517
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5555625790139065e-06,
+      "loss": 2.4595,
+      "step": 30518
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5555151706700382e-06,
+      "loss": 2.5292,
+      "step": 30519
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5554677623261694e-06,
+      "loss": 3.6494,
+      "step": 30520
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.555420353982301e-06,
+      "loss": 1.3682,
+      "step": 30521
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5553729456384322e-06,
+      "loss": 1.7971,
+      "step": 30522
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.555325537294564e-06,
+      "loss": 4.9591,
+      "step": 30523
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5552781289506954e-06,
+      "loss": 1.8227,
+      "step": 30524
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5552307206068267e-06,
+      "loss": 1.9543,
+      "step": 30525
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5551833122629584e-06,
+      "loss": 2.957,
+      "step": 30526
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5551359039190897e-06,
+      "loss": 2.447,
+      "step": 30527
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5550884955752212e-06,
+      "loss": 2.5574,
+      "step": 30528
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5550410872313529e-06,
+      "loss": 3.1605,
+      "step": 30529
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5549936788874841e-06,
+      "loss": 1.9987,
+      "step": 30530
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5549462705436156e-06,
+      "loss": 2.111,
+      "step": 30531
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5548988621997473e-06,
+      "loss": 1.3791,
+      "step": 30532
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5548514538558786e-06,
+      "loss": 3.2192,
+      "step": 30533
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5548040455120101e-06,
+      "loss": 2.2155,
+      "step": 30534
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5547566371681418e-06,
+      "loss": 1.4435,
+      "step": 30535
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.554709228824273e-06,
+      "loss": 3.5511,
+      "step": 30536
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5546618204804046e-06,
+      "loss": 1.912,
+      "step": 30537
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5546144121365363e-06,
+      "loss": 2.744,
+      "step": 30538
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5545670037926676e-06,
+      "loss": 3.329,
+      "step": 30539
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5545195954487989e-06,
+      "loss": 1.193,
+      "step": 30540
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5544721871049306e-06,
+      "loss": 2.4819,
+      "step": 30541
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.554424778761062e-06,
+      "loss": 1.9088,
+      "step": 30542
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5543773704171933e-06,
+      "loss": 2.7794,
+      "step": 30543
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5543299620733248e-06,
+      "loss": 3.077,
+      "step": 30544
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5542825537294565e-06,
+      "loss": 3.4682,
+      "step": 30545
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5542351453855878e-06,
+      "loss": 1.7751,
+      "step": 30546
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5541877370417193e-06,
+      "loss": 1.7465,
+      "step": 30547
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.554140328697851e-06,
+      "loss": 1.9882,
+      "step": 30548
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5540929203539823e-06,
+      "loss": 1.8489,
+      "step": 30549
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5540455120101138e-06,
+      "loss": 2.4949,
+      "step": 30550
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5539981036662455e-06,
+      "loss": 1.9655,
+      "step": 30551
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5539506953223768e-06,
+      "loss": 1.1535,
+      "step": 30552
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.553903286978508e-06,
+      "loss": 2.0702,
+      "step": 30553
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5538558786346397e-06,
+      "loss": 2.9731,
+      "step": 30554
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5538084702907712e-06,
+      "loss": 3.3458,
+      "step": 30555
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.5537610619469025e-06,
+      "loss": 3.2752,
+      "step": 30556
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5537136536030342e-06,
+      "loss": 2.3589,
+      "step": 30557
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5536662452591657e-06,
+      "loss": 1.9161,
+      "step": 30558
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.553618836915297e-06,
+      "loss": 1.875,
+      "step": 30559
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5535714285714287e-06,
+      "loss": 2.16,
+      "step": 30560
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5535240202275602e-06,
+      "loss": 2.0733,
+      "step": 30561
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5534766118836915e-06,
+      "loss": 3.7643,
+      "step": 30562
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5534292035398232e-06,
+      "loss": 1.5994,
+      "step": 30563
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5533817951959547e-06,
+      "loss": 2.5323,
+      "step": 30564
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.553334386852086e-06,
+      "loss": 1.7567,
+      "step": 30565
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5532869785082172e-06,
+      "loss": 1.6447,
+      "step": 30566
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.553239570164349e-06,
+      "loss": 2.0643,
+      "step": 30567
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5531921618204804e-06,
+      "loss": 2.5929,
+      "step": 30568
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5531447534766117e-06,
+      "loss": 2.8821,
+      "step": 30569
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5530973451327434e-06,
+      "loss": 2.6753,
+      "step": 30570
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.553049936788875e-06,
+      "loss": 2.6107,
+      "step": 30571
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5530025284450062e-06,
+      "loss": 2.9016,
+      "step": 30572
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5529551201011379e-06,
+      "loss": 2.7116,
+      "step": 30573
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5529077117572694e-06,
+      "loss": 1.9972,
+      "step": 30574
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5528603034134007e-06,
+      "loss": 3.0084,
+      "step": 30575
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5528128950695324e-06,
+      "loss": 1.2326,
+      "step": 30576
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5527654867256639e-06,
+      "loss": 3.5431,
+      "step": 30577
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5527180783817951e-06,
+      "loss": 2.8308,
+      "step": 30578
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5526706700379268e-06,
+      "loss": 2.3335,
+      "step": 30579
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5526232616940581e-06,
+      "loss": 1.8607,
+      "step": 30580
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5525758533501896e-06,
+      "loss": 2.6916,
+      "step": 30581
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5525284450063213e-06,
+      "loss": 1.5181,
+      "step": 30582
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5524810366624526e-06,
+      "loss": 1.2741,
+      "step": 30583
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.552433628318584e-06,
+      "loss": 1.9839,
+      "step": 30584
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5523862199747158e-06,
+      "loss": 2.5216,
+      "step": 30585
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.552338811630847e-06,
+      "loss": 2.4166,
+      "step": 30586
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5522914032869786e-06,
+      "loss": 2.741,
+      "step": 30587
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5522439949431098e-06,
+      "loss": 1.8174,
+      "step": 30588
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5521965865992416e-06,
+      "loss": 3.2141,
+      "step": 30589
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.552149178255373e-06,
+      "loss": 1.9189,
+      "step": 30590
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5521017699115043e-06,
+      "loss": 2.453,
+      "step": 30591
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.552054361567636e-06,
+      "loss": 2.0945,
+      "step": 30592
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5520069532237673e-06,
+      "loss": 1.947,
+      "step": 30593
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5519595448798988e-06,
+      "loss": 2.4313,
+      "step": 30594
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5519121365360305e-06,
+      "loss": 2.6722,
+      "step": 30595
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5518647281921618e-06,
+      "loss": 2.5008,
+      "step": 30596
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5518173198482933e-06,
+      "loss": 1.647,
+      "step": 30597
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.551769911504425e-06,
+      "loss": 1.8892,
+      "step": 30598
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5517225031605563e-06,
+      "loss": 3.2931,
+      "step": 30599
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5516750948166878e-06,
+      "loss": 1.7368,
+      "step": 30600
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5516276864728195e-06,
+      "loss": 3.275,
+      "step": 30601
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5515802781289507e-06,
+      "loss": 1.9788,
+      "step": 30602
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5515328697850822e-06,
+      "loss": 2.9886,
+      "step": 30603
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.551485461441214e-06,
+      "loss": 3.4633,
+      "step": 30604
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5514380530973452e-06,
+      "loss": 1.794,
+      "step": 30605
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5513906447534765e-06,
+      "loss": 2.6371,
+      "step": 30606
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5513432364096082e-06,
+      "loss": 1.2403,
+      "step": 30607
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5512958280657397e-06,
+      "loss": 1.3851,
+      "step": 30608
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.551248419721871e-06,
+      "loss": 3.2088,
+      "step": 30609
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5512010113780025e-06,
+      "loss": 1.4744,
+      "step": 30610
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5511536030341342e-06,
+      "loss": 3.0407,
+      "step": 30611
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5511061946902654e-06,
+      "loss": 2.4288,
+      "step": 30612
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.551058786346397e-06,
+      "loss": 2.6978,
+      "step": 30613
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5510113780025286e-06,
+      "loss": 1.1082,
+      "step": 30614
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.55096396965866e-06,
+      "loss": 2.0587,
+      "step": 30615
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5509165613147914e-06,
+      "loss": 2.5287,
+      "step": 30616
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5508691529709231e-06,
+      "loss": 1.1287,
+      "step": 30617
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5508217446270544e-06,
+      "loss": 1.485,
+      "step": 30618
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5507743362831857e-06,
+      "loss": 3.9766,
+      "step": 30619
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.5507269279393174e-06,
+      "loss": 2.9861,
+      "step": 30620
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5506795195954489e-06,
+      "loss": 2.0795,
+      "step": 30621
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5506321112515802e-06,
+      "loss": 2.3704,
+      "step": 30622
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5505847029077119e-06,
+      "loss": 1.7929,
+      "step": 30623
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5505372945638434e-06,
+      "loss": 1.6879,
+      "step": 30624
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5504898862199746e-06,
+      "loss": 2.5988,
+      "step": 30625
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5504424778761063e-06,
+      "loss": 1.9628,
+      "step": 30626
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5503950695322378e-06,
+      "loss": 3.3416,
+      "step": 30627
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5503476611883691e-06,
+      "loss": 2.354,
+      "step": 30628
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5503002528445008e-06,
+      "loss": 3.4009,
+      "step": 30629
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5502528445006323e-06,
+      "loss": 1.9287,
+      "step": 30630
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5502054361567636e-06,
+      "loss": 1.7984,
+      "step": 30631
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5501580278128949e-06,
+      "loss": 2.4744,
+      "step": 30632
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5501106194690266e-06,
+      "loss": 1.5305,
+      "step": 30633
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.550063211125158e-06,
+      "loss": 2.6812,
+      "step": 30634
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5500158027812893e-06,
+      "loss": 2.1878,
+      "step": 30635
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.549968394437421e-06,
+      "loss": 1.5398,
+      "step": 30636
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5499209860935525e-06,
+      "loss": 3.0451,
+      "step": 30637
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5498735777496838e-06,
+      "loss": 1.888,
+      "step": 30638
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5498261694058155e-06,
+      "loss": 1.658,
+      "step": 30639
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.549778761061947e-06,
+      "loss": 1.67,
+      "step": 30640
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5497313527180783e-06,
+      "loss": 3.006,
+      "step": 30641
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.54968394437421e-06,
+      "loss": 2.6374,
+      "step": 30642
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5496365360303415e-06,
+      "loss": 2.6775,
+      "step": 30643
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5495891276864728e-06,
+      "loss": 2.9277,
+      "step": 30644
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5495417193426045e-06,
+      "loss": 2.0591,
+      "step": 30645
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5494943109987358e-06,
+      "loss": 3.3367,
+      "step": 30646
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5494469026548673e-06,
+      "loss": 2.93,
+      "step": 30647
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.549399494310999e-06,
+      "loss": 3.5581,
+      "step": 30648
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5493520859671302e-06,
+      "loss": 1.0428,
+      "step": 30649
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5493046776232617e-06,
+      "loss": 2.2835,
+      "step": 30650
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5492572692793934e-06,
+      "loss": 2.169,
+      "step": 30651
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5492098609355247e-06,
+      "loss": 1.9821,
+      "step": 30652
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5491624525916562e-06,
+      "loss": 2.5453,
+      "step": 30653
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5491150442477875e-06,
+      "loss": 3.3507,
+      "step": 30654
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5490676359039192e-06,
+      "loss": 2.0744,
+      "step": 30655
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5490202275600505e-06,
+      "loss": 2.8937,
+      "step": 30656
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.548972819216182e-06,
+      "loss": 2.2825,
+      "step": 30657
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5489254108723137e-06,
+      "loss": 3.7624,
+      "step": 30658
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.548878002528445e-06,
+      "loss": 1.8096,
+      "step": 30659
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5488305941845764e-06,
+      "loss": 1.5808,
+      "step": 30660
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5487831858407081e-06,
+      "loss": 2.5349,
+      "step": 30661
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5487357774968394e-06,
+      "loss": 2.9154,
+      "step": 30662
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.548688369152971e-06,
+      "loss": 4.2295,
+      "step": 30663
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5486409608091026e-06,
+      "loss": 1.7748,
+      "step": 30664
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.548593552465234e-06,
+      "loss": 2.5511,
+      "step": 30665
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5485461441213654e-06,
+      "loss": 2.1844,
+      "step": 30666
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.548498735777497e-06,
+      "loss": 1.9634,
+      "step": 30667
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5484513274336284e-06,
+      "loss": 1.4675,
+      "step": 30668
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5484039190897597e-06,
+      "loss": 2.6544,
+      "step": 30669
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5483565107458914e-06,
+      "loss": 1.5034,
+      "step": 30670
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5483091024020229e-06,
+      "loss": 2.6463,
+      "step": 30671
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5482616940581541e-06,
+      "loss": 2.4529,
+      "step": 30672
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5482142857142858e-06,
+      "loss": 2.2534,
+      "step": 30673
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5481668773704173e-06,
+      "loss": 3.1607,
+      "step": 30674
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5481194690265486e-06,
+      "loss": 2.5756,
+      "step": 30675
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.54807206068268e-06,
+      "loss": 2.9768,
+      "step": 30676
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5480246523388118e-06,
+      "loss": 2.0813,
+      "step": 30677
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.547977243994943e-06,
+      "loss": 2.2968,
+      "step": 30678
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5479298356510746e-06,
+      "loss": 1.5791,
+      "step": 30679
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5478824273072063e-06,
+      "loss": 1.506,
+      "step": 30680
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5478350189633376e-06,
+      "loss": 3.0656,
+      "step": 30681
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5477876106194688e-06,
+      "loss": 2.0859,
+      "step": 30682
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.5477402022756005e-06,
+      "loss": 2.4526,
+      "step": 30683
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.547692793931732e-06,
+      "loss": 2.3554,
+      "step": 30684
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5476453855878633e-06,
+      "loss": 2.7884,
+      "step": 30685
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.547597977243995e-06,
+      "loss": 3.4497,
+      "step": 30686
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5475505689001265e-06,
+      "loss": 1.3593,
+      "step": 30687
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5475031605562578e-06,
+      "loss": 4.0182,
+      "step": 30688
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5474557522123895e-06,
+      "loss": 4.6449,
+      "step": 30689
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.547408343868521e-06,
+      "loss": 2.4881,
+      "step": 30690
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5473609355246523e-06,
+      "loss": 3.2219,
+      "step": 30691
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.547313527180784e-06,
+      "loss": 2.6037,
+      "step": 30692
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5472661188369155e-06,
+      "loss": 2.5782,
+      "step": 30693
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5472187104930467e-06,
+      "loss": 1.8166,
+      "step": 30694
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5471713021491785e-06,
+      "loss": 2.5152,
+      "step": 30695
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5471238938053097e-06,
+      "loss": 2.3821,
+      "step": 30696
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5470764854614412e-06,
+      "loss": 1.4849,
+      "step": 30697
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5470290771175725e-06,
+      "loss": 1.7045,
+      "step": 30698
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5469816687737042e-06,
+      "loss": 2.6896,
+      "step": 30699
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5469342604298357e-06,
+      "loss": 2.6646,
+      "step": 30700
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.546886852085967e-06,
+      "loss": 1.9923,
+      "step": 30701
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5468394437420987e-06,
+      "loss": 3.9109,
+      "step": 30702
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5467920353982302e-06,
+      "loss": 2.1626,
+      "step": 30703
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5467446270543615e-06,
+      "loss": 3.8162,
+      "step": 30704
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5466972187104932e-06,
+      "loss": 2.3168,
+      "step": 30705
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5466498103666247e-06,
+      "loss": 2.1177,
+      "step": 30706
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.546602402022756e-06,
+      "loss": 3.8483,
+      "step": 30707
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5465549936788876e-06,
+      "loss": 2.6284,
+      "step": 30708
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.546507585335019e-06,
+      "loss": 1.2582,
+      "step": 30709
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5464601769911504e-06,
+      "loss": 1.4095,
+      "step": 30710
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5464127686472821e-06,
+      "loss": 1.6096,
+      "step": 30711
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5463653603034134e-06,
+      "loss": 1.7766,
+      "step": 30712
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5463179519595449e-06,
+      "loss": 1.2551,
+      "step": 30713
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5462705436156766e-06,
+      "loss": 1.2428,
+      "step": 30714
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5462231352718079e-06,
+      "loss": 2.8049,
+      "step": 30715
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5461757269279394e-06,
+      "loss": 3.2952,
+      "step": 30716
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.546128318584071e-06,
+      "loss": 3.1309,
+      "step": 30717
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5460809102402023e-06,
+      "loss": 2.226,
+      "step": 30718
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5460335018963338e-06,
+      "loss": 1.7259,
+      "step": 30719
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5459860935524651e-06,
+      "loss": 1.7038,
+      "step": 30720
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5459386852085968e-06,
+      "loss": 1.5631,
+      "step": 30721
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545891276864728e-06,
+      "loss": 0.9108,
+      "step": 30722
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5458438685208596e-06,
+      "loss": 0.6799,
+      "step": 30723
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5457964601769913e-06,
+      "loss": 2.1021,
+      "step": 30724
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5457490518331226e-06,
+      "loss": 1.7785,
+      "step": 30725
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545701643489254e-06,
+      "loss": 2.2328,
+      "step": 30726
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5456542351453858e-06,
+      "loss": 2.3948,
+      "step": 30727
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545606826801517e-06,
+      "loss": 1.2522,
+      "step": 30728
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5455594184576486e-06,
+      "loss": 2.7504,
+      "step": 30729
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5455120101137803e-06,
+      "loss": 4.0651,
+      "step": 30730
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5454646017699115e-06,
+      "loss": 2.437,
+      "step": 30731
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545417193426043e-06,
+      "loss": 2.8335,
+      "step": 30732
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5453697850821747e-06,
+      "loss": 1.7336,
+      "step": 30733
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545322376738306e-06,
+      "loss": 4.1932,
+      "step": 30734
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5452749683944373e-06,
+      "loss": 2.0549,
+      "step": 30735
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545227560050569e-06,
+      "loss": 1.329,
+      "step": 30736
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5451801517067005e-06,
+      "loss": 1.8987,
+      "step": 30737
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5451327433628318e-06,
+      "loss": 1.5747,
+      "step": 30738
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5450853350189635e-06,
+      "loss": 2.9923,
+      "step": 30739
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.545037926675095e-06,
+      "loss": 1.8168,
+      "step": 30740
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5449905183312262e-06,
+      "loss": 3.1408,
+      "step": 30741
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5449431099873577e-06,
+      "loss": 1.0467,
+      "step": 30742
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5448957016434894e-06,
+      "loss": 1.99,
+      "step": 30743
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5448482932996207e-06,
+      "loss": 1.7134,
+      "step": 30744
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.5448008849557522e-06,
+      "loss": 2.944,
+      "step": 30745
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.544753476611884e-06,
+      "loss": 1.5173,
+      "step": 30746
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5447060682680152e-06,
+      "loss": 1.6339,
+      "step": 30747
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5446586599241465e-06,
+      "loss": 1.8664,
+      "step": 30748
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5446112515802782e-06,
+      "loss": 2.6861,
+      "step": 30749
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5445638432364097e-06,
+      "loss": 3.3192,
+      "step": 30750
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.544516434892541e-06,
+      "loss": 1.716,
+      "step": 30751
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5444690265486727e-06,
+      "loss": 2.0584,
+      "step": 30752
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5444216182048042e-06,
+      "loss": 2.6965,
+      "step": 30753
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5443742098609354e-06,
+      "loss": 2.6019,
+      "step": 30754
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5443268015170671e-06,
+      "loss": 2.0326,
+      "step": 30755
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5442793931731986e-06,
+      "loss": 3.0201,
+      "step": 30756
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.54423198482933e-06,
+      "loss": 1.926,
+      "step": 30757
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5441845764854616e-06,
+      "loss": 2.5929,
+      "step": 30758
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.544137168141593e-06,
+      "loss": 1.9143,
+      "step": 30759
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5440897597977244e-06,
+      "loss": 1.0051,
+      "step": 30760
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.544042351453856e-06,
+      "loss": 3.3549,
+      "step": 30761
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5439949431099874e-06,
+      "loss": 2.9823,
+      "step": 30762
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5439475347661189e-06,
+      "loss": 2.703,
+      "step": 30763
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5439001264222501e-06,
+      "loss": 2.6282,
+      "step": 30764
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5438527180783818e-06,
+      "loss": 3.0812,
+      "step": 30765
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5438053097345133e-06,
+      "loss": 2.3108,
+      "step": 30766
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5437579013906446e-06,
+      "loss": 2.1915,
+      "step": 30767
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5437104930467763e-06,
+      "loss": 2.5048,
+      "step": 30768
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5436630847029078e-06,
+      "loss": 3.7385,
+      "step": 30769
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.543615676359039e-06,
+      "loss": 2.5871,
+      "step": 30770
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5435682680151708e-06,
+      "loss": 0.8106,
+      "step": 30771
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5435208596713023e-06,
+      "loss": 1.5906,
+      "step": 30772
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5434734513274336e-06,
+      "loss": 2.2608,
+      "step": 30773
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5434260429835653e-06,
+      "loss": 1.9194,
+      "step": 30774
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5433786346396966e-06,
+      "loss": 1.066,
+      "step": 30775
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.543331226295828e-06,
+      "loss": 1.844,
+      "step": 30776
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5432838179519598e-06,
+      "loss": 2.5014,
+      "step": 30777
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.543236409608091e-06,
+      "loss": 1.8186,
+      "step": 30778
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5431890012642225e-06,
+      "loss": 1.6062,
+      "step": 30779
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5431415929203542e-06,
+      "loss": 2.5987,
+      "step": 30780
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5430941845764855e-06,
+      "loss": 1.7904,
+      "step": 30781
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.543046776232617e-06,
+      "loss": 2.9527,
+      "step": 30782
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5429993678887487e-06,
+      "loss": 1.7556,
+      "step": 30783
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.54295195954488e-06,
+      "loss": 2.013,
+      "step": 30784
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5429045512010113e-06,
+      "loss": 4.0272,
+      "step": 30785
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5428571428571428e-06,
+      "loss": 2.6113,
+      "step": 30786
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5428097345132745e-06,
+      "loss": 2.3983,
+      "step": 30787
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5427623261694057e-06,
+      "loss": 2.024,
+      "step": 30788
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5427149178255372e-06,
+      "loss": 3.7196,
+      "step": 30789
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.542667509481669e-06,
+      "loss": 1.9264,
+      "step": 30790
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5426201011378002e-06,
+      "loss": 1.6031,
+      "step": 30791
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5425726927939317e-06,
+      "loss": 3.7202,
+      "step": 30792
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5425252844500634e-06,
+      "loss": 1.3318,
+      "step": 30793
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5424778761061947e-06,
+      "loss": 2.3268,
+      "step": 30794
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5424304677623262e-06,
+      "loss": 1.9328,
+      "step": 30795
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5423830594184579e-06,
+      "loss": 2.3572,
+      "step": 30796
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5423356510745892e-06,
+      "loss": 4.4436,
+      "step": 30797
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5422882427307205e-06,
+      "loss": 3.5861,
+      "step": 30798
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5422408343868522e-06,
+      "loss": 2.3193,
+      "step": 30799
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5421934260429836e-06,
+      "loss": 3.2979,
+      "step": 30800
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.542146017699115e-06,
+      "loss": 2.0902,
+      "step": 30801
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5420986093552466e-06,
+      "loss": 3.1448,
+      "step": 30802
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5420512010113781e-06,
+      "loss": 1.9557,
+      "step": 30803
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5420037926675094e-06,
+      "loss": 2.0882,
+      "step": 30804
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5419563843236411e-06,
+      "loss": 2.8998,
+      "step": 30805
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5419089759797726e-06,
+      "loss": 2.3588,
+      "step": 30806
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5418615676359039e-06,
+      "loss": 2.0765,
+      "step": 30807
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5418141592920354e-06,
+      "loss": 2.8531,
+      "step": 30808
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.541766750948167e-06,
+      "loss": 1.8525,
+      "step": 30809
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.5417193426042984e-06,
+      "loss": 2.7619,
+      "step": 30810
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5416719342604296e-06,
+      "loss": 3.4758,
+      "step": 30811
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5416245259165613e-06,
+      "loss": 2.0898,
+      "step": 30812
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5415771175726928e-06,
+      "loss": 3.2941,
+      "step": 30813
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5415297092288241e-06,
+      "loss": 2.6622,
+      "step": 30814
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5414823008849558e-06,
+      "loss": 2.5857,
+      "step": 30815
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5414348925410873e-06,
+      "loss": 1.1154,
+      "step": 30816
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5413874841972186e-06,
+      "loss": 2.4692,
+      "step": 30817
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5413400758533503e-06,
+      "loss": 2.0252,
+      "step": 30818
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5412926675094818e-06,
+      "loss": 1.6966,
+      "step": 30819
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.541245259165613e-06,
+      "loss": 1.8101,
+      "step": 30820
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5411978508217448e-06,
+      "loss": 3.2599,
+      "step": 30821
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5411504424778763e-06,
+      "loss": 3.0388,
+      "step": 30822
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5411030341340075e-06,
+      "loss": 2.7519,
+      "step": 30823
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5410556257901392e-06,
+      "loss": 1.4413,
+      "step": 30824
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5410082174462705e-06,
+      "loss": 3.0222,
+      "step": 30825
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.540960809102402e-06,
+      "loss": 3.4224,
+      "step": 30826
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5409134007585333e-06,
+      "loss": 2.2282,
+      "step": 30827
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.540865992414665e-06,
+      "loss": 2.2104,
+      "step": 30828
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5408185840707965e-06,
+      "loss": 1.9729,
+      "step": 30829
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5407711757269278e-06,
+      "loss": 1.9714,
+      "step": 30830
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5407237673830595e-06,
+      "loss": 2.9807,
+      "step": 30831
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.540676359039191e-06,
+      "loss": 1.8348,
+      "step": 30832
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5406289506953223e-06,
+      "loss": 1.962,
+      "step": 30833
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.540581542351454e-06,
+      "loss": 1.381,
+      "step": 30834
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5405341340075855e-06,
+      "loss": 1.6282,
+      "step": 30835
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5404867256637167e-06,
+      "loss": 2.127,
+      "step": 30836
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5404393173198484e-06,
+      "loss": 3.6422,
+      "step": 30837
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5403919089759797e-06,
+      "loss": 1.1917,
+      "step": 30838
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5403445006321112e-06,
+      "loss": 2.5224,
+      "step": 30839
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.540297092288243e-06,
+      "loss": 1.6791,
+      "step": 30840
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5402496839443742e-06,
+      "loss": 1.7947,
+      "step": 30841
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5402022756005057e-06,
+      "loss": 1.2894,
+      "step": 30842
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5401548672566374e-06,
+      "loss": 4.3576,
+      "step": 30843
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5401074589127687e-06,
+      "loss": 2.9153,
+      "step": 30844
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5400600505689002e-06,
+      "loss": 0.7445,
+      "step": 30845
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5400126422250319e-06,
+      "loss": 2.9807,
+      "step": 30846
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5399652338811631e-06,
+      "loss": 1.8133,
+      "step": 30847
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5399178255372946e-06,
+      "loss": 2.2374,
+      "step": 30848
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.539870417193426e-06,
+      "loss": 1.5835,
+      "step": 30849
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5398230088495576e-06,
+      "loss": 2.4641,
+      "step": 30850
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.539775600505689e-06,
+      "loss": 1.9325,
+      "step": 30851
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5397281921618204e-06,
+      "loss": 2.5799,
+      "step": 30852
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.539680783817952e-06,
+      "loss": 1.8371,
+      "step": 30853
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5396333754740834e-06,
+      "loss": 1.6214,
+      "step": 30854
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5395859671302149e-06,
+      "loss": 3.7126,
+      "step": 30855
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5395385587863466e-06,
+      "loss": 2.5121,
+      "step": 30856
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5394911504424779e-06,
+      "loss": 2.2889,
+      "step": 30857
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5394437420986093e-06,
+      "loss": 3.6047,
+      "step": 30858
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.539396333754741e-06,
+      "loss": 2.4248,
+      "step": 30859
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5393489254108723e-06,
+      "loss": 3.7754,
+      "step": 30860
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5393015170670038e-06,
+      "loss": 3.2157,
+      "step": 30861
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5392541087231355e-06,
+      "loss": 1.7918,
+      "step": 30862
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5392067003792668e-06,
+      "loss": 2.314,
+      "step": 30863
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.539159292035398e-06,
+      "loss": 0.892,
+      "step": 30864
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5391118836915298e-06,
+      "loss": 1.7273,
+      "step": 30865
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5390644753476613e-06,
+      "loss": 2.1964,
+      "step": 30866
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5390170670037926e-06,
+      "loss": 1.3399,
+      "step": 30867
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5389696586599243e-06,
+      "loss": 1.764,
+      "step": 30868
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5389222503160558e-06,
+      "loss": 2.7057,
+      "step": 30869
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.538874841972187e-06,
+      "loss": 2.3991,
+      "step": 30870
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5388274336283185e-06,
+      "loss": 3.797,
+      "step": 30871
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5387800252844502e-06,
+      "loss": 1.702,
+      "step": 30872
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.5387326169405815e-06,
+      "loss": 2.6133,
+      "step": 30873
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.538685208596713e-06,
+      "loss": 1.9393,
+      "step": 30874
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5386378002528447e-06,
+      "loss": 1.6378,
+      "step": 30875
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.538590391908976e-06,
+      "loss": 2.4425,
+      "step": 30876
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5385429835651073e-06,
+      "loss": 2.9654,
+      "step": 30877
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.538495575221239e-06,
+      "loss": 3.4864,
+      "step": 30878
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5384481668773705e-06,
+      "loss": 2.2276,
+      "step": 30879
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5384007585335018e-06,
+      "loss": 2.2838,
+      "step": 30880
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5383533501896335e-06,
+      "loss": 2.7622,
+      "step": 30881
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.538305941845765e-06,
+      "loss": 2.6315,
+      "step": 30882
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5382585335018962e-06,
+      "loss": 3.0168,
+      "step": 30883
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.538211125158028e-06,
+      "loss": 1.9074,
+      "step": 30884
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5381637168141594e-06,
+      "loss": 2.2547,
+      "step": 30885
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5381163084702907e-06,
+      "loss": 1.2094,
+      "step": 30886
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5380689001264224e-06,
+      "loss": 1.3996,
+      "step": 30887
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.538021491782554e-06,
+      "loss": 1.4084,
+      "step": 30888
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5379740834386852e-06,
+      "loss": 2.493,
+      "step": 30889
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5379266750948169e-06,
+      "loss": 3.8554,
+      "step": 30890
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5378792667509482e-06,
+      "loss": 2.0529,
+      "step": 30891
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5378318584070797e-06,
+      "loss": 2.1674,
+      "step": 30892
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.537784450063211e-06,
+      "loss": 1.7063,
+      "step": 30893
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5377370417193426e-06,
+      "loss": 2.5998,
+      "step": 30894
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5376896333754741e-06,
+      "loss": 1.0698,
+      "step": 30895
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5376422250316054e-06,
+      "loss": 1.6476,
+      "step": 30896
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5375948166877371e-06,
+      "loss": 3.4818,
+      "step": 30897
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5375474083438686e-06,
+      "loss": 1.324,
+      "step": 30898
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5374999999999999e-06,
+      "loss": 2.3703,
+      "step": 30899
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5374525916561316e-06,
+      "loss": 3.7219,
+      "step": 30900
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.537405183312263e-06,
+      "loss": 1.5848,
+      "step": 30901
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5373577749683944e-06,
+      "loss": 2.1403,
+      "step": 30902
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.537310366624526e-06,
+      "loss": 3.0127,
+      "step": 30903
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5372629582806574e-06,
+      "loss": 3.4055,
+      "step": 30904
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5372155499367888e-06,
+      "loss": 3.2392,
+      "step": 30905
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5371681415929205e-06,
+      "loss": 2.255,
+      "step": 30906
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5371207332490518e-06,
+      "loss": 2.8856,
+      "step": 30907
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5370733249051833e-06,
+      "loss": 1.792,
+      "step": 30908
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.537025916561315e-06,
+      "loss": 1.514,
+      "step": 30909
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5369785082174463e-06,
+      "loss": 3.031,
+      "step": 30910
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5369310998735778e-06,
+      "loss": 2.5365,
+      "step": 30911
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5368836915297095e-06,
+      "loss": 3.3786,
+      "step": 30912
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5368362831858408e-06,
+      "loss": 1.6599,
+      "step": 30913
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5367888748419723e-06,
+      "loss": 1.8619,
+      "step": 30914
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5367414664981036e-06,
+      "loss": 2.7264,
+      "step": 30915
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5366940581542353e-06,
+      "loss": 2.2791,
+      "step": 30916
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5366466498103665e-06,
+      "loss": 1.5602,
+      "step": 30917
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.536599241466498e-06,
+      "loss": 1.6085,
+      "step": 30918
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5365518331226297e-06,
+      "loss": 2.7209,
+      "step": 30919
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.536504424778761e-06,
+      "loss": 2.4655,
+      "step": 30920
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5364570164348925e-06,
+      "loss": 1.6184,
+      "step": 30921
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5364096080910242e-06,
+      "loss": 2.0533,
+      "step": 30922
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5363621997471555e-06,
+      "loss": 2.9973,
+      "step": 30923
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.536314791403287e-06,
+      "loss": 1.8974,
+      "step": 30924
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5362673830594187e-06,
+      "loss": 1.8873,
+      "step": 30925
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.53621997471555e-06,
+      "loss": 1.6871,
+      "step": 30926
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5361725663716812e-06,
+      "loss": 1.1035,
+      "step": 30927
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5361251580278132e-06,
+      "loss": 2.7853,
+      "step": 30928
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5360777496839444e-06,
+      "loss": 2.3927,
+      "step": 30929
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5360303413400757e-06,
+      "loss": 2.1835,
+      "step": 30930
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5359829329962074e-06,
+      "loss": 2.4811,
+      "step": 30931
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.535935524652339e-06,
+      "loss": 2.0488,
+      "step": 30932
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5358881163084702e-06,
+      "loss": 2.1105,
+      "step": 30933
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.535840707964602e-06,
+      "loss": 3.1014,
+      "step": 30934
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5357932996207334e-06,
+      "loss": 0.9572,
+      "step": 30935
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5357458912768647e-06,
+      "loss": 2.6268,
+      "step": 30936
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5356984829329962e-06,
+      "loss": 2.7654,
+      "step": 30937
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5356510745891279e-06,
+      "loss": 2.491,
+      "step": 30938
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5356036662452592e-06,
+      "loss": 2.2478,
+      "step": 30939
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5355562579013904e-06,
+      "loss": 1.3764,
+      "step": 30940
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5355088495575221e-06,
+      "loss": 1.5808,
+      "step": 30941
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5354614412136536e-06,
+      "loss": 2.067,
+      "step": 30942
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.535414032869785e-06,
+      "loss": 1.6904,
+      "step": 30943
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5353666245259166e-06,
+      "loss": 1.2841,
+      "step": 30944
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5353192161820481e-06,
+      "loss": 3.0103,
+      "step": 30945
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5352718078381794e-06,
+      "loss": 3.1846,
+      "step": 30946
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.535224399494311e-06,
+      "loss": 2.265,
+      "step": 30947
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5351769911504426e-06,
+      "loss": 3.268,
+      "step": 30948
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5351295828065739e-06,
+      "loss": 2.8577,
+      "step": 30949
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5350821744627056e-06,
+      "loss": 0.8308,
+      "step": 30950
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.535034766118837e-06,
+      "loss": 2.6392,
+      "step": 30951
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5349873577749683e-06,
+      "loss": 2.4462,
+      "step": 30952
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5349399494311e-06,
+      "loss": 1.8179,
+      "step": 30953
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5348925410872313e-06,
+      "loss": 2.4605,
+      "step": 30954
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5348451327433628e-06,
+      "loss": 2.1014,
+      "step": 30955
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5347977243994945e-06,
+      "loss": 1.8369,
+      "step": 30956
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5347503160556258e-06,
+      "loss": 1.3044,
+      "step": 30957
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5347029077117573e-06,
+      "loss": 2.5419,
+      "step": 30958
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5346554993678886e-06,
+      "loss": 2.3886,
+      "step": 30959
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5346080910240203e-06,
+      "loss": 2.3804,
+      "step": 30960
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5345606826801518e-06,
+      "loss": 1.8103,
+      "step": 30961
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.534513274336283e-06,
+      "loss": 1.9491,
+      "step": 30962
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5344658659924148e-06,
+      "loss": 1.8442,
+      "step": 30963
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5344184576485462e-06,
+      "loss": 3.0133,
+      "step": 30964
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5343710493046775e-06,
+      "loss": 1.3057,
+      "step": 30965
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5343236409608092e-06,
+      "loss": 1.5072,
+      "step": 30966
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5342762326169405e-06,
+      "loss": 1.6076,
+      "step": 30967
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.534228824273072e-06,
+      "loss": 2.7172,
+      "step": 30968
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5341814159292037e-06,
+      "loss": 2.9897,
+      "step": 30969
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.534134007585335e-06,
+      "loss": 3.9368,
+      "step": 30970
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5340865992414665e-06,
+      "loss": 2.309,
+      "step": 30971
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5340391908975982e-06,
+      "loss": 2.4275,
+      "step": 30972
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5339917825537295e-06,
+      "loss": 1.2039,
+      "step": 30973
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.533944374209861e-06,
+      "loss": 2.8924,
+      "step": 30974
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5338969658659927e-06,
+      "loss": 2.2693,
+      "step": 30975
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.533849557522124e-06,
+      "loss": 3.5541,
+      "step": 30976
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5338021491782554e-06,
+      "loss": 2.8742,
+      "step": 30977
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5337547408343871e-06,
+      "loss": 2.1866,
+      "step": 30978
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5337073324905184e-06,
+      "loss": 2.1444,
+      "step": 30979
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5336599241466497e-06,
+      "loss": 3.0153,
+      "step": 30980
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5336125158027812e-06,
+      "loss": 4.6566,
+      "step": 30981
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.533565107458913e-06,
+      "loss": 2.2795,
+      "step": 30982
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5335176991150442e-06,
+      "loss": 2.2523,
+      "step": 30983
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5334702907711757e-06,
+      "loss": 2.1672,
+      "step": 30984
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5334228824273074e-06,
+      "loss": 2.3312,
+      "step": 30985
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5333754740834387e-06,
+      "loss": 2.4293,
+      "step": 30986
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5333280657395701e-06,
+      "loss": 2.4498,
+      "step": 30987
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5332806573957018e-06,
+      "loss": 2.337,
+      "step": 30988
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5332332490518331e-06,
+      "loss": 1.8729,
+      "step": 30989
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5331858407079646e-06,
+      "loss": 1.322,
+      "step": 30990
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5331384323640963e-06,
+      "loss": 1.9579,
+      "step": 30991
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5330910240202276e-06,
+      "loss": 2.9979,
+      "step": 30992
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5330436156763589e-06,
+      "loss": 1.2603,
+      "step": 30993
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5329962073324906e-06,
+      "loss": 1.6373,
+      "step": 30994
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.532948798988622e-06,
+      "loss": 2.7616,
+      "step": 30995
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5329013906447534e-06,
+      "loss": 3.2459,
+      "step": 30996
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.532853982300885e-06,
+      "loss": 4.0907,
+      "step": 30997
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5328065739570166e-06,
+      "loss": 1.5402,
+      "step": 30998
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5327591656131478e-06,
+      "loss": 2.2037,
+      "step": 30999
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.5327117572692795e-06,
+      "loss": 1.5346,
+      "step": 31000
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.532664348925411e-06,
+      "loss": 2.2324,
+      "step": 31001
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5326169405815423e-06,
+      "loss": 2.0551,
+      "step": 31002
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5325695322376738e-06,
+      "loss": 2.4043,
+      "step": 31003
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5325221238938055e-06,
+      "loss": 2.5985,
+      "step": 31004
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5324747155499368e-06,
+      "loss": 2.1265,
+      "step": 31005
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.532427307206068e-06,
+      "loss": 1.8358,
+      "step": 31006
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5323798988621998e-06,
+      "loss": 2.3614,
+      "step": 31007
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5323324905183313e-06,
+      "loss": 3.1441,
+      "step": 31008
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5322850821744625e-06,
+      "loss": 1.4301,
+      "step": 31009
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5322376738305943e-06,
+      "loss": 2.1254,
+      "step": 31010
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5321902654867257e-06,
+      "loss": 0.5415,
+      "step": 31011
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.532142857142857e-06,
+      "loss": 2.2342,
+      "step": 31012
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5320954487989887e-06,
+      "loss": 2.0848,
+      "step": 31013
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5320480404551202e-06,
+      "loss": 1.6004,
+      "step": 31014
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5320006321112515e-06,
+      "loss": 2.4776,
+      "step": 31015
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5319532237673832e-06,
+      "loss": 2.4984,
+      "step": 31016
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5319058154235147e-06,
+      "loss": 3.5413,
+      "step": 31017
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.531858407079646e-06,
+      "loss": 2.063,
+      "step": 31018
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5318109987357777e-06,
+      "loss": 2.8052,
+      "step": 31019
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.531763590391909e-06,
+      "loss": 1.7602,
+      "step": 31020
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5317161820480405e-06,
+      "loss": 2.6749,
+      "step": 31021
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5316687737041722e-06,
+      "loss": 2.7971,
+      "step": 31022
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5316213653603034e-06,
+      "loss": 1.0945,
+      "step": 31023
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.531573957016435e-06,
+      "loss": 1.7863,
+      "step": 31024
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5315265486725662e-06,
+      "loss": 2.8379,
+      "step": 31025
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.531479140328698e-06,
+      "loss": 2.6618,
+      "step": 31026
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5314317319848294e-06,
+      "loss": 2.01,
+      "step": 31027
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5313843236409607e-06,
+      "loss": 2.9798,
+      "step": 31028
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5313369152970924e-06,
+      "loss": 3.4538,
+      "step": 31029
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5312895069532239e-06,
+      "loss": 1.6742,
+      "step": 31030
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5312420986093552e-06,
+      "loss": 1.6843,
+      "step": 31031
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5311946902654869e-06,
+      "loss": 2.6015,
+      "step": 31032
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5311472819216181e-06,
+      "loss": 2.9418,
+      "step": 31033
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5310998735777496e-06,
+      "loss": 1.6065,
+      "step": 31034
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5310524652338813e-06,
+      "loss": 2.3528,
+      "step": 31035
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5310050568900126e-06,
+      "loss": 2.8739,
+      "step": 31036
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5309576485461441e-06,
+      "loss": 1.6532,
+      "step": 31037
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5309102402022758e-06,
+      "loss": 2.3483,
+      "step": 31038
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.530862831858407e-06,
+      "loss": 4.1959,
+      "step": 31039
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5308154235145386e-06,
+      "loss": 2.5859,
+      "step": 31040
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5307680151706703e-06,
+      "loss": 3.6116,
+      "step": 31041
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5307206068268016e-06,
+      "loss": 2.1421,
+      "step": 31042
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.530673198482933e-06,
+      "loss": 1.9596,
+      "step": 31043
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5306257901390648e-06,
+      "loss": 2.4791,
+      "step": 31044
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.530578381795196e-06,
+      "loss": 2.5274,
+      "step": 31045
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5305309734513273e-06,
+      "loss": 1.9603,
+      "step": 31046
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5304835651074588e-06,
+      "loss": 3.1389,
+      "step": 31047
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5304361567635905e-06,
+      "loss": 1.4057,
+      "step": 31048
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5303887484197218e-06,
+      "loss": 2.877,
+      "step": 31049
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5303413400758533e-06,
+      "loss": 2.7005,
+      "step": 31050
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.530293931731985e-06,
+      "loss": 1.0875,
+      "step": 31051
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5302465233881163e-06,
+      "loss": 2.3061,
+      "step": 31052
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5301991150442478e-06,
+      "loss": 2.638,
+      "step": 31053
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5301517067003795e-06,
+      "loss": 2.927,
+      "step": 31054
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5301042983565108e-06,
+      "loss": 1.5601,
+      "step": 31055
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5300568900126423e-06,
+      "loss": 3.4233,
+      "step": 31056
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.530009481668774e-06,
+      "loss": 2.3931,
+      "step": 31057
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5299620733249052e-06,
+      "loss": 2.8242,
+      "step": 31058
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5299146649810365e-06,
+      "loss": 1.4627,
+      "step": 31059
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5298672566371682e-06,
+      "loss": 2.3877,
+      "step": 31060
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5298198482932997e-06,
+      "loss": 2.3871,
+      "step": 31061
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.529772439949431e-06,
+      "loss": 1.4988,
+      "step": 31062
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5297250316055627e-06,
+      "loss": 1.7328,
+      "step": 31063
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5296776232616942e-06,
+      "loss": 1.9345,
+      "step": 31064
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5296302149178255e-06,
+      "loss": 1.457,
+      "step": 31065
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5295828065739572e-06,
+      "loss": 2.7078,
+      "step": 31066
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5295353982300887e-06,
+      "loss": 3.085,
+      "step": 31067
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.52948798988622e-06,
+      "loss": 2.224,
+      "step": 31068
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5294405815423512e-06,
+      "loss": 1.7088,
+      "step": 31069
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5293931731984831e-06,
+      "loss": 3.164,
+      "step": 31070
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5293457648546144e-06,
+      "loss": 1.678,
+      "step": 31071
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5292983565107457e-06,
+      "loss": 2.3825,
+      "step": 31072
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5292509481668774e-06,
+      "loss": 1.2941,
+      "step": 31073
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.529203539823009e-06,
+      "loss": 2.531,
+      "step": 31074
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5291561314791402e-06,
+      "loss": 1.7681,
+      "step": 31075
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5291087231352719e-06,
+      "loss": 3.5644,
+      "step": 31076
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5290613147914034e-06,
+      "loss": 2.4304,
+      "step": 31077
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5290139064475347e-06,
+      "loss": 1.9631,
+      "step": 31078
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5289664981036664e-06,
+      "loss": 2.6001,
+      "step": 31079
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5289190897597979e-06,
+      "loss": 2.2132,
+      "step": 31080
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5288716814159291e-06,
+      "loss": 2.6859,
+      "step": 31081
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5288242730720608e-06,
+      "loss": 2.5414,
+      "step": 31082
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5287768647281921e-06,
+      "loss": 3.0751,
+      "step": 31083
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5287294563843236e-06,
+      "loss": 1.6735,
+      "step": 31084
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5286820480404553e-06,
+      "loss": 2.0956,
+      "step": 31085
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5286346396965866e-06,
+      "loss": 2.9193,
+      "step": 31086
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.528587231352718e-06,
+      "loss": 2.3324,
+      "step": 31087
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5285398230088498e-06,
+      "loss": 2.41,
+      "step": 31088
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.528492414664981e-06,
+      "loss": 2.0677,
+      "step": 31089
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5284450063211126e-06,
+      "loss": 2.3625,
+      "step": 31090
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5283975979772438e-06,
+      "loss": 1.9242,
+      "step": 31091
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5283501896333756e-06,
+      "loss": 3.4259,
+      "step": 31092
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.528302781289507e-06,
+      "loss": 0.7514,
+      "step": 31093
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5282553729456383e-06,
+      "loss": 2.3021,
+      "step": 31094
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.52820796460177e-06,
+      "loss": 3.4597,
+      "step": 31095
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5281605562579013e-06,
+      "loss": 2.2305,
+      "step": 31096
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5281131479140328e-06,
+      "loss": 2.6122,
+      "step": 31097
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5280657395701645e-06,
+      "loss": 3.1968,
+      "step": 31098
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5280183312262958e-06,
+      "loss": 1.4076,
+      "step": 31099
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5279709228824273e-06,
+      "loss": 2.311,
+      "step": 31100
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.527923514538559e-06,
+      "loss": 2.7731,
+      "step": 31101
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5278761061946903e-06,
+      "loss": 1.6415,
+      "step": 31102
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5278286978508218e-06,
+      "loss": 2.3661,
+      "step": 31103
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5277812895069535e-06,
+      "loss": 1.8502,
+      "step": 31104
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5277338811630847e-06,
+      "loss": 2.9127,
+      "step": 31105
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5276864728192162e-06,
+      "loss": 2.1702,
+      "step": 31106
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.527639064475348e-06,
+      "loss": 3.0829,
+      "step": 31107
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5275916561314792e-06,
+      "loss": 2.9902,
+      "step": 31108
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5275442477876105e-06,
+      "loss": 2.2327,
+      "step": 31109
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5274968394437422e-06,
+      "loss": 1.6268,
+      "step": 31110
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5274494310998737e-06,
+      "loss": 2.9308,
+      "step": 31111
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.527402022756005e-06,
+      "loss": 3.5801,
+      "step": 31112
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5273546144121365e-06,
+      "loss": 2.5185,
+      "step": 31113
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5273072060682682e-06,
+      "loss": 2.222,
+      "step": 31114
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5272597977243994e-06,
+      "loss": 2.6032,
+      "step": 31115
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.527212389380531e-06,
+      "loss": 2.2163,
+      "step": 31116
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5271649810366626e-06,
+      "loss": 3.0482,
+      "step": 31117
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.527117572692794e-06,
+      "loss": 2.4295,
+      "step": 31118
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5270701643489254e-06,
+      "loss": 2.2882,
+      "step": 31119
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5270227560050571e-06,
+      "loss": 3.0898,
+      "step": 31120
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5269753476611884e-06,
+      "loss": 2.6569,
+      "step": 31121
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5269279393173197e-06,
+      "loss": 1.9182,
+      "step": 31122
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5268805309734514e-06,
+      "loss": 1.9478,
+      "step": 31123
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5268331226295829e-06,
+      "loss": 2.0219,
+      "step": 31124
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5267857142857142e-06,
+      "loss": 2.4737,
+      "step": 31125
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.5267383059418459e-06,
+      "loss": 3.2823,
+      "step": 31126
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5266908975979774e-06,
+      "loss": 2.3355,
+      "step": 31127
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5266434892541086e-06,
+      "loss": 1.1042,
+      "step": 31128
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5265960809102403e-06,
+      "loss": 1.8997,
+      "step": 31129
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5265486725663718e-06,
+      "loss": 1.7199,
+      "step": 31130
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5265012642225031e-06,
+      "loss": 2.3618,
+      "step": 31131
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5264538558786348e-06,
+      "loss": 2.3076,
+      "step": 31132
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5264064475347663e-06,
+      "loss": 1.4974,
+      "step": 31133
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5263590391908976e-06,
+      "loss": 2.7059,
+      "step": 31134
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5263116308470289e-06,
+      "loss": 3.9467,
+      "step": 31135
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5262642225031606e-06,
+      "loss": 2.0513,
+      "step": 31136
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.526216814159292e-06,
+      "loss": 2.5584,
+      "step": 31137
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5261694058154233e-06,
+      "loss": 1.8636,
+      "step": 31138
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.526121997471555e-06,
+      "loss": 3.7534,
+      "step": 31139
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5260745891276865e-06,
+      "loss": 2.7876,
+      "step": 31140
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5260271807838178e-06,
+      "loss": 2.2237,
+      "step": 31141
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5259797724399495e-06,
+      "loss": 2.0553,
+      "step": 31142
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.525932364096081e-06,
+      "loss": 2.3059,
+      "step": 31143
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5258849557522123e-06,
+      "loss": 2.3076,
+      "step": 31144
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.525837547408344e-06,
+      "loss": 1.7419,
+      "step": 31145
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5257901390644755e-06,
+      "loss": 3.1231,
+      "step": 31146
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5257427307206068e-06,
+      "loss": 1.2966,
+      "step": 31147
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5256953223767385e-06,
+      "loss": 2.2367,
+      "step": 31148
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5256479140328698e-06,
+      "loss": 1.591,
+      "step": 31149
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5256005056890013e-06,
+      "loss": 2.4647,
+      "step": 31150
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.525553097345133e-06,
+      "loss": 3.1805,
+      "step": 31151
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5255056890012642e-06,
+      "loss": 2.6509,
+      "step": 31152
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5254582806573957e-06,
+      "loss": 1.4747,
+      "step": 31153
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5254108723135274e-06,
+      "loss": 2.3477,
+      "step": 31154
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5253634639696587e-06,
+      "loss": 1.5579,
+      "step": 31155
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5253160556257902e-06,
+      "loss": 3.9401,
+      "step": 31156
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5252686472819215e-06,
+      "loss": 3.4642,
+      "step": 31157
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5252212389380532e-06,
+      "loss": 2.5783,
+      "step": 31158
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5251738305941847e-06,
+      "loss": 2.6411,
+      "step": 31159
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.525126422250316e-06,
+      "loss": 2.5742,
+      "step": 31160
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5250790139064477e-06,
+      "loss": 1.2337,
+      "step": 31161
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.525031605562579e-06,
+      "loss": 1.9674,
+      "step": 31162
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5249841972187104e-06,
+      "loss": 2.4347,
+      "step": 31163
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5249367888748421e-06,
+      "loss": 2.4424,
+      "step": 31164
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5248893805309734e-06,
+      "loss": 2.9748,
+      "step": 31165
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.524841972187105e-06,
+      "loss": 1.2521,
+      "step": 31166
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5247945638432366e-06,
+      "loss": 3.0425,
+      "step": 31167
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.524747155499368e-06,
+      "loss": 2.0997,
+      "step": 31168
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5246997471554994e-06,
+      "loss": 1.4742,
+      "step": 31169
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.524652338811631e-06,
+      "loss": 2.1969,
+      "step": 31170
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5246049304677624e-06,
+      "loss": 1.773,
+      "step": 31171
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5245575221238939e-06,
+      "loss": 1.7711,
+      "step": 31172
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5245101137800256e-06,
+      "loss": 2.6577,
+      "step": 31173
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5244627054361569e-06,
+      "loss": 3.2851,
+      "step": 31174
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5244152970922881e-06,
+      "loss": 2.0889,
+      "step": 31175
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5243678887484198e-06,
+      "loss": 3.3078,
+      "step": 31176
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5243204804045513e-06,
+      "loss": 3.3677,
+      "step": 31177
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5242730720606826e-06,
+      "loss": 2.1619,
+      "step": 31178
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.524225663716814e-06,
+      "loss": 1.6354,
+      "step": 31179
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5241782553729458e-06,
+      "loss": 2.7365,
+      "step": 31180
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.524130847029077e-06,
+      "loss": 1.8602,
+      "step": 31181
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5240834386852086e-06,
+      "loss": 2.7404,
+      "step": 31182
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5240360303413403e-06,
+      "loss": 2.4937,
+      "step": 31183
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5239886219974716e-06,
+      "loss": 1.3912,
+      "step": 31184
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.523941213653603e-06,
+      "loss": 3.4563,
+      "step": 31185
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5238938053097348e-06,
+      "loss": 1.3969,
+      "step": 31186
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.523846396965866e-06,
+      "loss": 1.6386,
+      "step": 31187
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5237989886219973e-06,
+      "loss": 2.8735,
+      "step": 31188
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.523751580278129e-06,
+      "loss": 2.1784,
+      "step": 31189
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.5237041719342605e-06,
+      "loss": 3.5693,
+      "step": 31190
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5236567635903918e-06,
+      "loss": 2.0451,
+      "step": 31191
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5236093552465235e-06,
+      "loss": 1.696,
+      "step": 31192
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.523561946902655e-06,
+      "loss": 2.3428,
+      "step": 31193
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5235145385587863e-06,
+      "loss": 2.2359,
+      "step": 31194
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.523467130214918e-06,
+      "loss": 2.5972,
+      "step": 31195
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5234197218710495e-06,
+      "loss": 1.8239,
+      "step": 31196
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5233723135271807e-06,
+      "loss": 2.5665,
+      "step": 31197
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5233249051833125e-06,
+      "loss": 3.0493,
+      "step": 31198
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.523277496839444e-06,
+      "loss": 3.2745,
+      "step": 31199
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5232300884955752e-06,
+      "loss": 2.4844,
+      "step": 31200
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5231826801517065e-06,
+      "loss": 2.2588,
+      "step": 31201
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5231352718078382e-06,
+      "loss": 1.3783,
+      "step": 31202
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5230878634639697e-06,
+      "loss": 2.354,
+      "step": 31203
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.523040455120101e-06,
+      "loss": 1.9559,
+      "step": 31204
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5229930467762327e-06,
+      "loss": 1.9589,
+      "step": 31205
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5229456384323642e-06,
+      "loss": 2.9214,
+      "step": 31206
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5228982300884955e-06,
+      "loss": 1.0464,
+      "step": 31207
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5228508217446272e-06,
+      "loss": 4.2,
+      "step": 31208
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5228034134007587e-06,
+      "loss": 2.4231,
+      "step": 31209
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.52275600505689e-06,
+      "loss": 1.7984,
+      "step": 31210
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5227085967130216e-06,
+      "loss": 2.0244,
+      "step": 31211
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.522661188369153e-06,
+      "loss": 2.8346,
+      "step": 31212
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5226137800252844e-06,
+      "loss": 2.7469,
+      "step": 31213
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5225663716814161e-06,
+      "loss": 2.4987,
+      "step": 31214
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5225189633375474e-06,
+      "loss": 3.0573,
+      "step": 31215
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5224715549936789e-06,
+      "loss": 1.9826,
+      "step": 31216
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5224241466498106e-06,
+      "loss": 2.8362,
+      "step": 31217
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5223767383059419e-06,
+      "loss": 2.2809,
+      "step": 31218
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5223293299620734e-06,
+      "loss": 1.8993,
+      "step": 31219
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.522281921618205e-06,
+      "loss": 2.8887,
+      "step": 31220
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5222345132743363e-06,
+      "loss": 3.1151,
+      "step": 31221
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5221871049304678e-06,
+      "loss": 2.6973,
+      "step": 31222
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5221396965865991e-06,
+      "loss": 2.2687,
+      "step": 31223
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5220922882427308e-06,
+      "loss": 1.7381,
+      "step": 31224
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.522044879898862e-06,
+      "loss": 1.6121,
+      "step": 31225
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5219974715549936e-06,
+      "loss": 2.0316,
+      "step": 31226
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5219500632111253e-06,
+      "loss": 2.1285,
+      "step": 31227
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5219026548672566e-06,
+      "loss": 2.9406,
+      "step": 31228
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.521855246523388e-06,
+      "loss": 2.71,
+      "step": 31229
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5218078381795198e-06,
+      "loss": 2.1551,
+      "step": 31230
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.521760429835651e-06,
+      "loss": 1.793,
+      "step": 31231
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5217130214917826e-06,
+      "loss": 2.556,
+      "step": 31232
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5216656131479143e-06,
+      "loss": 2.4259,
+      "step": 31233
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5216182048040455e-06,
+      "loss": 2.7302,
+      "step": 31234
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.521570796460177e-06,
+      "loss": 1.8789,
+      "step": 31235
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5215233881163087e-06,
+      "loss": 4.6113,
+      "step": 31236
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.52147597977244e-06,
+      "loss": 1.6261,
+      "step": 31237
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5214285714285713e-06,
+      "loss": 1.7342,
+      "step": 31238
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.521381163084703e-06,
+      "loss": 2.8038,
+      "step": 31239
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5213337547408345e-06,
+      "loss": 0.8278,
+      "step": 31240
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5212863463969658e-06,
+      "loss": 1.7285,
+      "step": 31241
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5212389380530975e-06,
+      "loss": 2.4239,
+      "step": 31242
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.521191529709229e-06,
+      "loss": 2.0952,
+      "step": 31243
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5211441213653602e-06,
+      "loss": 3.473,
+      "step": 31244
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5210967130214917e-06,
+      "loss": 1.5922,
+      "step": 31245
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5210493046776234e-06,
+      "loss": 3.3357,
+      "step": 31246
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5210018963337547e-06,
+      "loss": 2.1344,
+      "step": 31247
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5209544879898862e-06,
+      "loss": 3.9232,
+      "step": 31248
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.520907079646018e-06,
+      "loss": 2.4402,
+      "step": 31249
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5208596713021492e-06,
+      "loss": 3.3857,
+      "step": 31250
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5208122629582805e-06,
+      "loss": 2.8058,
+      "step": 31251
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5207648546144122e-06,
+      "loss": 2.6751,
+      "step": 31252
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.5207174462705437e-06,
+      "loss": 3.1213,
+      "step": 31253
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.520670037926675e-06,
+      "loss": 2.1361,
+      "step": 31254
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5206226295828067e-06,
+      "loss": 2.2262,
+      "step": 31255
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5205752212389382e-06,
+      "loss": 3.0025,
+      "step": 31256
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5205278128950694e-06,
+      "loss": 2.0168,
+      "step": 31257
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5204804045512011e-06,
+      "loss": 2.546,
+      "step": 31258
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5204329962073326e-06,
+      "loss": 2.3435,
+      "step": 31259
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.520385587863464e-06,
+      "loss": 2.803,
+      "step": 31260
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5203381795195956e-06,
+      "loss": 2.7662,
+      "step": 31261
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.520290771175727e-06,
+      "loss": 1.3841,
+      "step": 31262
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5202433628318584e-06,
+      "loss": 2.8777,
+      "step": 31263
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5201959544879897e-06,
+      "loss": 2.2026,
+      "step": 31264
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5201485461441214e-06,
+      "loss": 1.3626,
+      "step": 31265
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5201011378002529e-06,
+      "loss": 1.7995,
+      "step": 31266
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5200537294563841e-06,
+      "loss": 3.8497,
+      "step": 31267
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5200063211125158e-06,
+      "loss": 3.7339,
+      "step": 31268
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5199589127686473e-06,
+      "loss": 1.9791,
+      "step": 31269
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5199115044247786e-06,
+      "loss": 3.3564,
+      "step": 31270
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5198640960809103e-06,
+      "loss": 2.5903,
+      "step": 31271
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5198166877370418e-06,
+      "loss": 1.8218,
+      "step": 31272
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.519769279393173e-06,
+      "loss": 1.9505,
+      "step": 31273
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5197218710493048e-06,
+      "loss": 1.2102,
+      "step": 31274
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5196744627054363e-06,
+      "loss": 1.1038,
+      "step": 31275
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5196270543615676e-06,
+      "loss": 1.5092,
+      "step": 31276
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5195796460176993e-06,
+      "loss": 2.6858,
+      "step": 31277
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5195322376738306e-06,
+      "loss": 3.3684,
+      "step": 31278
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.519484829329962e-06,
+      "loss": 2.9379,
+      "step": 31279
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5194374209860938e-06,
+      "loss": 2.8587,
+      "step": 31280
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.519390012642225e-06,
+      "loss": 2.0802,
+      "step": 31281
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5193426042983565e-06,
+      "loss": 1.3221,
+      "step": 31282
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5192951959544882e-06,
+      "loss": 1.4603,
+      "step": 31283
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5192477876106195e-06,
+      "loss": 2.5171,
+      "step": 31284
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.519200379266751e-06,
+      "loss": 0.9315,
+      "step": 31285
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5191529709228823e-06,
+      "loss": 2.3024,
+      "step": 31286
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.519105562579014e-06,
+      "loss": 1.488,
+      "step": 31287
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5190581542351455e-06,
+      "loss": 3.0162,
+      "step": 31288
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5190107458912768e-06,
+      "loss": 2.7354,
+      "step": 31289
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5189633375474085e-06,
+      "loss": 1.7359,
+      "step": 31290
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5189159292035397e-06,
+      "loss": 2.7161,
+      "step": 31291
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5188685208596712e-06,
+      "loss": 2.8035,
+      "step": 31292
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.518821112515803e-06,
+      "loss": 2.069,
+      "step": 31293
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5187737041719342e-06,
+      "loss": 1.3434,
+      "step": 31294
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5187262958280657e-06,
+      "loss": 1.619,
+      "step": 31295
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5186788874841974e-06,
+      "loss": 3.695,
+      "step": 31296
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5186314791403287e-06,
+      "loss": 2.1847,
+      "step": 31297
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5185840707964602e-06,
+      "loss": 3.4406,
+      "step": 31298
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5185366624525919e-06,
+      "loss": 1.9995,
+      "step": 31299
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5184892541087232e-06,
+      "loss": 2.5752,
+      "step": 31300
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5184418457648547e-06,
+      "loss": 1.6069,
+      "step": 31301
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5183944374209864e-06,
+      "loss": 2.3371,
+      "step": 31302
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5183470290771176e-06,
+      "loss": 2.1191,
+      "step": 31303
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.518299620733249e-06,
+      "loss": 2.6782,
+      "step": 31304
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5182522123893806e-06,
+      "loss": 1.7875,
+      "step": 31305
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5182048040455121e-06,
+      "loss": 2.9958,
+      "step": 31306
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5181573957016434e-06,
+      "loss": 2.5091,
+      "step": 31307
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.518109987357775e-06,
+      "loss": 1.4433,
+      "step": 31308
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5180625790139066e-06,
+      "loss": 2.6508,
+      "step": 31309
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5180151706700379e-06,
+      "loss": 1.3179,
+      "step": 31310
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5179677623261694e-06,
+      "loss": 2.4362,
+      "step": 31311
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.517920353982301e-06,
+      "loss": 3.7694,
+      "step": 31312
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5178729456384324e-06,
+      "loss": 2.6006,
+      "step": 31313
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5178255372945639e-06,
+      "loss": 2.9693,
+      "step": 31314
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5177781289506956e-06,
+      "loss": 2.3623,
+      "step": 31315
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.5177307206068268e-06,
+      "loss": 2.159,
+      "step": 31316
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5176833122629581e-06,
+      "loss": 1.9321,
+      "step": 31317
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5176359039190898e-06,
+      "loss": 2.8705,
+      "step": 31318
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5175884955752213e-06,
+      "loss": 2.3057,
+      "step": 31319
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5175410872313526e-06,
+      "loss": 2.7007,
+      "step": 31320
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5174936788874843e-06,
+      "loss": 1.4484,
+      "step": 31321
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5174462705436158e-06,
+      "loss": 3.3944,
+      "step": 31322
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.517398862199747e-06,
+      "loss": 2.4158,
+      "step": 31323
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5173514538558788e-06,
+      "loss": 2.3281,
+      "step": 31324
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5173040455120103e-06,
+      "loss": 2.6652,
+      "step": 31325
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5172566371681415e-06,
+      "loss": 1.9628,
+      "step": 31326
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5172092288242732e-06,
+      "loss": 1.8577,
+      "step": 31327
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5171618204804047e-06,
+      "loss": 2.8634,
+      "step": 31328
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.517114412136536e-06,
+      "loss": 3.0914,
+      "step": 31329
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5170670037926673e-06,
+      "loss": 2.3737,
+      "step": 31330
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.517019595448799e-06,
+      "loss": 3.014,
+      "step": 31331
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5169721871049305e-06,
+      "loss": 1.4452,
+      "step": 31332
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5169247787610618e-06,
+      "loss": 1.7059,
+      "step": 31333
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5168773704171935e-06,
+      "loss": 0.6032,
+      "step": 31334
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.516829962073325e-06,
+      "loss": 1.6249,
+      "step": 31335
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5167825537294563e-06,
+      "loss": 2.9913,
+      "step": 31336
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.516735145385588e-06,
+      "loss": 1.9916,
+      "step": 31337
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5166877370417195e-06,
+      "loss": 2.4722,
+      "step": 31338
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5166403286978507e-06,
+      "loss": 2.2833,
+      "step": 31339
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5165929203539824e-06,
+      "loss": 2.4189,
+      "step": 31340
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.516545512010114e-06,
+      "loss": 2.0568,
+      "step": 31341
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5164981036662452e-06,
+      "loss": 2.4617,
+      "step": 31342
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.516450695322377e-06,
+      "loss": 2.2568,
+      "step": 31343
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5164032869785082e-06,
+      "loss": 2.7493,
+      "step": 31344
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5163558786346397e-06,
+      "loss": 2.1707,
+      "step": 31345
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5163084702907714e-06,
+      "loss": 1.7672,
+      "step": 31346
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5162610619469027e-06,
+      "loss": 1.6122,
+      "step": 31347
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5162136536030342e-06,
+      "loss": 1.4814,
+      "step": 31348
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5161662452591659e-06,
+      "loss": 1.4215,
+      "step": 31349
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5161188369152971e-06,
+      "loss": 3.2711,
+      "step": 31350
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5160714285714286e-06,
+      "loss": 1.4601,
+      "step": 31351
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.51602402022756e-06,
+      "loss": 3.4174,
+      "step": 31352
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5159766118836916e-06,
+      "loss": 3.4116,
+      "step": 31353
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.515929203539823e-06,
+      "loss": 3.1711,
+      "step": 31354
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5158817951959544e-06,
+      "loss": 2.83,
+      "step": 31355
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.515834386852086e-06,
+      "loss": 2.0051,
+      "step": 31356
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5157869785082174e-06,
+      "loss": 2.7558,
+      "step": 31357
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5157395701643489e-06,
+      "loss": 1.7656,
+      "step": 31358
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5156921618204806e-06,
+      "loss": 2.8315,
+      "step": 31359
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5156447534766119e-06,
+      "loss": 1.7518,
+      "step": 31360
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5155973451327433e-06,
+      "loss": 0.8454,
+      "step": 31361
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.515549936788875e-06,
+      "loss": 1.4131,
+      "step": 31362
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5155025284450063e-06,
+      "loss": 1.2532,
+      "step": 31363
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5154551201011378e-06,
+      "loss": 1.3131,
+      "step": 31364
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5154077117572695e-06,
+      "loss": 2.3773,
+      "step": 31365
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5153603034134008e-06,
+      "loss": 1.6586,
+      "step": 31366
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.515312895069532e-06,
+      "loss": 3.3521,
+      "step": 31367
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5152654867256638e-06,
+      "loss": 1.4593,
+      "step": 31368
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5152180783817953e-06,
+      "loss": 2.2114,
+      "step": 31369
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5151706700379266e-06,
+      "loss": 2.2712,
+      "step": 31370
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5151232616940583e-06,
+      "loss": 1.5117,
+      "step": 31371
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5150758533501898e-06,
+      "loss": 2.4904,
+      "step": 31372
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.515028445006321e-06,
+      "loss": 2.2776,
+      "step": 31373
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5149810366624525e-06,
+      "loss": 2.1677,
+      "step": 31374
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5149336283185842e-06,
+      "loss": 1.844,
+      "step": 31375
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5148862199747155e-06,
+      "loss": 2.3327,
+      "step": 31376
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.514838811630847e-06,
+      "loss": 2.737,
+      "step": 31377
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5147914032869787e-06,
+      "loss": 1.7939,
+      "step": 31378
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.51474399494311e-06,
+      "loss": 2.3003,
+      "step": 31379
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.5146965865992413e-06,
+      "loss": 1.8357,
+      "step": 31380
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.514649178255373e-06,
+      "loss": 1.6737,
+      "step": 31381
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5146017699115045e-06,
+      "loss": 4.2587,
+      "step": 31382
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5145543615676358e-06,
+      "loss": 3.1728,
+      "step": 31383
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5145069532237675e-06,
+      "loss": 2.2886,
+      "step": 31384
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.514459544879899e-06,
+      "loss": 2.1527,
+      "step": 31385
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5144121365360302e-06,
+      "loss": 1.1345,
+      "step": 31386
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.514364728192162e-06,
+      "loss": 2.589,
+      "step": 31387
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5143173198482934e-06,
+      "loss": 2.4297,
+      "step": 31388
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5142699115044247e-06,
+      "loss": 2.1163,
+      "step": 31389
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5142225031605564e-06,
+      "loss": 1.6062,
+      "step": 31390
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.514175094816688e-06,
+      "loss": 1.6415,
+      "step": 31391
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5141276864728192e-06,
+      "loss": 2.6576,
+      "step": 31392
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5140802781289509e-06,
+      "loss": 2.3672,
+      "step": 31393
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5140328697850822e-06,
+      "loss": 1.6373,
+      "step": 31394
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5139854614412137e-06,
+      "loss": 3.1925,
+      "step": 31395
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.513938053097345e-06,
+      "loss": 1.9332,
+      "step": 31396
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5138906447534766e-06,
+      "loss": 3.2471,
+      "step": 31397
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5138432364096081e-06,
+      "loss": 2.7213,
+      "step": 31398
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5137958280657394e-06,
+      "loss": 1.6735,
+      "step": 31399
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5137484197218711e-06,
+      "loss": 3.6852,
+      "step": 31400
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5137010113780026e-06,
+      "loss": 2.9711,
+      "step": 31401
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5136536030341339e-06,
+      "loss": 1.2926,
+      "step": 31402
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5136061946902656e-06,
+      "loss": 2.07,
+      "step": 31403
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.513558786346397e-06,
+      "loss": 2.5296,
+      "step": 31404
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5135113780025284e-06,
+      "loss": 2.8837,
+      "step": 31405
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.51346396965866e-06,
+      "loss": 2.4035,
+      "step": 31406
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5134165613147914e-06,
+      "loss": 1.7646,
+      "step": 31407
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5133691529709228e-06,
+      "loss": 1.9655,
+      "step": 31408
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5133217446270545e-06,
+      "loss": 3.3393,
+      "step": 31409
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5132743362831858e-06,
+      "loss": 2.7817,
+      "step": 31410
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5132269279393173e-06,
+      "loss": 2.4723,
+      "step": 31411
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.513179519595449e-06,
+      "loss": 1.7497,
+      "step": 31412
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5131321112515803e-06,
+      "loss": 2.5945,
+      "step": 31413
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5130847029077118e-06,
+      "loss": 2.5251,
+      "step": 31414
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5130372945638435e-06,
+      "loss": 2.0976,
+      "step": 31415
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5129898862199748e-06,
+      "loss": 2.6417,
+      "step": 31416
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5129424778761063e-06,
+      "loss": 2.3442,
+      "step": 31417
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5128950695322376e-06,
+      "loss": 2.4437,
+      "step": 31418
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5128476611883693e-06,
+      "loss": 2.3137,
+      "step": 31419
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5128002528445005e-06,
+      "loss": 1.9084,
+      "step": 31420
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.512752844500632e-06,
+      "loss": 2.2504,
+      "step": 31421
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5127054361567637e-06,
+      "loss": 2.3137,
+      "step": 31422
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.512658027812895e-06,
+      "loss": 2.8908,
+      "step": 31423
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5126106194690265e-06,
+      "loss": 2.3154,
+      "step": 31424
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5125632111251582e-06,
+      "loss": 2.2176,
+      "step": 31425
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5125158027812895e-06,
+      "loss": 1.5585,
+      "step": 31426
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.512468394437421e-06,
+      "loss": 1.7855,
+      "step": 31427
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5124209860935527e-06,
+      "loss": 1.9278,
+      "step": 31428
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.512373577749684e-06,
+      "loss": 4.0217,
+      "step": 31429
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5123261694058155e-06,
+      "loss": 3.2426,
+      "step": 31430
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5122787610619472e-06,
+      "loss": 2.0893,
+      "step": 31431
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5122313527180784e-06,
+      "loss": 1.6891,
+      "step": 31432
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5121839443742097e-06,
+      "loss": 3.6178,
+      "step": 31433
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5121365360303414e-06,
+      "loss": 1.4603,
+      "step": 31434
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.512089127686473e-06,
+      "loss": 2.1124,
+      "step": 31435
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5120417193426042e-06,
+      "loss": 3.0462,
+      "step": 31436
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.511994310998736e-06,
+      "loss": 2.6722,
+      "step": 31437
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5119469026548674e-06,
+      "loss": 2.9309,
+      "step": 31438
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5118994943109987e-06,
+      "loss": 3.2989,
+      "step": 31439
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5118520859671302e-06,
+      "loss": 2.7289,
+      "step": 31440
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5118046776232619e-06,
+      "loss": 2.9185,
+      "step": 31441
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5117572692793932e-06,
+      "loss": 0.616,
+      "step": 31442
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.5117098609355246e-06,
+      "loss": 1.3498,
+      "step": 31443
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5116624525916564e-06,
+      "loss": 2.464,
+      "step": 31444
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5116150442477876e-06,
+      "loss": 2.3535,
+      "step": 31445
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.511567635903919e-06,
+      "loss": 2.104,
+      "step": 31446
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5115202275600506e-06,
+      "loss": 2.9189,
+      "step": 31447
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5114728192161821e-06,
+      "loss": 3.3634,
+      "step": 31448
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5114254108723134e-06,
+      "loss": 1.3756,
+      "step": 31449
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.511378002528445e-06,
+      "loss": 3.3486,
+      "step": 31450
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5113305941845766e-06,
+      "loss": 2.8525,
+      "step": 31451
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5112831858407079e-06,
+      "loss": 1.2704,
+      "step": 31452
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5112357774968396e-06,
+      "loss": 1.7398,
+      "step": 31453
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.511188369152971e-06,
+      "loss": 2.6644,
+      "step": 31454
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5111409608091023e-06,
+      "loss": 2.501,
+      "step": 31455
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.511093552465234e-06,
+      "loss": 2.1836,
+      "step": 31456
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5110461441213655e-06,
+      "loss": 1.9809,
+      "step": 31457
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5109987357774968e-06,
+      "loss": 1.1165,
+      "step": 31458
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5109513274336285e-06,
+      "loss": 3.0986,
+      "step": 31459
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5109039190897598e-06,
+      "loss": 2.1223,
+      "step": 31460
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5108565107458913e-06,
+      "loss": 2.1052,
+      "step": 31461
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5108091024020226e-06,
+      "loss": 1.8504,
+      "step": 31462
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5107616940581543e-06,
+      "loss": 2.4655,
+      "step": 31463
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5107142857142858e-06,
+      "loss": 1.8818,
+      "step": 31464
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.510666877370417e-06,
+      "loss": 1.8513,
+      "step": 31465
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5106194690265488e-06,
+      "loss": 1.6897,
+      "step": 31466
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5105720606826802e-06,
+      "loss": 2.3292,
+      "step": 31467
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5105246523388115e-06,
+      "loss": 2.3333,
+      "step": 31468
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5104772439949432e-06,
+      "loss": 2.1992,
+      "step": 31469
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5104298356510747e-06,
+      "loss": 2.0498,
+      "step": 31470
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.510382427307206e-06,
+      "loss": 1.5046,
+      "step": 31471
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5103350189633377e-06,
+      "loss": 2.3879,
+      "step": 31472
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.510287610619469e-06,
+      "loss": 1.9267,
+      "step": 31473
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5102402022756005e-06,
+      "loss": 2.2802,
+      "step": 31474
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5101927939317322e-06,
+      "loss": 1.5271,
+      "step": 31475
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5101453855878635e-06,
+      "loss": 3.3304,
+      "step": 31476
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.510097977243995e-06,
+      "loss": 2.0028,
+      "step": 31477
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5100505689001267e-06,
+      "loss": 0.3267,
+      "step": 31478
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.510003160556258e-06,
+      "loss": 1.8347,
+      "step": 31479
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5099557522123894e-06,
+      "loss": 2.4101,
+      "step": 31480
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5099083438685211e-06,
+      "loss": 3.6455,
+      "step": 31481
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5098609355246524e-06,
+      "loss": 2.3502,
+      "step": 31482
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5098135271807837e-06,
+      "loss": 2.5196,
+      "step": 31483
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5097661188369152e-06,
+      "loss": 0.8825,
+      "step": 31484
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.509718710493047e-06,
+      "loss": 2.1964,
+      "step": 31485
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5096713021491782e-06,
+      "loss": 1.4182,
+      "step": 31486
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5096238938053097e-06,
+      "loss": 1.4516,
+      "step": 31487
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5095764854614414e-06,
+      "loss": 1.6794,
+      "step": 31488
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5095290771175727e-06,
+      "loss": 3.3553,
+      "step": 31489
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5094816687737041e-06,
+      "loss": 2.8576,
+      "step": 31490
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5094342604298358e-06,
+      "loss": 2.4518,
+      "step": 31491
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5093868520859671e-06,
+      "loss": 1.8912,
+      "step": 31492
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5093394437420986e-06,
+      "loss": 2.1383,
+      "step": 31493
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5092920353982303e-06,
+      "loss": 1.033,
+      "step": 31494
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5092446270543616e-06,
+      "loss": 2.3797,
+      "step": 31495
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5091972187104929e-06,
+      "loss": 1.3415,
+      "step": 31496
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5091498103666246e-06,
+      "loss": 1.3802,
+      "step": 31497
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.509102402022756e-06,
+      "loss": 1.6484,
+      "step": 31498
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5090549936788874e-06,
+      "loss": 2.8925,
+      "step": 31499
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.509007585335019e-06,
+      "loss": 3.7758,
+      "step": 31500
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5089601769911506e-06,
+      "loss": 2.5807,
+      "step": 31501
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5089127686472818e-06,
+      "loss": 2.1284,
+      "step": 31502
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5088653603034135e-06,
+      "loss": 1.2885,
+      "step": 31503
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.508817951959545e-06,
+      "loss": 2.8158,
+      "step": 31504
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5087705436156763e-06,
+      "loss": 0.9919,
+      "step": 31505
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.5087231352718078e-06,
+      "loss": 2.3344,
+      "step": 31506
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5086757269279395e-06,
+      "loss": 1.5594,
+      "step": 31507
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5086283185840708e-06,
+      "loss": 3.6292,
+      "step": 31508
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.508580910240202e-06,
+      "loss": 3.0777,
+      "step": 31509
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5085335018963338e-06,
+      "loss": 1.5349,
+      "step": 31510
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5084860935524653e-06,
+      "loss": 4.1062,
+      "step": 31511
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5084386852085965e-06,
+      "loss": 0.7524,
+      "step": 31512
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5083912768647283e-06,
+      "loss": 0.3423,
+      "step": 31513
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5083438685208597e-06,
+      "loss": 2.3398,
+      "step": 31514
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.508296460176991e-06,
+      "loss": 1.8037,
+      "step": 31515
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5082490518331227e-06,
+      "loss": 2.263,
+      "step": 31516
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5082016434892542e-06,
+      "loss": 4.5899,
+      "step": 31517
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5081542351453855e-06,
+      "loss": 2.3268,
+      "step": 31518
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5081068268015172e-06,
+      "loss": 3.2361,
+      "step": 31519
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5080594184576487e-06,
+      "loss": 3.0223,
+      "step": 31520
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.50801201011378e-06,
+      "loss": 2.5972,
+      "step": 31521
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5079646017699117e-06,
+      "loss": 1.6812,
+      "step": 31522
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.507917193426043e-06,
+      "loss": 2.3069,
+      "step": 31523
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5078697850821745e-06,
+      "loss": 1.1288,
+      "step": 31524
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5078223767383062e-06,
+      "loss": 3.5058,
+      "step": 31525
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5077749683944374e-06,
+      "loss": 3.6375,
+      "step": 31526
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.507727560050569e-06,
+      "loss": 1.7988,
+      "step": 31527
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5076801517067002e-06,
+      "loss": 2.7864,
+      "step": 31528
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.507632743362832e-06,
+      "loss": 3.0855,
+      "step": 31529
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5075853350189634e-06,
+      "loss": 1.7772,
+      "step": 31530
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5075379266750947e-06,
+      "loss": 2.7014,
+      "step": 31531
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5074905183312264e-06,
+      "loss": 2.2403,
+      "step": 31532
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5074431099873579e-06,
+      "loss": 1.4546,
+      "step": 31533
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5073957016434892e-06,
+      "loss": 1.1397,
+      "step": 31534
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5073482932996209e-06,
+      "loss": 2.0194,
+      "step": 31535
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5073008849557521e-06,
+      "loss": 3.4697,
+      "step": 31536
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5072534766118836e-06,
+      "loss": 3.2234,
+      "step": 31537
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5072060682680153e-06,
+      "loss": 2.0262,
+      "step": 31538
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5071586599241466e-06,
+      "loss": 2.0876,
+      "step": 31539
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5071112515802781e-06,
+      "loss": 1.6108,
+      "step": 31540
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5070638432364098e-06,
+      "loss": 2.3069,
+      "step": 31541
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.507016434892541e-06,
+      "loss": 2.1357,
+      "step": 31542
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5069690265486726e-06,
+      "loss": 2.9793,
+      "step": 31543
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5069216182048043e-06,
+      "loss": 2.3188,
+      "step": 31544
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5068742098609356e-06,
+      "loss": 2.1853,
+      "step": 31545
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.506826801517067e-06,
+      "loss": 3.2044,
+      "step": 31546
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5067793931731988e-06,
+      "loss": 1.9157,
+      "step": 31547
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.50673198482933e-06,
+      "loss": 3.1449,
+      "step": 31548
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5066845764854613e-06,
+      "loss": 1.6587,
+      "step": 31549
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5066371681415928e-06,
+      "loss": 2.5224,
+      "step": 31550
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5065897597977245e-06,
+      "loss": 3.21,
+      "step": 31551
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5065423514538558e-06,
+      "loss": 0.9645,
+      "step": 31552
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5064949431099873e-06,
+      "loss": 2.3501,
+      "step": 31553
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.506447534766119e-06,
+      "loss": 2.742,
+      "step": 31554
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5064001264222503e-06,
+      "loss": 2.8066,
+      "step": 31555
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5063527180783818e-06,
+      "loss": 3.2916,
+      "step": 31556
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5063053097345135e-06,
+      "loss": 2.9413,
+      "step": 31557
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5062579013906448e-06,
+      "loss": 1.7597,
+      "step": 31558
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5062104930467763e-06,
+      "loss": 2.3233,
+      "step": 31559
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.506163084702908e-06,
+      "loss": 2.2222,
+      "step": 31560
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5061156763590392e-06,
+      "loss": 2.3765,
+      "step": 31561
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5060682680151705e-06,
+      "loss": 2.6954,
+      "step": 31562
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5060208596713022e-06,
+      "loss": 1.2417,
+      "step": 31563
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5059734513274337e-06,
+      "loss": 1.8327,
+      "step": 31564
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.505926042983565e-06,
+      "loss": 3.0318,
+      "step": 31565
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5058786346396967e-06,
+      "loss": 2.469,
+      "step": 31566
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5058312262958282e-06,
+      "loss": 2.6184,
+      "step": 31567
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5057838179519595e-06,
+      "loss": 2.3081,
+      "step": 31568
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5057364096080912e-06,
+      "loss": 3.4685,
+      "step": 31569
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.5056890012642227e-06,
+      "loss": 3.08,
+      "step": 31570
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.505641592920354e-06,
+      "loss": 1.6647,
+      "step": 31571
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5055941845764854e-06,
+      "loss": 1.9848,
+      "step": 31572
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5055467762326171e-06,
+      "loss": 2.528,
+      "step": 31573
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5054993678887484e-06,
+      "loss": 2.7303,
+      "step": 31574
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5054519595448797e-06,
+      "loss": 1.8172,
+      "step": 31575
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5054045512010114e-06,
+      "loss": 2.794,
+      "step": 31576
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.505357142857143e-06,
+      "loss": 2.6262,
+      "step": 31577
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5053097345132742e-06,
+      "loss": 1.9493,
+      "step": 31578
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5052623261694059e-06,
+      "loss": 2.9215,
+      "step": 31579
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5052149178255374e-06,
+      "loss": 2.4212,
+      "step": 31580
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5051675094816687e-06,
+      "loss": 2.0927,
+      "step": 31581
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5051201011378004e-06,
+      "loss": 2.1898,
+      "step": 31582
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5050726927939319e-06,
+      "loss": 1.6711,
+      "step": 31583
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5050252844500631e-06,
+      "loss": 1.836,
+      "step": 31584
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5049778761061948e-06,
+      "loss": 1.7095,
+      "step": 31585
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5049304677623263e-06,
+      "loss": 2.3826,
+      "step": 31586
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5048830594184576e-06,
+      "loss": 3.4269,
+      "step": 31587
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5048356510745893e-06,
+      "loss": 2.6668,
+      "step": 31588
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5047882427307206e-06,
+      "loss": 1.9573,
+      "step": 31589
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.504740834386852e-06,
+      "loss": 2.759,
+      "step": 31590
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5046934260429838e-06,
+      "loss": 3.2642,
+      "step": 31591
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.504646017699115e-06,
+      "loss": 1.4095,
+      "step": 31592
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5045986093552466e-06,
+      "loss": 2.6957,
+      "step": 31593
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5045512010113778e-06,
+      "loss": 2.6191,
+      "step": 31594
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5045037926675096e-06,
+      "loss": 3.1654,
+      "step": 31595
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.504456384323641e-06,
+      "loss": 3.1417,
+      "step": 31596
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5044089759797723e-06,
+      "loss": 2.2235,
+      "step": 31597
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.504361567635904e-06,
+      "loss": 2.76,
+      "step": 31598
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5043141592920355e-06,
+      "loss": 2.5328,
+      "step": 31599
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5042667509481668e-06,
+      "loss": 2.6146,
+      "step": 31600
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5042193426042985e-06,
+      "loss": 4.034,
+      "step": 31601
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5041719342604298e-06,
+      "loss": 2.1913,
+      "step": 31602
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5041245259165613e-06,
+      "loss": 1.0348,
+      "step": 31603
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.504077117572693e-06,
+      "loss": 0.8355,
+      "step": 31604
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5040297092288243e-06,
+      "loss": 0.7989,
+      "step": 31605
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5039823008849558e-06,
+      "loss": 1.1768,
+      "step": 31606
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5039348925410875e-06,
+      "loss": 2.1608,
+      "step": 31607
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5038874841972187e-06,
+      "loss": 2.6406,
+      "step": 31608
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5038400758533502e-06,
+      "loss": 1.7056,
+      "step": 31609
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.503792667509482e-06,
+      "loss": 2.8208,
+      "step": 31610
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5037452591656132e-06,
+      "loss": 1.484,
+      "step": 31611
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5036978508217447e-06,
+      "loss": 2.1907,
+      "step": 31612
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5036504424778764e-06,
+      "loss": 3.4646,
+      "step": 31613
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5036030341340077e-06,
+      "loss": 1.9834,
+      "step": 31614
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.503555625790139e-06,
+      "loss": 3.2009,
+      "step": 31615
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5035082174462705e-06,
+      "loss": 2.8857,
+      "step": 31616
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5034608091024022e-06,
+      "loss": 1.581,
+      "step": 31617
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5034134007585334e-06,
+      "loss": 2.1683,
+      "step": 31618
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.503365992414665e-06,
+      "loss": 2.3549,
+      "step": 31619
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5033185840707966e-06,
+      "loss": 2.1153,
+      "step": 31620
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.503271175726928e-06,
+      "loss": 2.3237,
+      "step": 31621
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5032237673830594e-06,
+      "loss": 1.2331,
+      "step": 31622
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5031763590391911e-06,
+      "loss": 1.8373,
+      "step": 31623
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5031289506953224e-06,
+      "loss": 3.0255,
+      "step": 31624
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5030815423514537e-06,
+      "loss": 2.6075,
+      "step": 31625
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5030341340075856e-06,
+      "loss": 2.5117,
+      "step": 31626
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5029867256637169e-06,
+      "loss": 1.0319,
+      "step": 31627
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5029393173198482e-06,
+      "loss": 3.0736,
+      "step": 31628
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5028919089759799e-06,
+      "loss": 2.5189,
+      "step": 31629
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5028445006321114e-06,
+      "loss": 3.1705,
+      "step": 31630
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5027970922882426e-06,
+      "loss": 2.6721,
+      "step": 31631
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5027496839443743e-06,
+      "loss": 1.8443,
+      "step": 31632
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.5027022756005058e-06,
+      "loss": 1.7248,
+      "step": 31633
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5026548672566371e-06,
+      "loss": 2.6174,
+      "step": 31634
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5026074589127688e-06,
+      "loss": 1.6986,
+      "step": 31635
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5025600505689003e-06,
+      "loss": 1.6757,
+      "step": 31636
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5025126422250316e-06,
+      "loss": 2.4681,
+      "step": 31637
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5024652338811629e-06,
+      "loss": 2.8378,
+      "step": 31638
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5024178255372946e-06,
+      "loss": 2.0552,
+      "step": 31639
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.502370417193426e-06,
+      "loss": 2.2123,
+      "step": 31640
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5023230088495573e-06,
+      "loss": 1.4082,
+      "step": 31641
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.502275600505689e-06,
+      "loss": 2.9733,
+      "step": 31642
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5022281921618205e-06,
+      "loss": 2.1164,
+      "step": 31643
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5021807838179518e-06,
+      "loss": 2.2934,
+      "step": 31644
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5021333754740835e-06,
+      "loss": 2.7177,
+      "step": 31645
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.502085967130215e-06,
+      "loss": 1.678,
+      "step": 31646
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5020385587863463e-06,
+      "loss": 2.7321,
+      "step": 31647
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.501991150442478e-06,
+      "loss": 0.9388,
+      "step": 31648
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5019437420986095e-06,
+      "loss": 2.2294,
+      "step": 31649
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5018963337547408e-06,
+      "loss": 3.0623,
+      "step": 31650
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5018489254108725e-06,
+      "loss": 2.1655,
+      "step": 31651
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5018015170670038e-06,
+      "loss": 2.6498,
+      "step": 31652
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5017541087231353e-06,
+      "loss": 1.8644,
+      "step": 31653
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.501706700379267e-06,
+      "loss": 2.5753,
+      "step": 31654
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5016592920353982e-06,
+      "loss": 2.1637,
+      "step": 31655
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5016118836915297e-06,
+      "loss": 1.5372,
+      "step": 31656
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5015644753476614e-06,
+      "loss": 0.8303,
+      "step": 31657
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5015170670037927e-06,
+      "loss": 0.8952,
+      "step": 31658
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5014696586599242e-06,
+      "loss": 2.7189,
+      "step": 31659
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5014222503160555e-06,
+      "loss": 1.8011,
+      "step": 31660
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5013748419721872e-06,
+      "loss": 2.3844,
+      "step": 31661
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5013274336283187e-06,
+      "loss": 2.0723,
+      "step": 31662
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.50128002528445e-06,
+      "loss": 2.7909,
+      "step": 31663
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5012326169405817e-06,
+      "loss": 1.7958,
+      "step": 31664
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.501185208596713e-06,
+      "loss": 1.3762,
+      "step": 31665
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.47104795737122557,
+      "eval_loss": 2.237438917160034,
+      "eval_runtime": 156.5742,
+      "eval_samples_per_second": 17.979,
+      "eval_steps_per_second": 4.496,
+      "step": 31665
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5011378002528444e-06,
+      "loss": 1.4782,
+      "step": 31666
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5010903919089761e-06,
+      "loss": 2.104,
+      "step": 31667
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5010429835651074e-06,
+      "loss": 2.8056,
+      "step": 31668
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.500995575221239e-06,
+      "loss": 1.7242,
+      "step": 31669
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5009481668773706e-06,
+      "loss": 1.507,
+      "step": 31670
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.500900758533502e-06,
+      "loss": 2.5773,
+      "step": 31671
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5008533501896334e-06,
+      "loss": 1.9899,
+      "step": 31672
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.500805941845765e-06,
+      "loss": 2.4585,
+      "step": 31673
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5007585335018964e-06,
+      "loss": 3.0926,
+      "step": 31674
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5007111251580279e-06,
+      "loss": 3.6767,
+      "step": 31675
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5006637168141596e-06,
+      "loss": 2.9931,
+      "step": 31676
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5006163084702909e-06,
+      "loss": 2.3249,
+      "step": 31677
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5005689001264221e-06,
+      "loss": 2.636,
+      "step": 31678
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5005214917825538e-06,
+      "loss": 1.9469,
+      "step": 31679
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5004740834386853e-06,
+      "loss": 1.9129,
+      "step": 31680
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5004266750948166e-06,
+      "loss": 1.4488,
+      "step": 31681
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.500379266750948e-06,
+      "loss": 2.4576,
+      "step": 31682
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5003318584070798e-06,
+      "loss": 1.5325,
+      "step": 31683
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.500284450063211e-06,
+      "loss": 0.9679,
+      "step": 31684
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5002370417193426e-06,
+      "loss": 2.3617,
+      "step": 31685
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5001896333754743e-06,
+      "loss": 3.1579,
+      "step": 31686
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5001422250316056e-06,
+      "loss": 1.9123,
+      "step": 31687
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.500094816687737e-06,
+      "loss": 2.118,
+      "step": 31688
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5000474083438688e-06,
+      "loss": 2.8642,
+      "step": 31689
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.5e-06,
+      "loss": 3.203,
+      "step": 31690
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.4999525916561315e-06,
+      "loss": 1.2973,
+      "step": 31691
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.499905183312263e-06,
+      "loss": 3.8398,
+      "step": 31692
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.4998577749683945e-06,
+      "loss": 1.702,
+      "step": 31693
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.499810366624526e-06,
+      "loss": 2.0191,
+      "step": 31694
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.4997629582806575e-06,
+      "loss": 2.4088,
+      "step": 31695
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.499715549936789e-06,
+      "loss": 3.1864,
+      "step": 31696
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4996681415929203e-06,
+      "loss": 1.5636,
+      "step": 31697
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4996207332490518e-06,
+      "loss": 2.9346,
+      "step": 31698
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4995733249051835e-06,
+      "loss": 1.8923,
+      "step": 31699
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4995259165613147e-06,
+      "loss": 2.0983,
+      "step": 31700
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4994785082174462e-06,
+      "loss": 2.5603,
+      "step": 31701
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.499431099873578e-06,
+      "loss": 3.0606,
+      "step": 31702
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4993836915297092e-06,
+      "loss": 2.5514,
+      "step": 31703
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4993362831858407e-06,
+      "loss": 2.7772,
+      "step": 31704
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4992888748419722e-06,
+      "loss": 2.6218,
+      "step": 31705
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4992414664981037e-06,
+      "loss": 1.6874,
+      "step": 31706
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4991940581542352e-06,
+      "loss": 1.9988,
+      "step": 31707
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4991466498103667e-06,
+      "loss": 1.6478,
+      "step": 31708
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4990992414664982e-06,
+      "loss": 2.7454,
+      "step": 31709
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4990518331226297e-06,
+      "loss": 2.3845,
+      "step": 31710
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.499004424778761e-06,
+      "loss": 1.7432,
+      "step": 31711
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4989570164348927e-06,
+      "loss": 1.3729,
+      "step": 31712
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4989096080910241e-06,
+      "loss": 2.9403,
+      "step": 31713
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4988621997471554e-06,
+      "loss": 2.1825,
+      "step": 31714
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4988147914032871e-06,
+      "loss": 2.8463,
+      "step": 31715
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4987673830594186e-06,
+      "loss": 1.6205,
+      "step": 31716
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.49871997471555e-06,
+      "loss": 2.6565,
+      "step": 31717
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4986725663716814e-06,
+      "loss": 1.9983,
+      "step": 31718
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4986251580278129e-06,
+      "loss": 2.5143,
+      "step": 31719
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4985777496839444e-06,
+      "loss": 1.9692,
+      "step": 31720
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4985303413400759e-06,
+      "loss": 3.2899,
+      "step": 31721
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4984829329962074e-06,
+      "loss": 3.3879,
+      "step": 31722
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4984355246523389e-06,
+      "loss": 2.4021,
+      "step": 31723
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4983881163084703e-06,
+      "loss": 2.5404,
+      "step": 31724
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4983407079646018e-06,
+      "loss": 3.6605,
+      "step": 31725
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4982932996207333e-06,
+      "loss": 2.242,
+      "step": 31726
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4982458912768648e-06,
+      "loss": 2.3279,
+      "step": 31727
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4981984829329963e-06,
+      "loss": 2.7649,
+      "step": 31728
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4981510745891278e-06,
+      "loss": 2.2219,
+      "step": 31729
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.498103666245259e-06,
+      "loss": 2.1429,
+      "step": 31730
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4980562579013906e-06,
+      "loss": 1.5025,
+      "step": 31731
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4980088495575223e-06,
+      "loss": 2.1971,
+      "step": 31732
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4979614412136536e-06,
+      "loss": 2.2415,
+      "step": 31733
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497914032869785e-06,
+      "loss": 2.4425,
+      "step": 31734
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4978666245259168e-06,
+      "loss": 2.9239,
+      "step": 31735
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497819216182048e-06,
+      "loss": 1.3941,
+      "step": 31736
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4977718078381795e-06,
+      "loss": 2.8682,
+      "step": 31737
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497724399494311e-06,
+      "loss": 2.1134,
+      "step": 31738
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4976769911504425e-06,
+      "loss": 3.0785,
+      "step": 31739
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497629582806574e-06,
+      "loss": 1.5544,
+      "step": 31740
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4975821744627055e-06,
+      "loss": 2.6347,
+      "step": 31741
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497534766118837e-06,
+      "loss": 2.5418,
+      "step": 31742
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4974873577749685e-06,
+      "loss": 3.3602,
+      "step": 31743
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4974399494310998e-06,
+      "loss": 3.7753,
+      "step": 31744
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4973925410872315e-06,
+      "loss": 2.1945,
+      "step": 31745
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497345132743363e-06,
+      "loss": 1.6025,
+      "step": 31746
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4972977243994942e-06,
+      "loss": 2.2687,
+      "step": 31747
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.497250316055626e-06,
+      "loss": 1.7636,
+      "step": 31748
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4972029077117574e-06,
+      "loss": 1.8434,
+      "step": 31749
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4971554993678887e-06,
+      "loss": 2.9693,
+      "step": 31750
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4971080910240202e-06,
+      "loss": 2.8163,
+      "step": 31751
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4970606826801517e-06,
+      "loss": 0.8249,
+      "step": 31752
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4970132743362832e-06,
+      "loss": 2.7969,
+      "step": 31753
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4969658659924147e-06,
+      "loss": 2.5677,
+      "step": 31754
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4969184576485462e-06,
+      "loss": 3.3767,
+      "step": 31755
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4968710493046777e-06,
+      "loss": 2.7526,
+      "step": 31756
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4968236409608092e-06,
+      "loss": 2.4437,
+      "step": 31757
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4967762326169407e-06,
+      "loss": 2.1529,
+      "step": 31758
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 1.4967288242730722e-06,
+      "loss": 3.4079,
+      "step": 31759
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4966814159292036e-06,
+      "loss": 2.1833,
+      "step": 31760
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.496634007585335e-06,
+      "loss": 2.7884,
+      "step": 31761
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4965865992414666e-06,
+      "loss": 3.6353,
+      "step": 31762
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.496539190897598e-06,
+      "loss": 1.5119,
+      "step": 31763
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4964917825537294e-06,
+      "loss": 1.224,
+      "step": 31764
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.496444374209861e-06,
+      "loss": 2.8384,
+      "step": 31765
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4963969658659924e-06,
+      "loss": 3.5894,
+      "step": 31766
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4963495575221239e-06,
+      "loss": 1.9126,
+      "step": 31767
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4963021491782554e-06,
+      "loss": 2.7913,
+      "step": 31768
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4962547408343869e-06,
+      "loss": 2.9367,
+      "step": 31769
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4962073324905184e-06,
+      "loss": 1.8312,
+      "step": 31770
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4961599241466498e-06,
+      "loss": 3.2583,
+      "step": 31771
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4961125158027813e-06,
+      "loss": 2.7044,
+      "step": 31772
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4960651074589128e-06,
+      "loss": 0.5709,
+      "step": 31773
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4960176991150441e-06,
+      "loss": 2.3056,
+      "step": 31774
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4959702907711758e-06,
+      "loss": 2.5833,
+      "step": 31775
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4959228824273073e-06,
+      "loss": 1.6391,
+      "step": 31776
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4958754740834386e-06,
+      "loss": 2.3264,
+      "step": 31777
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4958280657395703e-06,
+      "loss": 2.4737,
+      "step": 31778
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4957806573957018e-06,
+      "loss": 2.8113,
+      "step": 31779
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.495733249051833e-06,
+      "loss": 2.0091,
+      "step": 31780
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4956858407079646e-06,
+      "loss": 2.8569,
+      "step": 31781
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4956384323640963e-06,
+      "loss": 2.4467,
+      "step": 31782
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4955910240202275e-06,
+      "loss": 1.7311,
+      "step": 31783
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.495543615676359e-06,
+      "loss": 1.8567,
+      "step": 31784
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4954962073324905e-06,
+      "loss": 2.1192,
+      "step": 31785
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.495448798988622e-06,
+      "loss": 1.8958,
+      "step": 31786
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4954013906447535e-06,
+      "loss": 2.7374,
+      "step": 31787
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.495353982300885e-06,
+      "loss": 1.6162,
+      "step": 31788
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4953065739570165e-06,
+      "loss": 3.5872,
+      "step": 31789
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.495259165613148e-06,
+      "loss": 2.2353,
+      "step": 31790
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4952117572692795e-06,
+      "loss": 4.1866,
+      "step": 31791
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.495164348925411e-06,
+      "loss": 1.3828,
+      "step": 31792
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4951169405815425e-06,
+      "loss": 2.0343,
+      "step": 31793
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4950695322376737e-06,
+      "loss": 3.2007,
+      "step": 31794
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4950221238938054e-06,
+      "loss": 2.8806,
+      "step": 31795
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4949747155499367e-06,
+      "loss": 2.4403,
+      "step": 31796
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4949273072060682e-06,
+      "loss": 1.2561,
+      "step": 31797
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4948798988622e-06,
+      "loss": 2.3964,
+      "step": 31798
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4948324905183312e-06,
+      "loss": 1.7859,
+      "step": 31799
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4947850821744627e-06,
+      "loss": 2.2454,
+      "step": 31800
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4947376738305942e-06,
+      "loss": 2.4003,
+      "step": 31801
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4946902654867257e-06,
+      "loss": 1.5162,
+      "step": 31802
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4946428571428572e-06,
+      "loss": 2.478,
+      "step": 31803
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4945954487989887e-06,
+      "loss": 1.832,
+      "step": 31804
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4945480404551202e-06,
+      "loss": 1.8041,
+      "step": 31805
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4945006321112516e-06,
+      "loss": 2.775,
+      "step": 31806
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.494453223767383e-06,
+      "loss": 2.0761,
+      "step": 31807
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4944058154235146e-06,
+      "loss": 3.1372,
+      "step": 31808
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4943584070796461e-06,
+      "loss": 1.8749,
+      "step": 31809
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4943109987357774e-06,
+      "loss": 1.4621,
+      "step": 31810
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4942635903919091e-06,
+      "loss": 3.6778,
+      "step": 31811
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4942161820480406e-06,
+      "loss": 2.9666,
+      "step": 31812
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4941687737041719e-06,
+      "loss": 2.1652,
+      "step": 31813
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4941213653603034e-06,
+      "loss": 2.038,
+      "step": 31814
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.494073957016435e-06,
+      "loss": 2.3052,
+      "step": 31815
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4940265486725664e-06,
+      "loss": 2.4192,
+      "step": 31816
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4939791403286979e-06,
+      "loss": 3.2673,
+      "step": 31817
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4939317319848293e-06,
+      "loss": 3.2726,
+      "step": 31818
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4938843236409608e-06,
+      "loss": 3.1176,
+      "step": 31819
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4938369152970923e-06,
+      "loss": 1.087,
+      "step": 31820
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4937895069532238e-06,
+      "loss": 2.1455,
+      "step": 31821
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4937420986093553e-06,
+      "loss": 1.6537,
+      "step": 31822
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 1.4936946902654868e-06,
+      "loss": 2.5731,
+      "step": 31823
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4936472819216183e-06,
+      "loss": 2.8475,
+      "step": 31824
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4935998735777498e-06,
+      "loss": 1.1356,
+      "step": 31825
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4935524652338813e-06,
+      "loss": 3.1796,
+      "step": 31826
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4935050568900126e-06,
+      "loss": 1.8961,
+      "step": 31827
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4934576485461443e-06,
+      "loss": 1.2601,
+      "step": 31828
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4934102402022755e-06,
+      "loss": 1.9419,
+      "step": 31829
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.493362831858407e-06,
+      "loss": 3.1194,
+      "step": 31830
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4933154235145387e-06,
+      "loss": 3.165,
+      "step": 31831
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.49326801517067e-06,
+      "loss": 2.5216,
+      "step": 31832
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4932206068268015e-06,
+      "loss": 1.4119,
+      "step": 31833
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.493173198482933e-06,
+      "loss": 2.2397,
+      "step": 31834
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4931257901390645e-06,
+      "loss": 2.3051,
+      "step": 31835
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.493078381795196e-06,
+      "loss": 1.5997,
+      "step": 31836
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4930309734513275e-06,
+      "loss": 2.333,
+      "step": 31837
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.492983565107459e-06,
+      "loss": 1.8692,
+      "step": 31838
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4929361567635905e-06,
+      "loss": 3.08,
+      "step": 31839
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4928887484197217e-06,
+      "loss": 2.0033,
+      "step": 31840
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4928413400758535e-06,
+      "loss": 1.7935,
+      "step": 31841
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.492793931731985e-06,
+      "loss": 1.8019,
+      "step": 31842
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4927465233881162e-06,
+      "loss": 1.3015,
+      "step": 31843
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.492699115044248e-06,
+      "loss": 3.0365,
+      "step": 31844
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4926517067003794e-06,
+      "loss": 2.684,
+      "step": 31845
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4926042983565107e-06,
+      "loss": 2.6411,
+      "step": 31846
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4925568900126422e-06,
+      "loss": 1.0935,
+      "step": 31847
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.492509481668774e-06,
+      "loss": 3.2817,
+      "step": 31848
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4924620733249052e-06,
+      "loss": 3.5719,
+      "step": 31849
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4924146649810367e-06,
+      "loss": 1.3263,
+      "step": 31850
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4923672566371682e-06,
+      "loss": 2.792,
+      "step": 31851
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4923198482932997e-06,
+      "loss": 2.4084,
+      "step": 31852
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4922724399494311e-06,
+      "loss": 1.8798,
+      "step": 31853
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4922250316055626e-06,
+      "loss": 1.0017,
+      "step": 31854
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4921776232616941e-06,
+      "loss": 0.5524,
+      "step": 31855
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4921302149178256e-06,
+      "loss": 1.3233,
+      "step": 31856
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4920828065739571e-06,
+      "loss": 2.8593,
+      "step": 31857
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4920353982300886e-06,
+      "loss": 1.8168,
+      "step": 31858
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.49198798988622e-06,
+      "loss": 3.0226,
+      "step": 31859
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4919405815423514e-06,
+      "loss": 2.1447,
+      "step": 31860
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.491893173198483e-06,
+      "loss": 1.397,
+      "step": 31861
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4918457648546144e-06,
+      "loss": 2.8373,
+      "step": 31862
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4917983565107459e-06,
+      "loss": 2.3245,
+      "step": 31863
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4917509481668776e-06,
+      "loss": 2.6123,
+      "step": 31864
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4917035398230088e-06,
+      "loss": 2.456,
+      "step": 31865
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4916561314791403e-06,
+      "loss": 3.8889,
+      "step": 31866
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4916087231352718e-06,
+      "loss": 2.0922,
+      "step": 31867
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4915613147914033e-06,
+      "loss": 2.9386,
+      "step": 31868
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4915139064475348e-06,
+      "loss": 1.3922,
+      "step": 31869
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4914664981036663e-06,
+      "loss": 2.2469,
+      "step": 31870
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4914190897597978e-06,
+      "loss": 2.6887,
+      "step": 31871
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4913716814159293e-06,
+      "loss": 2.6808,
+      "step": 31872
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4913242730720606e-06,
+      "loss": 2.4999,
+      "step": 31873
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4912768647281923e-06,
+      "loss": 3.1223,
+      "step": 31874
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4912294563843238e-06,
+      "loss": 3.8507,
+      "step": 31875
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.491182048040455e-06,
+      "loss": 2.1327,
+      "step": 31876
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4911346396965867e-06,
+      "loss": 2.3112,
+      "step": 31877
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4910872313527182e-06,
+      "loss": 2.0197,
+      "step": 31878
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4910398230088495e-06,
+      "loss": 2.8717,
+      "step": 31879
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.490992414664981e-06,
+      "loss": 2.5687,
+      "step": 31880
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4909450063211127e-06,
+      "loss": 3.3399,
+      "step": 31881
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.490897597977244e-06,
+      "loss": 3.2568,
+      "step": 31882
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4908501896333755e-06,
+      "loss": 1.8147,
+      "step": 31883
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.490802781289507e-06,
+      "loss": 1.7029,
+      "step": 31884
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.4907553729456385e-06,
+      "loss": 3.5654,
+      "step": 31885
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 1.49070796460177e-06,
+      "loss": 1.7866,
+      "step": 31886
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4906605562579015e-06,
+      "loss": 1.9664,
+      "step": 31887
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.490613147914033e-06,
+      "loss": 2.0343,
+      "step": 31888
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4905657395701644e-06,
+      "loss": 1.9756,
+      "step": 31889
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.490518331226296e-06,
+      "loss": 0.9829,
+      "step": 31890
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4904709228824274e-06,
+      "loss": 3.6442,
+      "step": 31891
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.490423514538559e-06,
+      "loss": 2.4041,
+      "step": 31892
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4903761061946902e-06,
+      "loss": 1.8856,
+      "step": 31893
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.490328697850822e-06,
+      "loss": 1.7931,
+      "step": 31894
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4902812895069532e-06,
+      "loss": 2.341,
+      "step": 31895
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4902338811630847e-06,
+      "loss": 2.1227,
+      "step": 31896
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4901864728192164e-06,
+      "loss": 4.5393,
+      "step": 31897
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4901390644753477e-06,
+      "loss": 3.3165,
+      "step": 31898
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4900916561314792e-06,
+      "loss": 1.9692,
+      "step": 31899
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4900442477876106e-06,
+      "loss": 2.6232,
+      "step": 31900
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4899968394437421e-06,
+      "loss": 1.3996,
+      "step": 31901
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4899494310998736e-06,
+      "loss": 1.7701,
+      "step": 31902
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4899020227560051e-06,
+      "loss": 3.322,
+      "step": 31903
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4898546144121366e-06,
+      "loss": 3.5316,
+      "step": 31904
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.489807206068268e-06,
+      "loss": 1.6974,
+      "step": 31905
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4897597977243994e-06,
+      "loss": 3.5993,
+      "step": 31906
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.489712389380531e-06,
+      "loss": 3.1959,
+      "step": 31907
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4896649810366626e-06,
+      "loss": 2.7363,
+      "step": 31908
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4896175726927939e-06,
+      "loss": 2.2967,
+      "step": 31909
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4895701643489254e-06,
+      "loss": 4.7767,
+      "step": 31910
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.489522756005057e-06,
+      "loss": 1.5206,
+      "step": 31911
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4894753476611883e-06,
+      "loss": 2.6604,
+      "step": 31912
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4894279393173198e-06,
+      "loss": 2.3774,
+      "step": 31913
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4893805309734513e-06,
+      "loss": 2.2612,
+      "step": 31914
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4893331226295828e-06,
+      "loss": 2.0487,
+      "step": 31915
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4892857142857143e-06,
+      "loss": 1.2245,
+      "step": 31916
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4892383059418458e-06,
+      "loss": 2.3204,
+      "step": 31917
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4891908975979773e-06,
+      "loss": 4.7007,
+      "step": 31918
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4891434892541088e-06,
+      "loss": 1.8823,
+      "step": 31919
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4890960809102403e-06,
+      "loss": 1.3433,
+      "step": 31920
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4890486725663718e-06,
+      "loss": 2.2561,
+      "step": 31921
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4890012642225033e-06,
+      "loss": 2.0615,
+      "step": 31922
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4889538558786345e-06,
+      "loss": 1.7492,
+      "step": 31923
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4889064475347662e-06,
+      "loss": 1.1502,
+      "step": 31924
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4888590391908975e-06,
+      "loss": 1.8022,
+      "step": 31925
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.488811630847029e-06,
+      "loss": 2.1591,
+      "step": 31926
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4887642225031607e-06,
+      "loss": 1.4242,
+      "step": 31927
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.488716814159292e-06,
+      "loss": 2.7911,
+      "step": 31928
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4886694058154235e-06,
+      "loss": 2.0923,
+      "step": 31929
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.488621997471555e-06,
+      "loss": 2.4558,
+      "step": 31930
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4885745891276865e-06,
+      "loss": 2.4781,
+      "step": 31931
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.488527180783818e-06,
+      "loss": 1.8066,
+      "step": 31932
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4884797724399495e-06,
+      "loss": 1.9532,
+      "step": 31933
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.488432364096081e-06,
+      "loss": 1.9763,
+      "step": 31934
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4883849557522124e-06,
+      "loss": 2.4715,
+      "step": 31935
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4883375474083437e-06,
+      "loss": 2.6368,
+      "step": 31936
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4882901390644754e-06,
+      "loss": 1.7257,
+      "step": 31937
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.488242730720607e-06,
+      "loss": 1.2052,
+      "step": 31938
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4881953223767382e-06,
+      "loss": 3.3956,
+      "step": 31939
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.48814791403287e-06,
+      "loss": 2.1931,
+      "step": 31940
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4881005056890014e-06,
+      "loss": 2.1836,
+      "step": 31941
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4880530973451327e-06,
+      "loss": 1.9269,
+      "step": 31942
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4880056890012642e-06,
+      "loss": 1.8694,
+      "step": 31943
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4879582806573959e-06,
+      "loss": 2.6572,
+      "step": 31944
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4879108723135272e-06,
+      "loss": 1.7371,
+      "step": 31945
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4878634639696586e-06,
+      "loss": 1.1628,
+      "step": 31946
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4878160556257901e-06,
+      "loss": 1.4325,
+      "step": 31947
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4877686472819216e-06,
+      "loss": 3.2766,
+      "step": 31948
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 1.4877212389380531e-06,
+      "loss": 1.7814,
+      "step": 31949
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4876738305941846e-06,
+      "loss": 2.2682,
+      "step": 31950
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4876264222503161e-06,
+      "loss": 2.3301,
+      "step": 31951
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4875790139064476e-06,
+      "loss": 2.1447,
+      "step": 31952
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.487531605562579e-06,
+      "loss": 0.8933,
+      "step": 31953
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4874841972187106e-06,
+      "loss": 1.8312,
+      "step": 31954
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.487436788874842e-06,
+      "loss": 1.8524,
+      "step": 31955
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4873893805309734e-06,
+      "loss": 2.6074,
+      "step": 31956
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.487341972187105e-06,
+      "loss": 2.7282,
+      "step": 31957
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4872945638432363e-06,
+      "loss": 2.7792,
+      "step": 31958
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4872471554993678e-06,
+      "loss": 1.6512,
+      "step": 31959
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4871997471554995e-06,
+      "loss": 2.3897,
+      "step": 31960
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4871523388116308e-06,
+      "loss": 1.9606,
+      "step": 31961
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4871049304677623e-06,
+      "loss": 3.5168,
+      "step": 31962
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4870575221238938e-06,
+      "loss": 2.2992,
+      "step": 31963
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4870101137800253e-06,
+      "loss": 1.7935,
+      "step": 31964
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4869627054361568e-06,
+      "loss": 0.7238,
+      "step": 31965
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4869152970922883e-06,
+      "loss": 3.6652,
+      "step": 31966
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4868678887484198e-06,
+      "loss": 3.9923,
+      "step": 31967
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4868204804045513e-06,
+      "loss": 2.3838,
+      "step": 31968
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4867730720606825e-06,
+      "loss": 2.0934,
+      "step": 31969
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4867256637168142e-06,
+      "loss": 2.0993,
+      "step": 31970
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4866782553729457e-06,
+      "loss": 3.3628,
+      "step": 31971
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.486630847029077e-06,
+      "loss": 2.2913,
+      "step": 31972
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4865834386852087e-06,
+      "loss": 3.6533,
+      "step": 31973
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4865360303413402e-06,
+      "loss": 1.6349,
+      "step": 31974
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4864886219974715e-06,
+      "loss": 2.9347,
+      "step": 31975
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.486441213653603e-06,
+      "loss": 3.8705,
+      "step": 31976
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4863938053097347e-06,
+      "loss": 2.1343,
+      "step": 31977
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.486346396965866e-06,
+      "loss": 3.7856,
+      "step": 31978
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4862989886219975e-06,
+      "loss": 1.802,
+      "step": 31979
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.486251580278129e-06,
+      "loss": 1.6168,
+      "step": 31980
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4862041719342604e-06,
+      "loss": 1.658,
+      "step": 31981
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.486156763590392e-06,
+      "loss": 2.0953,
+      "step": 31982
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4861093552465234e-06,
+      "loss": 2.3,
+      "step": 31983
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.486061946902655e-06,
+      "loss": 2.8799,
+      "step": 31984
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4860145385587864e-06,
+      "loss": 3.6281,
+      "step": 31985
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.485967130214918e-06,
+      "loss": 2.1901,
+      "step": 31986
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4859197218710494e-06,
+      "loss": 1.8045,
+      "step": 31987
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.485872313527181e-06,
+      "loss": 3.224,
+      "step": 31988
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4858249051833122e-06,
+      "loss": 2.7632,
+      "step": 31989
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4857774968394439e-06,
+      "loss": 1.5334,
+      "step": 31990
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4857300884955752e-06,
+      "loss": 1.9395,
+      "step": 31991
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4856826801517067e-06,
+      "loss": 3.3501,
+      "step": 31992
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4856352718078384e-06,
+      "loss": 2.2438,
+      "step": 31993
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4855878634639696e-06,
+      "loss": 2.3369,
+      "step": 31994
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4855404551201011e-06,
+      "loss": 3.4855,
+      "step": 31995
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4854930467762326e-06,
+      "loss": 4.1832,
+      "step": 31996
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4854456384323641e-06,
+      "loss": 2.1759,
+      "step": 31997
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4853982300884956e-06,
+      "loss": 2.0926,
+      "step": 31998
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.485350821744627e-06,
+      "loss": 2.3044,
+      "step": 31999
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4853034134007586e-06,
+      "loss": 2.8142,
+      "step": 32000
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.48525600505689e-06,
+      "loss": 1.9921,
+      "step": 32001
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4852085967130214e-06,
+      "loss": 3.4979,
+      "step": 32002
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.485161188369153e-06,
+      "loss": 3.1529,
+      "step": 32003
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4851137800252846e-06,
+      "loss": 2.5863,
+      "step": 32004
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4850663716814158e-06,
+      "loss": 2.7577,
+      "step": 32005
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4850189633375475e-06,
+      "loss": 2.9577,
+      "step": 32006
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.484971554993679e-06,
+      "loss": 1.9512,
+      "step": 32007
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4849241466498103e-06,
+      "loss": 2.3012,
+      "step": 32008
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4848767383059418e-06,
+      "loss": 2.5668,
+      "step": 32009
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4848293299620735e-06,
+      "loss": 1.6907,
+      "step": 32010
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4847819216182048e-06,
+      "loss": 2.9887,
+      "step": 32011
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4847345132743363e-06,
+      "loss": 1.0933,
+      "step": 32012
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 1.4846871049304678e-06,
+      "loss": 1.4441,
+      "step": 32013
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4846396965865993e-06,
+      "loss": 1.6293,
+      "step": 32014
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4845922882427308e-06,
+      "loss": 2.6107,
+      "step": 32015
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4845448798988623e-06,
+      "loss": 1.731,
+      "step": 32016
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4844974715549937e-06,
+      "loss": 2.924,
+      "step": 32017
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4844500632111252e-06,
+      "loss": 2.8671,
+      "step": 32018
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4844026548672567e-06,
+      "loss": 2.5559,
+      "step": 32019
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4843552465233882e-06,
+      "loss": 2.4691,
+      "step": 32020
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4843078381795197e-06,
+      "loss": 2.0335,
+      "step": 32021
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.484260429835651e-06,
+      "loss": 2.6211,
+      "step": 32022
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4842130214917827e-06,
+      "loss": 2.6853,
+      "step": 32023
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.484165613147914e-06,
+      "loss": 1.6087,
+      "step": 32024
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4841182048040455e-06,
+      "loss": 1.2841,
+      "step": 32025
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4840707964601772e-06,
+      "loss": 2.3043,
+      "step": 32026
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4840233881163085e-06,
+      "loss": 3.1917,
+      "step": 32027
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.48397597977244e-06,
+      "loss": 2.3754,
+      "step": 32028
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4839285714285714e-06,
+      "loss": 3.2246,
+      "step": 32029
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.483881163084703e-06,
+      "loss": 1.6944,
+      "step": 32030
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4838337547408344e-06,
+      "loss": 1.5761,
+      "step": 32031
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.483786346396966e-06,
+      "loss": 2.61,
+      "step": 32032
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4837389380530974e-06,
+      "loss": 2.4918,
+      "step": 32033
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.483691529709229e-06,
+      "loss": 2.2333,
+      "step": 32034
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4836441213653602e-06,
+      "loss": 2.2436,
+      "step": 32035
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4835967130214919e-06,
+      "loss": 1.7278,
+      "step": 32036
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4835493046776234e-06,
+      "loss": 0.6984,
+      "step": 32037
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4835018963337547e-06,
+      "loss": 1.5486,
+      "step": 32038
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4834544879898861e-06,
+      "loss": 2.2759,
+      "step": 32039
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4834070796460179e-06,
+      "loss": 2.1461,
+      "step": 32040
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4833596713021491e-06,
+      "loss": 2.1764,
+      "step": 32041
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4833122629582806e-06,
+      "loss": 2.9049,
+      "step": 32042
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4832648546144123e-06,
+      "loss": 3.1276,
+      "step": 32043
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4832174462705436e-06,
+      "loss": 2.2598,
+      "step": 32044
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.483170037926675e-06,
+      "loss": 1.4173,
+      "step": 32045
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4831226295828066e-06,
+      "loss": 2.6183,
+      "step": 32046
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.483075221238938e-06,
+      "loss": 1.958,
+      "step": 32047
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4830278128950696e-06,
+      "loss": 1.6385,
+      "step": 32048
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.482980404551201e-06,
+      "loss": 2.7239,
+      "step": 32049
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4829329962073326e-06,
+      "loss": 2.3213,
+      "step": 32050
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.482885587863464e-06,
+      "loss": 1.0768,
+      "step": 32051
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4828381795195953e-06,
+      "loss": 3.2086,
+      "step": 32052
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.482790771175727e-06,
+      "loss": 2.3787,
+      "step": 32053
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4827433628318585e-06,
+      "loss": 2.4056,
+      "step": 32054
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4826959544879898e-06,
+      "loss": 2.0635,
+      "step": 32055
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4826485461441215e-06,
+      "loss": 1.3348,
+      "step": 32056
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4826011378002528e-06,
+      "loss": 2.4147,
+      "step": 32057
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4825537294563843e-06,
+      "loss": 2.6481,
+      "step": 32058
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4825063211125158e-06,
+      "loss": 1.5157,
+      "step": 32059
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4824589127686473e-06,
+      "loss": 1.1102,
+      "step": 32060
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4824115044247788e-06,
+      "loss": 2.2975,
+      "step": 32061
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4823640960809103e-06,
+      "loss": 1.5941,
+      "step": 32062
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4823166877370417e-06,
+      "loss": 2.624,
+      "step": 32063
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4822692793931732e-06,
+      "loss": 0.4765,
+      "step": 32064
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4822218710493047e-06,
+      "loss": 3.2578,
+      "step": 32065
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4821744627054362e-06,
+      "loss": 1.9061,
+      "step": 32066
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4821270543615677e-06,
+      "loss": 3.1259,
+      "step": 32067
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.482079646017699e-06,
+      "loss": 2.5013,
+      "step": 32068
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4820322376738307e-06,
+      "loss": 2.7427,
+      "step": 32069
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4819848293299622e-06,
+      "loss": 2.9499,
+      "step": 32070
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4819374209860935e-06,
+      "loss": 1.7683,
+      "step": 32071
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.481890012642225e-06,
+      "loss": 1.9654,
+      "step": 32072
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4818426042983567e-06,
+      "loss": 2.0642,
+      "step": 32073
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.481795195954488e-06,
+      "loss": 2.9661,
+      "step": 32074
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4817477876106194e-06,
+      "loss": 1.7392,
+      "step": 32075
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 1.4817003792667511e-06,
+      "loss": 2.815,
+      "step": 32076
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4816529709228824e-06,
+      "loss": 2.9538,
+      "step": 32077
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.481605562579014e-06,
+      "loss": 1.2392,
+      "step": 32078
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4815581542351454e-06,
+      "loss": 2.6478,
+      "step": 32079
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.481510745891277e-06,
+      "loss": 3.4024,
+      "step": 32080
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4814633375474084e-06,
+      "loss": 2.3658,
+      "step": 32081
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4814159292035399e-06,
+      "loss": 2.2294,
+      "step": 32082
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4813685208596714e-06,
+      "loss": 1.9781,
+      "step": 32083
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4813211125158029e-06,
+      "loss": 1.8748,
+      "step": 32084
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4812737041719342e-06,
+      "loss": 2.9135,
+      "step": 32085
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4812262958280659e-06,
+      "loss": 2.2532,
+      "step": 32086
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4811788874841973e-06,
+      "loss": 2.9037,
+      "step": 32087
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4811314791403286e-06,
+      "loss": 2.2633,
+      "step": 32088
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4810840707964603e-06,
+      "loss": 4.0779,
+      "step": 32089
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4810366624525916e-06,
+      "loss": 2.5774,
+      "step": 32090
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4809892541087231e-06,
+      "loss": 3.0794,
+      "step": 32091
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4809418457648546e-06,
+      "loss": 1.5346,
+      "step": 32092
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.480894437420986e-06,
+      "loss": 1.5285,
+      "step": 32093
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4808470290771176e-06,
+      "loss": 2.2016,
+      "step": 32094
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.480799620733249e-06,
+      "loss": 1.9035,
+      "step": 32095
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4807522123893806e-06,
+      "loss": 1.7801,
+      "step": 32096
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.480704804045512e-06,
+      "loss": 2.0591,
+      "step": 32097
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4806573957016436e-06,
+      "loss": 3.2491,
+      "step": 32098
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.480609987357775e-06,
+      "loss": 1.5993,
+      "step": 32099
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4805625790139065e-06,
+      "loss": 2.1986,
+      "step": 32100
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4805151706700378e-06,
+      "loss": 1.4245,
+      "step": 32101
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4804677623261695e-06,
+      "loss": 1.4915,
+      "step": 32102
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.480420353982301e-06,
+      "loss": 2.5577,
+      "step": 32103
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4803729456384323e-06,
+      "loss": 0.7814,
+      "step": 32104
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4803255372945638e-06,
+      "loss": 2.1702,
+      "step": 32105
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4802781289506955e-06,
+      "loss": 2.5106,
+      "step": 32106
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4802307206068268e-06,
+      "loss": 2.0873,
+      "step": 32107
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4801833122629583e-06,
+      "loss": 0.9956,
+      "step": 32108
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.48013590391909e-06,
+      "loss": 0.9427,
+      "step": 32109
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4800884955752212e-06,
+      "loss": 2.2322,
+      "step": 32110
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4800410872313527e-06,
+      "loss": 1.7477,
+      "step": 32111
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4799936788874842e-06,
+      "loss": 2.8485,
+      "step": 32112
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4799462705436157e-06,
+      "loss": 2.5839,
+      "step": 32113
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4798988621997472e-06,
+      "loss": 2.0683,
+      "step": 32114
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4798514538558787e-06,
+      "loss": 2.3777,
+      "step": 32115
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4798040455120102e-06,
+      "loss": 3.0175,
+      "step": 32116
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4797566371681417e-06,
+      "loss": 2.4285,
+      "step": 32117
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.479709228824273e-06,
+      "loss": 3.429,
+      "step": 32118
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4796618204804047e-06,
+      "loss": 1.6024,
+      "step": 32119
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4796144121365362e-06,
+      "loss": 2.6418,
+      "step": 32120
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4795670037926674e-06,
+      "loss": 2.403,
+      "step": 32121
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4795195954487992e-06,
+      "loss": 1.9297,
+      "step": 32122
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4794721871049304e-06,
+      "loss": 1.5777,
+      "step": 32123
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.479424778761062e-06,
+      "loss": 1.5778,
+      "step": 32124
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4793773704171934e-06,
+      "loss": 2.2691,
+      "step": 32125
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.479329962073325e-06,
+      "loss": 3.3153,
+      "step": 32126
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4792825537294564e-06,
+      "loss": 1.1428,
+      "step": 32127
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4792351453855879e-06,
+      "loss": 3.6733,
+      "step": 32128
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4791877370417194e-06,
+      "loss": 3.9122,
+      "step": 32129
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4791403286978509e-06,
+      "loss": 2.0446,
+      "step": 32130
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4790929203539824e-06,
+      "loss": 2.3961,
+      "step": 32131
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4790455120101139e-06,
+      "loss": 1.302,
+      "step": 32132
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4789981036662454e-06,
+      "loss": 2.8304,
+      "step": 32133
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4789506953223766e-06,
+      "loss": 2.5987,
+      "step": 32134
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4789032869785083e-06,
+      "loss": 2.6089,
+      "step": 32135
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4788558786346398e-06,
+      "loss": 2.741,
+      "step": 32136
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4788084702907711e-06,
+      "loss": 1.6433,
+      "step": 32137
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4787610619469026e-06,
+      "loss": 1.7428,
+      "step": 32138
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4787136536030343e-06,
+      "loss": 2.247,
+      "step": 32139
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4786662452591656e-06,
+      "loss": 2.1668,
+      "step": 32140
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.478618836915297e-06,
+      "loss": 2.3617,
+      "step": 32141
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4785714285714288e-06,
+      "loss": 2.7572,
+      "step": 32142
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.47852402022756e-06,
+      "loss": 2.0859,
+      "step": 32143
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4784766118836916e-06,
+      "loss": 3.3183,
+      "step": 32144
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.478429203539823e-06,
+      "loss": 3.0127,
+      "step": 32145
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4783817951959545e-06,
+      "loss": 2.2624,
+      "step": 32146
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.478334386852086e-06,
+      "loss": 1.0808,
+      "step": 32147
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4782869785082175e-06,
+      "loss": 1.8697,
+      "step": 32148
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.478239570164349e-06,
+      "loss": 2.2819,
+      "step": 32149
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4781921618204805e-06,
+      "loss": 1.791,
+      "step": 32150
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4781447534766118e-06,
+      "loss": 0.9623,
+      "step": 32151
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4780973451327435e-06,
+      "loss": 2.5727,
+      "step": 32152
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.478049936788875e-06,
+      "loss": 1.4047,
+      "step": 32153
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4780025284450063e-06,
+      "loss": 2.7961,
+      "step": 32154
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.477955120101138e-06,
+      "loss": 1.6056,
+      "step": 32155
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4779077117572693e-06,
+      "loss": 1.4853,
+      "step": 32156
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4778603034134007e-06,
+      "loss": 2.5839,
+      "step": 32157
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4778128950695322e-06,
+      "loss": 2.8265,
+      "step": 32158
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4777654867256637e-06,
+      "loss": 2.6018,
+      "step": 32159
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4777180783817952e-06,
+      "loss": 1.8169,
+      "step": 32160
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4776706700379267e-06,
+      "loss": 1.7312,
+      "step": 32161
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4776232616940582e-06,
+      "loss": 2.4732,
+      "step": 32162
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4775758533501897e-06,
+      "loss": 2.7015,
+      "step": 32163
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4775284450063212e-06,
+      "loss": 1.6412,
+      "step": 32164
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4774810366624527e-06,
+      "loss": 2.6933,
+      "step": 32165
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4774336283185842e-06,
+      "loss": 2.5,
+      "step": 32166
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4773862199747155e-06,
+      "loss": 2.1325,
+      "step": 32167
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4773388116308472e-06,
+      "loss": 1.4217,
+      "step": 32168
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4772914032869786e-06,
+      "loss": 2.371,
+      "step": 32169
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.47724399494311e-06,
+      "loss": 3.0497,
+      "step": 32170
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4771965865992414e-06,
+      "loss": 2.8746,
+      "step": 32171
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4771491782553731e-06,
+      "loss": 1.5963,
+      "step": 32172
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4771017699115044e-06,
+      "loss": 1.6073,
+      "step": 32173
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.477054361567636e-06,
+      "loss": 1.9548,
+      "step": 32174
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4770069532237676e-06,
+      "loss": 2.2461,
+      "step": 32175
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4769595448798989e-06,
+      "loss": 3.7537,
+      "step": 32176
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4769121365360304e-06,
+      "loss": 3.0389,
+      "step": 32177
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4768647281921619e-06,
+      "loss": 1.4759,
+      "step": 32178
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4768173198482934e-06,
+      "loss": 2.9039,
+      "step": 32179
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4767699115044249e-06,
+      "loss": 1.9971,
+      "step": 32180
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4767225031605561e-06,
+      "loss": 2.3821,
+      "step": 32181
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4766750948166878e-06,
+      "loss": 1.3116,
+      "step": 32182
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4766276864728193e-06,
+      "loss": 1.3735,
+      "step": 32183
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4765802781289506e-06,
+      "loss": 2.614,
+      "step": 32184
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4765328697850823e-06,
+      "loss": 1.2873,
+      "step": 32185
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4764854614412138e-06,
+      "loss": 2.4764,
+      "step": 32186
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.476438053097345e-06,
+      "loss": 2.7204,
+      "step": 32187
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4763906447534766e-06,
+      "loss": 3.2465,
+      "step": 32188
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.476343236409608e-06,
+      "loss": 2.8431,
+      "step": 32189
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4762958280657396e-06,
+      "loss": 0.9491,
+      "step": 32190
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.476248419721871e-06,
+      "loss": 2.3419,
+      "step": 32191
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4762010113780025e-06,
+      "loss": 1.8631,
+      "step": 32192
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.476153603034134e-06,
+      "loss": 3.4839,
+      "step": 32193
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4761061946902655e-06,
+      "loss": 2.51,
+      "step": 32194
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.476058786346397e-06,
+      "loss": 2.7528,
+      "step": 32195
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4760113780025285e-06,
+      "loss": 2.9223,
+      "step": 32196
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.47596396965866e-06,
+      "loss": 1.577,
+      "step": 32197
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4759165613147915e-06,
+      "loss": 2.1069,
+      "step": 32198
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.475869152970923e-06,
+      "loss": 2.796,
+      "step": 32199
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4758217446270543e-06,
+      "loss": 3.6326,
+      "step": 32200
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4757743362831858e-06,
+      "loss": 3.1629,
+      "step": 32201
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4757269279393175e-06,
+      "loss": 2.8193,
+      "step": 32202
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 1.4756795195954487e-06,
+      "loss": 1.3273,
+      "step": 32203
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4756321112515802e-06,
+      "loss": 1.5306,
+      "step": 32204
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.475584702907712e-06,
+      "loss": 2.9477,
+      "step": 32205
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4755372945638432e-06,
+      "loss": 2.0527,
+      "step": 32206
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4754898862199747e-06,
+      "loss": 1.2055,
+      "step": 32207
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4754424778761062e-06,
+      "loss": 2.6572,
+      "step": 32208
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4753950695322377e-06,
+      "loss": 2.0334,
+      "step": 32209
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4753476611883692e-06,
+      "loss": 1.5515,
+      "step": 32210
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4753002528445007e-06,
+      "loss": 2.5325,
+      "step": 32211
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4752528445006322e-06,
+      "loss": 2.279,
+      "step": 32212
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4752054361567637e-06,
+      "loss": 1.9322,
+      "step": 32213
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.475158027812895e-06,
+      "loss": 2.5071,
+      "step": 32214
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4751106194690267e-06,
+      "loss": 1.8406,
+      "step": 32215
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4750632111251581e-06,
+      "loss": 1.5159,
+      "step": 32216
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4750158027812894e-06,
+      "loss": 2.9304,
+      "step": 32217
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4749683944374211e-06,
+      "loss": 1.1433,
+      "step": 32218
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4749209860935526e-06,
+      "loss": 1.6075,
+      "step": 32219
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.474873577749684e-06,
+      "loss": 1.8707,
+      "step": 32220
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4748261694058154e-06,
+      "loss": 2.9934,
+      "step": 32221
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4747787610619469e-06,
+      "loss": 2.1526,
+      "step": 32222
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4747313527180784e-06,
+      "loss": 2.0334,
+      "step": 32223
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4746839443742099e-06,
+      "loss": 2.2722,
+      "step": 32224
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4746365360303414e-06,
+      "loss": 3.1235,
+      "step": 32225
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4745891276864729e-06,
+      "loss": 1.2598,
+      "step": 32226
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4745417193426043e-06,
+      "loss": 1.7008,
+      "step": 32227
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4744943109987358e-06,
+      "loss": 2.0561,
+      "step": 32228
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4744469026548673e-06,
+      "loss": 1.8537,
+      "step": 32229
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4743994943109988e-06,
+      "loss": 2.3776,
+      "step": 32230
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4743520859671303e-06,
+      "loss": 2.8494,
+      "step": 32231
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4743046776232618e-06,
+      "loss": 2.2944,
+      "step": 32232
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.474257269279393e-06,
+      "loss": 1.8101,
+      "step": 32233
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4742098609355246e-06,
+      "loss": 1.814,
+      "step": 32234
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4741624525916563e-06,
+      "loss": 1.9778,
+      "step": 32235
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4741150442477876e-06,
+      "loss": 2.0128,
+      "step": 32236
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.474067635903919e-06,
+      "loss": 2.1639,
+      "step": 32237
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4740202275600508e-06,
+      "loss": 2.2117,
+      "step": 32238
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.473972819216182e-06,
+      "loss": 2.3883,
+      "step": 32239
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4739254108723135e-06,
+      "loss": 1.4896,
+      "step": 32240
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.473878002528445e-06,
+      "loss": 2.3529,
+      "step": 32241
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4738305941845765e-06,
+      "loss": 1.338,
+      "step": 32242
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.473783185840708e-06,
+      "loss": 2.2057,
+      "step": 32243
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4737357774968395e-06,
+      "loss": 1.9719,
+      "step": 32244
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.473688369152971e-06,
+      "loss": 2.5595,
+      "step": 32245
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4736409608091025e-06,
+      "loss": 2.9299,
+      "step": 32246
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4735935524652338e-06,
+      "loss": 2.3049,
+      "step": 32247
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4735461441213655e-06,
+      "loss": 2.4403,
+      "step": 32248
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.473498735777497e-06,
+      "loss": 1.2905,
+      "step": 32249
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4734513274336282e-06,
+      "loss": 1.8844,
+      "step": 32250
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.47340391908976e-06,
+      "loss": 2.8657,
+      "step": 32251
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4733565107458914e-06,
+      "loss": 1.2465,
+      "step": 32252
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4733091024020227e-06,
+      "loss": 1.3306,
+      "step": 32253
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4732616940581542e-06,
+      "loss": 2.6324,
+      "step": 32254
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4732142857142857e-06,
+      "loss": 1.8214,
+      "step": 32255
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4731668773704172e-06,
+      "loss": 2.1402,
+      "step": 32256
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4731194690265487e-06,
+      "loss": 2.7991,
+      "step": 32257
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4730720606826802e-06,
+      "loss": 1.6282,
+      "step": 32258
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4730246523388117e-06,
+      "loss": 2.0948,
+      "step": 32259
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4729772439949432e-06,
+      "loss": 1.943,
+      "step": 32260
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4729298356510747e-06,
+      "loss": 2.0849,
+      "step": 32261
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4728824273072062e-06,
+      "loss": 3.1682,
+      "step": 32262
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4728350189633376e-06,
+      "loss": 1.2162,
+      "step": 32263
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4727876106194691e-06,
+      "loss": 1.4458,
+      "step": 32264
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.4727402022756006e-06,
+      "loss": 2.401,
+      "step": 32265
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 1.472692793931732e-06,
+      "loss": 1.037,
+      "step": 32266
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4726453855878634e-06,
+      "loss": 2.3943,
+      "step": 32267
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.472597977243995e-06,
+      "loss": 1.4751,
+      "step": 32268
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4725505689001264e-06,
+      "loss": 0.8865,
+      "step": 32269
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4725031605562579e-06,
+      "loss": 1.6971,
+      "step": 32270
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4724557522123896e-06,
+      "loss": 2.9646,
+      "step": 32271
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4724083438685209e-06,
+      "loss": 2.4969,
+      "step": 32272
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4723609355246524e-06,
+      "loss": 2.3726,
+      "step": 32273
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4723135271807838e-06,
+      "loss": 2.7325,
+      "step": 32274
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4722661188369153e-06,
+      "loss": 1.3508,
+      "step": 32275
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4722187104930468e-06,
+      "loss": 1.5003,
+      "step": 32276
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4721713021491783e-06,
+      "loss": 3.4717,
+      "step": 32277
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4721238938053098e-06,
+      "loss": 3.1743,
+      "step": 32278
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4720764854614413e-06,
+      "loss": 1.498,
+      "step": 32279
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4720290771175726e-06,
+      "loss": 2.1074,
+      "step": 32280
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4719816687737043e-06,
+      "loss": 1.6613,
+      "step": 32281
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4719342604298358e-06,
+      "loss": 2.3507,
+      "step": 32282
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471886852085967e-06,
+      "loss": 2.61,
+      "step": 32283
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4718394437420988e-06,
+      "loss": 2.2512,
+      "step": 32284
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4717920353982303e-06,
+      "loss": 2.0029,
+      "step": 32285
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4717446270543615e-06,
+      "loss": 1.5763,
+      "step": 32286
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471697218710493e-06,
+      "loss": 1.912,
+      "step": 32287
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4716498103666245e-06,
+      "loss": 2.4362,
+      "step": 32288
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471602402022756e-06,
+      "loss": 2.3228,
+      "step": 32289
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4715549936788875e-06,
+      "loss": 2.0407,
+      "step": 32290
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471507585335019e-06,
+      "loss": 2.0471,
+      "step": 32291
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4714601769911505e-06,
+      "loss": 2.1467,
+      "step": 32292
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471412768647282e-06,
+      "loss": 1.4277,
+      "step": 32293
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4713653603034135e-06,
+      "loss": 3.2058,
+      "step": 32294
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471317951959545e-06,
+      "loss": 2.0015,
+      "step": 32295
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4712705436156765e-06,
+      "loss": 2.9198,
+      "step": 32296
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471223135271808e-06,
+      "loss": 2.3498,
+      "step": 32297
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4711757269279394e-06,
+      "loss": 3.0784,
+      "step": 32298
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4711283185840707e-06,
+      "loss": 3.0154,
+      "step": 32299
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4710809102402022e-06,
+      "loss": 2.2109,
+      "step": 32300
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.471033501896334e-06,
+      "loss": 2.8856,
+      "step": 32301
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4709860935524652e-06,
+      "loss": 1.5472,
+      "step": 32302
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4709386852085967e-06,
+      "loss": 2.0812,
+      "step": 32303
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4708912768647284e-06,
+      "loss": 1.265,
+      "step": 32304
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4708438685208597e-06,
+      "loss": 2.8152,
+      "step": 32305
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4707964601769912e-06,
+      "loss": 0.5828,
+      "step": 32306
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4707490518331227e-06,
+      "loss": 1.5486,
+      "step": 32307
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4707016434892542e-06,
+      "loss": 1.8158,
+      "step": 32308
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4706542351453856e-06,
+      "loss": 2.2177,
+      "step": 32309
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.470606826801517e-06,
+      "loss": 2.7281,
+      "step": 32310
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4705594184576486e-06,
+      "loss": 1.7695,
+      "step": 32311
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4705120101137801e-06,
+      "loss": 2.7442,
+      "step": 32312
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4704646017699114e-06,
+      "loss": 3.0175,
+      "step": 32313
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4704171934260431e-06,
+      "loss": 3.3971,
+      "step": 32314
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4703697850821746e-06,
+      "loss": 2.6193,
+      "step": 32315
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4703223767383059e-06,
+      "loss": 2.41,
+      "step": 32316
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4702749683944374e-06,
+      "loss": 2.2579,
+      "step": 32317
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.470227560050569e-06,
+      "loss": 2.7227,
+      "step": 32318
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4701801517067004e-06,
+      "loss": 2.6516,
+      "step": 32319
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4701327433628319e-06,
+      "loss": 2.2459,
+      "step": 32320
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4700853350189633e-06,
+      "loss": 2.3593,
+      "step": 32321
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4700379266750948e-06,
+      "loss": 3.065,
+      "step": 32322
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4699905183312263e-06,
+      "loss": 1.9412,
+      "step": 32323
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4699431099873578e-06,
+      "loss": 1.6676,
+      "step": 32324
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4698957016434893e-06,
+      "loss": 2.0016,
+      "step": 32325
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4698482932996208e-06,
+      "loss": 2.5638,
+      "step": 32326
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4698008849557523e-06,
+      "loss": 3.4788,
+      "step": 32327
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4697534766118838e-06,
+      "loss": 2.0741,
+      "step": 32328
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 1.4697060682680153e-06,
+      "loss": 3.6466,
+      "step": 32329
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4696586599241466e-06,
+      "loss": 1.8604,
+      "step": 32330
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4696112515802783e-06,
+      "loss": 2.4333,
+      "step": 32331
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4695638432364095e-06,
+      "loss": 2.3662,
+      "step": 32332
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.469516434892541e-06,
+      "loss": 1.4641,
+      "step": 32333
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4694690265486727e-06,
+      "loss": 3.147,
+      "step": 32334
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.469421618204804e-06,
+      "loss": 1.9197,
+      "step": 32335
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4693742098609355e-06,
+      "loss": 4.4445,
+      "step": 32336
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.469326801517067e-06,
+      "loss": 2.4016,
+      "step": 32337
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4692793931731985e-06,
+      "loss": 2.7173,
+      "step": 32338
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.46923198482933e-06,
+      "loss": 1.7863,
+      "step": 32339
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4691845764854615e-06,
+      "loss": 4.0255,
+      "step": 32340
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.469137168141593e-06,
+      "loss": 1.7968,
+      "step": 32341
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4690897597977245e-06,
+      "loss": 2.2312,
+      "step": 32342
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4690423514538557e-06,
+      "loss": 2.3614,
+      "step": 32343
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4689949431099875e-06,
+      "loss": 2.1374,
+      "step": 32344
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.468947534766119e-06,
+      "loss": 1.2301,
+      "step": 32345
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4689001264222502e-06,
+      "loss": 1.8279,
+      "step": 32346
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.468852718078382e-06,
+      "loss": 2.7074,
+      "step": 32347
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4688053097345134e-06,
+      "loss": 1.8602,
+      "step": 32348
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4687579013906447e-06,
+      "loss": 1.7738,
+      "step": 32349
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4687104930467762e-06,
+      "loss": 2.7787,
+      "step": 32350
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4686630847029077e-06,
+      "loss": 3.0506,
+      "step": 32351
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4686156763590392e-06,
+      "loss": 1.7347,
+      "step": 32352
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4685682680151707e-06,
+      "loss": 2.9787,
+      "step": 32353
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4685208596713022e-06,
+      "loss": 2.6016,
+      "step": 32354
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4684734513274337e-06,
+      "loss": 2.7872,
+      "step": 32355
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4684260429835651e-06,
+      "loss": 0.9927,
+      "step": 32356
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4683786346396966e-06,
+      "loss": 1.5144,
+      "step": 32357
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4683312262958281e-06,
+      "loss": 1.8798,
+      "step": 32358
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4682838179519596e-06,
+      "loss": 2.5483,
+      "step": 32359
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4682364096080911e-06,
+      "loss": 1.9932,
+      "step": 32360
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4681890012642226e-06,
+      "loss": 3.1275,
+      "step": 32361
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4681415929203539e-06,
+      "loss": 2.7666,
+      "step": 32362
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4680941845764854e-06,
+      "loss": 1.2343,
+      "step": 32363
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.468046776232617e-06,
+      "loss": 3.6415,
+      "step": 32364
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4679993678887484e-06,
+      "loss": 3.0658,
+      "step": 32365
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4679519595448799e-06,
+      "loss": 4.1882,
+      "step": 32366
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4679045512010116e-06,
+      "loss": 2.7122,
+      "step": 32367
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4678571428571428e-06,
+      "loss": 1.7986,
+      "step": 32368
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4678097345132743e-06,
+      "loss": 1.896,
+      "step": 32369
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4677623261694058e-06,
+      "loss": 2.7956,
+      "step": 32370
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4677149178255373e-06,
+      "loss": 2.4418,
+      "step": 32371
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4676675094816688e-06,
+      "loss": 2.4007,
+      "step": 32372
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4676201011378003e-06,
+      "loss": 2.9774,
+      "step": 32373
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4675726927939318e-06,
+      "loss": 2.496,
+      "step": 32374
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4675252844500633e-06,
+      "loss": 3.2643,
+      "step": 32375
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4674778761061946e-06,
+      "loss": 2.1206,
+      "step": 32376
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4674304677623263e-06,
+      "loss": 1.3675,
+      "step": 32377
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4673830594184578e-06,
+      "loss": 3.2413,
+      "step": 32378
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.467335651074589e-06,
+      "loss": 2.0815,
+      "step": 32379
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4672882427307207e-06,
+      "loss": 1.1629,
+      "step": 32380
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4672408343868522e-06,
+      "loss": 1.3886,
+      "step": 32381
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4671934260429835e-06,
+      "loss": 2.0393,
+      "step": 32382
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.467146017699115e-06,
+      "loss": 2.5057,
+      "step": 32383
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4670986093552465e-06,
+      "loss": 1.5815,
+      "step": 32384
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.467051201011378e-06,
+      "loss": 2.2507,
+      "step": 32385
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4670037926675095e-06,
+      "loss": 1.7328,
+      "step": 32386
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.466956384323641e-06,
+      "loss": 1.3474,
+      "step": 32387
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4669089759797725e-06,
+      "loss": 2.898,
+      "step": 32388
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.466861567635904e-06,
+      "loss": 2.8149,
+      "step": 32389
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4668141592920355e-06,
+      "loss": 1.8073,
+      "step": 32390
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.466766750948167e-06,
+      "loss": 2.2904,
+      "step": 32391
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.4667193426042984e-06,
+      "loss": 0.9628,
+      "step": 32392
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 1.46667193426043e-06,
+      "loss": 1.7616,
+      "step": 32393
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4666245259165614e-06,
+      "loss": 2.9055,
+      "step": 32394
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4665771175726927e-06,
+      "loss": 1.6875,
+      "step": 32395
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4665297092288242e-06,
+      "loss": 3.3667,
+      "step": 32396
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.466482300884956e-06,
+      "loss": 2.9122,
+      "step": 32397
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4664348925410872e-06,
+      "loss": 1.3988,
+      "step": 32398
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4663874841972187e-06,
+      "loss": 2.0422,
+      "step": 32399
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4663400758533504e-06,
+      "loss": 1.0247,
+      "step": 32400
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4662926675094817e-06,
+      "loss": 1.9906,
+      "step": 32401
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4662452591656132e-06,
+      "loss": 2.5746,
+      "step": 32402
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4661978508217446e-06,
+      "loss": 2.3208,
+      "step": 32403
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4661504424778761e-06,
+      "loss": 3.5031,
+      "step": 32404
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4661030341340076e-06,
+      "loss": 2.926,
+      "step": 32405
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4660556257901391e-06,
+      "loss": 2.8877,
+      "step": 32406
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4660082174462706e-06,
+      "loss": 3.0462,
+      "step": 32407
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.465960809102402e-06,
+      "loss": 2.3317,
+      "step": 32408
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4659134007585334e-06,
+      "loss": 1.723,
+      "step": 32409
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.465865992414665e-06,
+      "loss": 1.3669,
+      "step": 32410
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4658185840707966e-06,
+      "loss": 2.3556,
+      "step": 32411
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4657711757269279e-06,
+      "loss": 2.4163,
+      "step": 32412
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4657237673830596e-06,
+      "loss": 1.7504,
+      "step": 32413
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.465676359039191e-06,
+      "loss": 3.2355,
+      "step": 32414
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4656289506953223e-06,
+      "loss": 2.5788,
+      "step": 32415
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4655815423514538e-06,
+      "loss": 2.1763,
+      "step": 32416
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4655341340075853e-06,
+      "loss": 3.2213,
+      "step": 32417
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4654867256637168e-06,
+      "loss": 4.1524,
+      "step": 32418
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4654393173198483e-06,
+      "loss": 2.4076,
+      "step": 32419
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4653919089759798e-06,
+      "loss": 2.258,
+      "step": 32420
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4653445006321113e-06,
+      "loss": 3.3309,
+      "step": 32421
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4652970922882428e-06,
+      "loss": 0.9705,
+      "step": 32422
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4652496839443743e-06,
+      "loss": 2.3528,
+      "step": 32423
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4652022756005058e-06,
+      "loss": 1.7726,
+      "step": 32424
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4651548672566373e-06,
+      "loss": 2.1494,
+      "step": 32425
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4651074589127688e-06,
+      "loss": 2.615,
+      "step": 32426
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4650600505689002e-06,
+      "loss": 2.2375,
+      "step": 32427
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4650126422250315e-06,
+      "loss": 1.5445,
+      "step": 32428
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464965233881163e-06,
+      "loss": 2.1337,
+      "step": 32429
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4649178255372947e-06,
+      "loss": 2.9029,
+      "step": 32430
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464870417193426e-06,
+      "loss": 2.6966,
+      "step": 32431
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4648230088495575e-06,
+      "loss": 3.5916,
+      "step": 32432
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4647756005056892e-06,
+      "loss": 1.5651,
+      "step": 32433
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4647281921618205e-06,
+      "loss": 2.4412,
+      "step": 32434
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464680783817952e-06,
+      "loss": 2.5165,
+      "step": 32435
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4646333754740835e-06,
+      "loss": 2.5432,
+      "step": 32436
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464585967130215e-06,
+      "loss": 1.2744,
+      "step": 32437
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4645385587863464e-06,
+      "loss": 2.487,
+      "step": 32438
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464491150442478e-06,
+      "loss": 1.7528,
+      "step": 32439
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4644437420986094e-06,
+      "loss": 1.7256,
+      "step": 32440
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464396333754741e-06,
+      "loss": 1.2681,
+      "step": 32441
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4643489254108722e-06,
+      "loss": 2.3216,
+      "step": 32442
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.464301517067004e-06,
+      "loss": 2.349,
+      "step": 32443
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4642541087231354e-06,
+      "loss": 3.4767,
+      "step": 32444
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4642067003792667e-06,
+      "loss": 2.1424,
+      "step": 32445
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4641592920353984e-06,
+      "loss": 1.8617,
+      "step": 32446
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4641118836915299e-06,
+      "loss": 2.8092,
+      "step": 32447
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4640644753476612e-06,
+      "loss": 3.6868,
+      "step": 32448
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4640170670037926e-06,
+      "loss": 2.0085,
+      "step": 32449
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4639696586599241e-06,
+      "loss": 2.4865,
+      "step": 32450
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4639222503160556e-06,
+      "loss": 3.1792,
+      "step": 32451
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4638748419721871e-06,
+      "loss": 2.2884,
+      "step": 32452
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4638274336283186e-06,
+      "loss": 4.0197,
+      "step": 32453
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4637800252844501e-06,
+      "loss": 2.6347,
+      "step": 32454
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.4637326169405816e-06,
+      "loss": 2.0273,
+      "step": 32455
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 1.463685208596713e-06,
+      "loss": 2.0601,
+      "step": 32456
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4636378002528446e-06,
+      "loss": 2.6452,
+      "step": 32457
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.463590391908976e-06,
+      "loss": 2.1856,
+      "step": 32458
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4635429835651074e-06,
+      "loss": 1.7123,
+      "step": 32459
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.463495575221239e-06,
+      "loss": 2.497,
+      "step": 32460
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4634481668773703e-06,
+      "loss": 2.878,
+      "step": 32461
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4634007585335018e-06,
+      "loss": 1.1209,
+      "step": 32462
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4633533501896335e-06,
+      "loss": 3.4242,
+      "step": 32463
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4633059418457648e-06,
+      "loss": 2.9195,
+      "step": 32464
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4632585335018963e-06,
+      "loss": 2.4814,
+      "step": 32465
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4632111251580278e-06,
+      "loss": 0.7295,
+      "step": 32466
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4631637168141593e-06,
+      "loss": 2.3943,
+      "step": 32467
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4631163084702908e-06,
+      "loss": 2.7812,
+      "step": 32468
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4630689001264223e-06,
+      "loss": 2.5857,
+      "step": 32469
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4630214917825538e-06,
+      "loss": 1.2314,
+      "step": 32470
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4629740834386853e-06,
+      "loss": 2.0718,
+      "step": 32471
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4629266750948165e-06,
+      "loss": 3.1034,
+      "step": 32472
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4628792667509482e-06,
+      "loss": 1.9671,
+      "step": 32473
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4628318584070797e-06,
+      "loss": 2.1235,
+      "step": 32474
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462784450063211e-06,
+      "loss": 1.4509,
+      "step": 32475
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4627370417193427e-06,
+      "loss": 2.4691,
+      "step": 32476
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4626896333754742e-06,
+      "loss": 1.032,
+      "step": 32477
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4626422250316055e-06,
+      "loss": 3.288,
+      "step": 32478
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462594816687737e-06,
+      "loss": 2.6662,
+      "step": 32479
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4625474083438687e-06,
+      "loss": 3.1389,
+      "step": 32480
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4625e-06,
+      "loss": 2.0935,
+      "step": 32481
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4624525916561315e-06,
+      "loss": 3.146,
+      "step": 32482
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462405183312263e-06,
+      "loss": 1.1055,
+      "step": 32483
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4623577749683945e-06,
+      "loss": 3.6189,
+      "step": 32484
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462310366624526e-06,
+      "loss": 1.8516,
+      "step": 32485
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4622629582806574e-06,
+      "loss": 2.4853,
+      "step": 32486
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462215549936789e-06,
+      "loss": 1.7147,
+      "step": 32487
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4621681415929204e-06,
+      "loss": 2.0099,
+      "step": 32488
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462120733249052e-06,
+      "loss": 1.2127,
+      "step": 32489
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4620733249051834e-06,
+      "loss": 1.7797,
+      "step": 32490
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.462025916561315e-06,
+      "loss": 3.1749,
+      "step": 32491
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4619785082174462e-06,
+      "loss": 2.03,
+      "step": 32492
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4619310998735779e-06,
+      "loss": 1.5777,
+      "step": 32493
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4618836915297092e-06,
+      "loss": 2.1615,
+      "step": 32494
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4618362831858407e-06,
+      "loss": 1.2997,
+      "step": 32495
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4617888748419724e-06,
+      "loss": 1.7408,
+      "step": 32496
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4617414664981036e-06,
+      "loss": 2.1831,
+      "step": 32497
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4616940581542351e-06,
+      "loss": 1.5647,
+      "step": 32498
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4616466498103666e-06,
+      "loss": 2.428,
+      "step": 32499
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4615992414664981e-06,
+      "loss": 1.0933,
+      "step": 32500
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4615518331226296e-06,
+      "loss": 3.0439,
+      "step": 32501
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.461504424778761e-06,
+      "loss": 2.2362,
+      "step": 32502
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4614570164348926e-06,
+      "loss": 2.7273,
+      "step": 32503
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.461409608091024e-06,
+      "loss": 2.8083,
+      "step": 32504
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4613621997471554e-06,
+      "loss": 2.6369,
+      "step": 32505
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.461314791403287e-06,
+      "loss": 1.8811,
+      "step": 32506
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4612673830594186e-06,
+      "loss": 2.3553,
+      "step": 32507
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4612199747155498e-06,
+      "loss": 2.3596,
+      "step": 32508
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4611725663716815e-06,
+      "loss": 2.8251,
+      "step": 32509
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.461125158027813e-06,
+      "loss": 2.4387,
+      "step": 32510
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4610777496839443e-06,
+      "loss": 2.52,
+      "step": 32511
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4610303413400758e-06,
+      "loss": 2.4287,
+      "step": 32512
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4609829329962075e-06,
+      "loss": 0.9997,
+      "step": 32513
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4609355246523388e-06,
+      "loss": 1.4771,
+      "step": 32514
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4608881163084703e-06,
+      "loss": 1.9386,
+      "step": 32515
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4608407079646018e-06,
+      "loss": 2.9219,
+      "step": 32516
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4607932996207333e-06,
+      "loss": 1.7556,
+      "step": 32517
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4607458912768648e-06,
+      "loss": 1.664,
+      "step": 32518
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 1.4606984829329963e-06,
+      "loss": 3.7945,
+      "step": 32519
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4606510745891277e-06,
+      "loss": 2.2976,
+      "step": 32520
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4606036662452592e-06,
+      "loss": 0.9439,
+      "step": 32521
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4605562579013907e-06,
+      "loss": 1.2189,
+      "step": 32522
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4605088495575222e-06,
+      "loss": 2.2981,
+      "step": 32523
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4604614412136537e-06,
+      "loss": 2.9098,
+      "step": 32524
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.460414032869785e-06,
+      "loss": 2.6904,
+      "step": 32525
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4603666245259167e-06,
+      "loss": 1.4743,
+      "step": 32526
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.460319216182048e-06,
+      "loss": 3.1795,
+      "step": 32527
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4602718078381795e-06,
+      "loss": 2.3089,
+      "step": 32528
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4602243994943112e-06,
+      "loss": 1.2115,
+      "step": 32529
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4601769911504425e-06,
+      "loss": 2.9951,
+      "step": 32530
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.460129582806574e-06,
+      "loss": 2.8759,
+      "step": 32531
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4600821744627054e-06,
+      "loss": 2.0656,
+      "step": 32532
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.460034766118837e-06,
+      "loss": 2.9645,
+      "step": 32533
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4599873577749684e-06,
+      "loss": 2.3419,
+      "step": 32534
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4599399494311e-06,
+      "loss": 3.0572,
+      "step": 32535
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4598925410872314e-06,
+      "loss": 2.3102,
+      "step": 32536
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.459845132743363e-06,
+      "loss": 2.381,
+      "step": 32537
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4597977243994942e-06,
+      "loss": 3.3954,
+      "step": 32538
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4597503160556259e-06,
+      "loss": 2.9875,
+      "step": 32539
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4597029077117574e-06,
+      "loss": 2.2797,
+      "step": 32540
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4596554993678887e-06,
+      "loss": 2.4046,
+      "step": 32541
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4596080910240204e-06,
+      "loss": 1.3697,
+      "step": 32542
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4595606826801519e-06,
+      "loss": 2.2578,
+      "step": 32543
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4595132743362831e-06,
+      "loss": 2.5506,
+      "step": 32544
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4594658659924146e-06,
+      "loss": 2.5523,
+      "step": 32545
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4594184576485463e-06,
+      "loss": 1.8964,
+      "step": 32546
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4593710493046776e-06,
+      "loss": 3.4987,
+      "step": 32547
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.459323640960809e-06,
+      "loss": 1.4317,
+      "step": 32548
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4592762326169406e-06,
+      "loss": 2.5352,
+      "step": 32549
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.459228824273072e-06,
+      "loss": 2.4546,
+      "step": 32550
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4591814159292036e-06,
+      "loss": 3.0435,
+      "step": 32551
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.459134007585335e-06,
+      "loss": 1.3954,
+      "step": 32552
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4590865992414666e-06,
+      "loss": 1.2874,
+      "step": 32553
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.459039190897598e-06,
+      "loss": 3.2745,
+      "step": 32554
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4589917825537295e-06,
+      "loss": 1.3313,
+      "step": 32555
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.458944374209861e-06,
+      "loss": 3.2595,
+      "step": 32556
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4588969658659925e-06,
+      "loss": 2.856,
+      "step": 32557
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4588495575221238e-06,
+      "loss": 2.9105,
+      "step": 32558
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4588021491782555e-06,
+      "loss": 2.2186,
+      "step": 32559
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4587547408343868e-06,
+      "loss": 2.1824,
+      "step": 32560
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4587073324905183e-06,
+      "loss": 2.5863,
+      "step": 32561
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.45865992414665e-06,
+      "loss": 2.0887,
+      "step": 32562
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4586125158027813e-06,
+      "loss": 1.8493,
+      "step": 32563
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4585651074589128e-06,
+      "loss": 1.7811,
+      "step": 32564
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4585176991150443e-06,
+      "loss": 2.8386,
+      "step": 32565
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4584702907711757e-06,
+      "loss": 2.2607,
+      "step": 32566
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4584228824273072e-06,
+      "loss": 0.8239,
+      "step": 32567
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4583754740834387e-06,
+      "loss": 0.829,
+      "step": 32568
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4583280657395702e-06,
+      "loss": 2.2075,
+      "step": 32569
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4582806573957017e-06,
+      "loss": 2.1025,
+      "step": 32570
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.458233249051833e-06,
+      "loss": 2.0936,
+      "step": 32571
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4581858407079647e-06,
+      "loss": 1.9705,
+      "step": 32572
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4581384323640962e-06,
+      "loss": 2.6813,
+      "step": 32573
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4580910240202275e-06,
+      "loss": 3.2339,
+      "step": 32574
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4580436156763592e-06,
+      "loss": 1.4845,
+      "step": 32575
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4579962073324907e-06,
+      "loss": 0.9558,
+      "step": 32576
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.457948798988622e-06,
+      "loss": 3.2865,
+      "step": 32577
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4579013906447534e-06,
+      "loss": 2.3183,
+      "step": 32578
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4578539823008851e-06,
+      "loss": 3.0815,
+      "step": 32579
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4578065739570164e-06,
+      "loss": 3.4296,
+      "step": 32580
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.457759165613148e-06,
+      "loss": 2.1206,
+      "step": 32581
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.4577117572692794e-06,
+      "loss": 2.0869,
+      "step": 32582
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 1.457664348925411e-06,
+      "loss": 2.7569,
+      "step": 32583
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4576169405815424e-06,
+      "loss": 2.2284,
+      "step": 32584
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4575695322376739e-06,
+      "loss": 2.5493,
+      "step": 32585
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4575221238938054e-06,
+      "loss": 2.3706,
+      "step": 32586
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4574747155499369e-06,
+      "loss": 2.279,
+      "step": 32587
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4574273072060684e-06,
+      "loss": 1.8418,
+      "step": 32588
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4573798988621999e-06,
+      "loss": 3.0402,
+      "step": 32589
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4573324905183314e-06,
+      "loss": 2.0637,
+      "step": 32590
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4572850821744626e-06,
+      "loss": 2.7771,
+      "step": 32591
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4572376738305943e-06,
+      "loss": 1.6432,
+      "step": 32592
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4571902654867256e-06,
+      "loss": 2.3443,
+      "step": 32593
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4571428571428571e-06,
+      "loss": 2.7273,
+      "step": 32594
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4570954487989888e-06,
+      "loss": 3.3854,
+      "step": 32595
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.45704804045512e-06,
+      "loss": 2.6527,
+      "step": 32596
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4570006321112516e-06,
+      "loss": 1.6083,
+      "step": 32597
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.456953223767383e-06,
+      "loss": 3.2466,
+      "step": 32598
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4569058154235146e-06,
+      "loss": 3.0736,
+      "step": 32599
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.456858407079646e-06,
+      "loss": 2.4617,
+      "step": 32600
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4568109987357776e-06,
+      "loss": 2.5023,
+      "step": 32601
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.456763590391909e-06,
+      "loss": 1.9292,
+      "step": 32602
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4567161820480405e-06,
+      "loss": 2.1362,
+      "step": 32603
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4566687737041718e-06,
+      "loss": 2.5553,
+      "step": 32604
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4566213653603035e-06,
+      "loss": 1.7272,
+      "step": 32605
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.456573957016435e-06,
+      "loss": 2.1348,
+      "step": 32606
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4565265486725663e-06,
+      "loss": 2.9233,
+      "step": 32607
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4564791403286978e-06,
+      "loss": 2.6542,
+      "step": 32608
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4564317319848295e-06,
+      "loss": 1.675,
+      "step": 32609
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4563843236409608e-06,
+      "loss": 1.3048,
+      "step": 32610
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4563369152970923e-06,
+      "loss": 2.508,
+      "step": 32611
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.456289506953224e-06,
+      "loss": 1.9767,
+      "step": 32612
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4562420986093552e-06,
+      "loss": 2.7688,
+      "step": 32613
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4561946902654867e-06,
+      "loss": 3.0254,
+      "step": 32614
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4561472819216182e-06,
+      "loss": 2.2652,
+      "step": 32615
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4560998735777497e-06,
+      "loss": 2.2425,
+      "step": 32616
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4560524652338812e-06,
+      "loss": 2.2402,
+      "step": 32617
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4560050568900127e-06,
+      "loss": 4.4488,
+      "step": 32618
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4559576485461442e-06,
+      "loss": 1.313,
+      "step": 32619
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4559102402022757e-06,
+      "loss": 2.0821,
+      "step": 32620
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.455862831858407e-06,
+      "loss": 1.9705,
+      "step": 32621
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4558154235145387e-06,
+      "loss": 1.9955,
+      "step": 32622
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4557680151706702e-06,
+      "loss": 3.527,
+      "step": 32623
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4557206068268014e-06,
+      "loss": 1.9154,
+      "step": 32624
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4556731984829332e-06,
+      "loss": 1.4793,
+      "step": 32625
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4556257901390644e-06,
+      "loss": 2.278,
+      "step": 32626
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.455578381795196e-06,
+      "loss": 1.7486,
+      "step": 32627
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4555309734513274e-06,
+      "loss": 2.876,
+      "step": 32628
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.455483565107459e-06,
+      "loss": 2.3477,
+      "step": 32629
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4554361567635904e-06,
+      "loss": 2.7686,
+      "step": 32630
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.455388748419722e-06,
+      "loss": 3.6635,
+      "step": 32631
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4553413400758534e-06,
+      "loss": 2.3311,
+      "step": 32632
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4552939317319849e-06,
+      "loss": 2.243,
+      "step": 32633
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4552465233881164e-06,
+      "loss": 1.1983,
+      "step": 32634
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4551991150442479e-06,
+      "loss": 2.7495,
+      "step": 32635
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4551517067003794e-06,
+      "loss": 2.0716,
+      "step": 32636
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4551042983565106e-06,
+      "loss": 1.8693,
+      "step": 32637
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4550568900126423e-06,
+      "loss": 0.7985,
+      "step": 32638
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4550094816687738e-06,
+      "loss": 2.3707,
+      "step": 32639
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4549620733249051e-06,
+      "loss": 2.3727,
+      "step": 32640
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4549146649810366e-06,
+      "loss": 2.1263,
+      "step": 32641
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4548672566371683e-06,
+      "loss": 2.0543,
+      "step": 32642
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4548198482932996e-06,
+      "loss": 1.1476,
+      "step": 32643
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.454772439949431e-06,
+      "loss": 3.2375,
+      "step": 32644
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.4547250316055628e-06,
+      "loss": 2.1662,
+      "step": 32645
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 1.454677623261694e-06,
+      "loss": 2.0619,
+      "step": 32646
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4546302149178256e-06,
+      "loss": 1.994,
+      "step": 32647
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.454582806573957e-06,
+      "loss": 3.147,
+      "step": 32648
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4545353982300885e-06,
+      "loss": 2.2097,
+      "step": 32649
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.45448798988622e-06,
+      "loss": 1.5128,
+      "step": 32650
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4544405815423515e-06,
+      "loss": 1.5311,
+      "step": 32651
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.454393173198483e-06,
+      "loss": 2.5496,
+      "step": 32652
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4543457648546145e-06,
+      "loss": 3.9717,
+      "step": 32653
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4542983565107458e-06,
+      "loss": 3.1435,
+      "step": 32654
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4542509481668775e-06,
+      "loss": 2.0472,
+      "step": 32655
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.454203539823009e-06,
+      "loss": 2.2989,
+      "step": 32656
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4541561314791403e-06,
+      "loss": 1.727,
+      "step": 32657
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.454108723135272e-06,
+      "loss": 2.8069,
+      "step": 32658
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4540613147914033e-06,
+      "loss": 2.6344,
+      "step": 32659
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4540139064475347e-06,
+      "loss": 2.5481,
+      "step": 32660
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4539664981036662e-06,
+      "loss": 3.9176,
+      "step": 32661
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4539190897597977e-06,
+      "loss": 3.0785,
+      "step": 32662
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4538716814159292e-06,
+      "loss": 2.5849,
+      "step": 32663
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4538242730720607e-06,
+      "loss": 3.3573,
+      "step": 32664
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4537768647281922e-06,
+      "loss": 1.3104,
+      "step": 32665
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4537294563843237e-06,
+      "loss": 3.0524,
+      "step": 32666
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4536820480404552e-06,
+      "loss": 1.333,
+      "step": 32667
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4536346396965867e-06,
+      "loss": 1.7551,
+      "step": 32668
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4535872313527182e-06,
+      "loss": 2.3737,
+      "step": 32669
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4535398230088495e-06,
+      "loss": 1.8471,
+      "step": 32670
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4534924146649812e-06,
+      "loss": 2.3906,
+      "step": 32671
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4534450063211126e-06,
+      "loss": 1.4815,
+      "step": 32672
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.453397597977244e-06,
+      "loss": 1.7537,
+      "step": 32673
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4533501896333754e-06,
+      "loss": 3.4943,
+      "step": 32674
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4533027812895071e-06,
+      "loss": 1.1143,
+      "step": 32675
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4532553729456384e-06,
+      "loss": 1.9433,
+      "step": 32676
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.45320796460177e-06,
+      "loss": 3.329,
+      "step": 32677
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4531605562579016e-06,
+      "loss": 2.2459,
+      "step": 32678
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4531131479140329e-06,
+      "loss": 2.2197,
+      "step": 32679
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4530657395701644e-06,
+      "loss": 2.1098,
+      "step": 32680
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4530183312262959e-06,
+      "loss": 4.6955,
+      "step": 32681
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4529709228824274e-06,
+      "loss": 1.6151,
+      "step": 32682
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4529235145385589e-06,
+      "loss": 1.6187,
+      "step": 32683
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4528761061946903e-06,
+      "loss": 2.6857,
+      "step": 32684
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4528286978508218e-06,
+      "loss": 2.1683,
+      "step": 32685
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4527812895069533e-06,
+      "loss": 2.8828,
+      "step": 32686
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4527338811630846e-06,
+      "loss": 1.6582,
+      "step": 32687
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4526864728192163e-06,
+      "loss": 2.4566,
+      "step": 32688
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4526390644753478e-06,
+      "loss": 2.3521,
+      "step": 32689
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452591656131479e-06,
+      "loss": 2.0267,
+      "step": 32690
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4525442477876108e-06,
+      "loss": 3.1451,
+      "step": 32691
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452496839443742e-06,
+      "loss": 2.8485,
+      "step": 32692
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4524494310998736e-06,
+      "loss": 1.7116,
+      "step": 32693
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452402022756005e-06,
+      "loss": 2.4065,
+      "step": 32694
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4523546144121365e-06,
+      "loss": 1.7017,
+      "step": 32695
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452307206068268e-06,
+      "loss": 1.9745,
+      "step": 32696
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4522597977243995e-06,
+      "loss": 1.6598,
+      "step": 32697
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452212389380531e-06,
+      "loss": 2.5836,
+      "step": 32698
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4521649810366625e-06,
+      "loss": 2.9401,
+      "step": 32699
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452117572692794e-06,
+      "loss": 1.2799,
+      "step": 32700
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4520701643489255e-06,
+      "loss": 2.5338,
+      "step": 32701
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.452022756005057e-06,
+      "loss": 1.6735,
+      "step": 32702
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4519753476611883e-06,
+      "loss": 2.1235,
+      "step": 32703
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.45192793931732e-06,
+      "loss": 3.6464,
+      "step": 32704
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4518805309734515e-06,
+      "loss": 2.203,
+      "step": 32705
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4518331226295827e-06,
+      "loss": 1.3172,
+      "step": 32706
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4517857142857142e-06,
+      "loss": 0.6671,
+      "step": 32707
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.451738305941846e-06,
+      "loss": 2.7856,
+      "step": 32708
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 1.4516908975979772e-06,
+      "loss": 2.4652,
+      "step": 32709
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4516434892541087e-06,
+      "loss": 1.9984,
+      "step": 32710
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4515960809102404e-06,
+      "loss": 3.2114,
+      "step": 32711
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4515486725663717e-06,
+      "loss": 2.2025,
+      "step": 32712
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4515012642225032e-06,
+      "loss": 2.9704,
+      "step": 32713
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4514538558786347e-06,
+      "loss": 3.1582,
+      "step": 32714
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4514064475347662e-06,
+      "loss": 3.3497,
+      "step": 32715
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4513590391908977e-06,
+      "loss": 1.5894,
+      "step": 32716
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4513116308470292e-06,
+      "loss": 2.7007,
+      "step": 32717
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4512642225031607e-06,
+      "loss": 2.3057,
+      "step": 32718
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4512168141592921e-06,
+      "loss": 2.0035,
+      "step": 32719
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4511694058154234e-06,
+      "loss": 1.5392,
+      "step": 32720
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4511219974715551e-06,
+      "loss": 2.446,
+      "step": 32721
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4510745891276866e-06,
+      "loss": 1.9624,
+      "step": 32722
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.451027180783818e-06,
+      "loss": 1.8578,
+      "step": 32723
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4509797724399496e-06,
+      "loss": 2.3802,
+      "step": 32724
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4509323640960809e-06,
+      "loss": 3.4742,
+      "step": 32725
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4508849557522124e-06,
+      "loss": 3.076,
+      "step": 32726
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4508375474083439e-06,
+      "loss": 2.4468,
+      "step": 32727
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4507901390644754e-06,
+      "loss": 4.0185,
+      "step": 32728
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4507427307206069e-06,
+      "loss": 1.4369,
+      "step": 32729
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4506953223767383e-06,
+      "loss": 2.2631,
+      "step": 32730
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4506479140328698e-06,
+      "loss": 1.7335,
+      "step": 32731
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4506005056890013e-06,
+      "loss": 1.8721,
+      "step": 32732
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4505530973451328e-06,
+      "loss": 0.7646,
+      "step": 32733
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4505056890012643e-06,
+      "loss": 2.7296,
+      "step": 32734
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4504582806573958e-06,
+      "loss": 2.7372,
+      "step": 32735
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.450410872313527e-06,
+      "loss": 1.2709,
+      "step": 32736
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4503634639696586e-06,
+      "loss": 0.7016,
+      "step": 32737
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4503160556257903e-06,
+      "loss": 1.4228,
+      "step": 32738
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4502686472819216e-06,
+      "loss": 3.5951,
+      "step": 32739
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.450221238938053e-06,
+      "loss": 4.2175,
+      "step": 32740
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4501738305941848e-06,
+      "loss": 1.4224,
+      "step": 32741
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.450126422250316e-06,
+      "loss": 2.5859,
+      "step": 32742
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4500790139064475e-06,
+      "loss": 1.5979,
+      "step": 32743
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.450031605562579e-06,
+      "loss": 3.4029,
+      "step": 32744
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4499841972187105e-06,
+      "loss": 1.1854,
+      "step": 32745
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.449936788874842e-06,
+      "loss": 1.7423,
+      "step": 32746
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4498893805309735e-06,
+      "loss": 2.2949,
+      "step": 32747
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.449841972187105e-06,
+      "loss": 1.6123,
+      "step": 32748
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4497945638432365e-06,
+      "loss": 2.7895,
+      "step": 32749
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4497471554993678e-06,
+      "loss": 2.4892,
+      "step": 32750
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4496997471554995e-06,
+      "loss": 0.4966,
+      "step": 32751
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.449652338811631e-06,
+      "loss": 2.0298,
+      "step": 32752
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4496049304677622e-06,
+      "loss": 2.1108,
+      "step": 32753
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.449557522123894e-06,
+      "loss": 1.3894,
+      "step": 32754
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4495101137800254e-06,
+      "loss": 2.7358,
+      "step": 32755
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4494627054361567e-06,
+      "loss": 1.8798,
+      "step": 32756
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4494152970922882e-06,
+      "loss": 2.3455,
+      "step": 32757
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4493678887484197e-06,
+      "loss": 2.9155,
+      "step": 32758
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4493204804045512e-06,
+      "loss": 2.4508,
+      "step": 32759
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4492730720606827e-06,
+      "loss": 1.5945,
+      "step": 32760
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4492256637168142e-06,
+      "loss": 3.0683,
+      "step": 32761
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4491782553729457e-06,
+      "loss": 1.3949,
+      "step": 32762
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4491308470290772e-06,
+      "loss": 2.5495,
+      "step": 32763
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4490834386852087e-06,
+      "loss": 3.0097,
+      "step": 32764
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4490360303413402e-06,
+      "loss": 2.064,
+      "step": 32765
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4489886219974716e-06,
+      "loss": 2.257,
+      "step": 32766
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4489412136536031e-06,
+      "loss": 2.1693,
+      "step": 32767
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4488938053097346e-06,
+      "loss": 3.3521,
+      "step": 32768
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.448846396965866e-06,
+      "loss": 2.8393,
+      "step": 32769
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4487989886219974e-06,
+      "loss": 1.7998,
+      "step": 32770
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.448751580278129e-06,
+      "loss": 1.865,
+      "step": 32771
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4487041719342604e-06,
+      "loss": 3.663,
+      "step": 32772
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 1.4486567635903919e-06,
+      "loss": 2.8104,
+      "step": 32773
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4486093552465236e-06,
+      "loss": 2.5985,
+      "step": 32774
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4485619469026549e-06,
+      "loss": 1.3141,
+      "step": 32775
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4485145385587864e-06,
+      "loss": 1.7916,
+      "step": 32776
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4484671302149178e-06,
+      "loss": 2.7319,
+      "step": 32777
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4484197218710493e-06,
+      "loss": 1.2742,
+      "step": 32778
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4483723135271808e-06,
+      "loss": 1.5481,
+      "step": 32779
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4483249051833123e-06,
+      "loss": 2.2263,
+      "step": 32780
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4482774968394438e-06,
+      "loss": 2.1619,
+      "step": 32781
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4482300884955753e-06,
+      "loss": 2.0404,
+      "step": 32782
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4481826801517066e-06,
+      "loss": 1.8813,
+      "step": 32783
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4481352718078383e-06,
+      "loss": 1.6173,
+      "step": 32784
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4480878634639698e-06,
+      "loss": 2.7817,
+      "step": 32785
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.448040455120101e-06,
+      "loss": 3.4051,
+      "step": 32786
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4479930467762328e-06,
+      "loss": 3.1154,
+      "step": 32787
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4479456384323643e-06,
+      "loss": 2.7581,
+      "step": 32788
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4478982300884955e-06,
+      "loss": 2.3671,
+      "step": 32789
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.447850821744627e-06,
+      "loss": 1.8975,
+      "step": 32790
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4478034134007585e-06,
+      "loss": 2.978,
+      "step": 32791
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.44775600505689e-06,
+      "loss": 1.3182,
+      "step": 32792
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4477085967130215e-06,
+      "loss": 2.7611,
+      "step": 32793
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.447661188369153e-06,
+      "loss": 2.2912,
+      "step": 32794
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4476137800252845e-06,
+      "loss": 1.8182,
+      "step": 32795
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.447566371681416e-06,
+      "loss": 1.3792,
+      "step": 32796
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4475189633375475e-06,
+      "loss": 3.5158,
+      "step": 32797
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.447471554993679e-06,
+      "loss": 2.3529,
+      "step": 32798
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4474241466498103e-06,
+      "loss": 2.8722,
+      "step": 32799
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.447376738305942e-06,
+      "loss": 3.1997,
+      "step": 32800
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4473293299620734e-06,
+      "loss": 2.2203,
+      "step": 32801
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4472819216182047e-06,
+      "loss": 2.1963,
+      "step": 32802
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4472345132743362e-06,
+      "loss": 2.5055,
+      "step": 32803
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.447187104930468e-06,
+      "loss": 2.6402,
+      "step": 32804
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4471396965865992e-06,
+      "loss": 1.8022,
+      "step": 32805
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4470922882427307e-06,
+      "loss": 1.6381,
+      "step": 32806
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4470448798988624e-06,
+      "loss": 1.9135,
+      "step": 32807
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4469974715549937e-06,
+      "loss": 2.593,
+      "step": 32808
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4469500632111252e-06,
+      "loss": 1.7018,
+      "step": 32809
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4469026548672567e-06,
+      "loss": 2.0929,
+      "step": 32810
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4468552465233882e-06,
+      "loss": 2.9066,
+      "step": 32811
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4468078381795196e-06,
+      "loss": 4.0662,
+      "step": 32812
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4467604298356511e-06,
+      "loss": 2.3113,
+      "step": 32813
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4467130214917826e-06,
+      "loss": 2.312,
+      "step": 32814
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4466656131479141e-06,
+      "loss": 0.555,
+      "step": 32815
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4466182048040454e-06,
+      "loss": 3.1785,
+      "step": 32816
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4465707964601771e-06,
+      "loss": 3.1585,
+      "step": 32817
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4465233881163086e-06,
+      "loss": 3.3729,
+      "step": 32818
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4464759797724399e-06,
+      "loss": 2.9536,
+      "step": 32819
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4464285714285716e-06,
+      "loss": 1.7185,
+      "step": 32820
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4463811630847029e-06,
+      "loss": 2.0392,
+      "step": 32821
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4463337547408344e-06,
+      "loss": 3.0086,
+      "step": 32822
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4462863463969659e-06,
+      "loss": 0.7526,
+      "step": 32823
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4462389380530973e-06,
+      "loss": 1.8163,
+      "step": 32824
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4461915297092288e-06,
+      "loss": 0.7995,
+      "step": 32825
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4461441213653603e-06,
+      "loss": 1.6642,
+      "step": 32826
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4460967130214918e-06,
+      "loss": 3.102,
+      "step": 32827
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4460493046776233e-06,
+      "loss": 2.8117,
+      "step": 32828
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4460018963337548e-06,
+      "loss": 0.8419,
+      "step": 32829
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4459544879898863e-06,
+      "loss": 4.068,
+      "step": 32830
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4459070796460178e-06,
+      "loss": 3.6869,
+      "step": 32831
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.445859671302149e-06,
+      "loss": 0.6906,
+      "step": 32832
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4458122629582808e-06,
+      "loss": 1.7322,
+      "step": 32833
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4457648546144123e-06,
+      "loss": 1.9023,
+      "step": 32834
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.4457174462705435e-06,
+      "loss": 1.1643,
+      "step": 32835
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 1.445670037926675e-06,
+      "loss": 0.9016,
+      "step": 32836
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4456226295828067e-06,
+      "loss": 2.0271,
+      "step": 32837
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.445575221238938e-06,
+      "loss": 2.8053,
+      "step": 32838
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4455278128950695e-06,
+      "loss": 2.0238,
+      "step": 32839
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4454804045512012e-06,
+      "loss": 1.3694,
+      "step": 32840
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4454329962073325e-06,
+      "loss": 1.8209,
+      "step": 32841
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.445385587863464e-06,
+      "loss": 3.3466,
+      "step": 32842
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4453381795195955e-06,
+      "loss": 2.7106,
+      "step": 32843
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.445290771175727e-06,
+      "loss": 2.1504,
+      "step": 32844
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4452433628318585e-06,
+      "loss": 2.6013,
+      "step": 32845
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.44519595448799e-06,
+      "loss": 2.9995,
+      "step": 32846
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4451485461441215e-06,
+      "loss": 3.2663,
+      "step": 32847
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.445101137800253e-06,
+      "loss": 1.8299,
+      "step": 32848
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4450537294563842e-06,
+      "loss": 1.8328,
+      "step": 32849
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.445006321112516e-06,
+      "loss": 1.6205,
+      "step": 32850
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4449589127686474e-06,
+      "loss": 2.1313,
+      "step": 32851
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4449115044247787e-06,
+      "loss": 2.8318,
+      "step": 32852
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4448640960809104e-06,
+      "loss": 2.0051,
+      "step": 32853
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4448166877370417e-06,
+      "loss": 4.1048,
+      "step": 32854
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4447692793931732e-06,
+      "loss": 0.6288,
+      "step": 32855
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4447218710493047e-06,
+      "loss": 4.5831,
+      "step": 32856
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4446744627054362e-06,
+      "loss": 1.4014,
+      "step": 32857
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4446270543615677e-06,
+      "loss": 2.849,
+      "step": 32858
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4445796460176991e-06,
+      "loss": 1.8148,
+      "step": 32859
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4445322376738306e-06,
+      "loss": 2.2016,
+      "step": 32860
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4444848293299621e-06,
+      "loss": 1.8549,
+      "step": 32861
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4444374209860936e-06,
+      "loss": 2.1583,
+      "step": 32862
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4443900126422251e-06,
+      "loss": 1.7563,
+      "step": 32863
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4443426042983566e-06,
+      "loss": 3.6108,
+      "step": 32864
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4442951959544879e-06,
+      "loss": 3.1377,
+      "step": 32865
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4442477876106196e-06,
+      "loss": 2.2408,
+      "step": 32866
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.444200379266751e-06,
+      "loss": 2.2088,
+      "step": 32867
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4441529709228824e-06,
+      "loss": 2.9318,
+      "step": 32868
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4441055625790139e-06,
+      "loss": 3.1743,
+      "step": 32869
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4440581542351456e-06,
+      "loss": 2.7123,
+      "step": 32870
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4440107458912768e-06,
+      "loss": 3.4023,
+      "step": 32871
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4439633375474083e-06,
+      "loss": 1.6119,
+      "step": 32872
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.44391592920354e-06,
+      "loss": 2.982,
+      "step": 32873
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4438685208596713e-06,
+      "loss": 1.4215,
+      "step": 32874
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4438211125158028e-06,
+      "loss": 1.6739,
+      "step": 32875
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4437737041719343e-06,
+      "loss": 2.5824,
+      "step": 32876
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4437262958280658e-06,
+      "loss": 1.8651,
+      "step": 32877
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4436788874841973e-06,
+      "loss": 2.9055,
+      "step": 32878
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4436314791403286e-06,
+      "loss": 2.8581,
+      "step": 32879
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4435840707964603e-06,
+      "loss": 1.7781,
+      "step": 32880
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4435366624525918e-06,
+      "loss": 2.5425,
+      "step": 32881
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.443489254108723e-06,
+      "loss": 4.6147,
+      "step": 32882
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4434418457648547e-06,
+      "loss": 1.0917,
+      "step": 32883
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4433944374209862e-06,
+      "loss": 1.8861,
+      "step": 32884
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4433470290771175e-06,
+      "loss": 1.9292,
+      "step": 32885
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.443299620733249e-06,
+      "loss": 2.0693,
+      "step": 32886
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4432522123893805e-06,
+      "loss": 2.7417,
+      "step": 32887
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.443204804045512e-06,
+      "loss": 1.1428,
+      "step": 32888
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4431573957016435e-06,
+      "loss": 2.2704,
+      "step": 32889
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.443109987357775e-06,
+      "loss": 1.0211,
+      "step": 32890
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4430625790139065e-06,
+      "loss": 2.1195,
+      "step": 32891
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.443015170670038e-06,
+      "loss": 1.9645,
+      "step": 32892
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4429677623261695e-06,
+      "loss": 2.1136,
+      "step": 32893
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.442920353982301e-06,
+      "loss": 1.3385,
+      "step": 32894
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4428729456384324e-06,
+      "loss": 3.6391,
+      "step": 32895
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.442825537294564e-06,
+      "loss": 4.3408,
+      "step": 32896
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4427781289506954e-06,
+      "loss": 2.9879,
+      "step": 32897
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4427307206068267e-06,
+      "loss": 1.725,
+      "step": 32898
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 1.4426833122629582e-06,
+      "loss": 3.1979,
+      "step": 32899
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.44263590391909e-06,
+      "loss": 2.5202,
+      "step": 32900
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4425884955752212e-06,
+      "loss": 2.754,
+      "step": 32901
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4425410872313527e-06,
+      "loss": 1.5423,
+      "step": 32902
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4424936788874844e-06,
+      "loss": 2.0348,
+      "step": 32903
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4424462705436157e-06,
+      "loss": 2.3884,
+      "step": 32904
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4423988621997472e-06,
+      "loss": 3.3812,
+      "step": 32905
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4423514538558786e-06,
+      "loss": 0.6107,
+      "step": 32906
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4423040455120101e-06,
+      "loss": 3.4562,
+      "step": 32907
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4422566371681416e-06,
+      "loss": 3.4675,
+      "step": 32908
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4422092288242731e-06,
+      "loss": 2.8855,
+      "step": 32909
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4421618204804046e-06,
+      "loss": 1.0509,
+      "step": 32910
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.442114412136536e-06,
+      "loss": 2.0305,
+      "step": 32911
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4420670037926674e-06,
+      "loss": 2.333,
+      "step": 32912
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.442019595448799e-06,
+      "loss": 2.5082,
+      "step": 32913
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4419721871049306e-06,
+      "loss": 1.6671,
+      "step": 32914
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4419247787610619e-06,
+      "loss": 2.0452,
+      "step": 32915
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4418773704171936e-06,
+      "loss": 1.8805,
+      "step": 32916
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.441829962073325e-06,
+      "loss": 2.2791,
+      "step": 32917
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4417825537294563e-06,
+      "loss": 2.7991,
+      "step": 32918
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4417351453855878e-06,
+      "loss": 1.2736,
+      "step": 32919
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4416877370417193e-06,
+      "loss": 1.8972,
+      "step": 32920
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4416403286978508e-06,
+      "loss": 2.5886,
+      "step": 32921
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4415929203539823e-06,
+      "loss": 1.8144,
+      "step": 32922
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4415455120101138e-06,
+      "loss": 2.5605,
+      "step": 32923
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4414981036662453e-06,
+      "loss": 1.7938,
+      "step": 32924
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4414506953223768e-06,
+      "loss": 2.9266,
+      "step": 32925
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4414032869785083e-06,
+      "loss": 1.7804,
+      "step": 32926
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4413558786346398e-06,
+      "loss": 2.4474,
+      "step": 32927
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4413084702907713e-06,
+      "loss": 3.0821,
+      "step": 32928
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4412610619469028e-06,
+      "loss": 2.0052,
+      "step": 32929
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4412136536030342e-06,
+      "loss": 1.0976,
+      "step": 32930
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4411662452591655e-06,
+      "loss": 2.3298,
+      "step": 32931
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.441118836915297e-06,
+      "loss": 2.3173,
+      "step": 32932
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4410714285714287e-06,
+      "loss": 1.1315,
+      "step": 32933
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.44102402022756e-06,
+      "loss": 3.4126,
+      "step": 32934
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4409766118836915e-06,
+      "loss": 1.1118,
+      "step": 32935
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4409292035398232e-06,
+      "loss": 1.9119,
+      "step": 32936
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4408817951959545e-06,
+      "loss": 2.4516,
+      "step": 32937
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.440834386852086e-06,
+      "loss": 1.4766,
+      "step": 32938
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4407869785082175e-06,
+      "loss": 1.2763,
+      "step": 32939
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.440739570164349e-06,
+      "loss": 2.353,
+      "step": 32940
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4406921618204804e-06,
+      "loss": 1.4212,
+      "step": 32941
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.440644753476612e-06,
+      "loss": 3.2152,
+      "step": 32942
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4405973451327434e-06,
+      "loss": 2.0264,
+      "step": 32943
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.440549936788875e-06,
+      "loss": 3.5508,
+      "step": 32944
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4405025284450062e-06,
+      "loss": 1.635,
+      "step": 32945
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.440455120101138e-06,
+      "loss": 1.7554,
+      "step": 32946
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4404077117572694e-06,
+      "loss": 1.1085,
+      "step": 32947
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4403603034134007e-06,
+      "loss": 2.0041,
+      "step": 32948
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4403128950695324e-06,
+      "loss": 3.4359,
+      "step": 32949
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4402654867256639e-06,
+      "loss": 2.4989,
+      "step": 32950
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4402180783817952e-06,
+      "loss": 1.3726,
+      "step": 32951
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4401706700379266e-06,
+      "loss": 1.493,
+      "step": 32952
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4401232616940581e-06,
+      "loss": 1.6977,
+      "step": 32953
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4400758533501896e-06,
+      "loss": 2.1071,
+      "step": 32954
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4400284450063211e-06,
+      "loss": 1.8877,
+      "step": 32955
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4399810366624526e-06,
+      "loss": 3.3243,
+      "step": 32956
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4399336283185841e-06,
+      "loss": 2.1738,
+      "step": 32957
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4398862199747156e-06,
+      "loss": 1.7821,
+      "step": 32958
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.439838811630847e-06,
+      "loss": 1.0392,
+      "step": 32959
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4397914032869786e-06,
+      "loss": 3.1238,
+      "step": 32960
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.43974399494311e-06,
+      "loss": 2.6188,
+      "step": 32961
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.4396965865992416e-06,
+      "loss": 1.8735,
+      "step": 32962
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.439649178255373e-06,
+      "loss": 1.6937,
+      "step": 32963
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4396017699115043e-06,
+      "loss": 3.4938,
+      "step": 32964
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4395543615676358e-06,
+      "loss": 1.4571,
+      "step": 32965
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4395069532237675e-06,
+      "loss": 2.5392,
+      "step": 32966
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4394595448798988e-06,
+      "loss": 2.1806,
+      "step": 32967
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4394121365360303e-06,
+      "loss": 1.2712,
+      "step": 32968
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.439364728192162e-06,
+      "loss": 2.6685,
+      "step": 32969
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4393173198482933e-06,
+      "loss": 2.0528,
+      "step": 32970
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4392699115044248e-06,
+      "loss": 2.6398,
+      "step": 32971
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4392225031605563e-06,
+      "loss": 2.5368,
+      "step": 32972
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4391750948166878e-06,
+      "loss": 2.7616,
+      "step": 32973
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4391276864728193e-06,
+      "loss": 1.8172,
+      "step": 32974
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4390802781289508e-06,
+      "loss": 1.6951,
+      "step": 32975
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4390328697850822e-06,
+      "loss": 2.5418,
+      "step": 32976
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4389854614412137e-06,
+      "loss": 2.5374,
+      "step": 32977
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.438938053097345e-06,
+      "loss": 1.9876,
+      "step": 32978
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4388906447534767e-06,
+      "loss": 2.0156,
+      "step": 32979
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4388432364096082e-06,
+      "loss": 2.0837,
+      "step": 32980
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4387958280657395e-06,
+      "loss": 1.6933,
+      "step": 32981
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4387484197218712e-06,
+      "loss": 1.7213,
+      "step": 32982
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4387010113780027e-06,
+      "loss": 2.2912,
+      "step": 32983
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.438653603034134e-06,
+      "loss": 1.9565,
+      "step": 32984
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4386061946902655e-06,
+      "loss": 1.0934,
+      "step": 32985
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.438558786346397e-06,
+      "loss": 2.8206,
+      "step": 32986
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4385113780025285e-06,
+      "loss": 1.5832,
+      "step": 32987
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.43846396965866e-06,
+      "loss": 2.1744,
+      "step": 32988
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4384165613147914e-06,
+      "loss": 2.3727,
+      "step": 32989
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.438369152970923e-06,
+      "loss": 2.3362,
+      "step": 32990
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4383217446270544e-06,
+      "loss": 2.2867,
+      "step": 32991
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.438274336283186e-06,
+      "loss": 1.7835,
+      "step": 32992
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4382269279393174e-06,
+      "loss": 2.7265,
+      "step": 32993
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.438179519595449e-06,
+      "loss": 1.497,
+      "step": 32994
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4381321112515804e-06,
+      "loss": 2.9978,
+      "step": 32995
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4380847029077119e-06,
+      "loss": 1.106,
+      "step": 32996
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4380372945638432e-06,
+      "loss": 1.6618,
+      "step": 32997
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4379898862199747e-06,
+      "loss": 2.221,
+      "step": 32998
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4379424778761064e-06,
+      "loss": 2.2676,
+      "step": 32999
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4378950695322376e-06,
+      "loss": 2.2482,
+      "step": 33000
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4378476611883691e-06,
+      "loss": 1.9995,
+      "step": 33001
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4378002528445008e-06,
+      "loss": 3.1962,
+      "step": 33002
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4377528445006321e-06,
+      "loss": 3.0821,
+      "step": 33003
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4377054361567636e-06,
+      "loss": 3.5528,
+      "step": 33004
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.437658027812895e-06,
+      "loss": 1.4496,
+      "step": 33005
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4376106194690266e-06,
+      "loss": 2.7684,
+      "step": 33006
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.437563211125158e-06,
+      "loss": 1.4099,
+      "step": 33007
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4375158027812894e-06,
+      "loss": 1.465,
+      "step": 33008
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.437468394437421e-06,
+      "loss": 1.7208,
+      "step": 33009
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4374209860935526e-06,
+      "loss": 3.8863,
+      "step": 33010
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4373735777496838e-06,
+      "loss": 1.7354,
+      "step": 33011
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4373261694058155e-06,
+      "loss": 3.7555,
+      "step": 33012
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.437278761061947e-06,
+      "loss": 2.9149,
+      "step": 33013
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4372313527180783e-06,
+      "loss": 1.5381,
+      "step": 33014
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4371839443742098e-06,
+      "loss": 2.2256,
+      "step": 33015
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4371365360303415e-06,
+      "loss": 0.9519,
+      "step": 33016
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4370891276864728e-06,
+      "loss": 1.3092,
+      "step": 33017
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4370417193426043e-06,
+      "loss": 2.317,
+      "step": 33018
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4369943109987358e-06,
+      "loss": 2.2378,
+      "step": 33019
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4369469026548673e-06,
+      "loss": 1.8512,
+      "step": 33020
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4368994943109988e-06,
+      "loss": 1.7694,
+      "step": 33021
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4368520859671303e-06,
+      "loss": 2.8111,
+      "step": 33022
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4368046776232617e-06,
+      "loss": 2.6649,
+      "step": 33023
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4367572692793932e-06,
+      "loss": 1.646,
+      "step": 33024
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4367098609355247e-06,
+      "loss": 1.7847,
+      "step": 33025
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 1.4366624525916562e-06,
+      "loss": 2.8192,
+      "step": 33026
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4366150442477877e-06,
+      "loss": 1.602,
+      "step": 33027
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.436567635903919e-06,
+      "loss": 2.2424,
+      "step": 33028
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4365202275600507e-06,
+      "loss": 1.6811,
+      "step": 33029
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.436472819216182e-06,
+      "loss": 4.233,
+      "step": 33030
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4364254108723135e-06,
+      "loss": 1.1373,
+      "step": 33031
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4363780025284452e-06,
+      "loss": 2.7291,
+      "step": 33032
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4363305941845765e-06,
+      "loss": 1.7406,
+      "step": 33033
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.436283185840708e-06,
+      "loss": 2.9213,
+      "step": 33034
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4362357774968394e-06,
+      "loss": 1.5816,
+      "step": 33035
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.436188369152971e-06,
+      "loss": 1.9902,
+      "step": 33036
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4361409608091024e-06,
+      "loss": 2.3122,
+      "step": 33037
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.436093552465234e-06,
+      "loss": 1.2523,
+      "step": 33038
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4360461441213654e-06,
+      "loss": 1.2017,
+      "step": 33039
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.435998735777497e-06,
+      "loss": 2.8122,
+      "step": 33040
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4359513274336282e-06,
+      "loss": 0.8847,
+      "step": 33041
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4359039190897599e-06,
+      "loss": 1.7561,
+      "step": 33042
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4358565107458914e-06,
+      "loss": 2.2622,
+      "step": 33043
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4358091024020227e-06,
+      "loss": 2.9233,
+      "step": 33044
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4357616940581544e-06,
+      "loss": 1.9601,
+      "step": 33045
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4357142857142859e-06,
+      "loss": 2.4803,
+      "step": 33046
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4356668773704171e-06,
+      "loss": 2.8441,
+      "step": 33047
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4356194690265486e-06,
+      "loss": 1.62,
+      "step": 33048
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4355720606826803e-06,
+      "loss": 2.5997,
+      "step": 33049
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4355246523388116e-06,
+      "loss": 1.7768,
+      "step": 33050
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.435477243994943e-06,
+      "loss": 3.0763,
+      "step": 33051
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4354298356510746e-06,
+      "loss": 1.3977,
+      "step": 33052
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.435382427307206e-06,
+      "loss": 1.6172,
+      "step": 33053
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4353350189633376e-06,
+      "loss": 3.5718,
+      "step": 33054
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.435287610619469e-06,
+      "loss": 2.1984,
+      "step": 33055
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4352402022756006e-06,
+      "loss": 3.562,
+      "step": 33056
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.435192793931732e-06,
+      "loss": 2.6738,
+      "step": 33057
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4351453855878635e-06,
+      "loss": 1.6475,
+      "step": 33058
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.435097977243995e-06,
+      "loss": 2.0802,
+      "step": 33059
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4350505689001265e-06,
+      "loss": 2.5642,
+      "step": 33060
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4350031605562578e-06,
+      "loss": 2.7236,
+      "step": 33061
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4349557522123895e-06,
+      "loss": 3.1847,
+      "step": 33062
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4349083438685208e-06,
+      "loss": 2.2676,
+      "step": 33063
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4348609355246523e-06,
+      "loss": 3.0147,
+      "step": 33064
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.434813527180784e-06,
+      "loss": 1.6375,
+      "step": 33065
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4347661188369153e-06,
+      "loss": 1.5223,
+      "step": 33066
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4347187104930468e-06,
+      "loss": 1.9657,
+      "step": 33067
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4346713021491783e-06,
+      "loss": 2.2439,
+      "step": 33068
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4346238938053098e-06,
+      "loss": 2.8414,
+      "step": 33069
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4345764854614412e-06,
+      "loss": 3.1012,
+      "step": 33070
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4345290771175727e-06,
+      "loss": 1.7662,
+      "step": 33071
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4344816687737042e-06,
+      "loss": 2.0448,
+      "step": 33072
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4344342604298357e-06,
+      "loss": 2.3547,
+      "step": 33073
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.434386852085967e-06,
+      "loss": 2.272,
+      "step": 33074
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4343394437420987e-06,
+      "loss": 2.4822,
+      "step": 33075
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4342920353982302e-06,
+      "loss": 1.5405,
+      "step": 33076
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4342446270543615e-06,
+      "loss": 1.1864,
+      "step": 33077
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4341972187104932e-06,
+      "loss": 2.2745,
+      "step": 33078
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4341498103666247e-06,
+      "loss": 2.8947,
+      "step": 33079
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.434102402022756e-06,
+      "loss": 2.9308,
+      "step": 33080
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4340549936788874e-06,
+      "loss": 1.5794,
+      "step": 33081
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4340075853350191e-06,
+      "loss": 2.8275,
+      "step": 33082
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4339601769911504e-06,
+      "loss": 1.8666,
+      "step": 33083
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.433912768647282e-06,
+      "loss": 2.0571,
+      "step": 33084
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4338653603034134e-06,
+      "loss": 1.3695,
+      "step": 33085
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.433817951959545e-06,
+      "loss": 2.2683,
+      "step": 33086
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4337705436156764e-06,
+      "loss": 2.3126,
+      "step": 33087
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4337231352718079e-06,
+      "loss": 3.0765,
+      "step": 33088
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 1.4336757269279394e-06,
+      "loss": 2.4906,
+      "step": 33089
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4336283185840709e-06,
+      "loss": 2.2993,
+      "step": 33090
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4335809102402024e-06,
+      "loss": 3.2221,
+      "step": 33091
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4335335018963339e-06,
+      "loss": 1.9431,
+      "step": 33092
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4334860935524654e-06,
+      "loss": 2.2675,
+      "step": 33093
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4334386852085966e-06,
+      "loss": 1.4143,
+      "step": 33094
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4333912768647283e-06,
+      "loss": 1.8882,
+      "step": 33095
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4333438685208596e-06,
+      "loss": 3.2021,
+      "step": 33096
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4332964601769911e-06,
+      "loss": 2.3548,
+      "step": 33097
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4332490518331228e-06,
+      "loss": 2.3144,
+      "step": 33098
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.433201643489254e-06,
+      "loss": 2.9748,
+      "step": 33099
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4331542351453856e-06,
+      "loss": 1.4826,
+      "step": 33100
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.433106826801517e-06,
+      "loss": 2.5877,
+      "step": 33101
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4330594184576486e-06,
+      "loss": 1.9106,
+      "step": 33102
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.43301201011378e-06,
+      "loss": 2.1076,
+      "step": 33103
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4329646017699116e-06,
+      "loss": 1.9329,
+      "step": 33104
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.432917193426043e-06,
+      "loss": 2.2116,
+      "step": 33105
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4328697850821745e-06,
+      "loss": 2.4458,
+      "step": 33106
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4328223767383058e-06,
+      "loss": 2.0451,
+      "step": 33107
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4327749683944375e-06,
+      "loss": 2.0303,
+      "step": 33108
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.432727560050569e-06,
+      "loss": 3.3288,
+      "step": 33109
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4326801517067003e-06,
+      "loss": 2.6603,
+      "step": 33110
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.432632743362832e-06,
+      "loss": 4.0217,
+      "step": 33111
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4325853350189635e-06,
+      "loss": 2.8891,
+      "step": 33112
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4325379266750948e-06,
+      "loss": 2.5451,
+      "step": 33113
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4324905183312263e-06,
+      "loss": 2.6874,
+      "step": 33114
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.432443109987358e-06,
+      "loss": 2.4994,
+      "step": 33115
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4323957016434892e-06,
+      "loss": 2.5194,
+      "step": 33116
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4323482932996207e-06,
+      "loss": 2.1599,
+      "step": 33117
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4323008849557522e-06,
+      "loss": 3.4466,
+      "step": 33118
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4322534766118837e-06,
+      "loss": 3.0842,
+      "step": 33119
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4322060682680152e-06,
+      "loss": 1.4246,
+      "step": 33120
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4321586599241467e-06,
+      "loss": 2.9047,
+      "step": 33121
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4321112515802782e-06,
+      "loss": 1.6346,
+      "step": 33122
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4320638432364097e-06,
+      "loss": 2.3906,
+      "step": 33123
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4320164348925412e-06,
+      "loss": 2.3992,
+      "step": 33124
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4319690265486727e-06,
+      "loss": 2.2176,
+      "step": 33125
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4319216182048042e-06,
+      "loss": 1.3868,
+      "step": 33126
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4318742098609354e-06,
+      "loss": 2.0742,
+      "step": 33127
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4318268015170672e-06,
+      "loss": 2.2093,
+      "step": 33128
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4317793931731984e-06,
+      "loss": 3.0022,
+      "step": 33129
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.43173198482933e-06,
+      "loss": 1.9813,
+      "step": 33130
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4316845764854616e-06,
+      "loss": 1.6305,
+      "step": 33131
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.431637168141593e-06,
+      "loss": 2.0024,
+      "step": 33132
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4315897597977244e-06,
+      "loss": 2.7447,
+      "step": 33133
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.431542351453856e-06,
+      "loss": 2.6602,
+      "step": 33134
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4314949431099874e-06,
+      "loss": 1.2867,
+      "step": 33135
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4314475347661189e-06,
+      "loss": 2.8371,
+      "step": 33136
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4314001264222504e-06,
+      "loss": 2.1234,
+      "step": 33137
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4313527180783819e-06,
+      "loss": 1.5607,
+      "step": 33138
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4313053097345134e-06,
+      "loss": 1.5105,
+      "step": 33139
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4312579013906446e-06,
+      "loss": 3.1411,
+      "step": 33140
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4312104930467763e-06,
+      "loss": 2.6311,
+      "step": 33141
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4311630847029078e-06,
+      "loss": 2.1345,
+      "step": 33142
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4311156763590391e-06,
+      "loss": 1.9847,
+      "step": 33143
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4310682680151708e-06,
+      "loss": 1.6657,
+      "step": 33144
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4310208596713023e-06,
+      "loss": 2.2753,
+      "step": 33145
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4309734513274336e-06,
+      "loss": 2.771,
+      "step": 33146
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.430926042983565e-06,
+      "loss": 1.6866,
+      "step": 33147
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4308786346396968e-06,
+      "loss": 3.5643,
+      "step": 33148
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.430831226295828e-06,
+      "loss": 1.65,
+      "step": 33149
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4307838179519596e-06,
+      "loss": 1.6626,
+      "step": 33150
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.430736409608091e-06,
+      "loss": 2.8606,
+      "step": 33151
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.4306890012642225e-06,
+      "loss": 2.1744,
+      "step": 33152
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 1.430641592920354e-06,
+      "loss": 2.2655,
+      "step": 33153
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4305941845764855e-06,
+      "loss": 2.8437,
+      "step": 33154
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.430546776232617e-06,
+      "loss": 2.1856,
+      "step": 33155
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4304993678887485e-06,
+      "loss": 2.6867,
+      "step": 33156
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4304519595448798e-06,
+      "loss": 2.2753,
+      "step": 33157
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4304045512010115e-06,
+      "loss": 1.6789,
+      "step": 33158
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.430357142857143e-06,
+      "loss": 2.8866,
+      "step": 33159
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4303097345132743e-06,
+      "loss": 3.0535,
+      "step": 33160
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.430262326169406e-06,
+      "loss": 0.9878,
+      "step": 33161
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4302149178255373e-06,
+      "loss": 1.9239,
+      "step": 33162
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4301675094816687e-06,
+      "loss": 2.2629,
+      "step": 33163
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4301201011378002e-06,
+      "loss": 2.7775,
+      "step": 33164
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4300726927939317e-06,
+      "loss": 2.1566,
+      "step": 33165
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4300252844500632e-06,
+      "loss": 2.588,
+      "step": 33166
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4299778761061947e-06,
+      "loss": 1.1995,
+      "step": 33167
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4299304677623262e-06,
+      "loss": 2.1101,
+      "step": 33168
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4298830594184577e-06,
+      "loss": 1.8675,
+      "step": 33169
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4298356510745892e-06,
+      "loss": 1.3151,
+      "step": 33170
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4297882427307207e-06,
+      "loss": 3.5472,
+      "step": 33171
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4297408343868522e-06,
+      "loss": 2.3936,
+      "step": 33172
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4296934260429835e-06,
+      "loss": 1.3925,
+      "step": 33173
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4296460176991152e-06,
+      "loss": 2.2434,
+      "step": 33174
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4295986093552467e-06,
+      "loss": 1.8639,
+      "step": 33175
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.429551201011378e-06,
+      "loss": 1.7461,
+      "step": 33176
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4295037926675094e-06,
+      "loss": 2.0712,
+      "step": 33177
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4294563843236411e-06,
+      "loss": 3.2827,
+      "step": 33178
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4294089759797724e-06,
+      "loss": 2.8475,
+      "step": 33179
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.429361567635904e-06,
+      "loss": 2.5612,
+      "step": 33180
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4293141592920356e-06,
+      "loss": 3.6201,
+      "step": 33181
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4292667509481669e-06,
+      "loss": 2.9969,
+      "step": 33182
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4292193426042984e-06,
+      "loss": 2.1054,
+      "step": 33183
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4291719342604299e-06,
+      "loss": 2.7742,
+      "step": 33184
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4291245259165614e-06,
+      "loss": 1.3687,
+      "step": 33185
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4290771175726929e-06,
+      "loss": 2.1249,
+      "step": 33186
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4290297092288243e-06,
+      "loss": 1.918,
+      "step": 33187
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4289823008849558e-06,
+      "loss": 0.8839,
+      "step": 33188
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4289348925410873e-06,
+      "loss": 2.4739,
+      "step": 33189
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4288874841972186e-06,
+      "loss": 1.5744,
+      "step": 33190
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4288400758533503e-06,
+      "loss": 3.9173,
+      "step": 33191
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4287926675094818e-06,
+      "loss": 2.8841,
+      "step": 33192
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428745259165613e-06,
+      "loss": 1.2943,
+      "step": 33193
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4286978508217448e-06,
+      "loss": 2.5384,
+      "step": 33194
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428650442477876e-06,
+      "loss": 1.378,
+      "step": 33195
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4286030341340076e-06,
+      "loss": 1.9755,
+      "step": 33196
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428555625790139e-06,
+      "loss": 1.5694,
+      "step": 33197
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4285082174462705e-06,
+      "loss": 2.1194,
+      "step": 33198
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428460809102402e-06,
+      "loss": 2.4777,
+      "step": 33199
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4284134007585335e-06,
+      "loss": 1.7231,
+      "step": 33200
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428365992414665e-06,
+      "loss": 1.3042,
+      "step": 33201
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4283185840707965e-06,
+      "loss": 3.4241,
+      "step": 33202
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428271175726928e-06,
+      "loss": 2.1596,
+      "step": 33203
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4282237673830595e-06,
+      "loss": 1.2219,
+      "step": 33204
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428176359039191e-06,
+      "loss": 1.6505,
+      "step": 33205
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4281289506953223e-06,
+      "loss": 3.27,
+      "step": 33206
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.428081542351454e-06,
+      "loss": 2.3606,
+      "step": 33207
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4280341340075855e-06,
+      "loss": 2.1104,
+      "step": 33208
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4279867256637167e-06,
+      "loss": 3.4867,
+      "step": 33209
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4279393173198482e-06,
+      "loss": 2.5594,
+      "step": 33210
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.42789190897598e-06,
+      "loss": 3.1469,
+      "step": 33211
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4278445006321112e-06,
+      "loss": 3.2157,
+      "step": 33212
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4277970922882427e-06,
+      "loss": 2.6164,
+      "step": 33213
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4277496839443744e-06,
+      "loss": 2.3098,
+      "step": 33214
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4277022756005057e-06,
+      "loss": 1.4961,
+      "step": 33215
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 1.4276548672566372e-06,
+      "loss": 1.2157,
+      "step": 33216
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4276074589127687e-06,
+      "loss": 1.7368,
+      "step": 33217
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4275600505689002e-06,
+      "loss": 2.3824,
+      "step": 33218
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4275126422250317e-06,
+      "loss": 1.4273,
+      "step": 33219
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4274652338811632e-06,
+      "loss": 2.0968,
+      "step": 33220
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4274178255372947e-06,
+      "loss": 1.4675,
+      "step": 33221
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4273704171934261e-06,
+      "loss": 1.538,
+      "step": 33222
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4273230088495574e-06,
+      "loss": 1.7906,
+      "step": 33223
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4272756005056891e-06,
+      "loss": 2.986,
+      "step": 33224
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4272281921618206e-06,
+      "loss": 2.7093,
+      "step": 33225
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.427180783817952e-06,
+      "loss": 2.444,
+      "step": 33226
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4271333754740836e-06,
+      "loss": 2.3129,
+      "step": 33227
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4270859671302149e-06,
+      "loss": 1.4902,
+      "step": 33228
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4270385587863464e-06,
+      "loss": 0.8234,
+      "step": 33229
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4269911504424779e-06,
+      "loss": 2.7945,
+      "step": 33230
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4269437420986094e-06,
+      "loss": 3.1035,
+      "step": 33231
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4268963337547409e-06,
+      "loss": 2.1651,
+      "step": 33232
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4268489254108723e-06,
+      "loss": 2.6911,
+      "step": 33233
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4268015170670038e-06,
+      "loss": 2.3363,
+      "step": 33234
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4267541087231353e-06,
+      "loss": 3.0882,
+      "step": 33235
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4267067003792666e-06,
+      "loss": 4.1443,
+      "step": 33236
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4266592920353983e-06,
+      "loss": 1.7128,
+      "step": 33237
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4266118836915298e-06,
+      "loss": 2.1217,
+      "step": 33238
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.426564475347661e-06,
+      "loss": 1.6122,
+      "step": 33239
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4265170670037928e-06,
+      "loss": 3.9355,
+      "step": 33240
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4264696586599243e-06,
+      "loss": 1.9885,
+      "step": 33241
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4264222503160556e-06,
+      "loss": 1.8855,
+      "step": 33242
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.426374841972187e-06,
+      "loss": 2.7953,
+      "step": 33243
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4263274336283188e-06,
+      "loss": 2.9055,
+      "step": 33244
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.42628002528445e-06,
+      "loss": 2.6361,
+      "step": 33245
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4262326169405815e-06,
+      "loss": 1.0739,
+      "step": 33246
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.426185208596713e-06,
+      "loss": 2.4189,
+      "step": 33247
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4261378002528445e-06,
+      "loss": 1.6416,
+      "step": 33248
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.426090391908976e-06,
+      "loss": 0.9464,
+      "step": 33249
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4260429835651075e-06,
+      "loss": 3.1091,
+      "step": 33250
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.425995575221239e-06,
+      "loss": 2.6914,
+      "step": 33251
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4259481668773705e-06,
+      "loss": 0.9902,
+      "step": 33252
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.425900758533502e-06,
+      "loss": 2.2762,
+      "step": 33253
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4258533501896335e-06,
+      "loss": 2.718,
+      "step": 33254
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.425805941845765e-06,
+      "loss": 2.8182,
+      "step": 33255
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4257585335018962e-06,
+      "loss": 3.3502,
+      "step": 33256
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.425711125158028e-06,
+      "loss": 1.9329,
+      "step": 33257
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4256637168141592e-06,
+      "loss": 2.1538,
+      "step": 33258
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4256163084702907e-06,
+      "loss": 1.3779,
+      "step": 33259
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4255689001264224e-06,
+      "loss": 1.4991,
+      "step": 33260
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4255214917825537e-06,
+      "loss": 3.3014,
+      "step": 33261
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4254740834386852e-06,
+      "loss": 1.2878,
+      "step": 33262
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4254266750948167e-06,
+      "loss": 2.0626,
+      "step": 33263
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4253792667509482e-06,
+      "loss": 2.3381,
+      "step": 33264
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4253318584070797e-06,
+      "loss": 2.2696,
+      "step": 33265
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4252844500632112e-06,
+      "loss": 1.5653,
+      "step": 33266
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4252370417193427e-06,
+      "loss": 2.691,
+      "step": 33267
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4251896333754742e-06,
+      "loss": 3.0181,
+      "step": 33268
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4251422250316054e-06,
+      "loss": 1.9042,
+      "step": 33269
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4250948166877371e-06,
+      "loss": 3.0128,
+      "step": 33270
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4250474083438686e-06,
+      "loss": 1.5007,
+      "step": 33271
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.425e-06,
+      "loss": 2.7442,
+      "step": 33272
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4249525916561316e-06,
+      "loss": 1.0139,
+      "step": 33273
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.424905183312263e-06,
+      "loss": 1.2299,
+      "step": 33274
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4248577749683944e-06,
+      "loss": 2.0689,
+      "step": 33275
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4248103666245259e-06,
+      "loss": 2.6712,
+      "step": 33276
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4247629582806576e-06,
+      "loss": 2.4127,
+      "step": 33277
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4247155499367889e-06,
+      "loss": 2.0218,
+      "step": 33278
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 1.4246681415929204e-06,
+      "loss": 1.6,
+      "step": 33279
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4246207332490518e-06,
+      "loss": 2.0206,
+      "step": 33280
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4245733249051833e-06,
+      "loss": 2.8996,
+      "step": 33281
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4245259165613148e-06,
+      "loss": 1.5789,
+      "step": 33282
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4244785082174463e-06,
+      "loss": 2.3591,
+      "step": 33283
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4244310998735778e-06,
+      "loss": 1.7342,
+      "step": 33284
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4243836915297093e-06,
+      "loss": 2.1364,
+      "step": 33285
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4243362831858406e-06,
+      "loss": 2.6723,
+      "step": 33286
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4242888748419723e-06,
+      "loss": 2.2678,
+      "step": 33287
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4242414664981038e-06,
+      "loss": 1.2822,
+      "step": 33288
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.424194058154235e-06,
+      "loss": 2.9583,
+      "step": 33289
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4241466498103668e-06,
+      "loss": 1.0106,
+      "step": 33290
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.424099241466498e-06,
+      "loss": 1.8028,
+      "step": 33291
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4240518331226295e-06,
+      "loss": 2.1275,
+      "step": 33292
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.424004424778761e-06,
+      "loss": 2.624,
+      "step": 33293
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4239570164348925e-06,
+      "loss": 1.1589,
+      "step": 33294
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.423909608091024e-06,
+      "loss": 2.9343,
+      "step": 33295
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4238621997471555e-06,
+      "loss": 1.9016,
+      "step": 33296
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.423814791403287e-06,
+      "loss": 2.3694,
+      "step": 33297
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4237673830594185e-06,
+      "loss": 2.4134,
+      "step": 33298
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.42371997471555e-06,
+      "loss": 0.8052,
+      "step": 33299
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4236725663716815e-06,
+      "loss": 2.2548,
+      "step": 33300
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.423625158027813e-06,
+      "loss": 2.927,
+      "step": 33301
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4235777496839443e-06,
+      "loss": 2.591,
+      "step": 33302
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.423530341340076e-06,
+      "loss": 1.7501,
+      "step": 33303
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4234829329962074e-06,
+      "loss": 2.1432,
+      "step": 33304
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4234355246523387e-06,
+      "loss": 2.1782,
+      "step": 33305
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4233881163084702e-06,
+      "loss": 2.1506,
+      "step": 33306
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.423340707964602e-06,
+      "loss": 1.7386,
+      "step": 33307
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4232932996207332e-06,
+      "loss": 3.0619,
+      "step": 33308
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4232458912768647e-06,
+      "loss": 2.6335,
+      "step": 33309
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4231984829329964e-06,
+      "loss": 3.0592,
+      "step": 33310
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4231510745891277e-06,
+      "loss": 2.8939,
+      "step": 33311
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4231036662452592e-06,
+      "loss": 1.9586,
+      "step": 33312
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4230562579013907e-06,
+      "loss": 1.5095,
+      "step": 33313
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4230088495575222e-06,
+      "loss": 3.0078,
+      "step": 33314
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4229614412136536e-06,
+      "loss": 2.7821,
+      "step": 33315
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4229140328697851e-06,
+      "loss": 3.6505,
+      "step": 33316
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4228666245259166e-06,
+      "loss": 1.7907,
+      "step": 33317
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4228192161820481e-06,
+      "loss": 3.5975,
+      "step": 33318
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4227718078381794e-06,
+      "loss": 1.7091,
+      "step": 33319
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4227243994943111e-06,
+      "loss": 0.8288,
+      "step": 33320
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4226769911504426e-06,
+      "loss": 2.2348,
+      "step": 33321
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4226295828065739e-06,
+      "loss": 1.9704,
+      "step": 33322
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4225821744627056e-06,
+      "loss": 2.9179,
+      "step": 33323
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4225347661188369e-06,
+      "loss": 1.1362,
+      "step": 33324
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4224873577749684e-06,
+      "loss": 1.3041,
+      "step": 33325
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4224399494310999e-06,
+      "loss": 1.6624,
+      "step": 33326
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4223925410872313e-06,
+      "loss": 0.8812,
+      "step": 33327
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4223451327433628e-06,
+      "loss": 2.0626,
+      "step": 33328
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4222977243994943e-06,
+      "loss": 2.7152,
+      "step": 33329
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4222503160556258e-06,
+      "loss": 2.08,
+      "step": 33330
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4222029077117573e-06,
+      "loss": 3.8543,
+      "step": 33331
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4221554993678888e-06,
+      "loss": 2.1911,
+      "step": 33332
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4221080910240203e-06,
+      "loss": 2.7308,
+      "step": 33333
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4220606826801518e-06,
+      "loss": 2.0572,
+      "step": 33334
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.422013274336283e-06,
+      "loss": 1.5086,
+      "step": 33335
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4219658659924148e-06,
+      "loss": 1.7475,
+      "step": 33336
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4219184576485463e-06,
+      "loss": 2.2928,
+      "step": 33337
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4218710493046775e-06,
+      "loss": 1.4711,
+      "step": 33338
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.421823640960809e-06,
+      "loss": 0.7319,
+      "step": 33339
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4217762326169407e-06,
+      "loss": 1.6674,
+      "step": 33340
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.421728824273072e-06,
+      "loss": 3.6203,
+      "step": 33341
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4216814159292035e-06,
+      "loss": 2.2494,
+      "step": 33342
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 1.4216340075853352e-06,
+      "loss": 3.771,
+      "step": 33343
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4215865992414665e-06,
+      "loss": 1.3286,
+      "step": 33344
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.421539190897598e-06,
+      "loss": 2.5887,
+      "step": 33345
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4214917825537295e-06,
+      "loss": 3.0221,
+      "step": 33346
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.421444374209861e-06,
+      "loss": 2.5413,
+      "step": 33347
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4213969658659925e-06,
+      "loss": 1.2418,
+      "step": 33348
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.421349557522124e-06,
+      "loss": 1.7238,
+      "step": 33349
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4213021491782555e-06,
+      "loss": 1.3321,
+      "step": 33350
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.421254740834387e-06,
+      "loss": 1.0097,
+      "step": 33351
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4212073324905182e-06,
+      "loss": 1.8257,
+      "step": 33352
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.42115992414665e-06,
+      "loss": 2.4896,
+      "step": 33353
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4211125158027814e-06,
+      "loss": 2.5713,
+      "step": 33354
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4210651074589127e-06,
+      "loss": 2.8138,
+      "step": 33355
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4210176991150444e-06,
+      "loss": 1.7465,
+      "step": 33356
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4209702907711757e-06,
+      "loss": 2.4137,
+      "step": 33357
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4209228824273072e-06,
+      "loss": 1.7562,
+      "step": 33358
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4208754740834387e-06,
+      "loss": 1.6681,
+      "step": 33359
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4208280657395702e-06,
+      "loss": 3.1688,
+      "step": 33360
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4207806573957017e-06,
+      "loss": 1.8777,
+      "step": 33361
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4207332490518331e-06,
+      "loss": 1.897,
+      "step": 33362
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4206858407079646e-06,
+      "loss": 3.471,
+      "step": 33363
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4206384323640961e-06,
+      "loss": 1.317,
+      "step": 33364
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4205910240202276e-06,
+      "loss": 1.8212,
+      "step": 33365
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4205436156763591e-06,
+      "loss": 2.4801,
+      "step": 33366
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4204962073324906e-06,
+      "loss": 3.0815,
+      "step": 33367
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4204487989886219e-06,
+      "loss": 2.7278,
+      "step": 33368
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4204013906447536e-06,
+      "loss": 1.3783,
+      "step": 33369
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.420353982300885e-06,
+      "loss": 2.3468,
+      "step": 33370
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4203065739570164e-06,
+      "loss": 3.7751,
+      "step": 33371
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4202591656131479e-06,
+      "loss": 2.2953,
+      "step": 33372
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4202117572692796e-06,
+      "loss": 1.7623,
+      "step": 33373
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4201643489254108e-06,
+      "loss": 2.4628,
+      "step": 33374
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4201169405815423e-06,
+      "loss": 3.053,
+      "step": 33375
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.420069532237674e-06,
+      "loss": 3.1896,
+      "step": 33376
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4200221238938053e-06,
+      "loss": 3.0094,
+      "step": 33377
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4199747155499368e-06,
+      "loss": 1.1449,
+      "step": 33378
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4199273072060683e-06,
+      "loss": 1.3753,
+      "step": 33379
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4198798988621998e-06,
+      "loss": 1.9579,
+      "step": 33380
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4198324905183313e-06,
+      "loss": 1.4235,
+      "step": 33381
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4197850821744628e-06,
+      "loss": 1.6158,
+      "step": 33382
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4197376738305943e-06,
+      "loss": 4.1691,
+      "step": 33383
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4196902654867258e-06,
+      "loss": 3.1515,
+      "step": 33384
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.419642857142857e-06,
+      "loss": 2.1326,
+      "step": 33385
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4195954487989887e-06,
+      "loss": 1.7983,
+      "step": 33386
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4195480404551202e-06,
+      "loss": 1.9098,
+      "step": 33387
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4195006321112515e-06,
+      "loss": 2.6457,
+      "step": 33388
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4194532237673832e-06,
+      "loss": 2.0498,
+      "step": 33389
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4194058154235145e-06,
+      "loss": 1.9833,
+      "step": 33390
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.419358407079646e-06,
+      "loss": 3.0476,
+      "step": 33391
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4193109987357775e-06,
+      "loss": 3.0078,
+      "step": 33392
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.419263590391909e-06,
+      "loss": 2.0941,
+      "step": 33393
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4192161820480405e-06,
+      "loss": 2.1135,
+      "step": 33394
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.419168773704172e-06,
+      "loss": 1.4404,
+      "step": 33395
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4191213653603035e-06,
+      "loss": 2.2896,
+      "step": 33396
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.419073957016435e-06,
+      "loss": 2.0598,
+      "step": 33397
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4190265486725664e-06,
+      "loss": 3.8835,
+      "step": 33398
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.418979140328698e-06,
+      "loss": 3.3507,
+      "step": 33399
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4189317319848294e-06,
+      "loss": 3.2014,
+      "step": 33400
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4188843236409607e-06,
+      "loss": 2.2596,
+      "step": 33401
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4188369152970924e-06,
+      "loss": 2.1393,
+      "step": 33402
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.418789506953224e-06,
+      "loss": 2.6079,
+      "step": 33403
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4187420986093552e-06,
+      "loss": 2.568,
+      "step": 33404
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4186946902654867e-06,
+      "loss": 1.9088,
+      "step": 33405
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 1.4186472819216184e-06,
+      "loss": 2.5964,
+      "step": 33406
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4185998735777497e-06,
+      "loss": 3.703,
+      "step": 33407
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4185524652338812e-06,
+      "loss": 1.5377,
+      "step": 33408
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4185050568900129e-06,
+      "loss": 1.8887,
+      "step": 33409
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4184576485461441e-06,
+      "loss": 1.9843,
+      "step": 33410
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4184102402022756e-06,
+      "loss": 2.455,
+      "step": 33411
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4183628318584071e-06,
+      "loss": 2.6747,
+      "step": 33412
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4183154235145386e-06,
+      "loss": 2.9591,
+      "step": 33413
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.41826801517067e-06,
+      "loss": 2.0867,
+      "step": 33414
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4182206068268016e-06,
+      "loss": 1.4112,
+      "step": 33415
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.418173198482933e-06,
+      "loss": 2.7769,
+      "step": 33416
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4181257901390646e-06,
+      "loss": 1.2623,
+      "step": 33417
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4180783817951959e-06,
+      "loss": 1.8822,
+      "step": 33418
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4180309734513276e-06,
+      "loss": 2.3808,
+      "step": 33419
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.417983565107459e-06,
+      "loss": 3.263,
+      "step": 33420
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4179361567635903e-06,
+      "loss": 2.2064,
+      "step": 33421
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.417888748419722e-06,
+      "loss": 2.948,
+      "step": 33422
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4178413400758533e-06,
+      "loss": 1.9985,
+      "step": 33423
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4177939317319848e-06,
+      "loss": 3.8091,
+      "step": 33424
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4177465233881163e-06,
+      "loss": 1.8314,
+      "step": 33425
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4176991150442478e-06,
+      "loss": 2.3597,
+      "step": 33426
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4176517067003793e-06,
+      "loss": 2.8567,
+      "step": 33427
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4176042983565108e-06,
+      "loss": 0.9889,
+      "step": 33428
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4175568900126423e-06,
+      "loss": 2.2293,
+      "step": 33429
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4175094816687738e-06,
+      "loss": 0.9138,
+      "step": 33430
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4174620733249053e-06,
+      "loss": 2.7564,
+      "step": 33431
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4174146649810368e-06,
+      "loss": 1.1016,
+      "step": 33432
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4173672566371682e-06,
+      "loss": 3.097,
+      "step": 33433
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4173198482932995e-06,
+      "loss": 2.5212,
+      "step": 33434
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.417272439949431e-06,
+      "loss": 1.9434,
+      "step": 33435
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4172250316055627e-06,
+      "loss": 1.8604,
+      "step": 33436
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.417177623261694e-06,
+      "loss": 2.2305,
+      "step": 33437
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4171302149178255e-06,
+      "loss": 4.3573,
+      "step": 33438
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4170828065739572e-06,
+      "loss": 2.3447,
+      "step": 33439
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4170353982300885e-06,
+      "loss": 2.1855,
+      "step": 33440
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.41698798988622e-06,
+      "loss": 1.3613,
+      "step": 33441
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4169405815423515e-06,
+      "loss": 2.7608,
+      "step": 33442
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.416893173198483e-06,
+      "loss": 2.756,
+      "step": 33443
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4168457648546144e-06,
+      "loss": 2.761,
+      "step": 33444
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.416798356510746e-06,
+      "loss": 0.8893,
+      "step": 33445
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4167509481668774e-06,
+      "loss": 1.1883,
+      "step": 33446
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.416703539823009e-06,
+      "loss": 2.1671,
+      "step": 33447
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4166561314791402e-06,
+      "loss": 2.49,
+      "step": 33448
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.416608723135272e-06,
+      "loss": 1.5347,
+      "step": 33449
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4165613147914034e-06,
+      "loss": 3.3292,
+      "step": 33450
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4165139064475347e-06,
+      "loss": 2.8431,
+      "step": 33451
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4164664981036664e-06,
+      "loss": 3.4056,
+      "step": 33452
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4164190897597979e-06,
+      "loss": 2.3556,
+      "step": 33453
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4163716814159292e-06,
+      "loss": 1.9891,
+      "step": 33454
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4163242730720606e-06,
+      "loss": 2.1917,
+      "step": 33455
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4162768647281921e-06,
+      "loss": 2.08,
+      "step": 33456
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4162294563843236e-06,
+      "loss": 2.8838,
+      "step": 33457
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4161820480404551e-06,
+      "loss": 2.4616,
+      "step": 33458
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4161346396965866e-06,
+      "loss": 3.387,
+      "step": 33459
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4160872313527181e-06,
+      "loss": 1.8183,
+      "step": 33460
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4160398230088496e-06,
+      "loss": 2.1157,
+      "step": 33461
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.415992414664981e-06,
+      "loss": 1.6984,
+      "step": 33462
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4159450063211126e-06,
+      "loss": 1.9426,
+      "step": 33463
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.415897597977244e-06,
+      "loss": 0.9722,
+      "step": 33464
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4158501896333756e-06,
+      "loss": 2.0795,
+      "step": 33465
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.415802781289507e-06,
+      "loss": 2.3086,
+      "step": 33466
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4157553729456383e-06,
+      "loss": 3.0792,
+      "step": 33467
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4157079646017698e-06,
+      "loss": 2.0656,
+      "step": 33468
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 1.4156605562579015e-06,
+      "loss": 2.2387,
+      "step": 33469
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4156131479140328e-06,
+      "loss": 2.572,
+      "step": 33470
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4155657395701643e-06,
+      "loss": 3.6645,
+      "step": 33471
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.415518331226296e-06,
+      "loss": 3.4809,
+      "step": 33472
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4154709228824273e-06,
+      "loss": 0.9097,
+      "step": 33473
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4154235145385588e-06,
+      "loss": 2.3885,
+      "step": 33474
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4153761061946903e-06,
+      "loss": 3.1869,
+      "step": 33475
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4153286978508218e-06,
+      "loss": 2.1167,
+      "step": 33476
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4152812895069533e-06,
+      "loss": 1.5133,
+      "step": 33477
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4152338811630848e-06,
+      "loss": 3.0326,
+      "step": 33478
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4151864728192162e-06,
+      "loss": 2.8046,
+      "step": 33479
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4151390644753477e-06,
+      "loss": 1.7426,
+      "step": 33480
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.415091656131479e-06,
+      "loss": 3.1725,
+      "step": 33481
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4150442477876107e-06,
+      "loss": 2.4986,
+      "step": 33482
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4149968394437422e-06,
+      "loss": 2.6116,
+      "step": 33483
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4149494310998735e-06,
+      "loss": 3.2701,
+      "step": 33484
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4149020227560052e-06,
+      "loss": 1.745,
+      "step": 33485
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4148546144121367e-06,
+      "loss": 1.6596,
+      "step": 33486
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.414807206068268e-06,
+      "loss": 2.0715,
+      "step": 33487
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4147597977243995e-06,
+      "loss": 0.6206,
+      "step": 33488
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.414712389380531e-06,
+      "loss": 2.6872,
+      "step": 33489
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4146649810366625e-06,
+      "loss": 2.9611,
+      "step": 33490
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.414617572692794e-06,
+      "loss": 1.5481,
+      "step": 33491
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4145701643489254e-06,
+      "loss": 2.3196,
+      "step": 33492
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.414522756005057e-06,
+      "loss": 2.1124,
+      "step": 33493
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4144753476611884e-06,
+      "loss": 1.8958,
+      "step": 33494
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.41442793931732e-06,
+      "loss": 2.1312,
+      "step": 33495
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4143805309734514e-06,
+      "loss": 1.8142,
+      "step": 33496
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.414333122629583e-06,
+      "loss": 2.033,
+      "step": 33497
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4142857142857144e-06,
+      "loss": 2.5511,
+      "step": 33498
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4142383059418459e-06,
+      "loss": 1.6194,
+      "step": 33499
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4141908975979772e-06,
+      "loss": 3.3869,
+      "step": 33500
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4141434892541087e-06,
+      "loss": 2.5888,
+      "step": 33501
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4140960809102404e-06,
+      "loss": 1.9886,
+      "step": 33502
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4140486725663716e-06,
+      "loss": 0.8425,
+      "step": 33503
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4140012642225031e-06,
+      "loss": 0.9192,
+      "step": 33504
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4139538558786348e-06,
+      "loss": 1.2434,
+      "step": 33505
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4139064475347661e-06,
+      "loss": 2.2935,
+      "step": 33506
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4138590391908976e-06,
+      "loss": 2.3894,
+      "step": 33507
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.413811630847029e-06,
+      "loss": 3.3703,
+      "step": 33508
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4137642225031606e-06,
+      "loss": 2.2323,
+      "step": 33509
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.413716814159292e-06,
+      "loss": 2.53,
+      "step": 33510
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4136694058154236e-06,
+      "loss": 3.3898,
+      "step": 33511
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.413621997471555e-06,
+      "loss": 3.2989,
+      "step": 33512
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4135745891276866e-06,
+      "loss": 3.0467,
+      "step": 33513
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4135271807838178e-06,
+      "loss": 1.7192,
+      "step": 33514
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4134797724399495e-06,
+      "loss": 2.6679,
+      "step": 33515
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.413432364096081e-06,
+      "loss": 2.9698,
+      "step": 33516
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4133849557522123e-06,
+      "loss": 1.2811,
+      "step": 33517
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.413337547408344e-06,
+      "loss": 2.3088,
+      "step": 33518
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4132901390644755e-06,
+      "loss": 2.2434,
+      "step": 33519
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4132427307206068e-06,
+      "loss": 1.0653,
+      "step": 33520
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4131953223767383e-06,
+      "loss": 1.2523,
+      "step": 33521
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4131479140328698e-06,
+      "loss": 1.5109,
+      "step": 33522
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4131005056890013e-06,
+      "loss": 2.6201,
+      "step": 33523
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4130530973451328e-06,
+      "loss": 1.9669,
+      "step": 33524
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4130056890012643e-06,
+      "loss": 1.6527,
+      "step": 33525
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4129582806573957e-06,
+      "loss": 1.8331,
+      "step": 33526
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4129108723135272e-06,
+      "loss": 2.7637,
+      "step": 33527
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4128634639696587e-06,
+      "loss": 1.4548,
+      "step": 33528
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4128160556257902e-06,
+      "loss": 0.8497,
+      "step": 33529
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4127686472819217e-06,
+      "loss": 2.3065,
+      "step": 33530
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4127212389380532e-06,
+      "loss": 2.7306,
+      "step": 33531
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.4126738305941847e-06,
+      "loss": 0.894,
+      "step": 33532
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 1.412626422250316e-06,
+      "loss": 2.9137,
+      "step": 33533
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4125790139064475e-06,
+      "loss": 2.21,
+      "step": 33534
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4125316055625792e-06,
+      "loss": 1.3449,
+      "step": 33535
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4124841972187105e-06,
+      "loss": 2.389,
+      "step": 33536
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.412436788874842e-06,
+      "loss": 2.2425,
+      "step": 33537
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4123893805309737e-06,
+      "loss": 2.2226,
+      "step": 33538
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.412341972187105e-06,
+      "loss": 2.792,
+      "step": 33539
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4122945638432364e-06,
+      "loss": 0.969,
+      "step": 33540
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.412247155499368e-06,
+      "loss": 3.7863,
+      "step": 33541
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4121997471554994e-06,
+      "loss": 2.1597,
+      "step": 33542
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.412152338811631e-06,
+      "loss": 1.5348,
+      "step": 33543
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4121049304677624e-06,
+      "loss": 1.1831,
+      "step": 33544
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4120575221238939e-06,
+      "loss": 2.6985,
+      "step": 33545
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4120101137800254e-06,
+      "loss": 2.4706,
+      "step": 33546
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4119627054361567e-06,
+      "loss": 2.0509,
+      "step": 33547
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4119152970922884e-06,
+      "loss": 2.0263,
+      "step": 33548
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4118678887484199e-06,
+      "loss": 3.7459,
+      "step": 33549
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4118204804045511e-06,
+      "loss": 1.5684,
+      "step": 33550
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4117730720606828e-06,
+      "loss": 2.1553,
+      "step": 33551
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4117256637168143e-06,
+      "loss": 2.9054,
+      "step": 33552
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4116782553729456e-06,
+      "loss": 1.7577,
+      "step": 33553
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.411630847029077e-06,
+      "loss": 2.0037,
+      "step": 33554
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4115834386852086e-06,
+      "loss": 2.9414,
+      "step": 33555
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.41153603034134e-06,
+      "loss": 2.1767,
+      "step": 33556
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4114886219974716e-06,
+      "loss": 2.4491,
+      "step": 33557
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.411441213653603e-06,
+      "loss": 2.2646,
+      "step": 33558
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4113938053097346e-06,
+      "loss": 3.4337,
+      "step": 33559
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.411346396965866e-06,
+      "loss": 2.2234,
+      "step": 33560
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4112989886219975e-06,
+      "loss": 2.8934,
+      "step": 33561
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.411251580278129e-06,
+      "loss": 1.4151,
+      "step": 33562
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4112041719342605e-06,
+      "loss": 0.9293,
+      "step": 33563
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4111567635903918e-06,
+      "loss": 2.9472,
+      "step": 33564
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4111093552465235e-06,
+      "loss": 1.7342,
+      "step": 33565
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4110619469026548e-06,
+      "loss": 1.1818,
+      "step": 33566
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4110145385587863e-06,
+      "loss": 3.4784,
+      "step": 33567
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.410967130214918e-06,
+      "loss": 2.1743,
+      "step": 33568
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4109197218710493e-06,
+      "loss": 2.3333,
+      "step": 33569
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4108723135271808e-06,
+      "loss": 2.9996,
+      "step": 33570
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4108249051833123e-06,
+      "loss": 1.8296,
+      "step": 33571
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4107774968394438e-06,
+      "loss": 1.585,
+      "step": 33572
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4107300884955752e-06,
+      "loss": 1.4119,
+      "step": 33573
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4106826801517067e-06,
+      "loss": 1.2931,
+      "step": 33574
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4106352718078382e-06,
+      "loss": 2.2064,
+      "step": 33575
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4105878634639697e-06,
+      "loss": 1.7141,
+      "step": 33576
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.410540455120101e-06,
+      "loss": 2.0077,
+      "step": 33577
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4104930467762327e-06,
+      "loss": 2.0838,
+      "step": 33578
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4104456384323642e-06,
+      "loss": 2.2036,
+      "step": 33579
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4103982300884955e-06,
+      "loss": 2.0925,
+      "step": 33580
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4103508217446272e-06,
+      "loss": 3.0682,
+      "step": 33581
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4103034134007587e-06,
+      "loss": 2.728,
+      "step": 33582
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.41025600505689e-06,
+      "loss": 2.4321,
+      "step": 33583
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4102085967130214e-06,
+      "loss": 2.5016,
+      "step": 33584
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4101611883691531e-06,
+      "loss": 2.7766,
+      "step": 33585
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4101137800252844e-06,
+      "loss": 4.2613,
+      "step": 33586
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.410066371681416e-06,
+      "loss": 2.2682,
+      "step": 33587
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4100189633375474e-06,
+      "loss": 1.5058,
+      "step": 33588
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.409971554993679e-06,
+      "loss": 2.7182,
+      "step": 33589
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4099241466498104e-06,
+      "loss": 2.6824,
+      "step": 33590
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4098767383059419e-06,
+      "loss": 1.5909,
+      "step": 33591
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4098293299620734e-06,
+      "loss": 3.1786,
+      "step": 33592
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4097819216182049e-06,
+      "loss": 1.8233,
+      "step": 33593
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4097345132743364e-06,
+      "loss": 2.9427,
+      "step": 33594
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4096871049304679e-06,
+      "loss": 1.9284,
+      "step": 33595
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 1.4096396965865994e-06,
+      "loss": 2.3133,
+      "step": 33596
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4095922882427306e-06,
+      "loss": 2.6706,
+      "step": 33597
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4095448798988623e-06,
+      "loss": 3.1415,
+      "step": 33598
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4094974715549936e-06,
+      "loss": 1.1581,
+      "step": 33599
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4094500632111251e-06,
+      "loss": 2.2034,
+      "step": 33600
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4094026548672568e-06,
+      "loss": 2.9331,
+      "step": 33601
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.409355246523388e-06,
+      "loss": 1.7671,
+      "step": 33602
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4093078381795196e-06,
+      "loss": 2.1653,
+      "step": 33603
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.409260429835651e-06,
+      "loss": 2.7457,
+      "step": 33604
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4092130214917826e-06,
+      "loss": 1.5577,
+      "step": 33605
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.409165613147914e-06,
+      "loss": 2.0779,
+      "step": 33606
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4091182048040456e-06,
+      "loss": 1.7189,
+      "step": 33607
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.409070796460177e-06,
+      "loss": 2.0257,
+      "step": 33608
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4090233881163085e-06,
+      "loss": 2.0184,
+      "step": 33609
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4089759797724398e-06,
+      "loss": 1.6565,
+      "step": 33610
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4089285714285715e-06,
+      "loss": 2.018,
+      "step": 33611
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.408881163084703e-06,
+      "loss": 3.577,
+      "step": 33612
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4088337547408343e-06,
+      "loss": 1.9501,
+      "step": 33613
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.408786346396966e-06,
+      "loss": 4.0291,
+      "step": 33614
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4087389380530975e-06,
+      "loss": 1.3931,
+      "step": 33615
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4086915297092288e-06,
+      "loss": 2.5628,
+      "step": 33616
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4086441213653603e-06,
+      "loss": 2.7643,
+      "step": 33617
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.408596713021492e-06,
+      "loss": 2.643,
+      "step": 33618
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4085493046776232e-06,
+      "loss": 2.229,
+      "step": 33619
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4085018963337547e-06,
+      "loss": 1.6913,
+      "step": 33620
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4084544879898862e-06,
+      "loss": 1.4973,
+      "step": 33621
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4084070796460177e-06,
+      "loss": 3.258,
+      "step": 33622
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4083596713021492e-06,
+      "loss": 2.3548,
+      "step": 33623
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4083122629582807e-06,
+      "loss": 2.7896,
+      "step": 33624
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4082648546144122e-06,
+      "loss": 2.5474,
+      "step": 33625
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4082174462705437e-06,
+      "loss": 2.01,
+      "step": 33626
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4081700379266752e-06,
+      "loss": 2.0396,
+      "step": 33627
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4081226295828067e-06,
+      "loss": 1.0752,
+      "step": 33628
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4080752212389382e-06,
+      "loss": 1.6362,
+      "step": 33629
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4080278128950694e-06,
+      "loss": 2.5813,
+      "step": 33630
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4079804045512012e-06,
+      "loss": 2.3758,
+      "step": 33631
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4079329962073324e-06,
+      "loss": 3.4153,
+      "step": 33632
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.407885587863464e-06,
+      "loss": 2.1681,
+      "step": 33633
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4078381795195956e-06,
+      "loss": 2.9937,
+      "step": 33634
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.407790771175727e-06,
+      "loss": 2.2032,
+      "step": 33635
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4077433628318584e-06,
+      "loss": 2.4228,
+      "step": 33636
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.40769595448799e-06,
+      "loss": 1.5304,
+      "step": 33637
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4076485461441214e-06,
+      "loss": 2.0373,
+      "step": 33638
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4076011378002529e-06,
+      "loss": 0.4476,
+      "step": 33639
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4075537294563844e-06,
+      "loss": 1.9461,
+      "step": 33640
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4075063211125159e-06,
+      "loss": 2.4824,
+      "step": 33641
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4074589127686474e-06,
+      "loss": 1.8448,
+      "step": 33642
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4074115044247786e-06,
+      "loss": 1.5322,
+      "step": 33643
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4073640960809103e-06,
+      "loss": 2.5092,
+      "step": 33644
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4073166877370418e-06,
+      "loss": 2.1464,
+      "step": 33645
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4072692793931731e-06,
+      "loss": 3.4908,
+      "step": 33646
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4072218710493048e-06,
+      "loss": 2.2959,
+      "step": 33647
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4071744627054363e-06,
+      "loss": 2.3721,
+      "step": 33648
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4071270543615676e-06,
+      "loss": 2.7017,
+      "step": 33649
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.407079646017699e-06,
+      "loss": 2.6617,
+      "step": 33650
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4070322376738308e-06,
+      "loss": 3.281,
+      "step": 33651
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.406984829329962e-06,
+      "loss": 2.1029,
+      "step": 33652
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4069374209860936e-06,
+      "loss": 2.9321,
+      "step": 33653
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.406890012642225e-06,
+      "loss": 1.1169,
+      "step": 33654
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4068426042983565e-06,
+      "loss": 2.3785,
+      "step": 33655
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.406795195954488e-06,
+      "loss": 1.6994,
+      "step": 33656
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4067477876106195e-06,
+      "loss": 2.7487,
+      "step": 33657
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.406700379266751e-06,
+      "loss": 2.8645,
+      "step": 33658
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 1.4066529709228825e-06,
+      "loss": 1.3585,
+      "step": 33659
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.406605562579014e-06,
+      "loss": 2.5097,
+      "step": 33660
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4065581542351455e-06,
+      "loss": 2.6384,
+      "step": 33661
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.406510745891277e-06,
+      "loss": 2.3759,
+      "step": 33662
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4064633375474083e-06,
+      "loss": 2.3063,
+      "step": 33663
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.40641592920354e-06,
+      "loss": 1.5721,
+      "step": 33664
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4063685208596713e-06,
+      "loss": 3.3098,
+      "step": 33665
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4063211125158027e-06,
+      "loss": 2.2469,
+      "step": 33666
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4062737041719344e-06,
+      "loss": 3.1503,
+      "step": 33667
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4062262958280657e-06,
+      "loss": 2.1157,
+      "step": 33668
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4061788874841972e-06,
+      "loss": 1.8116,
+      "step": 33669
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4061314791403287e-06,
+      "loss": 2.9225,
+      "step": 33670
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4060840707964602e-06,
+      "loss": 2.5276,
+      "step": 33671
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4060366624525917e-06,
+      "loss": 2.1542,
+      "step": 33672
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4059892541087232e-06,
+      "loss": 2.4283,
+      "step": 33673
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4059418457648547e-06,
+      "loss": 1.6624,
+      "step": 33674
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4058944374209862e-06,
+      "loss": 2.477,
+      "step": 33675
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4058470290771175e-06,
+      "loss": 2.8169,
+      "step": 33676
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4057996207332492e-06,
+      "loss": 3.032,
+      "step": 33677
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4057522123893807e-06,
+      "loss": 2.3177,
+      "step": 33678
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.405704804045512e-06,
+      "loss": 0.877,
+      "step": 33679
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4056573957016436e-06,
+      "loss": 1.4484,
+      "step": 33680
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4056099873577751e-06,
+      "loss": 2.8096,
+      "step": 33681
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4055625790139064e-06,
+      "loss": 1.6853,
+      "step": 33682
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.405515170670038e-06,
+      "loss": 1.8111,
+      "step": 33683
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4054677623261694e-06,
+      "loss": 3.2929,
+      "step": 33684
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4054203539823009e-06,
+      "loss": 1.6328,
+      "step": 33685
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4053729456384324e-06,
+      "loss": 2.5893,
+      "step": 33686
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4053255372945639e-06,
+      "loss": 3.3594,
+      "step": 33687
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4052781289506954e-06,
+      "loss": 2.5662,
+      "step": 33688
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4052307206068269e-06,
+      "loss": 1.6244,
+      "step": 33689
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4051833122629583e-06,
+      "loss": 1.5766,
+      "step": 33690
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4051359039190898e-06,
+      "loss": 1.6352,
+      "step": 33691
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4050884955752213e-06,
+      "loss": 2.6495,
+      "step": 33692
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4050410872313528e-06,
+      "loss": 3.1557,
+      "step": 33693
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4049936788874843e-06,
+      "loss": 1.9992,
+      "step": 33694
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4049462705436156e-06,
+      "loss": 2.628,
+      "step": 33695
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.404898862199747e-06,
+      "loss": 2.5017,
+      "step": 33696
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4048514538558788e-06,
+      "loss": 2.3228,
+      "step": 33697
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.40480404551201e-06,
+      "loss": 1.9812,
+      "step": 33698
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4047566371681416e-06,
+      "loss": 2.9135,
+      "step": 33699
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4047092288242733e-06,
+      "loss": 3.4833,
+      "step": 33700
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4046618204804045e-06,
+      "loss": 3.0527,
+      "step": 33701
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.404614412136536e-06,
+      "loss": 2.9299,
+      "step": 33702
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4045670037926675e-06,
+      "loss": 3.2997,
+      "step": 33703
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.404519595448799e-06,
+      "loss": 2.8737,
+      "step": 33704
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4044721871049305e-06,
+      "loss": 2.0438,
+      "step": 33705
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4044247787610618e-06,
+      "loss": 2.015,
+      "step": 33706
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4043773704171935e-06,
+      "loss": 2.5924,
+      "step": 33707
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.404329962073325e-06,
+      "loss": 2.142,
+      "step": 33708
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4042825537294563e-06,
+      "loss": 1.0082,
+      "step": 33709
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.404235145385588e-06,
+      "loss": 1.5081,
+      "step": 33710
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4041877370417195e-06,
+      "loss": 2.4114,
+      "step": 33711
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4041403286978507e-06,
+      "loss": 1.8217,
+      "step": 33712
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4040929203539822e-06,
+      "loss": 2.4374,
+      "step": 33713
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.404045512010114e-06,
+      "loss": 1.6134,
+      "step": 33714
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4039981036662452e-06,
+      "loss": 2.1909,
+      "step": 33715
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4039506953223767e-06,
+      "loss": 2.7013,
+      "step": 33716
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4039032869785082e-06,
+      "loss": 2.3577,
+      "step": 33717
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4038558786346397e-06,
+      "loss": 2.1812,
+      "step": 33718
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4038084702907712e-06,
+      "loss": 3.6456,
+      "step": 33719
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4037610619469027e-06,
+      "loss": 2.5377,
+      "step": 33720
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4037136536030342e-06,
+      "loss": 1.7844,
+      "step": 33721
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4036662452591657e-06,
+      "loss": 3.3844,
+      "step": 33722
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 1.4036188369152972e-06,
+      "loss": 2.283,
+      "step": 33723
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4035714285714287e-06,
+      "loss": 1.9921,
+      "step": 33724
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4035240202275601e-06,
+      "loss": 2.5905,
+      "step": 33725
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4034766118836914e-06,
+      "loss": 1.5615,
+      "step": 33726
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4034292035398231e-06,
+      "loss": 3.1246,
+      "step": 33727
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4033817951959544e-06,
+      "loss": 3.1455,
+      "step": 33728
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.403334386852086e-06,
+      "loss": 1.4431,
+      "step": 33729
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4032869785082176e-06,
+      "loss": 2.9933,
+      "step": 33730
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4032395701643489e-06,
+      "loss": 2.2704,
+      "step": 33731
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4031921618204804e-06,
+      "loss": 3.2355,
+      "step": 33732
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4031447534766119e-06,
+      "loss": 1.7247,
+      "step": 33733
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4030973451327434e-06,
+      "loss": 2.8105,
+      "step": 33734
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4030499367888749e-06,
+      "loss": 1.4894,
+      "step": 33735
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4030025284450063e-06,
+      "loss": 1.7028,
+      "step": 33736
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4029551201011378e-06,
+      "loss": 1.2763,
+      "step": 33737
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4029077117572693e-06,
+      "loss": 2.6066,
+      "step": 33738
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4028603034134006e-06,
+      "loss": 2.121,
+      "step": 33739
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4028128950695323e-06,
+      "loss": 1.0379,
+      "step": 33740
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4027654867256638e-06,
+      "loss": 1.7485,
+      "step": 33741
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.402718078381795e-06,
+      "loss": 2.9292,
+      "step": 33742
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4026706700379268e-06,
+      "loss": 1.4014,
+      "step": 33743
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4026232616940583e-06,
+      "loss": 2.9573,
+      "step": 33744
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4025758533501896e-06,
+      "loss": 2.1465,
+      "step": 33745
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.402528445006321e-06,
+      "loss": 1.922,
+      "step": 33746
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4024810366624528e-06,
+      "loss": 1.4592,
+      "step": 33747
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.402433628318584e-06,
+      "loss": 2.7078,
+      "step": 33748
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4023862199747155e-06,
+      "loss": 1.7508,
+      "step": 33749
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.402338811630847e-06,
+      "loss": 1.7489,
+      "step": 33750
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4022914032869785e-06,
+      "loss": 2.0918,
+      "step": 33751
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.40224399494311e-06,
+      "loss": 3.1287,
+      "step": 33752
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4021965865992415e-06,
+      "loss": 2.1735,
+      "step": 33753
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.402149178255373e-06,
+      "loss": 2.2308,
+      "step": 33754
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4021017699115045e-06,
+      "loss": 0.7127,
+      "step": 33755
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.402054361567636e-06,
+      "loss": 1.3025,
+      "step": 33756
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4020069532237675e-06,
+      "loss": 1.8305,
+      "step": 33757
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.401959544879899e-06,
+      "loss": 1.2737,
+      "step": 33758
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4019121365360302e-06,
+      "loss": 2.1141,
+      "step": 33759
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.401864728192162e-06,
+      "loss": 3.0937,
+      "step": 33760
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4018173198482932e-06,
+      "loss": 4.5817,
+      "step": 33761
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4017699115044247e-06,
+      "loss": 3.0136,
+      "step": 33762
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4017225031605564e-06,
+      "loss": 1.5074,
+      "step": 33763
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4016750948166877e-06,
+      "loss": 1.8007,
+      "step": 33764
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4016276864728192e-06,
+      "loss": 1.6458,
+      "step": 33765
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4015802781289507e-06,
+      "loss": 3.2376,
+      "step": 33766
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4015328697850822e-06,
+      "loss": 2.0364,
+      "step": 33767
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4014854614412137e-06,
+      "loss": 3.1021,
+      "step": 33768
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4014380530973452e-06,
+      "loss": 1.0633,
+      "step": 33769
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4013906447534767e-06,
+      "loss": 2.8573,
+      "step": 33770
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4013432364096082e-06,
+      "loss": 1.6702,
+      "step": 33771
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4012958280657394e-06,
+      "loss": 1.3857,
+      "step": 33772
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4012484197218711e-06,
+      "loss": 3.1212,
+      "step": 33773
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4012010113780026e-06,
+      "loss": 2.3287,
+      "step": 33774
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.401153603034134e-06,
+      "loss": 1.5069,
+      "step": 33775
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4011061946902656e-06,
+      "loss": 2.3142,
+      "step": 33776
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.401058786346397e-06,
+      "loss": 2.5526,
+      "step": 33777
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4010113780025284e-06,
+      "loss": 2.3209,
+      "step": 33778
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4009639696586599e-06,
+      "loss": 2.2479,
+      "step": 33779
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4009165613147916e-06,
+      "loss": 2.2337,
+      "step": 33780
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4008691529709229e-06,
+      "loss": 2.925,
+      "step": 33781
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4008217446270544e-06,
+      "loss": 2.2591,
+      "step": 33782
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4007743362831858e-06,
+      "loss": 2.3993,
+      "step": 33783
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4007269279393173e-06,
+      "loss": 2.6471,
+      "step": 33784
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4006795195954488e-06,
+      "loss": 2.0024,
+      "step": 33785
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 1.4006321112515803e-06,
+      "loss": 2.5838,
+      "step": 33786
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4005847029077118e-06,
+      "loss": 1.2337,
+      "step": 33787
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4005372945638433e-06,
+      "loss": 2.4984,
+      "step": 33788
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4004898862199748e-06,
+      "loss": 2.8617,
+      "step": 33789
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4004424778761063e-06,
+      "loss": 1.7882,
+      "step": 33790
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4003950695322378e-06,
+      "loss": 2.1897,
+      "step": 33791
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.400347661188369e-06,
+      "loss": 2.3938,
+      "step": 33792
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4003002528445008e-06,
+      "loss": 2.6031,
+      "step": 33793
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.400252844500632e-06,
+      "loss": 2.5811,
+      "step": 33794
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4002054361567635e-06,
+      "loss": 2.6831,
+      "step": 33795
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4001580278128952e-06,
+      "loss": 3.032,
+      "step": 33796
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4001106194690265e-06,
+      "loss": 3.17,
+      "step": 33797
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.400063211125158e-06,
+      "loss": 1.4473,
+      "step": 33798
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.4000158027812895e-06,
+      "loss": 2.0765,
+      "step": 33799
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.399968394437421e-06,
+      "loss": 2.2522,
+      "step": 33800
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3999209860935525e-06,
+      "loss": 2.2599,
+      "step": 33801
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.399873577749684e-06,
+      "loss": 1.1037,
+      "step": 33802
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3998261694058155e-06,
+      "loss": 1.0216,
+      "step": 33803
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.399778761061947e-06,
+      "loss": 2.7832,
+      "step": 33804
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3997313527180783e-06,
+      "loss": 1.7971,
+      "step": 33805
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.39968394437421e-06,
+      "loss": 2.6464,
+      "step": 33806
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3996365360303414e-06,
+      "loss": 2.2855,
+      "step": 33807
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3995891276864727e-06,
+      "loss": 2.6462,
+      "step": 33808
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3995417193426044e-06,
+      "loss": 1.5376,
+      "step": 33809
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.399494310998736e-06,
+      "loss": 1.8189,
+      "step": 33810
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3994469026548672e-06,
+      "loss": 1.747,
+      "step": 33811
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3993994943109987e-06,
+      "loss": 4.1501,
+      "step": 33812
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3993520859671304e-06,
+      "loss": 2.1613,
+      "step": 33813
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3993046776232617e-06,
+      "loss": 1.3003,
+      "step": 33814
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3992572692793932e-06,
+      "loss": 2.9657,
+      "step": 33815
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3992098609355247e-06,
+      "loss": 1.0852,
+      "step": 33816
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3991624525916562e-06,
+      "loss": 1.8478,
+      "step": 33817
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3991150442477876e-06,
+      "loss": 2.8205,
+      "step": 33818
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3990676359039191e-06,
+      "loss": 1.9991,
+      "step": 33819
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3990202275600506e-06,
+      "loss": 2.7429,
+      "step": 33820
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3989728192161821e-06,
+      "loss": 3.0315,
+      "step": 33821
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3989254108723136e-06,
+      "loss": 2.3504,
+      "step": 33822
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3988780025284451e-06,
+      "loss": 3.1662,
+      "step": 33823
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3988305941845766e-06,
+      "loss": 3.4695,
+      "step": 33824
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3987831858407079e-06,
+      "loss": 2.0092,
+      "step": 33825
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3987357774968396e-06,
+      "loss": 2.5264,
+      "step": 33826
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3986883691529709e-06,
+      "loss": 1.2889,
+      "step": 33827
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3986409608091024e-06,
+      "loss": 1.9337,
+      "step": 33828
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.398593552465234e-06,
+      "loss": 2.2514,
+      "step": 33829
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3985461441213653e-06,
+      "loss": 2.6794,
+      "step": 33830
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3984987357774968e-06,
+      "loss": 2.2428,
+      "step": 33831
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3984513274336283e-06,
+      "loss": 2.7744,
+      "step": 33832
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3984039190897598e-06,
+      "loss": 1.6955,
+      "step": 33833
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3983565107458913e-06,
+      "loss": 2.397,
+      "step": 33834
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3983091024020228e-06,
+      "loss": 2.0639,
+      "step": 33835
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3982616940581543e-06,
+      "loss": 3.6774,
+      "step": 33836
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3982142857142858e-06,
+      "loss": 2.7141,
+      "step": 33837
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.398166877370417e-06,
+      "loss": 2.8894,
+      "step": 33838
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3981194690265488e-06,
+      "loss": 1.7464,
+      "step": 33839
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3980720606826803e-06,
+      "loss": 1.5866,
+      "step": 33840
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3980246523388115e-06,
+      "loss": 0.9138,
+      "step": 33841
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.397977243994943e-06,
+      "loss": 2.1165,
+      "step": 33842
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3979298356510747e-06,
+      "loss": 1.9369,
+      "step": 33843
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.397882427307206e-06,
+      "loss": 1.5875,
+      "step": 33844
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3978350189633375e-06,
+      "loss": 2.1888,
+      "step": 33845
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3977876106194692e-06,
+      "loss": 2.0557,
+      "step": 33846
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3977402022756005e-06,
+      "loss": 1.2024,
+      "step": 33847
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.397692793931732e-06,
+      "loss": 2.3108,
+      "step": 33848
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 1.3976453855878635e-06,
+      "loss": 2.8351,
+      "step": 33849
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.397597977243995e-06,
+      "loss": 2.4186,
+      "step": 33850
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3975505689001265e-06,
+      "loss": 1.4723,
+      "step": 33851
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.397503160556258e-06,
+      "loss": 3.5348,
+      "step": 33852
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3974557522123895e-06,
+      "loss": 1.7796,
+      "step": 33853
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.397408343868521e-06,
+      "loss": 2.4444,
+      "step": 33854
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3973609355246522e-06,
+      "loss": 2.9404,
+      "step": 33855
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.397313527180784e-06,
+      "loss": 2.7502,
+      "step": 33856
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3972661188369154e-06,
+      "loss": 2.286,
+      "step": 33857
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3972187104930467e-06,
+      "loss": 2.5348,
+      "step": 33858
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3971713021491784e-06,
+      "loss": 2.9228,
+      "step": 33859
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3971238938053097e-06,
+      "loss": 3.297,
+      "step": 33860
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3970764854614412e-06,
+      "loss": 2.5023,
+      "step": 33861
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3970290771175727e-06,
+      "loss": 1.2759,
+      "step": 33862
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3969816687737042e-06,
+      "loss": 2.925,
+      "step": 33863
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3969342604298357e-06,
+      "loss": 1.8347,
+      "step": 33864
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3968868520859671e-06,
+      "loss": 3.3673,
+      "step": 33865
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3968394437420986e-06,
+      "loss": 2.0161,
+      "step": 33866
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3967920353982301e-06,
+      "loss": 1.757,
+      "step": 33867
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3967446270543616e-06,
+      "loss": 2.3579,
+      "step": 33868
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3966972187104931e-06,
+      "loss": 1.1278,
+      "step": 33869
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3966498103666246e-06,
+      "loss": 2.0128,
+      "step": 33870
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3966024020227559e-06,
+      "loss": 1.4401,
+      "step": 33871
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3965549936788876e-06,
+      "loss": 2.2151,
+      "step": 33872
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.396507585335019e-06,
+      "loss": 2.3976,
+      "step": 33873
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3964601769911504e-06,
+      "loss": 1.0874,
+      "step": 33874
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3964127686472819e-06,
+      "loss": 0.9083,
+      "step": 33875
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3963653603034136e-06,
+      "loss": 2.6977,
+      "step": 33876
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3963179519595448e-06,
+      "loss": 1.5527,
+      "step": 33877
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3962705436156763e-06,
+      "loss": 1.9365,
+      "step": 33878
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.396223135271808e-06,
+      "loss": 1.9204,
+      "step": 33879
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3961757269279393e-06,
+      "loss": 1.9617,
+      "step": 33880
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3961283185840708e-06,
+      "loss": 2.6091,
+      "step": 33881
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3960809102402023e-06,
+      "loss": 1.5259,
+      "step": 33882
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3960335018963338e-06,
+      "loss": 1.2942,
+      "step": 33883
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3959860935524653e-06,
+      "loss": 2.2545,
+      "step": 33884
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3959386852085968e-06,
+      "loss": 3.8772,
+      "step": 33885
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3958912768647283e-06,
+      "loss": 1.1157,
+      "step": 33886
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3958438685208598e-06,
+      "loss": 2.4946,
+      "step": 33887
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.395796460176991e-06,
+      "loss": 2.5377,
+      "step": 33888
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3957490518331227e-06,
+      "loss": 3.0541,
+      "step": 33889
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3957016434892542e-06,
+      "loss": 2.3432,
+      "step": 33890
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3956542351453855e-06,
+      "loss": 1.4694,
+      "step": 33891
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3956068268015172e-06,
+      "loss": 2.2009,
+      "step": 33892
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3955594184576485e-06,
+      "loss": 1.8856,
+      "step": 33893
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.39551201011378e-06,
+      "loss": 2.5385,
+      "step": 33894
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3954646017699115e-06,
+      "loss": 2.1481,
+      "step": 33895
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.395417193426043e-06,
+      "loss": 2.9168,
+      "step": 33896
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3953697850821745e-06,
+      "loss": 1.6774,
+      "step": 33897
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.395322376738306e-06,
+      "loss": 1.7456,
+      "step": 33898
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3952749683944375e-06,
+      "loss": 1.5399,
+      "step": 33899
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.395227560050569e-06,
+      "loss": 2.883,
+      "step": 33900
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3951801517067004e-06,
+      "loss": 1.5417,
+      "step": 33901
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.395132743362832e-06,
+      "loss": 2.1925,
+      "step": 33902
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3950853350189634e-06,
+      "loss": 2.3385,
+      "step": 33903
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3950379266750947e-06,
+      "loss": 3.0746,
+      "step": 33904
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3949905183312264e-06,
+      "loss": 2.2429,
+      "step": 33905
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.394943109987358e-06,
+      "loss": 2.6572,
+      "step": 33906
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3948957016434892e-06,
+      "loss": 2.3342,
+      "step": 33907
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3948482932996207e-06,
+      "loss": 2.6142,
+      "step": 33908
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3948008849557524e-06,
+      "loss": 1.8158,
+      "step": 33909
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3947534766118837e-06,
+      "loss": 1.0063,
+      "step": 33910
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3947060682680152e-06,
+      "loss": 3.1848,
+      "step": 33911
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3946586599241469e-06,
+      "loss": 3.3247,
+      "step": 33912
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 1.3946112515802781e-06,
+      "loss": 3.3222,
+      "step": 33913
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3945638432364096e-06,
+      "loss": 3.7127,
+      "step": 33914
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3945164348925411e-06,
+      "loss": 2.3886,
+      "step": 33915
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3944690265486726e-06,
+      "loss": 1.7026,
+      "step": 33916
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.394421618204804e-06,
+      "loss": 1.718,
+      "step": 33917
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3943742098609356e-06,
+      "loss": 0.7938,
+      "step": 33918
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.394326801517067e-06,
+      "loss": 2.9358,
+      "step": 33919
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3942793931731986e-06,
+      "loss": 3.9083,
+      "step": 33920
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3942319848293299e-06,
+      "loss": 1.9749,
+      "step": 33921
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3941845764854616e-06,
+      "loss": 2.1395,
+      "step": 33922
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.394137168141593e-06,
+      "loss": 1.9896,
+      "step": 33923
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3940897597977243e-06,
+      "loss": 2.9759,
+      "step": 33924
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.394042351453856e-06,
+      "loss": 1.1447,
+      "step": 33925
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3939949431099873e-06,
+      "loss": 1.4826,
+      "step": 33926
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3939475347661188e-06,
+      "loss": 2.1847,
+      "step": 33927
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3939001264222503e-06,
+      "loss": 1.7901,
+      "step": 33928
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3938527180783818e-06,
+      "loss": 1.7559,
+      "step": 33929
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3938053097345133e-06,
+      "loss": 2.7367,
+      "step": 33930
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3937579013906448e-06,
+      "loss": 1.2002,
+      "step": 33931
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3937104930467763e-06,
+      "loss": 1.3642,
+      "step": 33932
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3936630847029078e-06,
+      "loss": 3.7926,
+      "step": 33933
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3936156763590393e-06,
+      "loss": 3.1382,
+      "step": 33934
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3935682680151708e-06,
+      "loss": 2.7344,
+      "step": 33935
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3935208596713022e-06,
+      "loss": 1.9446,
+      "step": 33936
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3934734513274335e-06,
+      "loss": 2.8917,
+      "step": 33937
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3934260429835652e-06,
+      "loss": 1.0559,
+      "step": 33938
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3933786346396967e-06,
+      "loss": 1.7096,
+      "step": 33939
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.393331226295828e-06,
+      "loss": 2.1265,
+      "step": 33940
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3932838179519595e-06,
+      "loss": 3.0736,
+      "step": 33941
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3932364096080912e-06,
+      "loss": 1.5613,
+      "step": 33942
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3931890012642225e-06,
+      "loss": 1.7729,
+      "step": 33943
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.393141592920354e-06,
+      "loss": 2.0345,
+      "step": 33944
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3930941845764857e-06,
+      "loss": 3.3727,
+      "step": 33945
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.393046776232617e-06,
+      "loss": 1.2954,
+      "step": 33946
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3929993678887484e-06,
+      "loss": 2.0713,
+      "step": 33947
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.39295195954488e-06,
+      "loss": 2.9742,
+      "step": 33948
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3929045512010114e-06,
+      "loss": 2.8311,
+      "step": 33949
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.392857142857143e-06,
+      "loss": 2.8025,
+      "step": 33950
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3928097345132744e-06,
+      "loss": 1.7567,
+      "step": 33951
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.392762326169406e-06,
+      "loss": 2.5264,
+      "step": 33952
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3927149178255374e-06,
+      "loss": 2.5385,
+      "step": 33953
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3926675094816687e-06,
+      "loss": 1.7034,
+      "step": 33954
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3926201011378004e-06,
+      "loss": 1.838,
+      "step": 33955
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3925726927939319e-06,
+      "loss": 1.9943,
+      "step": 33956
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3925252844500632e-06,
+      "loss": 3.4107,
+      "step": 33957
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3924778761061949e-06,
+      "loss": 2.4111,
+      "step": 33958
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3924304677623261e-06,
+      "loss": 0.7267,
+      "step": 33959
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3923830594184576e-06,
+      "loss": 2.2198,
+      "step": 33960
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3923356510745891e-06,
+      "loss": 1.7162,
+      "step": 33961
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3922882427307206e-06,
+      "loss": 1.8536,
+      "step": 33962
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3922408343868521e-06,
+      "loss": 1.9128,
+      "step": 33963
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3921934260429836e-06,
+      "loss": 2.1861,
+      "step": 33964
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.392146017699115e-06,
+      "loss": 1.5391,
+      "step": 33965
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3920986093552466e-06,
+      "loss": 1.619,
+      "step": 33966
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.392051201011378e-06,
+      "loss": 3.2655,
+      "step": 33967
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3920037926675096e-06,
+      "loss": 1.2785,
+      "step": 33968
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.391956384323641e-06,
+      "loss": 0.6384,
+      "step": 33969
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3919089759797723e-06,
+      "loss": 2.4229,
+      "step": 33970
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.391861567635904e-06,
+      "loss": 1.4586,
+      "step": 33971
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3918141592920355e-06,
+      "loss": 1.9472,
+      "step": 33972
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3917667509481668e-06,
+      "loss": 2.3174,
+      "step": 33973
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3917193426042983e-06,
+      "loss": 2.4401,
+      "step": 33974
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.39167193426043e-06,
+      "loss": 1.3721,
+      "step": 33975
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 1.3916245259165613e-06,
+      "loss": 1.7873,
+      "step": 33976
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3915771175726928e-06,
+      "loss": 2.6714,
+      "step": 33977
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3915297092288245e-06,
+      "loss": 2.3051,
+      "step": 33978
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3914823008849558e-06,
+      "loss": 0.8867,
+      "step": 33979
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3914348925410873e-06,
+      "loss": 2.1501,
+      "step": 33980
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3913874841972188e-06,
+      "loss": 3.8776,
+      "step": 33981
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3913400758533502e-06,
+      "loss": 2.1794,
+      "step": 33982
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3912926675094817e-06,
+      "loss": 1.1178,
+      "step": 33983
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.391245259165613e-06,
+      "loss": 2.3144,
+      "step": 33984
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3911978508217447e-06,
+      "loss": 0.9918,
+      "step": 33985
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3911504424778762e-06,
+      "loss": 1.982,
+      "step": 33986
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3911030341340075e-06,
+      "loss": 1.5996,
+      "step": 33987
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3910556257901392e-06,
+      "loss": 2.4891,
+      "step": 33988
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3910082174462707e-06,
+      "loss": 2.3617,
+      "step": 33989
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.390960809102402e-06,
+      "loss": 1.8478,
+      "step": 33990
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3909134007585335e-06,
+      "loss": 2.9515,
+      "step": 33991
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.390865992414665e-06,
+      "loss": 4.0045,
+      "step": 33992
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3908185840707965e-06,
+      "loss": 1.3333,
+      "step": 33993
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.390771175726928e-06,
+      "loss": 1.7998,
+      "step": 33994
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3907237673830594e-06,
+      "loss": 1.7791,
+      "step": 33995
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.390676359039191e-06,
+      "loss": 3.4171,
+      "step": 33996
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3906289506953224e-06,
+      "loss": 2.4856,
+      "step": 33997
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.390581542351454e-06,
+      "loss": 2.4255,
+      "step": 33998
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3905341340075854e-06,
+      "loss": 2.888,
+      "step": 33999
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.390486725663717e-06,
+      "loss": 2.2003,
+      "step": 34000
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3904393173198484e-06,
+      "loss": 2.7383,
+      "step": 34001
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3903919089759799e-06,
+      "loss": 3.6275,
+      "step": 34002
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3903445006321112e-06,
+      "loss": 1.6691,
+      "step": 34003
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3902970922882427e-06,
+      "loss": 1.3865,
+      "step": 34004
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3902496839443744e-06,
+      "loss": 1.7262,
+      "step": 34005
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3902022756005056e-06,
+      "loss": 1.6616,
+      "step": 34006
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3901548672566371e-06,
+      "loss": 3.4186,
+      "step": 34007
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3901074589127688e-06,
+      "loss": 2.9526,
+      "step": 34008
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3900600505689001e-06,
+      "loss": 1.1402,
+      "step": 34009
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3900126422250316e-06,
+      "loss": 1.879,
+      "step": 34010
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.389965233881163e-06,
+      "loss": 3.9367,
+      "step": 34011
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3899178255372946e-06,
+      "loss": 2.2501,
+      "step": 34012
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.389870417193426e-06,
+      "loss": 2.6205,
+      "step": 34013
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3898230088495576e-06,
+      "loss": 2.9817,
+      "step": 34014
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.389775600505689e-06,
+      "loss": 2.5535,
+      "step": 34015
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3897281921618206e-06,
+      "loss": 0.6577,
+      "step": 34016
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3896807838179518e-06,
+      "loss": 1.6933,
+      "step": 34017
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3896333754740835e-06,
+      "loss": 0.9358,
+      "step": 34018
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.389585967130215e-06,
+      "loss": 2.6978,
+      "step": 34019
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3895385587863463e-06,
+      "loss": 2.2205,
+      "step": 34020
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.389491150442478e-06,
+      "loss": 1.4243,
+      "step": 34021
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3894437420986095e-06,
+      "loss": 2.5417,
+      "step": 34022
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3893963337547408e-06,
+      "loss": 2.6209,
+      "step": 34023
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3893489254108723e-06,
+      "loss": 3.0766,
+      "step": 34024
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3893015170670038e-06,
+      "loss": 2.076,
+      "step": 34025
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3892541087231353e-06,
+      "loss": 2.05,
+      "step": 34026
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3892067003792668e-06,
+      "loss": 3.5235,
+      "step": 34027
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3891592920353983e-06,
+      "loss": 2.4798,
+      "step": 34028
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3891118836915297e-06,
+      "loss": 1.1906,
+      "step": 34029
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3890644753476612e-06,
+      "loss": 2.674,
+      "step": 34030
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3890170670037927e-06,
+      "loss": 2.2517,
+      "step": 34031
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3889696586599242e-06,
+      "loss": 2.2026,
+      "step": 34032
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3889222503160557e-06,
+      "loss": 1.5994,
+      "step": 34033
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3888748419721872e-06,
+      "loss": 1.7203,
+      "step": 34034
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3888274336283187e-06,
+      "loss": 2.3818,
+      "step": 34035
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.38878002528445e-06,
+      "loss": 3.817,
+      "step": 34036
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3887326169405815e-06,
+      "loss": 1.6254,
+      "step": 34037
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3886852085967132e-06,
+      "loss": 1.948,
+      "step": 34038
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 1.3886378002528445e-06,
+      "loss": 2.2592,
+      "step": 34039
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.388590391908976e-06,
+      "loss": 2.5828,
+      "step": 34040
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3885429835651077e-06,
+      "loss": 1.5739,
+      "step": 34041
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.388495575221239e-06,
+      "loss": 1.8763,
+      "step": 34042
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3884481668773704e-06,
+      "loss": 2.8415,
+      "step": 34043
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.388400758533502e-06,
+      "loss": 2.6649,
+      "step": 34044
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3883533501896334e-06,
+      "loss": 3.6124,
+      "step": 34045
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.388305941845765e-06,
+      "loss": 2.8967,
+      "step": 34046
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3882585335018964e-06,
+      "loss": 1.3103,
+      "step": 34047
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3882111251580279e-06,
+      "loss": 1.8745,
+      "step": 34048
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3881637168141594e-06,
+      "loss": 2.5364,
+      "step": 34049
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3881163084702907e-06,
+      "loss": 1.6223,
+      "step": 34050
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3880689001264224e-06,
+      "loss": 2.5753,
+      "step": 34051
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3880214917825539e-06,
+      "loss": 2.8545,
+      "step": 34052
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3879740834386851e-06,
+      "loss": 2.9778,
+      "step": 34053
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3879266750948168e-06,
+      "loss": 2.2744,
+      "step": 34054
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3878792667509483e-06,
+      "loss": 2.7166,
+      "step": 34055
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3878318584070796e-06,
+      "loss": 2.2143,
+      "step": 34056
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.387784450063211e-06,
+      "loss": 0.8176,
+      "step": 34057
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3877370417193426e-06,
+      "loss": 2.1842,
+      "step": 34058
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.387689633375474e-06,
+      "loss": 2.0636,
+      "step": 34059
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3876422250316056e-06,
+      "loss": 1.9738,
+      "step": 34060
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.387594816687737e-06,
+      "loss": 2.1247,
+      "step": 34061
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3875474083438686e-06,
+      "loss": 2.1647,
+      "step": 34062
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3875e-06,
+      "loss": 1.6381,
+      "step": 34063
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3874525916561315e-06,
+      "loss": 1.5368,
+      "step": 34064
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.387405183312263e-06,
+      "loss": 1.2836,
+      "step": 34065
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3873577749683945e-06,
+      "loss": 3.0534,
+      "step": 34066
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.387310366624526e-06,
+      "loss": 2.3204,
+      "step": 34067
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3872629582806575e-06,
+      "loss": 2.2179,
+      "step": 34068
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3872155499367888e-06,
+      "loss": 1.5141,
+      "step": 34069
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3871681415929203e-06,
+      "loss": 3.0162,
+      "step": 34070
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.387120733249052e-06,
+      "loss": 2.5475,
+      "step": 34071
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3870733249051833e-06,
+      "loss": 2.3628,
+      "step": 34072
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3870259165613148e-06,
+      "loss": 2.3046,
+      "step": 34073
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3869785082174465e-06,
+      "loss": 2.6133,
+      "step": 34074
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3869310998735778e-06,
+      "loss": 1.9652,
+      "step": 34075
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3868836915297092e-06,
+      "loss": 2.0585,
+      "step": 34076
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3868362831858407e-06,
+      "loss": 3.1139,
+      "step": 34077
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3867888748419722e-06,
+      "loss": 1.4845,
+      "step": 34078
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3867414664981037e-06,
+      "loss": 2.1065,
+      "step": 34079
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3866940581542352e-06,
+      "loss": 1.4459,
+      "step": 34080
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3866466498103667e-06,
+      "loss": 1.2094,
+      "step": 34081
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3865992414664982e-06,
+      "loss": 2.456,
+      "step": 34082
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3865518331226295e-06,
+      "loss": 2.2179,
+      "step": 34083
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3865044247787612e-06,
+      "loss": 1.9891,
+      "step": 34084
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3864570164348927e-06,
+      "loss": 0.9284,
+      "step": 34085
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.386409608091024e-06,
+      "loss": 2.1835,
+      "step": 34086
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3863621997471557e-06,
+      "loss": 1.3737,
+      "step": 34087
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3863147914032871e-06,
+      "loss": 2.1576,
+      "step": 34088
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3862673830594184e-06,
+      "loss": 3.6144,
+      "step": 34089
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.38621997471555e-06,
+      "loss": 1.271,
+      "step": 34090
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3861725663716814e-06,
+      "loss": 2.1834,
+      "step": 34091
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.386125158027813e-06,
+      "loss": 1.6016,
+      "step": 34092
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3860777496839444e-06,
+      "loss": 1.7331,
+      "step": 34093
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3860303413400759e-06,
+      "loss": 2.4571,
+      "step": 34094
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3859829329962074e-06,
+      "loss": 1.7035,
+      "step": 34095
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3859355246523389e-06,
+      "loss": 2.891,
+      "step": 34096
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3858881163084704e-06,
+      "loss": 2.6188,
+      "step": 34097
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3858407079646019e-06,
+      "loss": 1.8843,
+      "step": 34098
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3857932996207334e-06,
+      "loss": 2.3179,
+      "step": 34099
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3857458912768648e-06,
+      "loss": 2.5702,
+      "step": 34100
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3856984829329963e-06,
+      "loss": 1.8947,
+      "step": 34101
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3856510745891276e-06,
+      "loss": 2.3155,
+      "step": 34102
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 1.3856036662452591e-06,
+      "loss": 2.6457,
+      "step": 34103
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3855562579013908e-06,
+      "loss": 2.743,
+      "step": 34104
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.385508849557522e-06,
+      "loss": 1.7388,
+      "step": 34105
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3854614412136536e-06,
+      "loss": 1.4845,
+      "step": 34106
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3854140328697853e-06,
+      "loss": 1.306,
+      "step": 34107
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3853666245259166e-06,
+      "loss": 2.2429,
+      "step": 34108
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.385319216182048e-06,
+      "loss": 2.2266,
+      "step": 34109
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3852718078381796e-06,
+      "loss": 2.3048,
+      "step": 34110
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.385224399494311e-06,
+      "loss": 3.2181,
+      "step": 34111
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3851769911504425e-06,
+      "loss": 1.6424,
+      "step": 34112
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.385129582806574e-06,
+      "loss": 1.9151,
+      "step": 34113
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3850821744627055e-06,
+      "loss": 2.4109,
+      "step": 34114
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.385034766118837e-06,
+      "loss": 3.9308,
+      "step": 34115
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3849873577749683e-06,
+      "loss": 1.5381,
+      "step": 34116
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3849399494311e-06,
+      "loss": 1.4845,
+      "step": 34117
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3848925410872315e-06,
+      "loss": 2.4923,
+      "step": 34118
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3848451327433628e-06,
+      "loss": 3.1034,
+      "step": 34119
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3847977243994945e-06,
+      "loss": 1.9006,
+      "step": 34120
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3847503160556258e-06,
+      "loss": 1.7945,
+      "step": 34121
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3847029077117572e-06,
+      "loss": 2.6686,
+      "step": 34122
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3846554993678887e-06,
+      "loss": 4.1961,
+      "step": 34123
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3846080910240202e-06,
+      "loss": 1.8865,
+      "step": 34124
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3845606826801517e-06,
+      "loss": 2.2827,
+      "step": 34125
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3845132743362832e-06,
+      "loss": 1.9421,
+      "step": 34126
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3844658659924147e-06,
+      "loss": 3.9516,
+      "step": 34127
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3844184576485462e-06,
+      "loss": 1.9173,
+      "step": 34128
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3843710493046777e-06,
+      "loss": 2.3141,
+      "step": 34129
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3843236409608092e-06,
+      "loss": 1.9126,
+      "step": 34130
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3842762326169407e-06,
+      "loss": 1.7749,
+      "step": 34131
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.384228824273072e-06,
+      "loss": 2.6831,
+      "step": 34132
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3841814159292034e-06,
+      "loss": 2.7309,
+      "step": 34133
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3841340075853352e-06,
+      "loss": 2.0991,
+      "step": 34134
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3840865992414664e-06,
+      "loss": 2.8029,
+      "step": 34135
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.384039190897598e-06,
+      "loss": 0.8125,
+      "step": 34136
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3839917825537296e-06,
+      "loss": 1.2756,
+      "step": 34137
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.383944374209861e-06,
+      "loss": 2.8206,
+      "step": 34138
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3838969658659924e-06,
+      "loss": 2.8959,
+      "step": 34139
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.383849557522124e-06,
+      "loss": 3.9161,
+      "step": 34140
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3838021491782554e-06,
+      "loss": 1.4727,
+      "step": 34141
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3837547408343869e-06,
+      "loss": 2.5396,
+      "step": 34142
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3837073324905184e-06,
+      "loss": 1.8715,
+      "step": 34143
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3836599241466499e-06,
+      "loss": 1.7299,
+      "step": 34144
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3836125158027814e-06,
+      "loss": 4.1357,
+      "step": 34145
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3835651074589126e-06,
+      "loss": 1.9539,
+      "step": 34146
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3835176991150443e-06,
+      "loss": 2.2661,
+      "step": 34147
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3834702907711758e-06,
+      "loss": 1.2347,
+      "step": 34148
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3834228824273071e-06,
+      "loss": 1.4586,
+      "step": 34149
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3833754740834388e-06,
+      "loss": 2.092,
+      "step": 34150
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3833280657395703e-06,
+      "loss": 1.2142,
+      "step": 34151
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3832806573957016e-06,
+      "loss": 1.9649,
+      "step": 34152
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.383233249051833e-06,
+      "loss": 2.1395,
+      "step": 34153
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3831858407079646e-06,
+      "loss": 1.1114,
+      "step": 34154
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.383138432364096e-06,
+      "loss": 2.5128,
+      "step": 34155
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3830910240202276e-06,
+      "loss": 2.4392,
+      "step": 34156
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.383043615676359e-06,
+      "loss": 2.565,
+      "step": 34157
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3829962073324905e-06,
+      "loss": 2.9274,
+      "step": 34158
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.382948798988622e-06,
+      "loss": 2.3762,
+      "step": 34159
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3829013906447535e-06,
+      "loss": 2.1417,
+      "step": 34160
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.382853982300885e-06,
+      "loss": 2.2623,
+      "step": 34161
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3828065739570165e-06,
+      "loss": 2.1361,
+      "step": 34162
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.382759165613148e-06,
+      "loss": 2.694,
+      "step": 34163
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3827117572692795e-06,
+      "loss": 2.1249,
+      "step": 34164
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3826643489254108e-06,
+      "loss": 3.3097,
+      "step": 34165
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 1.3826169405815423e-06,
+      "loss": 2.0368,
+      "step": 34166
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.382569532237674e-06,
+      "loss": 2.352,
+      "step": 34167
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3825221238938053e-06,
+      "loss": 1.7773,
+      "step": 34168
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3824747155499367e-06,
+      "loss": 2.7528,
+      "step": 34169
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3824273072060684e-06,
+      "loss": 2.2033,
+      "step": 34170
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3823798988621997e-06,
+      "loss": 2.6936,
+      "step": 34171
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3823324905183312e-06,
+      "loss": 1.345,
+      "step": 34172
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3822850821744627e-06,
+      "loss": 3.1897,
+      "step": 34173
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3822376738305942e-06,
+      "loss": 1.7662,
+      "step": 34174
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3821902654867257e-06,
+      "loss": 1.3099,
+      "step": 34175
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3821428571428572e-06,
+      "loss": 2.735,
+      "step": 34176
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3820954487989887e-06,
+      "loss": 0.934,
+      "step": 34177
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3820480404551202e-06,
+      "loss": 2.2512,
+      "step": 34178
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3820006321112515e-06,
+      "loss": 2.4733,
+      "step": 34179
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3819532237673832e-06,
+      "loss": 1.3443,
+      "step": 34180
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3819058154235147e-06,
+      "loss": 2.8727,
+      "step": 34181
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.381858407079646e-06,
+      "loss": 2.5049,
+      "step": 34182
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3818109987357776e-06,
+      "loss": 1.9452,
+      "step": 34183
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3817635903919091e-06,
+      "loss": 2.7166,
+      "step": 34184
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3817161820480404e-06,
+      "loss": 2.5461,
+      "step": 34185
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.381668773704172e-06,
+      "loss": 2.3883,
+      "step": 34186
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3816213653603034e-06,
+      "loss": 2.4414,
+      "step": 34187
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3815739570164349e-06,
+      "loss": 2.5205,
+      "step": 34188
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3815265486725664e-06,
+      "loss": 1.9039,
+      "step": 34189
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3814791403286979e-06,
+      "loss": 3.0677,
+      "step": 34190
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3814317319848294e-06,
+      "loss": 2.2735,
+      "step": 34191
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3813843236409609e-06,
+      "loss": 2.9567,
+      "step": 34192
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3813369152970923e-06,
+      "loss": 2.6647,
+      "step": 34193
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3812895069532238e-06,
+      "loss": 1.5719,
+      "step": 34194
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3812420986093553e-06,
+      "loss": 1.9879,
+      "step": 34195
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3811946902654868e-06,
+      "loss": 2.0905,
+      "step": 34196
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3811472819216183e-06,
+      "loss": 2.1656,
+      "step": 34197
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3810998735777496e-06,
+      "loss": 2.2865,
+      "step": 34198
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.381052465233881e-06,
+      "loss": 1.2673,
+      "step": 34199
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3810050568900128e-06,
+      "loss": 3.9601,
+      "step": 34200
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.380957648546144e-06,
+      "loss": 1.4087,
+      "step": 34201
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3809102402022756e-06,
+      "loss": 2.2056,
+      "step": 34202
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3808628318584073e-06,
+      "loss": 2.8886,
+      "step": 34203
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3808154235145385e-06,
+      "loss": 1.0964,
+      "step": 34204
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.38076801517067e-06,
+      "loss": 2.1201,
+      "step": 34205
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3807206068268015e-06,
+      "loss": 1.0832,
+      "step": 34206
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.380673198482933e-06,
+      "loss": 1.2884,
+      "step": 34207
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3806257901390645e-06,
+      "loss": 3.8078,
+      "step": 34208
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.380578381795196e-06,
+      "loss": 0.9376,
+      "step": 34209
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3805309734513275e-06,
+      "loss": 2.8433,
+      "step": 34210
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.380483565107459e-06,
+      "loss": 2.3488,
+      "step": 34211
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3804361567635903e-06,
+      "loss": 2.5688,
+      "step": 34212
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.380388748419722e-06,
+      "loss": 1.4562,
+      "step": 34213
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3803413400758535e-06,
+      "loss": 2.4915,
+      "step": 34214
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3802939317319847e-06,
+      "loss": 1.9698,
+      "step": 34215
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3802465233881165e-06,
+      "loss": 2.0591,
+      "step": 34216
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.380199115044248e-06,
+      "loss": 1.5145,
+      "step": 34217
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3801517067003792e-06,
+      "loss": 3.4077,
+      "step": 34218
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3801042983565107e-06,
+      "loss": 2.4627,
+      "step": 34219
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3800568900126422e-06,
+      "loss": 2.2677,
+      "step": 34220
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3800094816687737e-06,
+      "loss": 1.5778,
+      "step": 34221
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3799620733249052e-06,
+      "loss": 1.9272,
+      "step": 34222
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3799146649810367e-06,
+      "loss": 1.3256,
+      "step": 34223
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3798672566371682e-06,
+      "loss": 2.7452,
+      "step": 34224
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3798198482932997e-06,
+      "loss": 1.6559,
+      "step": 34225
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3797724399494312e-06,
+      "loss": 2.0143,
+      "step": 34226
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3797250316055627e-06,
+      "loss": 1.7396,
+      "step": 34227
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3796776232616941e-06,
+      "loss": 2.4369,
+      "step": 34228
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 1.3796302149178256e-06,
+      "loss": 2.2175,
+      "step": 34229
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3795828065739571e-06,
+      "loss": 1.9811,
+      "step": 34230
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3795353982300884e-06,
+      "loss": 2.1525,
+      "step": 34231
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.37948798988622e-06,
+      "loss": 1.1691,
+      "step": 34232
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3794405815423516e-06,
+      "loss": 2.8027,
+      "step": 34233
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3793931731984829e-06,
+      "loss": 2.62,
+      "step": 34234
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3793457648546144e-06,
+      "loss": 1.944,
+      "step": 34235
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.379298356510746e-06,
+      "loss": 1.118,
+      "step": 34236
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3792509481668774e-06,
+      "loss": 2.8664,
+      "step": 34237
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3792035398230089e-06,
+      "loss": 2.329,
+      "step": 34238
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3791561314791404e-06,
+      "loss": 2.152,
+      "step": 34239
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3791087231352718e-06,
+      "loss": 1.7644,
+      "step": 34240
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3790613147914033e-06,
+      "loss": 1.502,
+      "step": 34241
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3790139064475348e-06,
+      "loss": 3.1857,
+      "step": 34242
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3789664981036663e-06,
+      "loss": 1.9761,
+      "step": 34243
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3789190897597978e-06,
+      "loss": 1.2241,
+      "step": 34244
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378871681415929e-06,
+      "loss": 2.2431,
+      "step": 34245
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3788242730720608e-06,
+      "loss": 2.4073,
+      "step": 34246
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3787768647281923e-06,
+      "loss": 1.969,
+      "step": 34247
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3787294563843236e-06,
+      "loss": 3.2254,
+      "step": 34248
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3786820480404553e-06,
+      "loss": 1.6517,
+      "step": 34249
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3786346396965868e-06,
+      "loss": 2.1657,
+      "step": 34250
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378587231352718e-06,
+      "loss": 2.4653,
+      "step": 34251
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3785398230088495e-06,
+      "loss": 1.492,
+      "step": 34252
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378492414664981e-06,
+      "loss": 1.9284,
+      "step": 34253
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3784450063211125e-06,
+      "loss": 1.7934,
+      "step": 34254
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378397597977244e-06,
+      "loss": 2.2756,
+      "step": 34255
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3783501896333755e-06,
+      "loss": 1.6139,
+      "step": 34256
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378302781289507e-06,
+      "loss": 1.8688,
+      "step": 34257
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3782553729456385e-06,
+      "loss": 2.2935,
+      "step": 34258
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.37820796460177e-06,
+      "loss": 2.2469,
+      "step": 34259
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3781605562579015e-06,
+      "loss": 2.611,
+      "step": 34260
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378113147914033e-06,
+      "loss": 2.5508,
+      "step": 34261
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3780657395701642e-06,
+      "loss": 3.1546,
+      "step": 34262
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.378018331226296e-06,
+      "loss": 1.8889,
+      "step": 34263
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3779709228824272e-06,
+      "loss": 0.8533,
+      "step": 34264
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3779235145385587e-06,
+      "loss": 2.4722,
+      "step": 34265
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3778761061946904e-06,
+      "loss": 0.8706,
+      "step": 34266
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3778286978508217e-06,
+      "loss": 2.7244,
+      "step": 34267
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3777812895069532e-06,
+      "loss": 2.6798,
+      "step": 34268
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3777338811630847e-06,
+      "loss": 1.9283,
+      "step": 34269
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3776864728192162e-06,
+      "loss": 3.4081,
+      "step": 34270
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3776390644753477e-06,
+      "loss": 2.4245,
+      "step": 34271
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3775916561314792e-06,
+      "loss": 2.0489,
+      "step": 34272
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3775442477876107e-06,
+      "loss": 3.3161,
+      "step": 34273
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3774968394437422e-06,
+      "loss": 1.0112,
+      "step": 34274
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3774494310998734e-06,
+      "loss": 1.6147,
+      "step": 34275
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3774020227560051e-06,
+      "loss": 2.9127,
+      "step": 34276
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3773546144121366e-06,
+      "loss": 2.9619,
+      "step": 34277
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.377307206068268e-06,
+      "loss": 2.7004,
+      "step": 34278
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3772597977243996e-06,
+      "loss": 2.4624,
+      "step": 34279
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.377212389380531e-06,
+      "loss": 2.4747,
+      "step": 34280
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3771649810366624e-06,
+      "loss": 1.3089,
+      "step": 34281
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3771175726927939e-06,
+      "loss": 1.3375,
+      "step": 34282
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3770701643489256e-06,
+      "loss": 1.4771,
+      "step": 34283
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3770227560050569e-06,
+      "loss": 2.4404,
+      "step": 34284
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3769753476611884e-06,
+      "loss": 2.9088,
+      "step": 34285
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3769279393173198e-06,
+      "loss": 3.2697,
+      "step": 34286
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3768805309734513e-06,
+      "loss": 3.9788,
+      "step": 34287
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3768331226295828e-06,
+      "loss": 1.7783,
+      "step": 34288
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3767857142857143e-06,
+      "loss": 2.5766,
+      "step": 34289
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3767383059418458e-06,
+      "loss": 1.4782,
+      "step": 34290
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3766908975979773e-06,
+      "loss": 3.0456,
+      "step": 34291
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3766434892541088e-06,
+      "loss": 1.8327,
+      "step": 34292
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 1.3765960809102403e-06,
+      "loss": 2.271,
+      "step": 34293
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3765486725663718e-06,
+      "loss": 2.8742,
+      "step": 34294
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.376501264222503e-06,
+      "loss": 1.0693,
+      "step": 34295
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3764538558786348e-06,
+      "loss": 2.1134,
+      "step": 34296
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.376406447534766e-06,
+      "loss": 2.0503,
+      "step": 34297
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3763590391908975e-06,
+      "loss": 3.0737,
+      "step": 34298
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3763116308470292e-06,
+      "loss": 1.5925,
+      "step": 34299
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3762642225031605e-06,
+      "loss": 1.5418,
+      "step": 34300
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.376216814159292e-06,
+      "loss": 1.4738,
+      "step": 34301
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3761694058154235e-06,
+      "loss": 3.5603,
+      "step": 34302
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.376121997471555e-06,
+      "loss": 1.7055,
+      "step": 34303
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3760745891276865e-06,
+      "loss": 2.6664,
+      "step": 34304
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.376027180783818e-06,
+      "loss": 1.4021,
+      "step": 34305
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3759797724399495e-06,
+      "loss": 1.9369,
+      "step": 34306
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.375932364096081e-06,
+      "loss": 3.0334,
+      "step": 34307
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3758849557522123e-06,
+      "loss": 3.8659,
+      "step": 34308
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.375837547408344e-06,
+      "loss": 3.2336,
+      "step": 34309
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3757901390644754e-06,
+      "loss": 1.6055,
+      "step": 34310
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3757427307206067e-06,
+      "loss": 1.3616,
+      "step": 34311
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3756953223767384e-06,
+      "loss": 1.5416,
+      "step": 34312
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.37564791403287e-06,
+      "loss": 1.7016,
+      "step": 34313
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3756005056890012e-06,
+      "loss": 2.253,
+      "step": 34314
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3755530973451327e-06,
+      "loss": 2.8549,
+      "step": 34315
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3755056890012644e-06,
+      "loss": 1.8805,
+      "step": 34316
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3754582806573957e-06,
+      "loss": 2.764,
+      "step": 34317
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3754108723135272e-06,
+      "loss": 3.1567,
+      "step": 34318
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3753634639696587e-06,
+      "loss": 2.0066,
+      "step": 34319
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3753160556257902e-06,
+      "loss": 2.6958,
+      "step": 34320
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3752686472819216e-06,
+      "loss": 3.3811,
+      "step": 34321
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3752212389380531e-06,
+      "loss": 2.877,
+      "step": 34322
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3751738305941846e-06,
+      "loss": 3.2248,
+      "step": 34323
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3751264222503161e-06,
+      "loss": 2.1109,
+      "step": 34324
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3750790139064476e-06,
+      "loss": 1.5911,
+      "step": 34325
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3750316055625791e-06,
+      "loss": 2.8425,
+      "step": 34326
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3749841972187106e-06,
+      "loss": 2.6926,
+      "step": 34327
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3749367888748419e-06,
+      "loss": 1.8188,
+      "step": 34328
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3748893805309736e-06,
+      "loss": 1.9178,
+      "step": 34329
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3748419721871049e-06,
+      "loss": 1.9815,
+      "step": 34330
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3747945638432364e-06,
+      "loss": 1.2181,
+      "step": 34331
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.374747155499368e-06,
+      "loss": 2.578,
+      "step": 34332
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3746997471554993e-06,
+      "loss": 2.8159,
+      "step": 34333
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3746523388116308e-06,
+      "loss": 2.4861,
+      "step": 34334
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3746049304677623e-06,
+      "loss": 1.9003,
+      "step": 34335
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3745575221238938e-06,
+      "loss": 2.0127,
+      "step": 34336
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3745101137800253e-06,
+      "loss": 2.6174,
+      "step": 34337
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3744627054361568e-06,
+      "loss": 2.1547,
+      "step": 34338
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3744152970922883e-06,
+      "loss": 2.4986,
+      "step": 34339
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3743678887484198e-06,
+      "loss": 2.5631,
+      "step": 34340
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.374320480404551e-06,
+      "loss": 2.1099,
+      "step": 34341
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3742730720606828e-06,
+      "loss": 3.1903,
+      "step": 34342
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3742256637168143e-06,
+      "loss": 1.9099,
+      "step": 34343
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3741782553729455e-06,
+      "loss": 1.2745,
+      "step": 34344
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3741308470290773e-06,
+      "loss": 0.2527,
+      "step": 34345
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3740834386852087e-06,
+      "loss": 1.9966,
+      "step": 34346
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.37403603034134e-06,
+      "loss": 1.9813,
+      "step": 34347
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3739886219974715e-06,
+      "loss": 1.2776,
+      "step": 34348
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3739412136536032e-06,
+      "loss": 2.3881,
+      "step": 34349
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3738938053097345e-06,
+      "loss": 2.526,
+      "step": 34350
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.373846396965866e-06,
+      "loss": 1.0164,
+      "step": 34351
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3737989886219975e-06,
+      "loss": 2.344,
+      "step": 34352
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.373751580278129e-06,
+      "loss": 1.5405,
+      "step": 34353
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3737041719342605e-06,
+      "loss": 2.3821,
+      "step": 34354
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.373656763590392e-06,
+      "loss": 2.0946,
+      "step": 34355
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 1.3736093552465235e-06,
+      "loss": 1.9371,
+      "step": 34356
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.373561946902655e-06,
+      "loss": 1.454,
+      "step": 34357
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3735145385587864e-06,
+      "loss": 2.824,
+      "step": 34358
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.373467130214918e-06,
+      "loss": 3.8928,
+      "step": 34359
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3734197218710494e-06,
+      "loss": 1.1264,
+      "step": 34360
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3733723135271807e-06,
+      "loss": 1.627,
+      "step": 34361
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3733249051833124e-06,
+      "loss": 3.3936,
+      "step": 34362
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3732774968394437e-06,
+      "loss": 1.3013,
+      "step": 34363
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3732300884955752e-06,
+      "loss": 2.0266,
+      "step": 34364
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3731826801517069e-06,
+      "loss": 1.7701,
+      "step": 34365
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3731352718078382e-06,
+      "loss": 1.1794,
+      "step": 34366
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3730878634639697e-06,
+      "loss": 2.5778,
+      "step": 34367
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3730404551201011e-06,
+      "loss": 1.976,
+      "step": 34368
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3729930467762326e-06,
+      "loss": 2.2634,
+      "step": 34369
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3729456384323641e-06,
+      "loss": 2.6968,
+      "step": 34370
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3728982300884956e-06,
+      "loss": 0.7623,
+      "step": 34371
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3728508217446271e-06,
+      "loss": 3.5372,
+      "step": 34372
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3728034134007586e-06,
+      "loss": 2.1148,
+      "step": 34373
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3727560050568899e-06,
+      "loss": 1.288,
+      "step": 34374
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3727085967130216e-06,
+      "loss": 1.6326,
+      "step": 34375
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.372661188369153e-06,
+      "loss": 1.765,
+      "step": 34376
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3726137800252844e-06,
+      "loss": 3.9373,
+      "step": 34377
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.372566371681416e-06,
+      "loss": 2.7655,
+      "step": 34378
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3725189633375476e-06,
+      "loss": 3.206,
+      "step": 34379
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3724715549936788e-06,
+      "loss": 2.3085,
+      "step": 34380
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3724241466498103e-06,
+      "loss": 1.4927,
+      "step": 34381
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.372376738305942e-06,
+      "loss": 3.0831,
+      "step": 34382
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3723293299620733e-06,
+      "loss": 1.9578,
+      "step": 34383
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3722819216182048e-06,
+      "loss": 2.1665,
+      "step": 34384
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3722345132743363e-06,
+      "loss": 1.3049,
+      "step": 34385
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3721871049304678e-06,
+      "loss": 2.8348,
+      "step": 34386
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3721396965865993e-06,
+      "loss": 1.3634,
+      "step": 34387
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3720922882427308e-06,
+      "loss": 2.3888,
+      "step": 34388
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3720448798988623e-06,
+      "loss": 1.8824,
+      "step": 34389
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3719974715549938e-06,
+      "loss": 2.0537,
+      "step": 34390
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3719500632111253e-06,
+      "loss": 2.0926,
+      "step": 34391
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3719026548672567e-06,
+      "loss": 1.6547,
+      "step": 34392
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3718552465233882e-06,
+      "loss": 1.7803,
+      "step": 34393
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3718078381795195e-06,
+      "loss": 1.1102,
+      "step": 34394
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3717604298356512e-06,
+      "loss": 2.3487,
+      "step": 34395
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3717130214917825e-06,
+      "loss": 2.087,
+      "step": 34396
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.371665613147914e-06,
+      "loss": 3.0616,
+      "step": 34397
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3716182048040457e-06,
+      "loss": 2.5698,
+      "step": 34398
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.371570796460177e-06,
+      "loss": 1.9536,
+      "step": 34399
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3715233881163085e-06,
+      "loss": 1.5983,
+      "step": 34400
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.37147597977244e-06,
+      "loss": 1.1297,
+      "step": 34401
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3714285714285715e-06,
+      "loss": 4.268,
+      "step": 34402
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.371381163084703e-06,
+      "loss": 0.9332,
+      "step": 34403
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3713337547408344e-06,
+      "loss": 2.5135,
+      "step": 34404
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.371286346396966e-06,
+      "loss": 2.0005,
+      "step": 34405
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3712389380530974e-06,
+      "loss": 0.9833,
+      "step": 34406
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3711915297092287e-06,
+      "loss": 3.1079,
+      "step": 34407
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3711441213653604e-06,
+      "loss": 1.7412,
+      "step": 34408
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.371096713021492e-06,
+      "loss": 1.6958,
+      "step": 34409
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3710493046776232e-06,
+      "loss": 1.5863,
+      "step": 34410
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3710018963337547e-06,
+      "loss": 2.6215,
+      "step": 34411
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3709544879898864e-06,
+      "loss": 3.7487,
+      "step": 34412
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3709070796460177e-06,
+      "loss": 0.8402,
+      "step": 34413
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3708596713021492e-06,
+      "loss": 0.9647,
+      "step": 34414
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3708122629582809e-06,
+      "loss": 0.728,
+      "step": 34415
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3707648546144121e-06,
+      "loss": 3.0955,
+      "step": 34416
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3707174462705436e-06,
+      "loss": 2.7485,
+      "step": 34417
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3706700379266751e-06,
+      "loss": 1.8087,
+      "step": 34418
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 1.3706226295828066e-06,
+      "loss": 0.7714,
+      "step": 34419
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.370575221238938e-06,
+      "loss": 2.3462,
+      "step": 34420
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3705278128950696e-06,
+      "loss": 1.4593,
+      "step": 34421
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.370480404551201e-06,
+      "loss": 3.9521,
+      "step": 34422
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3704329962073326e-06,
+      "loss": 1.7528,
+      "step": 34423
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3703855878634639e-06,
+      "loss": 2.9075,
+      "step": 34424
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3703381795195956e-06,
+      "loss": 2.2277,
+      "step": 34425
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.370290771175727e-06,
+      "loss": 1.717,
+      "step": 34426
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3702433628318583e-06,
+      "loss": 1.7878,
+      "step": 34427
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.37019595448799e-06,
+      "loss": 2.3085,
+      "step": 34428
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3701485461441213e-06,
+      "loss": 2.9162,
+      "step": 34429
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3701011378002528e-06,
+      "loss": 1.5008,
+      "step": 34430
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3700537294563843e-06,
+      "loss": 1.4934,
+      "step": 34431
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3700063211125158e-06,
+      "loss": 1.8472,
+      "step": 34432
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3699589127686473e-06,
+      "loss": 1.7783,
+      "step": 34433
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3699115044247788e-06,
+      "loss": 2.1461,
+      "step": 34434
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3698640960809103e-06,
+      "loss": 2.1223,
+      "step": 34435
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3698166877370418e-06,
+      "loss": 2.1622,
+      "step": 34436
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3697692793931733e-06,
+      "loss": 1.7432,
+      "step": 34437
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3697218710493048e-06,
+      "loss": 2.1894,
+      "step": 34438
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3696744627054362e-06,
+      "loss": 2.9216,
+      "step": 34439
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3696270543615675e-06,
+      "loss": 2.0543,
+      "step": 34440
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3695796460176992e-06,
+      "loss": 3.3649,
+      "step": 34441
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3695322376738307e-06,
+      "loss": 1.6085,
+      "step": 34442
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.369484829329962e-06,
+      "loss": 4.1615,
+      "step": 34443
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3694374209860935e-06,
+      "loss": 3.303,
+      "step": 34444
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3693900126422252e-06,
+      "loss": 3.7374,
+      "step": 34445
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3693426042983565e-06,
+      "loss": 2.4904,
+      "step": 34446
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.369295195954488e-06,
+      "loss": 1.6988,
+      "step": 34447
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3692477876106197e-06,
+      "loss": 2.1752,
+      "step": 34448
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.369200379266751e-06,
+      "loss": 2.7605,
+      "step": 34449
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3691529709228824e-06,
+      "loss": 1.363,
+      "step": 34450
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.369105562579014e-06,
+      "loss": 1.4374,
+      "step": 34451
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3690581542351454e-06,
+      "loss": 2.2118,
+      "step": 34452
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.369010745891277e-06,
+      "loss": 2.7417,
+      "step": 34453
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3689633375474084e-06,
+      "loss": 1.5356,
+      "step": 34454
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.36891592920354e-06,
+      "loss": 3.2377,
+      "step": 34455
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3688685208596714e-06,
+      "loss": 1.3738,
+      "step": 34456
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3688211125158027e-06,
+      "loss": 2.2782,
+      "step": 34457
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3687737041719344e-06,
+      "loss": 2.2447,
+      "step": 34458
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3687262958280659e-06,
+      "loss": 2.324,
+      "step": 34459
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3686788874841972e-06,
+      "loss": 3.0776,
+      "step": 34460
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3686314791403289e-06,
+      "loss": 1.514,
+      "step": 34461
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3685840707964601e-06,
+      "loss": 1.2086,
+      "step": 34462
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3685366624525916e-06,
+      "loss": 3.3846,
+      "step": 34463
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3684892541087231e-06,
+      "loss": 2.564,
+      "step": 34464
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3684418457648546e-06,
+      "loss": 1.3688,
+      "step": 34465
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3683944374209861e-06,
+      "loss": 2.987,
+      "step": 34466
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3683470290771176e-06,
+      "loss": 2.0843,
+      "step": 34467
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.368299620733249e-06,
+      "loss": 1.8593,
+      "step": 34468
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3682522123893806e-06,
+      "loss": 0.9309,
+      "step": 34469
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.368204804045512e-06,
+      "loss": 3.1043,
+      "step": 34470
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3681573957016436e-06,
+      "loss": 2.3448,
+      "step": 34471
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.368109987357775e-06,
+      "loss": 3.1594,
+      "step": 34472
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3680625790139063e-06,
+      "loss": 2.4438,
+      "step": 34473
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.368015170670038e-06,
+      "loss": 2.6277,
+      "step": 34474
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3679677623261695e-06,
+      "loss": 1.2575,
+      "step": 34475
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3679203539823008e-06,
+      "loss": 2.6469,
+      "step": 34476
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3678729456384323e-06,
+      "loss": 1.5035,
+      "step": 34477
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.367825537294564e-06,
+      "loss": 1.3153,
+      "step": 34478
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3677781289506953e-06,
+      "loss": 1.4765,
+      "step": 34479
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3677307206068268e-06,
+      "loss": 1.4316,
+      "step": 34480
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3676833122629585e-06,
+      "loss": 1.5502,
+      "step": 34481
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3676359039190898e-06,
+      "loss": 2.3364,
+      "step": 34482
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 1.3675884955752213e-06,
+      "loss": 1.3655,
+      "step": 34483
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3675410872313528e-06,
+      "loss": 1.5771,
+      "step": 34484
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3674936788874842e-06,
+      "loss": 2.159,
+      "step": 34485
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3674462705436157e-06,
+      "loss": 1.1951,
+      "step": 34486
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3673988621997472e-06,
+      "loss": 2.8052,
+      "step": 34487
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3673514538558787e-06,
+      "loss": 2.619,
+      "step": 34488
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3673040455120102e-06,
+      "loss": 2.7723,
+      "step": 34489
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3672566371681415e-06,
+      "loss": 2.231,
+      "step": 34490
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3672092288242732e-06,
+      "loss": 2.8023,
+      "step": 34491
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3671618204804047e-06,
+      "loss": 1.8628,
+      "step": 34492
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.367114412136536e-06,
+      "loss": 2.3675,
+      "step": 34493
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3670670037926677e-06,
+      "loss": 1.3409,
+      "step": 34494
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.367019595448799e-06,
+      "loss": 2.2248,
+      "step": 34495
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3669721871049305e-06,
+      "loss": 1.95,
+      "step": 34496
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.366924778761062e-06,
+      "loss": 3.2339,
+      "step": 34497
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3668773704171934e-06,
+      "loss": 2.5956,
+      "step": 34498
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.366829962073325e-06,
+      "loss": 1.4545,
+      "step": 34499
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3667825537294564e-06,
+      "loss": 1.1623,
+      "step": 34500
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.366735145385588e-06,
+      "loss": 2.3884,
+      "step": 34501
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3666877370417194e-06,
+      "loss": 1.7664,
+      "step": 34502
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.366640328697851e-06,
+      "loss": 1.7523,
+      "step": 34503
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3665929203539824e-06,
+      "loss": 1.546,
+      "step": 34504
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3665455120101139e-06,
+      "loss": 1.5028,
+      "step": 34505
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3664981036662452e-06,
+      "loss": 2.5342,
+      "step": 34506
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3664506953223769e-06,
+      "loss": 3.333,
+      "step": 34507
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3664032869785084e-06,
+      "loss": 1.8862,
+      "step": 34508
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3663558786346396e-06,
+      "loss": 1.8485,
+      "step": 34509
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3663084702907711e-06,
+      "loss": 2.42,
+      "step": 34510
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3662610619469028e-06,
+      "loss": 2.1293,
+      "step": 34511
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3662136536030341e-06,
+      "loss": 1.0351,
+      "step": 34512
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3661662452591656e-06,
+      "loss": 2.5011,
+      "step": 34513
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3661188369152973e-06,
+      "loss": 1.2829,
+      "step": 34514
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3660714285714286e-06,
+      "loss": 2.3365,
+      "step": 34515
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.36602402022756e-06,
+      "loss": 2.0391,
+      "step": 34516
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3659766118836916e-06,
+      "loss": 1.4853,
+      "step": 34517
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.365929203539823e-06,
+      "loss": 2.8481,
+      "step": 34518
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3658817951959546e-06,
+      "loss": 2.8899,
+      "step": 34519
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.365834386852086e-06,
+      "loss": 1.914,
+      "step": 34520
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3657869785082175e-06,
+      "loss": 2.3463,
+      "step": 34521
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.365739570164349e-06,
+      "loss": 2.6115,
+      "step": 34522
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3656921618204803e-06,
+      "loss": 2.0673,
+      "step": 34523
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.365644753476612e-06,
+      "loss": 3.07,
+      "step": 34524
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3655973451327435e-06,
+      "loss": 2.2038,
+      "step": 34525
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3655499367888748e-06,
+      "loss": 2.6399,
+      "step": 34526
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3655025284450065e-06,
+      "loss": 3.011,
+      "step": 34527
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3654551201011378e-06,
+      "loss": 3.0927,
+      "step": 34528
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3654077117572693e-06,
+      "loss": 2.4838,
+      "step": 34529
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3653603034134008e-06,
+      "loss": 1.9097,
+      "step": 34530
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3653128950695323e-06,
+      "loss": 2.1794,
+      "step": 34531
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3652654867256637e-06,
+      "loss": 3.4943,
+      "step": 34532
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3652180783817952e-06,
+      "loss": 3.2745,
+      "step": 34533
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3651706700379267e-06,
+      "loss": 2.8876,
+      "step": 34534
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3651232616940582e-06,
+      "loss": 2.053,
+      "step": 34535
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3650758533501897e-06,
+      "loss": 1.5646,
+      "step": 34536
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3650284450063212e-06,
+      "loss": 1.5742,
+      "step": 34537
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3649810366624527e-06,
+      "loss": 2.3544,
+      "step": 34538
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.364933628318584e-06,
+      "loss": 2.3082,
+      "step": 34539
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3648862199747155e-06,
+      "loss": 3.146,
+      "step": 34540
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3648388116308472e-06,
+      "loss": 1.6156,
+      "step": 34541
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3647914032869785e-06,
+      "loss": 2.5121,
+      "step": 34542
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.36474399494311e-06,
+      "loss": 2.4154,
+      "step": 34543
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3646965865992417e-06,
+      "loss": 1.7547,
+      "step": 34544
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.364649178255373e-06,
+      "loss": 1.6302,
+      "step": 34545
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 1.3646017699115044e-06,
+      "loss": 1.8562,
+      "step": 34546
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.364554361567636e-06,
+      "loss": 2.3315,
+      "step": 34547
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3645069532237674e-06,
+      "loss": 1.3741,
+      "step": 34548
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.364459544879899e-06,
+      "loss": 2.295,
+      "step": 34549
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3644121365360304e-06,
+      "loss": 1.411,
+      "step": 34550
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3643647281921619e-06,
+      "loss": 1.4601,
+      "step": 34551
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3643173198482934e-06,
+      "loss": 2.3442,
+      "step": 34552
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3642699115044247e-06,
+      "loss": 3.0871,
+      "step": 34553
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3642225031605564e-06,
+      "loss": 2.398,
+      "step": 34554
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3641750948166879e-06,
+      "loss": 3.1859,
+      "step": 34555
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3641276864728191e-06,
+      "loss": 1.1847,
+      "step": 34556
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3640802781289508e-06,
+      "loss": 3.3536,
+      "step": 34557
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3640328697850821e-06,
+      "loss": 2.2065,
+      "step": 34558
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3639854614412136e-06,
+      "loss": 3.1677,
+      "step": 34559
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.363938053097345e-06,
+      "loss": 3.6971,
+      "step": 34560
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3638906447534766e-06,
+      "loss": 0.5096,
+      "step": 34561
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.363843236409608e-06,
+      "loss": 1.7305,
+      "step": 34562
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3637958280657396e-06,
+      "loss": 1.5307,
+      "step": 34563
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.363748419721871e-06,
+      "loss": 1.3441,
+      "step": 34564
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3637010113780026e-06,
+      "loss": 1.674,
+      "step": 34565
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.363653603034134e-06,
+      "loss": 1.8352,
+      "step": 34566
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3636061946902655e-06,
+      "loss": 3.092,
+      "step": 34567
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.363558786346397e-06,
+      "loss": 1.6099,
+      "step": 34568
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3635113780025283e-06,
+      "loss": 1.8056,
+      "step": 34569
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.36346396965866e-06,
+      "loss": 4.081,
+      "step": 34570
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3634165613147915e-06,
+      "loss": 3.5959,
+      "step": 34571
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3633691529709228e-06,
+      "loss": 2.0237,
+      "step": 34572
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3633217446270543e-06,
+      "loss": 2.4918,
+      "step": 34573
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.363274336283186e-06,
+      "loss": 1.1665,
+      "step": 34574
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3632269279393173e-06,
+      "loss": 1.9972,
+      "step": 34575
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3631795195954488e-06,
+      "loss": 2.4675,
+      "step": 34576
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3631321112515805e-06,
+      "loss": 2.7998,
+      "step": 34577
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3630847029077118e-06,
+      "loss": 2.2803,
+      "step": 34578
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3630372945638432e-06,
+      "loss": 2.1641,
+      "step": 34579
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3629898862199747e-06,
+      "loss": 1.1746,
+      "step": 34580
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3629424778761062e-06,
+      "loss": 2.1065,
+      "step": 34581
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3628950695322377e-06,
+      "loss": 2.2193,
+      "step": 34582
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3628476611883692e-06,
+      "loss": 1.9112,
+      "step": 34583
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3628002528445007e-06,
+      "loss": 2.953,
+      "step": 34584
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3627528445006322e-06,
+      "loss": 1.227,
+      "step": 34585
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3627054361567635e-06,
+      "loss": 1.6986,
+      "step": 34586
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3626580278128952e-06,
+      "loss": 3.4346,
+      "step": 34587
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3626106194690267e-06,
+      "loss": 1.8319,
+      "step": 34588
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.362563211125158e-06,
+      "loss": 0.8208,
+      "step": 34589
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3625158027812897e-06,
+      "loss": 1.4745,
+      "step": 34590
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.362468394437421e-06,
+      "loss": 1.7793,
+      "step": 34591
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3624209860935524e-06,
+      "loss": 1.9374,
+      "step": 34592
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.362373577749684e-06,
+      "loss": 2.9492,
+      "step": 34593
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3623261694058154e-06,
+      "loss": 1.533,
+      "step": 34594
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.362278761061947e-06,
+      "loss": 1.1566,
+      "step": 34595
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3622313527180784e-06,
+      "loss": 2.1212,
+      "step": 34596
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3621839443742099e-06,
+      "loss": 3.1645,
+      "step": 34597
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3621365360303414e-06,
+      "loss": 1.6943,
+      "step": 34598
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3620891276864729e-06,
+      "loss": 2.6944,
+      "step": 34599
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3620417193426044e-06,
+      "loss": 2.0841,
+      "step": 34600
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3619943109987359e-06,
+      "loss": 2.5968,
+      "step": 34601
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3619469026548671e-06,
+      "loss": 3.08,
+      "step": 34602
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3618994943109988e-06,
+      "loss": 1.5775,
+      "step": 34603
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3618520859671303e-06,
+      "loss": 3.6518,
+      "step": 34604
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3618046776232616e-06,
+      "loss": 1.3106,
+      "step": 34605
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3617572692793931e-06,
+      "loss": 2.742,
+      "step": 34606
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3617098609355248e-06,
+      "loss": 1.2168,
+      "step": 34607
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.361662452591656e-06,
+      "loss": 2.0414,
+      "step": 34608
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 1.3616150442477876e-06,
+      "loss": 0.9017,
+      "step": 34609
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3615676359039193e-06,
+      "loss": 0.522,
+      "step": 34610
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3615202275600506e-06,
+      "loss": 1.9649,
+      "step": 34611
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.361472819216182e-06,
+      "loss": 2.1892,
+      "step": 34612
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3614254108723136e-06,
+      "loss": 2.7974,
+      "step": 34613
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.361378002528445e-06,
+      "loss": 2.5529,
+      "step": 34614
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3613305941845765e-06,
+      "loss": 1.9254,
+      "step": 34615
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.361283185840708e-06,
+      "loss": 2.1314,
+      "step": 34616
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3612357774968395e-06,
+      "loss": 3.0067,
+      "step": 34617
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.361188369152971e-06,
+      "loss": 2.9128,
+      "step": 34618
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3611409608091023e-06,
+      "loss": 2.7007,
+      "step": 34619
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.361093552465234e-06,
+      "loss": 2.6269,
+      "step": 34620
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3610461441213655e-06,
+      "loss": 3.4079,
+      "step": 34621
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3609987357774968e-06,
+      "loss": 3.3645,
+      "step": 34622
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3609513274336285e-06,
+      "loss": 1.4166,
+      "step": 34623
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3609039190897598e-06,
+      "loss": 2.7064,
+      "step": 34624
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3608565107458912e-06,
+      "loss": 2.0713,
+      "step": 34625
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3608091024020227e-06,
+      "loss": 2.7918,
+      "step": 34626
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3607616940581542e-06,
+      "loss": 1.718,
+      "step": 34627
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3607142857142857e-06,
+      "loss": 1.4633,
+      "step": 34628
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3606668773704172e-06,
+      "loss": 3.547,
+      "step": 34629
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3606194690265487e-06,
+      "loss": 1.0525,
+      "step": 34630
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3605720606826802e-06,
+      "loss": 2.086,
+      "step": 34631
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3605246523388117e-06,
+      "loss": 1.4309,
+      "step": 34632
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3604772439949432e-06,
+      "loss": 2.0325,
+      "step": 34633
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3604298356510747e-06,
+      "loss": 2.1817,
+      "step": 34634
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.360382427307206e-06,
+      "loss": 2.1005,
+      "step": 34635
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3603350189633377e-06,
+      "loss": 3.0494,
+      "step": 34636
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3602876106194692e-06,
+      "loss": 1.9214,
+      "step": 34637
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3602402022756004e-06,
+      "loss": 3.1908,
+      "step": 34638
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.360192793931732e-06,
+      "loss": 2.4403,
+      "step": 34639
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3601453855878636e-06,
+      "loss": 1.258,
+      "step": 34640
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.360097977243995e-06,
+      "loss": 2.8818,
+      "step": 34641
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3600505689001264e-06,
+      "loss": 2.7698,
+      "step": 34642
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3600031605562581e-06,
+      "loss": 3.46,
+      "step": 34643
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3599557522123894e-06,
+      "loss": 2.3753,
+      "step": 34644
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3599083438685209e-06,
+      "loss": 0.8982,
+      "step": 34645
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3598609355246524e-06,
+      "loss": 2.2985,
+      "step": 34646
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3598135271807839e-06,
+      "loss": 1.6422,
+      "step": 34647
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3597661188369154e-06,
+      "loss": 2.1997,
+      "step": 34648
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3597187104930468e-06,
+      "loss": 2.8118,
+      "step": 34649
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3596713021491783e-06,
+      "loss": 3.1049,
+      "step": 34650
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3596238938053098e-06,
+      "loss": 2.2064,
+      "step": 34651
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3595764854614411e-06,
+      "loss": 1.7553,
+      "step": 34652
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3595290771175728e-06,
+      "loss": 2.6917,
+      "step": 34653
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3594816687737043e-06,
+      "loss": 2.2046,
+      "step": 34654
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3594342604298356e-06,
+      "loss": 2.9395,
+      "step": 34655
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3593868520859673e-06,
+      "loss": 2.2485,
+      "step": 34656
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3593394437420986e-06,
+      "loss": 2.3786,
+      "step": 34657
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.35929203539823e-06,
+      "loss": 1.8345,
+      "step": 34658
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3592446270543616e-06,
+      "loss": 2.2649,
+      "step": 34659
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.359197218710493e-06,
+      "loss": 1.6047,
+      "step": 34660
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3591498103666245e-06,
+      "loss": 2.0496,
+      "step": 34661
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.359102402022756e-06,
+      "loss": 1.452,
+      "step": 34662
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3590549936788875e-06,
+      "loss": 4.2524,
+      "step": 34663
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.359007585335019e-06,
+      "loss": 1.8679,
+      "step": 34664
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3589601769911505e-06,
+      "loss": 1.0659,
+      "step": 34665
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.358912768647282e-06,
+      "loss": 1.107,
+      "step": 34666
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3588653603034135e-06,
+      "loss": 1.3415,
+      "step": 34667
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3588179519595448e-06,
+      "loss": 2.6271,
+      "step": 34668
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3587705436156765e-06,
+      "loss": 1.3358,
+      "step": 34669
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.358723135271808e-06,
+      "loss": 3.364,
+      "step": 34670
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3586757269279393e-06,
+      "loss": 1.8695,
+      "step": 34671
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3586283185840707e-06,
+      "loss": 1.6921,
+      "step": 34672
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 1.3585809102402024e-06,
+      "loss": 1.9446,
+      "step": 34673
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3585335018963337e-06,
+      "loss": 1.4974,
+      "step": 34674
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3584860935524652e-06,
+      "loss": 2.0168,
+      "step": 34675
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.358438685208597e-06,
+      "loss": 1.1694,
+      "step": 34676
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3583912768647282e-06,
+      "loss": 1.8201,
+      "step": 34677
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3583438685208597e-06,
+      "loss": 3.5346,
+      "step": 34678
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3582964601769912e-06,
+      "loss": 1.7323,
+      "step": 34679
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3582490518331227e-06,
+      "loss": 2.0796,
+      "step": 34680
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3582016434892542e-06,
+      "loss": 1.7892,
+      "step": 34681
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3581542351453855e-06,
+      "loss": 2.258,
+      "step": 34682
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3581068268015172e-06,
+      "loss": 2.508,
+      "step": 34683
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3580594184576487e-06,
+      "loss": 2.2952,
+      "step": 34684
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.35801201011378e-06,
+      "loss": 3.4414,
+      "step": 34685
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3579646017699116e-06,
+      "loss": 1.5841,
+      "step": 34686
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3579171934260431e-06,
+      "loss": 2.7617,
+      "step": 34687
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3578697850821744e-06,
+      "loss": 2.0689,
+      "step": 34688
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.357822376738306e-06,
+      "loss": 2.1372,
+      "step": 34689
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3577749683944374e-06,
+      "loss": 1.0989,
+      "step": 34690
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3577275600505689e-06,
+      "loss": 2.6012,
+      "step": 34691
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3576801517067004e-06,
+      "loss": 2.845,
+      "step": 34692
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3576327433628319e-06,
+      "loss": 1.2731,
+      "step": 34693
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3575853350189634e-06,
+      "loss": 3.0848,
+      "step": 34694
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3575379266750949e-06,
+      "loss": 1.1841,
+      "step": 34695
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3574905183312263e-06,
+      "loss": 1.2934,
+      "step": 34696
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3574431099873578e-06,
+      "loss": 1.3295,
+      "step": 34697
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3573957016434893e-06,
+      "loss": 1.6957,
+      "step": 34698
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3573482932996208e-06,
+      "loss": 2.5669,
+      "step": 34699
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3573008849557523e-06,
+      "loss": 3.4501,
+      "step": 34700
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3572534766118836e-06,
+      "loss": 1.7589,
+      "step": 34701
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.357206068268015e-06,
+      "loss": 2.8066,
+      "step": 34702
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3571586599241468e-06,
+      "loss": 3.5734,
+      "step": 34703
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.357111251580278e-06,
+      "loss": 2.3896,
+      "step": 34704
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3570638432364096e-06,
+      "loss": 1.9755,
+      "step": 34705
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3570164348925413e-06,
+      "loss": 1.5825,
+      "step": 34706
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3569690265486725e-06,
+      "loss": 1.085,
+      "step": 34707
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.356921618204804e-06,
+      "loss": 1.6742,
+      "step": 34708
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3568742098609355e-06,
+      "loss": 1.0388,
+      "step": 34709
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.356826801517067e-06,
+      "loss": 1.4926,
+      "step": 34710
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3567793931731985e-06,
+      "loss": 2.4145,
+      "step": 34711
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.35673198482933e-06,
+      "loss": 3.9242,
+      "step": 34712
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3566845764854615e-06,
+      "loss": 2.4784,
+      "step": 34713
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.356637168141593e-06,
+      "loss": 0.9776,
+      "step": 34714
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3565897597977243e-06,
+      "loss": 2.5352,
+      "step": 34715
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.356542351453856e-06,
+      "loss": 1.2247,
+      "step": 34716
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3564949431099875e-06,
+      "loss": 1.3772,
+      "step": 34717
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3564475347661187e-06,
+      "loss": 2.4704,
+      "step": 34718
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3564001264222505e-06,
+      "loss": 1.9587,
+      "step": 34719
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.356352718078382e-06,
+      "loss": 2.5625,
+      "step": 34720
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3563053097345132e-06,
+      "loss": 2.2193,
+      "step": 34721
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3562579013906447e-06,
+      "loss": 0.7918,
+      "step": 34722
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3562104930467762e-06,
+      "loss": 2.7501,
+      "step": 34723
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3561630847029077e-06,
+      "loss": 2.8771,
+      "step": 34724
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3561156763590392e-06,
+      "loss": 2.631,
+      "step": 34725
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3560682680151707e-06,
+      "loss": 3.9346,
+      "step": 34726
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3560208596713022e-06,
+      "loss": 1.7255,
+      "step": 34727
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3559734513274337e-06,
+      "loss": 3.4504,
+      "step": 34728
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3559260429835652e-06,
+      "loss": 2.7023,
+      "step": 34729
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3558786346396967e-06,
+      "loss": 2.3854,
+      "step": 34730
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3558312262958281e-06,
+      "loss": 2.5668,
+      "step": 34731
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3557838179519596e-06,
+      "loss": 2.5636,
+      "step": 34732
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3557364096080911e-06,
+      "loss": 2.9658,
+      "step": 34733
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3556890012642224e-06,
+      "loss": 2.3435,
+      "step": 34734
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.355641592920354e-06,
+      "loss": 2.7756,
+      "step": 34735
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 1.3555941845764856e-06,
+      "loss": 2.9973,
+      "step": 34736
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3555467762326169e-06,
+      "loss": 2.0794,
+      "step": 34737
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3554993678887484e-06,
+      "loss": 1.6498,
+      "step": 34738
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.35545195954488e-06,
+      "loss": 2.4872,
+      "step": 34739
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3554045512010114e-06,
+      "loss": 2.6868,
+      "step": 34740
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3553571428571429e-06,
+      "loss": 2.7252,
+      "step": 34741
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3553097345132744e-06,
+      "loss": 1.9968,
+      "step": 34742
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3552623261694058e-06,
+      "loss": 1.2266,
+      "step": 34743
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3552149178255373e-06,
+      "loss": 2.2504,
+      "step": 34744
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3551675094816688e-06,
+      "loss": 1.6335,
+      "step": 34745
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3551201011378003e-06,
+      "loss": 1.1106,
+      "step": 34746
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3550726927939318e-06,
+      "loss": 1.6624,
+      "step": 34747
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.355025284450063e-06,
+      "loss": 1.7415,
+      "step": 34748
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3549778761061948e-06,
+      "loss": 2.6299,
+      "step": 34749
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3549304677623263e-06,
+      "loss": 0.649,
+      "step": 34750
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3548830594184576e-06,
+      "loss": 3.0736,
+      "step": 34751
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3548356510745893e-06,
+      "loss": 1.2962,
+      "step": 34752
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3547882427307208e-06,
+      "loss": 2.0438,
+      "step": 34753
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.354740834386852e-06,
+      "loss": 2.7771,
+      "step": 34754
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3546934260429835e-06,
+      "loss": 2.649,
+      "step": 34755
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.354646017699115e-06,
+      "loss": 1.9131,
+      "step": 34756
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3545986093552465e-06,
+      "loss": 2.123,
+      "step": 34757
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.354551201011378e-06,
+      "loss": 2.8309,
+      "step": 34758
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3545037926675095e-06,
+      "loss": 2.2408,
+      "step": 34759
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.354456384323641e-06,
+      "loss": 2.3205,
+      "step": 34760
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3544089759797725e-06,
+      "loss": 2.0745,
+      "step": 34761
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.354361567635904e-06,
+      "loss": 1.9638,
+      "step": 34762
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3543141592920355e-06,
+      "loss": 1.985,
+      "step": 34763
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.354266750948167e-06,
+      "loss": 2.28,
+      "step": 34764
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3542193426042985e-06,
+      "loss": 1.3002,
+      "step": 34765
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.35417193426043e-06,
+      "loss": 1.7681,
+      "step": 34766
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3541245259165612e-06,
+      "loss": 1.6918,
+      "step": 34767
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3540771175726927e-06,
+      "loss": 0.692,
+      "step": 34768
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3540297092288244e-06,
+      "loss": 2.9151,
+      "step": 34769
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3539823008849557e-06,
+      "loss": 1.5766,
+      "step": 34770
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3539348925410872e-06,
+      "loss": 0.9778,
+      "step": 34771
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.353887484197219e-06,
+      "loss": 2.5097,
+      "step": 34772
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3538400758533502e-06,
+      "loss": 1.6432,
+      "step": 34773
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3537926675094817e-06,
+      "loss": 2.1892,
+      "step": 34774
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3537452591656132e-06,
+      "loss": 1.3744,
+      "step": 34775
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3536978508217447e-06,
+      "loss": 3.5224,
+      "step": 34776
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3536504424778762e-06,
+      "loss": 1.798,
+      "step": 34777
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3536030341340076e-06,
+      "loss": 2.0006,
+      "step": 34778
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3535556257901391e-06,
+      "loss": 4.2323,
+      "step": 34779
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3535082174462706e-06,
+      "loss": 1.9811,
+      "step": 34780
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.353460809102402e-06,
+      "loss": 2.1345,
+      "step": 34781
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3534134007585336e-06,
+      "loss": 1.8361,
+      "step": 34782
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3533659924146651e-06,
+      "loss": 3.0776,
+      "step": 34783
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3533185840707964e-06,
+      "loss": 4.3569,
+      "step": 34784
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.353271175726928e-06,
+      "loss": 2.4446,
+      "step": 34785
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3532237673830596e-06,
+      "loss": 4.5145,
+      "step": 34786
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3531763590391909e-06,
+      "loss": 2.7578,
+      "step": 34787
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3531289506953224e-06,
+      "loss": 2.5029,
+      "step": 34788
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3530815423514538e-06,
+      "loss": 2.644,
+      "step": 34789
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3530341340075853e-06,
+      "loss": 2.4062,
+      "step": 34790
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3529867256637168e-06,
+      "loss": 1.1278,
+      "step": 34791
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3529393173198483e-06,
+      "loss": 1.793,
+      "step": 34792
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3528919089759798e-06,
+      "loss": 3.7056,
+      "step": 34793
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3528445006321113e-06,
+      "loss": 2.48,
+      "step": 34794
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3527970922882428e-06,
+      "loss": 1.9166,
+      "step": 34795
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3527496839443743e-06,
+      "loss": 1.9332,
+      "step": 34796
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3527022756005058e-06,
+      "loss": 1.4104,
+      "step": 34797
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3526548672566373e-06,
+      "loss": 2.25,
+      "step": 34798
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 1.3526074589127688e-06,
+      "loss": 1.9115,
+      "step": 34799
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3525600505689e-06,
+      "loss": 1.7871,
+      "step": 34800
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3525126422250315e-06,
+      "loss": 1.0897,
+      "step": 34801
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3524652338811632e-06,
+      "loss": 1.3648,
+      "step": 34802
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3524178255372945e-06,
+      "loss": 1.6873,
+      "step": 34803
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.352370417193426e-06,
+      "loss": 1.1824,
+      "step": 34804
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3523230088495577e-06,
+      "loss": 2.3526,
+      "step": 34805
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.352275600505689e-06,
+      "loss": 2.559,
+      "step": 34806
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3522281921618205e-06,
+      "loss": 2.5631,
+      "step": 34807
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.352180783817952e-06,
+      "loss": 2.4334,
+      "step": 34808
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3521333754740835e-06,
+      "loss": 1.8198,
+      "step": 34809
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.352085967130215e-06,
+      "loss": 2.6623,
+      "step": 34810
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3520385587863463e-06,
+      "loss": 2.9977,
+      "step": 34811
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.351991150442478e-06,
+      "loss": 1.7417,
+      "step": 34812
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3519437420986094e-06,
+      "loss": 2.2572,
+      "step": 34813
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3518963337547407e-06,
+      "loss": 2.0477,
+      "step": 34814
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3518489254108724e-06,
+      "loss": 1.7079,
+      "step": 34815
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.351801517067004e-06,
+      "loss": 1.0322,
+      "step": 34816
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3517541087231352e-06,
+      "loss": 3.1234,
+      "step": 34817
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3517067003792667e-06,
+      "loss": 2.1402,
+      "step": 34818
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3516592920353984e-06,
+      "loss": 1.9753,
+      "step": 34819
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3516118836915297e-06,
+      "loss": 1.8125,
+      "step": 34820
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3515644753476612e-06,
+      "loss": 1.8915,
+      "step": 34821
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3515170670037927e-06,
+      "loss": 3.2629,
+      "step": 34822
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3514696586599242e-06,
+      "loss": 1.4956,
+      "step": 34823
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3514222503160557e-06,
+      "loss": 3.3922,
+      "step": 34824
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3513748419721871e-06,
+      "loss": 2.3694,
+      "step": 34825
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3513274336283186e-06,
+      "loss": 2.1189,
+      "step": 34826
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3512800252844501e-06,
+      "loss": 2.4775,
+      "step": 34827
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3512326169405816e-06,
+      "loss": 2.5751,
+      "step": 34828
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3511852085967131e-06,
+      "loss": 2.7714,
+      "step": 34829
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3511378002528446e-06,
+      "loss": 2.8517,
+      "step": 34830
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3510903919089759e-06,
+      "loss": 1.9678,
+      "step": 34831
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3510429835651076e-06,
+      "loss": 2.3754,
+      "step": 34832
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3509955752212389e-06,
+      "loss": 2.9392,
+      "step": 34833
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3509481668773704e-06,
+      "loss": 0.3457,
+      "step": 34834
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.350900758533502e-06,
+      "loss": 1.5665,
+      "step": 34835
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3508533501896333e-06,
+      "loss": 0.9075,
+      "step": 34836
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3508059418457648e-06,
+      "loss": 1.4039,
+      "step": 34837
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3507585335018963e-06,
+      "loss": 2.3355,
+      "step": 34838
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3507111251580278e-06,
+      "loss": 1.7987,
+      "step": 34839
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3506637168141593e-06,
+      "loss": 2.526,
+      "step": 34840
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3506163084702908e-06,
+      "loss": 1.9768,
+      "step": 34841
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3505689001264223e-06,
+      "loss": 1.8988,
+      "step": 34842
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3505214917825538e-06,
+      "loss": 1.287,
+      "step": 34843
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.350474083438685e-06,
+      "loss": 3.028,
+      "step": 34844
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3504266750948168e-06,
+      "loss": 2.5008,
+      "step": 34845
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3503792667509483e-06,
+      "loss": 2.3166,
+      "step": 34846
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3503318584070795e-06,
+      "loss": 1.1943,
+      "step": 34847
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3502844500632113e-06,
+      "loss": 2.7899,
+      "step": 34848
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3502370417193427e-06,
+      "loss": 1.5234,
+      "step": 34849
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.350189633375474e-06,
+      "loss": 1.3897,
+      "step": 34850
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3501422250316055e-06,
+      "loss": 1.2675,
+      "step": 34851
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3500948166877372e-06,
+      "loss": 4.0834,
+      "step": 34852
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3500474083438685e-06,
+      "loss": 1.5405,
+      "step": 34853
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.35e-06,
+      "loss": 3.0385,
+      "step": 34854
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3499525916561315e-06,
+      "loss": 2.1621,
+      "step": 34855
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.349905183312263e-06,
+      "loss": 1.7166,
+      "step": 34856
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3498577749683945e-06,
+      "loss": 3.2718,
+      "step": 34857
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.349810366624526e-06,
+      "loss": 1.2713,
+      "step": 34858
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3497629582806575e-06,
+      "loss": 2.4764,
+      "step": 34859
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.349715549936789e-06,
+      "loss": 3.1328,
+      "step": 34860
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3496681415929204e-06,
+      "loss": 0.7149,
+      "step": 34861
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.349620733249052e-06,
+      "loss": 2.3683,
+      "step": 34862
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 1.3495733249051834e-06,
+      "loss": 2.748,
+      "step": 34863
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3495259165613147e-06,
+      "loss": 1.9146,
+      "step": 34864
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3494785082174464e-06,
+      "loss": 2.0119,
+      "step": 34865
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3494310998735777e-06,
+      "loss": 3.4703,
+      "step": 34866
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3493836915297092e-06,
+      "loss": 1.6555,
+      "step": 34867
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3493362831858409e-06,
+      "loss": 2.3711,
+      "step": 34868
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3492888748419722e-06,
+      "loss": 2.7896,
+      "step": 34869
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3492414664981037e-06,
+      "loss": 2.0618,
+      "step": 34870
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3491940581542351e-06,
+      "loss": 3.4517,
+      "step": 34871
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3491466498103666e-06,
+      "loss": 2.0999,
+      "step": 34872
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3490992414664981e-06,
+      "loss": 3.3287,
+      "step": 34873
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3490518331226296e-06,
+      "loss": 1.7408,
+      "step": 34874
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3490044247787611e-06,
+      "loss": 2.163,
+      "step": 34875
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3489570164348926e-06,
+      "loss": 2.9907,
+      "step": 34876
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3489096080910239e-06,
+      "loss": 1.9515,
+      "step": 34877
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3488621997471556e-06,
+      "loss": 2.372,
+      "step": 34878
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.348814791403287e-06,
+      "loss": 3.4102,
+      "step": 34879
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3487673830594184e-06,
+      "loss": 2.0245,
+      "step": 34880
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.34871997471555e-06,
+      "loss": 1.3357,
+      "step": 34881
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3486725663716816e-06,
+      "loss": 1.827,
+      "step": 34882
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3486251580278128e-06,
+      "loss": 1.7182,
+      "step": 34883
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3485777496839443e-06,
+      "loss": 1.7982,
+      "step": 34884
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.348530341340076e-06,
+      "loss": 1.4585,
+      "step": 34885
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3484829329962073e-06,
+      "loss": 1.2131,
+      "step": 34886
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3484355246523388e-06,
+      "loss": 3.074,
+      "step": 34887
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3483881163084703e-06,
+      "loss": 1.7224,
+      "step": 34888
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3483407079646018e-06,
+      "loss": 2.642,
+      "step": 34889
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3482932996207333e-06,
+      "loss": 0.9561,
+      "step": 34890
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3482458912768648e-06,
+      "loss": 1.4498,
+      "step": 34891
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3481984829329963e-06,
+      "loss": 1.2444,
+      "step": 34892
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3481510745891278e-06,
+      "loss": 1.6597,
+      "step": 34893
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3481036662452593e-06,
+      "loss": 1.4949,
+      "step": 34894
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3480562579013907e-06,
+      "loss": 2.8635,
+      "step": 34895
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3480088495575222e-06,
+      "loss": 2.6188,
+      "step": 34896
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3479614412136535e-06,
+      "loss": 1.4852,
+      "step": 34897
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3479140328697852e-06,
+      "loss": 2.248,
+      "step": 34898
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3478666245259165e-06,
+      "loss": 1.2745,
+      "step": 34899
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.347819216182048e-06,
+      "loss": 1.7683,
+      "step": 34900
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3477718078381797e-06,
+      "loss": 2.4649,
+      "step": 34901
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.347724399494311e-06,
+      "loss": 1.7233,
+      "step": 34902
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3476769911504425e-06,
+      "loss": 2.2099,
+      "step": 34903
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.347629582806574e-06,
+      "loss": 0.3884,
+      "step": 34904
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3475821744627055e-06,
+      "loss": 3.3385,
+      "step": 34905
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.347534766118837e-06,
+      "loss": 1.9314,
+      "step": 34906
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3474873577749684e-06,
+      "loss": 1.0262,
+      "step": 34907
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3474399494311e-06,
+      "loss": 1.9665,
+      "step": 34908
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3473925410872314e-06,
+      "loss": 2.0433,
+      "step": 34909
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3473451327433627e-06,
+      "loss": 1.3621,
+      "step": 34910
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3472977243994944e-06,
+      "loss": 2.7641,
+      "step": 34911
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.347250316055626e-06,
+      "loss": 2.6893,
+      "step": 34912
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3472029077117572e-06,
+      "loss": 2.8894,
+      "step": 34913
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3471554993678889e-06,
+      "loss": 2.5317,
+      "step": 34914
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3471080910240204e-06,
+      "loss": 2.6091,
+      "step": 34915
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3470606826801517e-06,
+      "loss": 1.0737,
+      "step": 34916
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3470132743362832e-06,
+      "loss": 2.1503,
+      "step": 34917
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3469658659924149e-06,
+      "loss": 2.4487,
+      "step": 34918
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3469184576485461e-06,
+      "loss": 1.2879,
+      "step": 34919
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3468710493046776e-06,
+      "loss": 1.5213,
+      "step": 34920
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3468236409608091e-06,
+      "loss": 3.3052,
+      "step": 34921
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3467762326169406e-06,
+      "loss": 3.1463,
+      "step": 34922
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.346728824273072e-06,
+      "loss": 2.2899,
+      "step": 34923
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3466814159292036e-06,
+      "loss": 2.4839,
+      "step": 34924
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.346634007585335e-06,
+      "loss": 0.6662,
+      "step": 34925
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 1.3465865992414666e-06,
+      "loss": 1.7804,
+      "step": 34926
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.346539190897598e-06,
+      "loss": 2.8956,
+      "step": 34927
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3464917825537296e-06,
+      "loss": 2.0287,
+      "step": 34928
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.346444374209861e-06,
+      "loss": 2.3993,
+      "step": 34929
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3463969658659923e-06,
+      "loss": 2.3555,
+      "step": 34930
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.346349557522124e-06,
+      "loss": 1.1273,
+      "step": 34931
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3463021491782553e-06,
+      "loss": 2.4039,
+      "step": 34932
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3462547408343868e-06,
+      "loss": 1.0773,
+      "step": 34933
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3462073324905185e-06,
+      "loss": 1.8015,
+      "step": 34934
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3461599241466498e-06,
+      "loss": 2.1893,
+      "step": 34935
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3461125158027813e-06,
+      "loss": 1.3956,
+      "step": 34936
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3460651074589128e-06,
+      "loss": 1.0063,
+      "step": 34937
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3460176991150443e-06,
+      "loss": 2.2581,
+      "step": 34938
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3459702907711758e-06,
+      "loss": 2.8783,
+      "step": 34939
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3459228824273073e-06,
+      "loss": 1.4754,
+      "step": 34940
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3458754740834388e-06,
+      "loss": 1.6307,
+      "step": 34941
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3458280657395702e-06,
+      "loss": 2.2972,
+      "step": 34942
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3457806573957015e-06,
+      "loss": 2.224,
+      "step": 34943
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3457332490518332e-06,
+      "loss": 3.0941,
+      "step": 34944
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3456858407079647e-06,
+      "loss": 1.8823,
+      "step": 34945
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.345638432364096e-06,
+      "loss": 2.7727,
+      "step": 34946
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3455910240202277e-06,
+      "loss": 1.1035,
+      "step": 34947
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3455436156763592e-06,
+      "loss": 3.6192,
+      "step": 34948
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3454962073324905e-06,
+      "loss": 0.9606,
+      "step": 34949
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.345448798988622e-06,
+      "loss": 1.7581,
+      "step": 34950
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3454013906447537e-06,
+      "loss": 1.902,
+      "step": 34951
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.345353982300885e-06,
+      "loss": 3.0058,
+      "step": 34952
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3453065739570164e-06,
+      "loss": 1.9611,
+      "step": 34953
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.345259165613148e-06,
+      "loss": 1.9082,
+      "step": 34954
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3452117572692794e-06,
+      "loss": 1.5171,
+      "step": 34955
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.345164348925411e-06,
+      "loss": 1.3203,
+      "step": 34956
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3451169405815424e-06,
+      "loss": 1.3067,
+      "step": 34957
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.345069532237674e-06,
+      "loss": 1.193,
+      "step": 34958
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3450221238938054e-06,
+      "loss": 1.6177,
+      "step": 34959
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3449747155499367e-06,
+      "loss": 0.7416,
+      "step": 34960
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3449273072060684e-06,
+      "loss": 1.4933,
+      "step": 34961
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3448798988621999e-06,
+      "loss": 1.7735,
+      "step": 34962
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3448324905183312e-06,
+      "loss": 2.0331,
+      "step": 34963
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3447850821744629e-06,
+      "loss": 0.8936,
+      "step": 34964
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3447376738305941e-06,
+      "loss": 0.9977,
+      "step": 34965
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3446902654867256e-06,
+      "loss": 2.2942,
+      "step": 34966
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3446428571428571e-06,
+      "loss": 1.1849,
+      "step": 34967
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3445954487989886e-06,
+      "loss": 3.5899,
+      "step": 34968
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3445480404551201e-06,
+      "loss": 2.5054,
+      "step": 34969
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3445006321112516e-06,
+      "loss": 1.5784,
+      "step": 34970
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.344453223767383e-06,
+      "loss": 3.0362,
+      "step": 34971
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3444058154235146e-06,
+      "loss": 2.2121,
+      "step": 34972
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.344358407079646e-06,
+      "loss": 2.5135,
+      "step": 34973
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3443109987357776e-06,
+      "loss": 1.4882,
+      "step": 34974
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.344263590391909e-06,
+      "loss": 2.5654,
+      "step": 34975
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3442161820480403e-06,
+      "loss": 1.0111,
+      "step": 34976
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.344168773704172e-06,
+      "loss": 1.6194,
+      "step": 34977
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3441213653603035e-06,
+      "loss": 1.5748,
+      "step": 34978
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3440739570164348e-06,
+      "loss": 1.3342,
+      "step": 34979
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3440265486725663e-06,
+      "loss": 1.9158,
+      "step": 34980
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.343979140328698e-06,
+      "loss": 2.7184,
+      "step": 34981
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3439317319848293e-06,
+      "loss": 1.8255,
+      "step": 34982
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3438843236409608e-06,
+      "loss": 2.7183,
+      "step": 34983
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3438369152970925e-06,
+      "loss": 1.8865,
+      "step": 34984
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3437895069532238e-06,
+      "loss": 2.4562,
+      "step": 34985
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3437420986093553e-06,
+      "loss": 2.0576,
+      "step": 34986
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3436946902654868e-06,
+      "loss": 2.4019,
+      "step": 34987
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3436472819216182e-06,
+      "loss": 2.5618,
+      "step": 34988
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 1.3435998735777497e-06,
+      "loss": 2.5534,
+      "step": 34989
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3435524652338812e-06,
+      "loss": 2.8153,
+      "step": 34990
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3435050568900127e-06,
+      "loss": 2.8641,
+      "step": 34991
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3434576485461442e-06,
+      "loss": 1.9666,
+      "step": 34992
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3434102402022755e-06,
+      "loss": 2.2065,
+      "step": 34993
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3433628318584072e-06,
+      "loss": 2.6066,
+      "step": 34994
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3433154235145385e-06,
+      "loss": 2.4049,
+      "step": 34995
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.34326801517067e-06,
+      "loss": 3.0873,
+      "step": 34996
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3432206068268017e-06,
+      "loss": 2.5912,
+      "step": 34997
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.343173198482933e-06,
+      "loss": 1.9637,
+      "step": 34998
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3431257901390645e-06,
+      "loss": 1.5996,
+      "step": 34999
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.343078381795196e-06,
+      "loss": 2.0222,
+      "step": 35000
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3430309734513274e-06,
+      "loss": 1.3933,
+      "step": 35001
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.342983565107459e-06,
+      "loss": 1.694,
+      "step": 35002
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3429361567635904e-06,
+      "loss": 3.3,
+      "step": 35003
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.342888748419722e-06,
+      "loss": 1.1127,
+      "step": 35004
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3428413400758534e-06,
+      "loss": 0.9481,
+      "step": 35005
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3427939317319847e-06,
+      "loss": 2.3004,
+      "step": 35006
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3427465233881164e-06,
+      "loss": 3.4066,
+      "step": 35007
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3426991150442479e-06,
+      "loss": 1.6997,
+      "step": 35008
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3426517067003792e-06,
+      "loss": 3.4114,
+      "step": 35009
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3426042983565109e-06,
+      "loss": 1.2766,
+      "step": 35010
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3425568900126424e-06,
+      "loss": 1.2418,
+      "step": 35011
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3425094816687736e-06,
+      "loss": 2.396,
+      "step": 35012
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3424620733249051e-06,
+      "loss": 2.3835,
+      "step": 35013
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3424146649810368e-06,
+      "loss": 1.4068,
+      "step": 35014
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3423672566371681e-06,
+      "loss": 1.4543,
+      "step": 35015
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3423198482932996e-06,
+      "loss": 1.5788,
+      "step": 35016
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.342272439949431e-06,
+      "loss": 3.7775,
+      "step": 35017
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3422250316055626e-06,
+      "loss": 1.3574,
+      "step": 35018
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.342177623261694e-06,
+      "loss": 2.1847,
+      "step": 35019
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3421302149178256e-06,
+      "loss": 1.9386,
+      "step": 35020
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.342082806573957e-06,
+      "loss": 2.3245,
+      "step": 35021
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3420353982300886e-06,
+      "loss": 3.2096,
+      "step": 35022
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.34198798988622e-06,
+      "loss": 3.4888,
+      "step": 35023
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3419405815423515e-06,
+      "loss": 2.6297,
+      "step": 35024
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.341893173198483e-06,
+      "loss": 4.1331,
+      "step": 35025
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3418457648546143e-06,
+      "loss": 2.6162,
+      "step": 35026
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.341798356510746e-06,
+      "loss": 2.6591,
+      "step": 35027
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3417509481668773e-06,
+      "loss": 1.8487,
+      "step": 35028
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3417035398230088e-06,
+      "loss": 1.5947,
+      "step": 35029
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3416561314791405e-06,
+      "loss": 2.0425,
+      "step": 35030
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3416087231352718e-06,
+      "loss": 2.0543,
+      "step": 35031
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3415613147914033e-06,
+      "loss": 1.3233,
+      "step": 35032
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3415139064475348e-06,
+      "loss": 1.618,
+      "step": 35033
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3414664981036663e-06,
+      "loss": 2.464,
+      "step": 35034
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3414190897597977e-06,
+      "loss": 2.3585,
+      "step": 35035
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3413716814159292e-06,
+      "loss": 2.5045,
+      "step": 35036
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3413242730720607e-06,
+      "loss": 2.1057,
+      "step": 35037
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3412768647281922e-06,
+      "loss": 2.6478,
+      "step": 35038
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3412294563843235e-06,
+      "loss": 1.6107,
+      "step": 35039
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3411820480404552e-06,
+      "loss": 2.4499,
+      "step": 35040
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3411346396965867e-06,
+      "loss": 3.9699,
+      "step": 35041
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.341087231352718e-06,
+      "loss": 1.7713,
+      "step": 35042
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3410398230088497e-06,
+      "loss": 2.0212,
+      "step": 35043
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3409924146649812e-06,
+      "loss": 2.155,
+      "step": 35044
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3409450063211125e-06,
+      "loss": 2.701,
+      "step": 35045
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.340897597977244e-06,
+      "loss": 2.6536,
+      "step": 35046
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3408501896333757e-06,
+      "loss": 2.4921,
+      "step": 35047
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.340802781289507e-06,
+      "loss": 3.0175,
+      "step": 35048
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3407553729456384e-06,
+      "loss": 3.6613,
+      "step": 35049
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.34070796460177e-06,
+      "loss": 2.1096,
+      "step": 35050
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3406605562579014e-06,
+      "loss": 2.0967,
+      "step": 35051
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.340613147914033e-06,
+      "loss": 0.996,
+      "step": 35052
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 1.3405657395701644e-06,
+      "loss": 2.6342,
+      "step": 35053
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3405183312262959e-06,
+      "loss": 3.0016,
+      "step": 35054
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3404709228824274e-06,
+      "loss": 2.9003,
+      "step": 35055
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3404235145385589e-06,
+      "loss": 1.6991,
+      "step": 35056
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3403761061946904e-06,
+      "loss": 2.7164,
+      "step": 35057
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3403286978508219e-06,
+      "loss": 2.3378,
+      "step": 35058
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3402812895069531e-06,
+      "loss": 3.5551,
+      "step": 35059
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3402338811630848e-06,
+      "loss": 1.8791,
+      "step": 35060
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3401864728192161e-06,
+      "loss": 2.5179,
+      "step": 35061
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3401390644753476e-06,
+      "loss": 1.1964,
+      "step": 35062
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3400916561314793e-06,
+      "loss": 2.7942,
+      "step": 35063
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3400442477876106e-06,
+      "loss": 3.2459,
+      "step": 35064
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.339996839443742e-06,
+      "loss": 1.5264,
+      "step": 35065
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3399494310998736e-06,
+      "loss": 2.4516,
+      "step": 35066
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.339902022756005e-06,
+      "loss": 3.1197,
+      "step": 35067
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3398546144121366e-06,
+      "loss": 2.1103,
+      "step": 35068
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.339807206068268e-06,
+      "loss": 1.76,
+      "step": 35069
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3397597977243995e-06,
+      "loss": 2.2224,
+      "step": 35070
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.339712389380531e-06,
+      "loss": 2.3149,
+      "step": 35071
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3396649810366623e-06,
+      "loss": 3.1685,
+      "step": 35072
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.339617572692794e-06,
+      "loss": 1.7448,
+      "step": 35073
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3395701643489255e-06,
+      "loss": 1.261,
+      "step": 35074
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3395227560050568e-06,
+      "loss": 1.6201,
+      "step": 35075
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3394753476611885e-06,
+      "loss": 1.7401,
+      "step": 35076
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.33942793931732e-06,
+      "loss": 2.1533,
+      "step": 35077
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3393805309734513e-06,
+      "loss": 1.3624,
+      "step": 35078
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3393331226295828e-06,
+      "loss": 2.4025,
+      "step": 35079
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3392857142857145e-06,
+      "loss": 3.2738,
+      "step": 35080
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3392383059418458e-06,
+      "loss": 3.32,
+      "step": 35081
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3391908975979772e-06,
+      "loss": 0.5819,
+      "step": 35082
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3391434892541087e-06,
+      "loss": 1.5181,
+      "step": 35083
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3390960809102402e-06,
+      "loss": 1.3009,
+      "step": 35084
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3390486725663717e-06,
+      "loss": 3.5384,
+      "step": 35085
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3390012642225032e-06,
+      "loss": 1.7431,
+      "step": 35086
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3389538558786347e-06,
+      "loss": 1.0737,
+      "step": 35087
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3389064475347662e-06,
+      "loss": 2.8298,
+      "step": 35088
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3388590391908975e-06,
+      "loss": 1.3308,
+      "step": 35089
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3388116308470292e-06,
+      "loss": 2.5143,
+      "step": 35090
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3387642225031607e-06,
+      "loss": 2.3375,
+      "step": 35091
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.338716814159292e-06,
+      "loss": 2.8174,
+      "step": 35092
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3386694058154237e-06,
+      "loss": 3.1605,
+      "step": 35093
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.338621997471555e-06,
+      "loss": 1.8111,
+      "step": 35094
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3385745891276864e-06,
+      "loss": 1.6972,
+      "step": 35095
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.338527180783818e-06,
+      "loss": 1.4516,
+      "step": 35096
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3384797724399494e-06,
+      "loss": 1.3898,
+      "step": 35097
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.338432364096081e-06,
+      "loss": 2.9913,
+      "step": 35098
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3383849557522124e-06,
+      "loss": 2.0342,
+      "step": 35099
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3383375474083439e-06,
+      "loss": 3.5537,
+      "step": 35100
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3382901390644754e-06,
+      "loss": 1.6563,
+      "step": 35101
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3382427307206069e-06,
+      "loss": 2.9931,
+      "step": 35102
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3381953223767384e-06,
+      "loss": 2.5589,
+      "step": 35103
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3381479140328699e-06,
+      "loss": 1.9983,
+      "step": 35104
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3381005056890011e-06,
+      "loss": 0.9968,
+      "step": 35105
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3380530973451328e-06,
+      "loss": 2.3766,
+      "step": 35106
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3380056890012643e-06,
+      "loss": 2.3286,
+      "step": 35107
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3379582806573956e-06,
+      "loss": 2.5656,
+      "step": 35108
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3379108723135271e-06,
+      "loss": 1.3403,
+      "step": 35109
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3378634639696588e-06,
+      "loss": 2.4497,
+      "step": 35110
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.33781605562579e-06,
+      "loss": 2.462,
+      "step": 35111
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3377686472819216e-06,
+      "loss": 2.7648,
+      "step": 35112
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3377212389380533e-06,
+      "loss": 3.2881,
+      "step": 35113
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3376738305941846e-06,
+      "loss": 1.9487,
+      "step": 35114
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.337626422250316e-06,
+      "loss": 1.5508,
+      "step": 35115
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 1.3375790139064476e-06,
+      "loss": 2.5527,
+      "step": 35116
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.337531605562579e-06,
+      "loss": 2.679,
+      "step": 35117
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3374841972187105e-06,
+      "loss": 4.2548,
+      "step": 35118
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.337436788874842e-06,
+      "loss": 2.74,
+      "step": 35119
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3373893805309735e-06,
+      "loss": 4.0853,
+      "step": 35120
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.337341972187105e-06,
+      "loss": 1.8978,
+      "step": 35121
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3372945638432363e-06,
+      "loss": 2.5496,
+      "step": 35122
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.337247155499368e-06,
+      "loss": 1.7795,
+      "step": 35123
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3371997471554995e-06,
+      "loss": 2.8748,
+      "step": 35124
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3371523388116308e-06,
+      "loss": 2.1341,
+      "step": 35125
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3371049304677625e-06,
+      "loss": 1.6144,
+      "step": 35126
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3370575221238938e-06,
+      "loss": 2.4696,
+      "step": 35127
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3370101137800252e-06,
+      "loss": 1.4919,
+      "step": 35128
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3369627054361567e-06,
+      "loss": 2.7165,
+      "step": 35129
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3369152970922882e-06,
+      "loss": 2.1277,
+      "step": 35130
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3368678887484197e-06,
+      "loss": 1.8328,
+      "step": 35131
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3368204804045512e-06,
+      "loss": 1.3834,
+      "step": 35132
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3367730720606827e-06,
+      "loss": 0.9252,
+      "step": 35133
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3367256637168142e-06,
+      "loss": 1.1814,
+      "step": 35134
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3366782553729457e-06,
+      "loss": 2.482,
+      "step": 35135
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3366308470290772e-06,
+      "loss": 2.3637,
+      "step": 35136
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3365834386852087e-06,
+      "loss": 2.0048,
+      "step": 35137
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.33653603034134e-06,
+      "loss": 3.1603,
+      "step": 35138
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3364886219974717e-06,
+      "loss": 1.2835,
+      "step": 35139
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3364412136536032e-06,
+      "loss": 2.1905,
+      "step": 35140
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3363938053097344e-06,
+      "loss": 1.4832,
+      "step": 35141
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.336346396965866e-06,
+      "loss": 3.152,
+      "step": 35142
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3362989886219976e-06,
+      "loss": 2.6339,
+      "step": 35143
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.336251580278129e-06,
+      "loss": 0.8646,
+      "step": 35144
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3362041719342604e-06,
+      "loss": 1.8385,
+      "step": 35145
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3361567635903921e-06,
+      "loss": 1.2712,
+      "step": 35146
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3361093552465234e-06,
+      "loss": 2.8949,
+      "step": 35147
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3360619469026549e-06,
+      "loss": 2.7271,
+      "step": 35148
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3360145385587864e-06,
+      "loss": 2.0195,
+      "step": 35149
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3359671302149179e-06,
+      "loss": 1.2117,
+      "step": 35150
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3359197218710494e-06,
+      "loss": 2.134,
+      "step": 35151
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3358723135271808e-06,
+      "loss": 1.9583,
+      "step": 35152
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3358249051833123e-06,
+      "loss": 3.2132,
+      "step": 35153
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3357774968394438e-06,
+      "loss": 3.2741,
+      "step": 35154
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3357300884955751e-06,
+      "loss": 2.2189,
+      "step": 35155
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3356826801517068e-06,
+      "loss": 1.018,
+      "step": 35156
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3356352718078383e-06,
+      "loss": 3.5498,
+      "step": 35157
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3355878634639696e-06,
+      "loss": 1.4629,
+      "step": 35158
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3355404551201013e-06,
+      "loss": 2.6573,
+      "step": 35159
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3354930467762326e-06,
+      "loss": 2.3425,
+      "step": 35160
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.335445638432364e-06,
+      "loss": 3.0846,
+      "step": 35161
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3353982300884956e-06,
+      "loss": 2.1606,
+      "step": 35162
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.335350821744627e-06,
+      "loss": 2.3716,
+      "step": 35163
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3353034134007585e-06,
+      "loss": 2.1245,
+      "step": 35164
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.33525600505689e-06,
+      "loss": 1.8963,
+      "step": 35165
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3352085967130215e-06,
+      "loss": 2.3583,
+      "step": 35166
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.335161188369153e-06,
+      "loss": 1.8601,
+      "step": 35167
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3351137800252845e-06,
+      "loss": 2.513,
+      "step": 35168
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.335066371681416e-06,
+      "loss": 1.551,
+      "step": 35169
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3350189633375475e-06,
+      "loss": 1.5149,
+      "step": 35170
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3349715549936788e-06,
+      "loss": 1.6999,
+      "step": 35171
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3349241466498105e-06,
+      "loss": 2.6538,
+      "step": 35172
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.334876738305942e-06,
+      "loss": 2.2634,
+      "step": 35173
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3348293299620733e-06,
+      "loss": 3.245,
+      "step": 35174
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3347819216182047e-06,
+      "loss": 2.1155,
+      "step": 35175
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3347345132743364e-06,
+      "loss": 2.6606,
+      "step": 35176
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3346871049304677e-06,
+      "loss": 3.9061,
+      "step": 35177
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.3346396965865992e-06,
+      "loss": 3.4338,
+      "step": 35178
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 1.334592288242731e-06,
+      "loss": 2.0759,
+      "step": 35179
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3345448798988622e-06,
+      "loss": 3.6741,
+      "step": 35180
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3344974715549937e-06,
+      "loss": 1.5694,
+      "step": 35181
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3344500632111252e-06,
+      "loss": 0.9134,
+      "step": 35182
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3344026548672567e-06,
+      "loss": 2.5911,
+      "step": 35183
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3343552465233882e-06,
+      "loss": 1.3203,
+      "step": 35184
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3343078381795197e-06,
+      "loss": 1.1805,
+      "step": 35185
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3342604298356512e-06,
+      "loss": 1.8999,
+      "step": 35186
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3342130214917827e-06,
+      "loss": 0.4326,
+      "step": 35187
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.334165613147914e-06,
+      "loss": 1.7634,
+      "step": 35188
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3341182048040456e-06,
+      "loss": 2.5601,
+      "step": 35189
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3340707964601771e-06,
+      "loss": 2.499,
+      "step": 35190
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3340233881163084e-06,
+      "loss": 2.4023,
+      "step": 35191
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3339759797724401e-06,
+      "loss": 3.0166,
+      "step": 35192
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3339285714285714e-06,
+      "loss": 1.1971,
+      "step": 35193
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3338811630847029e-06,
+      "loss": 2.0487,
+      "step": 35194
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3338337547408344e-06,
+      "loss": 2.4217,
+      "step": 35195
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3337863463969659e-06,
+      "loss": 1.0527,
+      "step": 35196
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3337389380530974e-06,
+      "loss": 1.7156,
+      "step": 35197
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3336915297092289e-06,
+      "loss": 2.1079,
+      "step": 35198
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3336441213653603e-06,
+      "loss": 2.6824,
+      "step": 35199
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3335967130214918e-06,
+      "loss": 2.4991,
+      "step": 35200
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3335493046776233e-06,
+      "loss": 1.2931,
+      "step": 35201
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3335018963337548e-06,
+      "loss": 1.9919,
+      "step": 35202
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3334544879898863e-06,
+      "loss": 1.8675,
+      "step": 35203
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3334070796460176e-06,
+      "loss": 2.0931,
+      "step": 35204
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3333596713021493e-06,
+      "loss": 1.7476,
+      "step": 35205
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3333122629582808e-06,
+      "loss": 2.0707,
+      "step": 35206
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.333264854614412e-06,
+      "loss": 3.0481,
+      "step": 35207
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3332174462705436e-06,
+      "loss": 3.498,
+      "step": 35208
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3331700379266753e-06,
+      "loss": 1.6229,
+      "step": 35209
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3331226295828065e-06,
+      "loss": 3.1103,
+      "step": 35210
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.333075221238938e-06,
+      "loss": 3.9723,
+      "step": 35211
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3330278128950697e-06,
+      "loss": 4.1591,
+      "step": 35212
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.332980404551201e-06,
+      "loss": 2.5202,
+      "step": 35213
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3329329962073325e-06,
+      "loss": 2.2282,
+      "step": 35214
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.332885587863464e-06,
+      "loss": 1.6801,
+      "step": 35215
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3328381795195955e-06,
+      "loss": 1.5806,
+      "step": 35216
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.332790771175727e-06,
+      "loss": 1.8969,
+      "step": 35217
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3327433628318585e-06,
+      "loss": 2.2112,
+      "step": 35218
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.33269595448799e-06,
+      "loss": 2.5032,
+      "step": 35219
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3326485461441215e-06,
+      "loss": 0.4754,
+      "step": 35220
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3326011378002528e-06,
+      "loss": 2.6073,
+      "step": 35221
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3325537294563845e-06,
+      "loss": 2.3326,
+      "step": 35222
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.332506321112516e-06,
+      "loss": 0.7868,
+      "step": 35223
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3324589127686472e-06,
+      "loss": 4.2802,
+      "step": 35224
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.332411504424779e-06,
+      "loss": 2.0042,
+      "step": 35225
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3323640960809102e-06,
+      "loss": 1.8661,
+      "step": 35226
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3323166877370417e-06,
+      "loss": 2.8361,
+      "step": 35227
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3322692793931732e-06,
+      "loss": 2.3076,
+      "step": 35228
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3322218710493047e-06,
+      "loss": 2.887,
+      "step": 35229
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3321744627054362e-06,
+      "loss": 1.4109,
+      "step": 35230
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3321270543615677e-06,
+      "loss": 1.039,
+      "step": 35231
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3320796460176992e-06,
+      "loss": 1.7844,
+      "step": 35232
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3320322376738307e-06,
+      "loss": 2.6677,
+      "step": 35233
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3319848293299621e-06,
+      "loss": 2.8953,
+      "step": 35234
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3319374209860936e-06,
+      "loss": 2.3967,
+      "step": 35235
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3318900126422251e-06,
+      "loss": 1.8607,
+      "step": 35236
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3318426042983564e-06,
+      "loss": 2.5609,
+      "step": 35237
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.331795195954488e-06,
+      "loss": 2.7222,
+      "step": 35238
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3317477876106196e-06,
+      "loss": 0.9669,
+      "step": 35239
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3317003792667509e-06,
+      "loss": 1.9554,
+      "step": 35240
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3316529709228824e-06,
+      "loss": 1.5359,
+      "step": 35241
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.331605562579014e-06,
+      "loss": 2.2971,
+      "step": 35242
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 1.3315581542351454e-06,
+      "loss": 3.2433,
+      "step": 35243
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3315107458912769e-06,
+      "loss": 1.6204,
+      "step": 35244
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3314633375474084e-06,
+      "loss": 3.7049,
+      "step": 35245
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3314159292035398e-06,
+      "loss": 2.3969,
+      "step": 35246
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3313685208596713e-06,
+      "loss": 3.0199,
+      "step": 35247
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3313211125158028e-06,
+      "loss": 1.5987,
+      "step": 35248
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3312737041719343e-06,
+      "loss": 2.3027,
+      "step": 35249
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3312262958280658e-06,
+      "loss": 1.6854,
+      "step": 35250
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.331178887484197e-06,
+      "loss": 1.6187,
+      "step": 35251
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3311314791403288e-06,
+      "loss": 1.3919,
+      "step": 35252
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3310840707964603e-06,
+      "loss": 2.6618,
+      "step": 35253
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3310366624525916e-06,
+      "loss": 2.0404,
+      "step": 35254
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3309892541087233e-06,
+      "loss": 1.6738,
+      "step": 35255
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3309418457648548e-06,
+      "loss": 1.6123,
+      "step": 35256
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330894437420986e-06,
+      "loss": 3.9012,
+      "step": 35257
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3308470290771175e-06,
+      "loss": 2.0769,
+      "step": 35258
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330799620733249e-06,
+      "loss": 2.4492,
+      "step": 35259
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3307522123893805e-06,
+      "loss": 1.0935,
+      "step": 35260
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330704804045512e-06,
+      "loss": 2.5031,
+      "step": 35261
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3306573957016435e-06,
+      "loss": 2.2671,
+      "step": 35262
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330609987357775e-06,
+      "loss": 2.1314,
+      "step": 35263
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3305625790139065e-06,
+      "loss": 2.9105,
+      "step": 35264
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330515170670038e-06,
+      "loss": 3.8815,
+      "step": 35265
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3304677623261695e-06,
+      "loss": 1.8768,
+      "step": 35266
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330420353982301e-06,
+      "loss": 2.2759,
+      "step": 35267
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3303729456384325e-06,
+      "loss": 2.0795,
+      "step": 35268
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330325537294564e-06,
+      "loss": 1.1909,
+      "step": 35269
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3302781289506952e-06,
+      "loss": 2.1935,
+      "step": 35270
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3302307206068267e-06,
+      "loss": 1.7436,
+      "step": 35271
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3301833122629584e-06,
+      "loss": 2.4887,
+      "step": 35272
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3301359039190897e-06,
+      "loss": 1.7096,
+      "step": 35273
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3300884955752212e-06,
+      "loss": 2.2087,
+      "step": 35274
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.330041087231353e-06,
+      "loss": 1.796,
+      "step": 35275
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3299936788874842e-06,
+      "loss": 1.1052,
+      "step": 35276
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3299462705436157e-06,
+      "loss": 2.5121,
+      "step": 35277
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3298988621997472e-06,
+      "loss": 3.8272,
+      "step": 35278
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3298514538558787e-06,
+      "loss": 2.9312,
+      "step": 35279
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3298040455120102e-06,
+      "loss": 0.9644,
+      "step": 35280
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3297566371681416e-06,
+      "loss": 2.6738,
+      "step": 35281
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3297092288242731e-06,
+      "loss": 3.0922,
+      "step": 35282
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3296618204804046e-06,
+      "loss": 1.9356,
+      "step": 35283
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.329614412136536e-06,
+      "loss": 1.4094,
+      "step": 35284
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3295670037926676e-06,
+      "loss": 2.2891,
+      "step": 35285
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3295195954487991e-06,
+      "loss": 1.7507,
+      "step": 35286
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3294721871049304e-06,
+      "loss": 1.9499,
+      "step": 35287
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.329424778761062e-06,
+      "loss": 1.5723,
+      "step": 35288
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3293773704171936e-06,
+      "loss": 3.8207,
+      "step": 35289
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3293299620733249e-06,
+      "loss": 1.9937,
+      "step": 35290
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3292825537294564e-06,
+      "loss": 1.1181,
+      "step": 35291
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3292351453855878e-06,
+      "loss": 1.9644,
+      "step": 35292
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3291877370417193e-06,
+      "loss": 0.765,
+      "step": 35293
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3291403286978508e-06,
+      "loss": 3.6015,
+      "step": 35294
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3290929203539823e-06,
+      "loss": 2.4094,
+      "step": 35295
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3290455120101138e-06,
+      "loss": 1.8205,
+      "step": 35296
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3289981036662453e-06,
+      "loss": 2.1989,
+      "step": 35297
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3289506953223768e-06,
+      "loss": 3.862,
+      "step": 35298
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3289032869785083e-06,
+      "loss": 1.6977,
+      "step": 35299
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3288558786346398e-06,
+      "loss": 1.9053,
+      "step": 35300
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3288084702907713e-06,
+      "loss": 1.3374,
+      "step": 35301
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3287610619469028e-06,
+      "loss": 2.8483,
+      "step": 35302
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.328713653603034e-06,
+      "loss": 1.2121,
+      "step": 35303
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3286662452591655e-06,
+      "loss": 2.2739,
+      "step": 35304
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3286188369152972e-06,
+      "loss": 3.369,
+      "step": 35305
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 1.3285714285714285e-06,
+      "loss": 1.305,
+      "step": 35306
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.32852402022756e-06,
+      "loss": 2.7264,
+      "step": 35307
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3284766118836917e-06,
+      "loss": 3.1068,
+      "step": 35308
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.328429203539823e-06,
+      "loss": 1.5801,
+      "step": 35309
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3283817951959545e-06,
+      "loss": 2.2473,
+      "step": 35310
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.328334386852086e-06,
+      "loss": 2.815,
+      "step": 35311
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3282869785082175e-06,
+      "loss": 3.008,
+      "step": 35312
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.328239570164349e-06,
+      "loss": 2.7383,
+      "step": 35313
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3281921618204805e-06,
+      "loss": 1.5213,
+      "step": 35314
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.328144753476612e-06,
+      "loss": 4.3027,
+      "step": 35315
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3280973451327434e-06,
+      "loss": 2.9073,
+      "step": 35316
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3280499367888747e-06,
+      "loss": 2.5959,
+      "step": 35317
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3280025284450064e-06,
+      "loss": 2.7975,
+      "step": 35318
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.327955120101138e-06,
+      "loss": 1.3829,
+      "step": 35319
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3279077117572692e-06,
+      "loss": 1.5477,
+      "step": 35320
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.327860303413401e-06,
+      "loss": 1.9692,
+      "step": 35321
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3278128950695324e-06,
+      "loss": 1.7829,
+      "step": 35322
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3277654867256637e-06,
+      "loss": 3.3677,
+      "step": 35323
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3277180783817952e-06,
+      "loss": 2.3884,
+      "step": 35324
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3276706700379267e-06,
+      "loss": 1.7182,
+      "step": 35325
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3276232616940582e-06,
+      "loss": 2.9742,
+      "step": 35326
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3275758533501897e-06,
+      "loss": 2.1265,
+      "step": 35327
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3275284450063211e-06,
+      "loss": 2.6458,
+      "step": 35328
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3274810366624526e-06,
+      "loss": 2.3523,
+      "step": 35329
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3274336283185841e-06,
+      "loss": 0.6557,
+      "step": 35330
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3273862199747156e-06,
+      "loss": 2.0047,
+      "step": 35331
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3273388116308471e-06,
+      "loss": 1.6954,
+      "step": 35332
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3272914032869786e-06,
+      "loss": 2.2329,
+      "step": 35333
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.32724399494311e-06,
+      "loss": 2.1191,
+      "step": 35334
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3271965865992416e-06,
+      "loss": 3.2151,
+      "step": 35335
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3271491782553729e-06,
+      "loss": 1.1364,
+      "step": 35336
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3271017699115044e-06,
+      "loss": 1.9546,
+      "step": 35337
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.327054361567636e-06,
+      "loss": 1.6704,
+      "step": 35338
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3270069532237673e-06,
+      "loss": 3.497,
+      "step": 35339
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3269595448798988e-06,
+      "loss": 0.7172,
+      "step": 35340
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3269121365360305e-06,
+      "loss": 1.6543,
+      "step": 35341
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3268647281921618e-06,
+      "loss": 1.2462,
+      "step": 35342
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3268173198482933e-06,
+      "loss": 2.9619,
+      "step": 35343
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3267699115044248e-06,
+      "loss": 1.6642,
+      "step": 35344
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3267225031605563e-06,
+      "loss": 2.7974,
+      "step": 35345
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3266750948166878e-06,
+      "loss": 2.168,
+      "step": 35346
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3266276864728193e-06,
+      "loss": 2.6262,
+      "step": 35347
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3265802781289508e-06,
+      "loss": 2.0129,
+      "step": 35348
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3265328697850823e-06,
+      "loss": 2.5101,
+      "step": 35349
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3264854614412135e-06,
+      "loss": 1.4392,
+      "step": 35350
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3264380530973453e-06,
+      "loss": 2.486,
+      "step": 35351
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3263906447534767e-06,
+      "loss": 1.9335,
+      "step": 35352
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.326343236409608e-06,
+      "loss": 3.0674,
+      "step": 35353
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3262958280657397e-06,
+      "loss": 3.0349,
+      "step": 35354
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3262484197218712e-06,
+      "loss": 2.6115,
+      "step": 35355
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3262010113780025e-06,
+      "loss": 2.7155,
+      "step": 35356
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.326153603034134e-06,
+      "loss": 2.2563,
+      "step": 35357
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3261061946902655e-06,
+      "loss": 2.2575,
+      "step": 35358
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.326058786346397e-06,
+      "loss": 2.4421,
+      "step": 35359
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3260113780025285e-06,
+      "loss": 3.1134,
+      "step": 35360
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.32596396965866e-06,
+      "loss": 1.989,
+      "step": 35361
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3259165613147915e-06,
+      "loss": 1.8627,
+      "step": 35362
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.325869152970923e-06,
+      "loss": 2.4856,
+      "step": 35363
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3258217446270544e-06,
+      "loss": 2.3131,
+      "step": 35364
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.325774336283186e-06,
+      "loss": 3.589,
+      "step": 35365
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3257269279393174e-06,
+      "loss": 1.1884,
+      "step": 35366
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3256795195954487e-06,
+      "loss": 1.6601,
+      "step": 35367
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3256321112515804e-06,
+      "loss": 2.8675,
+      "step": 35368
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 1.3255847029077117e-06,
+      "loss": 1.1823,
+      "step": 35369
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3255372945638432e-06,
+      "loss": 2.7447,
+      "step": 35370
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3254898862199749e-06,
+      "loss": 2.0356,
+      "step": 35371
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3254424778761062e-06,
+      "loss": 1.7718,
+      "step": 35372
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3253950695322377e-06,
+      "loss": 2.7476,
+      "step": 35373
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3253476611883691e-06,
+      "loss": 1.0511,
+      "step": 35374
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3253002528445006e-06,
+      "loss": 2.386,
+      "step": 35375
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3252528445006321e-06,
+      "loss": 2.7098,
+      "step": 35376
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3252054361567636e-06,
+      "loss": 2.6129,
+      "step": 35377
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3251580278128951e-06,
+      "loss": 1.2996,
+      "step": 35378
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3251106194690266e-06,
+      "loss": 3.1299,
+      "step": 35379
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3250632111251579e-06,
+      "loss": 3.4506,
+      "step": 35380
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3250158027812896e-06,
+      "loss": 2.9986,
+      "step": 35381
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.324968394437421e-06,
+      "loss": 1.01,
+      "step": 35382
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3249209860935524e-06,
+      "loss": 2.5875,
+      "step": 35383
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.324873577749684e-06,
+      "loss": 2.2015,
+      "step": 35384
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3248261694058156e-06,
+      "loss": 1.3644,
+      "step": 35385
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3247787610619468e-06,
+      "loss": 0.9644,
+      "step": 35386
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3247313527180783e-06,
+      "loss": 2.6714,
+      "step": 35387
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.32468394437421e-06,
+      "loss": 1.1967,
+      "step": 35388
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3246365360303413e-06,
+      "loss": 2.164,
+      "step": 35389
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3245891276864728e-06,
+      "loss": 1.4572,
+      "step": 35390
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3245417193426043e-06,
+      "loss": 1.8707,
+      "step": 35391
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3244943109987358e-06,
+      "loss": 2.3827,
+      "step": 35392
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3244469026548673e-06,
+      "loss": 1.5433,
+      "step": 35393
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3243994943109988e-06,
+      "loss": 1.8436,
+      "step": 35394
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3243520859671303e-06,
+      "loss": 1.4243,
+      "step": 35395
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3243046776232618e-06,
+      "loss": 1.8805,
+      "step": 35396
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3242572692793933e-06,
+      "loss": 2.703,
+      "step": 35397
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3242098609355247e-06,
+      "loss": 2.783,
+      "step": 35398
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3241624525916562e-06,
+      "loss": 1.8571,
+      "step": 35399
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3241150442477875e-06,
+      "loss": 2.283,
+      "step": 35400
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3240676359039192e-06,
+      "loss": 2.664,
+      "step": 35401
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3240202275600505e-06,
+      "loss": 2.5066,
+      "step": 35402
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.323972819216182e-06,
+      "loss": 2.3921,
+      "step": 35403
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3239254108723137e-06,
+      "loss": 2.9867,
+      "step": 35404
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.323878002528445e-06,
+      "loss": 1.9353,
+      "step": 35405
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3238305941845765e-06,
+      "loss": 2.8587,
+      "step": 35406
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.323783185840708e-06,
+      "loss": 2.5852,
+      "step": 35407
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3237357774968395e-06,
+      "loss": 1.6854,
+      "step": 35408
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.323688369152971e-06,
+      "loss": 2.5869,
+      "step": 35409
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3236409608091024e-06,
+      "loss": 1.6845,
+      "step": 35410
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.323593552465234e-06,
+      "loss": 2.6522,
+      "step": 35411
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3235461441213654e-06,
+      "loss": 3.271,
+      "step": 35412
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3234987357774967e-06,
+      "loss": 1.3329,
+      "step": 35413
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3234513274336284e-06,
+      "loss": 2.4796,
+      "step": 35414
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.32340391908976e-06,
+      "loss": 3.3934,
+      "step": 35415
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3233565107458912e-06,
+      "loss": 1.5893,
+      "step": 35416
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3233091024020229e-06,
+      "loss": 2.2095,
+      "step": 35417
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3232616940581544e-06,
+      "loss": 1.3579,
+      "step": 35418
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3232142857142857e-06,
+      "loss": 0.4939,
+      "step": 35419
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3231668773704172e-06,
+      "loss": 1.4951,
+      "step": 35420
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3231194690265489e-06,
+      "loss": 1.4479,
+      "step": 35421
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3230720606826801e-06,
+      "loss": 1.6784,
+      "step": 35422
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3230246523388116e-06,
+      "loss": 2.0215,
+      "step": 35423
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3229772439949431e-06,
+      "loss": 2.0321,
+      "step": 35424
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3229298356510746e-06,
+      "loss": 2.6037,
+      "step": 35425
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.322882427307206e-06,
+      "loss": 1.3792,
+      "step": 35426
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3228350189633376e-06,
+      "loss": 1.2042,
+      "step": 35427
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.322787610619469e-06,
+      "loss": 1.9883,
+      "step": 35428
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3227402022756006e-06,
+      "loss": 2.6285,
+      "step": 35429
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.322692793931732e-06,
+      "loss": 2.54,
+      "step": 35430
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3226453855878636e-06,
+      "loss": 2.0349,
+      "step": 35431
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3225979772439948e-06,
+      "loss": 3.2284,
+      "step": 35432
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 1.3225505689001263e-06,
+      "loss": 2.937,
+      "step": 35433
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.322503160556258e-06,
+      "loss": 2.1156,
+      "step": 35434
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3224557522123893e-06,
+      "loss": 3.4908,
+      "step": 35435
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3224083438685208e-06,
+      "loss": 2.1181,
+      "step": 35436
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3223609355246525e-06,
+      "loss": 2.2813,
+      "step": 35437
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3223135271807838e-06,
+      "loss": 2.5113,
+      "step": 35438
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3222661188369153e-06,
+      "loss": 1.1583,
+      "step": 35439
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3222187104930468e-06,
+      "loss": 3.4262,
+      "step": 35440
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3221713021491783e-06,
+      "loss": 1.9082,
+      "step": 35441
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3221238938053098e-06,
+      "loss": 1.3347,
+      "step": 35442
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3220764854614413e-06,
+      "loss": 2.9702,
+      "step": 35443
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3220290771175728e-06,
+      "loss": 2.5186,
+      "step": 35444
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3219816687737042e-06,
+      "loss": 2.5823,
+      "step": 35445
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3219342604298355e-06,
+      "loss": 3.1419,
+      "step": 35446
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3218868520859672e-06,
+      "loss": 1.2341,
+      "step": 35447
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3218394437420987e-06,
+      "loss": 0.7812,
+      "step": 35448
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.32179203539823e-06,
+      "loss": 3.1012,
+      "step": 35449
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3217446270543617e-06,
+      "loss": 3.2224,
+      "step": 35450
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3216972187104932e-06,
+      "loss": 1.0315,
+      "step": 35451
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3216498103666245e-06,
+      "loss": 2.9986,
+      "step": 35452
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.321602402022756e-06,
+      "loss": 1.5519,
+      "step": 35453
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3215549936788875e-06,
+      "loss": 1.5085,
+      "step": 35454
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.321507585335019e-06,
+      "loss": 1.1669,
+      "step": 35455
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3214601769911504e-06,
+      "loss": 1.9944,
+      "step": 35456
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.321412768647282e-06,
+      "loss": 2.2084,
+      "step": 35457
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3213653603034134e-06,
+      "loss": 2.3339,
+      "step": 35458
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.321317951959545e-06,
+      "loss": 1.1819,
+      "step": 35459
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3212705436156764e-06,
+      "loss": 0.8876,
+      "step": 35460
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.321223135271808e-06,
+      "loss": 0.5687,
+      "step": 35461
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3211757269279394e-06,
+      "loss": 1.7553,
+      "step": 35462
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3211283185840709e-06,
+      "loss": 1.8019,
+      "step": 35463
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3210809102402024e-06,
+      "loss": 1.7528,
+      "step": 35464
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3210335018963337e-06,
+      "loss": 2.7243,
+      "step": 35465
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3209860935524652e-06,
+      "loss": 2.1887,
+      "step": 35466
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3209386852085969e-06,
+      "loss": 0.9787,
+      "step": 35467
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3208912768647281e-06,
+      "loss": 1.3281,
+      "step": 35468
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3208438685208596e-06,
+      "loss": 2.1569,
+      "step": 35469
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3207964601769913e-06,
+      "loss": 1.4242,
+      "step": 35470
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3207490518331226e-06,
+      "loss": 3.8774,
+      "step": 35471
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3207016434892541e-06,
+      "loss": 2.4959,
+      "step": 35472
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3206542351453856e-06,
+      "loss": 2.1253,
+      "step": 35473
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.320606826801517e-06,
+      "loss": 3.1817,
+      "step": 35474
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3205594184576486e-06,
+      "loss": 1.0623,
+      "step": 35475
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.32051201011378e-06,
+      "loss": 2.321,
+      "step": 35476
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3204646017699116e-06,
+      "loss": 2.6174,
+      "step": 35477
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.320417193426043e-06,
+      "loss": 2.5067,
+      "step": 35478
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3203697850821743e-06,
+      "loss": 4.1822,
+      "step": 35479
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.320322376738306e-06,
+      "loss": 2.4389,
+      "step": 35480
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3202749683944375e-06,
+      "loss": 1.4697,
+      "step": 35481
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3202275600505688e-06,
+      "loss": 1.9757,
+      "step": 35482
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3201801517067005e-06,
+      "loss": 2.8138,
+      "step": 35483
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.320132743362832e-06,
+      "loss": 1.9882,
+      "step": 35484
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3200853350189633e-06,
+      "loss": 1.8751,
+      "step": 35485
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3200379266750948e-06,
+      "loss": 1.5237,
+      "step": 35486
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3199905183312263e-06,
+      "loss": 1.3191,
+      "step": 35487
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3199431099873578e-06,
+      "loss": 4.7441,
+      "step": 35488
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3198957016434893e-06,
+      "loss": 1.7729,
+      "step": 35489
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3198482932996208e-06,
+      "loss": 1.4199,
+      "step": 35490
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3198008849557522e-06,
+      "loss": 1.1136,
+      "step": 35491
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3197534766118837e-06,
+      "loss": 2.9562,
+      "step": 35492
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3197060682680152e-06,
+      "loss": 1.3739,
+      "step": 35493
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3196586599241467e-06,
+      "loss": 1.968,
+      "step": 35494
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3196112515802782e-06,
+      "loss": 3.4149,
+      "step": 35495
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 1.3195638432364097e-06,
+      "loss": 2.552,
+      "step": 35496
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3195164348925412e-06,
+      "loss": 1.4286,
+      "step": 35497
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3194690265486725e-06,
+      "loss": 2.4499,
+      "step": 35498
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.319421618204804e-06,
+      "loss": 1.6582,
+      "step": 35499
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3193742098609357e-06,
+      "loss": 1.9925,
+      "step": 35500
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.319326801517067e-06,
+      "loss": 2.7037,
+      "step": 35501
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3192793931731985e-06,
+      "loss": 1.0952,
+      "step": 35502
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3192319848293302e-06,
+      "loss": 2.1447,
+      "step": 35503
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3191845764854614e-06,
+      "loss": 2.8902,
+      "step": 35504
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.319137168141593e-06,
+      "loss": 2.6359,
+      "step": 35505
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3190897597977244e-06,
+      "loss": 1.5889,
+      "step": 35506
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.319042351453856e-06,
+      "loss": 2.0511,
+      "step": 35507
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3189949431099874e-06,
+      "loss": 2.0206,
+      "step": 35508
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3189475347661187e-06,
+      "loss": 2.0041,
+      "step": 35509
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3189001264222504e-06,
+      "loss": 2.4217,
+      "step": 35510
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3188527180783819e-06,
+      "loss": 1.8548,
+      "step": 35511
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3188053097345132e-06,
+      "loss": 1.5364,
+      "step": 35512
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3187579013906449e-06,
+      "loss": 1.9546,
+      "step": 35513
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3187104930467764e-06,
+      "loss": 1.4685,
+      "step": 35514
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3186630847029076e-06,
+      "loss": 2.943,
+      "step": 35515
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3186156763590391e-06,
+      "loss": 2.6315,
+      "step": 35516
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3185682680151708e-06,
+      "loss": 1.8858,
+      "step": 35517
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3185208596713021e-06,
+      "loss": 1.8934,
+      "step": 35518
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3184734513274336e-06,
+      "loss": 3.5646,
+      "step": 35519
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.318426042983565e-06,
+      "loss": 3.0033,
+      "step": 35520
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3183786346396966e-06,
+      "loss": 2.9321,
+      "step": 35521
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.318331226295828e-06,
+      "loss": 1.3791,
+      "step": 35522
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3182838179519596e-06,
+      "loss": 2.792,
+      "step": 35523
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.318236409608091e-06,
+      "loss": 1.3654,
+      "step": 35524
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3181890012642226e-06,
+      "loss": 2.9737,
+      "step": 35525
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.318141592920354e-06,
+      "loss": 1.0932,
+      "step": 35526
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3180941845764855e-06,
+      "loss": 2.1885,
+      "step": 35527
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.318046776232617e-06,
+      "loss": 1.8721,
+      "step": 35528
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3179993678887483e-06,
+      "loss": 3.6727,
+      "step": 35529
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.31795195954488e-06,
+      "loss": 2.9364,
+      "step": 35530
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3179045512010113e-06,
+      "loss": 1.9464,
+      "step": 35531
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3178571428571428e-06,
+      "loss": 0.5946,
+      "step": 35532
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3178097345132745e-06,
+      "loss": 2.0569,
+      "step": 35533
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3177623261694058e-06,
+      "loss": 2.9124,
+      "step": 35534
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3177149178255373e-06,
+      "loss": 1.9049,
+      "step": 35535
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3176675094816688e-06,
+      "loss": 2.2306,
+      "step": 35536
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3176201011378003e-06,
+      "loss": 1.3346,
+      "step": 35537
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3175726927939317e-06,
+      "loss": 1.2492,
+      "step": 35538
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3175252844500632e-06,
+      "loss": 0.7122,
+      "step": 35539
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3174778761061947e-06,
+      "loss": 1.8888,
+      "step": 35540
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3174304677623262e-06,
+      "loss": 2.9525,
+      "step": 35541
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3173830594184575e-06,
+      "loss": 1.6625,
+      "step": 35542
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3173356510745892e-06,
+      "loss": 3.3296,
+      "step": 35543
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3172882427307207e-06,
+      "loss": 1.9243,
+      "step": 35544
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.317240834386852e-06,
+      "loss": 2.4813,
+      "step": 35545
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3171934260429837e-06,
+      "loss": 1.4765,
+      "step": 35546
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3171460176991152e-06,
+      "loss": 2.5545,
+      "step": 35547
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3170986093552465e-06,
+      "loss": 3.3389,
+      "step": 35548
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.317051201011378e-06,
+      "loss": 4.4227,
+      "step": 35549
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3170037926675097e-06,
+      "loss": 1.6728,
+      "step": 35550
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.316956384323641e-06,
+      "loss": 2.8112,
+      "step": 35551
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3169089759797724e-06,
+      "loss": 2.4643,
+      "step": 35552
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.316861567635904e-06,
+      "loss": 0.752,
+      "step": 35553
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3168141592920354e-06,
+      "loss": 0.4957,
+      "step": 35554
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.316766750948167e-06,
+      "loss": 1.8161,
+      "step": 35555
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3167193426042984e-06,
+      "loss": 2.5513,
+      "step": 35556
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3166719342604299e-06,
+      "loss": 2.872,
+      "step": 35557
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3166245259165614e-06,
+      "loss": 2.0601,
+      "step": 35558
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 1.3165771175726929e-06,
+      "loss": 2.6527,
+      "step": 35559
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3165297092288244e-06,
+      "loss": 1.6044,
+      "step": 35560
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3164823008849559e-06,
+      "loss": 2.0619,
+      "step": 35561
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3164348925410871e-06,
+      "loss": 0.7488,
+      "step": 35562
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3163874841972188e-06,
+      "loss": 2.8453,
+      "step": 35563
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3163400758533501e-06,
+      "loss": 1.9421,
+      "step": 35564
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3162926675094816e-06,
+      "loss": 1.9704,
+      "step": 35565
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3162452591656133e-06,
+      "loss": 3.8975,
+      "step": 35566
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3161978508217446e-06,
+      "loss": 2.6981,
+      "step": 35567
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.316150442477876e-06,
+      "loss": 2.1702,
+      "step": 35568
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3161030341340076e-06,
+      "loss": 3.5885,
+      "step": 35569
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.316055625790139e-06,
+      "loss": 2.6161,
+      "step": 35570
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3160082174462706e-06,
+      "loss": 2.4661,
+      "step": 35571
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.315960809102402e-06,
+      "loss": 2.819,
+      "step": 35572
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3159134007585335e-06,
+      "loss": 1.4681,
+      "step": 35573
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.315865992414665e-06,
+      "loss": 2.7834,
+      "step": 35574
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3158185840707963e-06,
+      "loss": 3.1415,
+      "step": 35575
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.315771175726928e-06,
+      "loss": 2.1856,
+      "step": 35576
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3157237673830595e-06,
+      "loss": 1.3115,
+      "step": 35577
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3156763590391908e-06,
+      "loss": 2.2897,
+      "step": 35578
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3156289506953225e-06,
+      "loss": 1.9603,
+      "step": 35579
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.315581542351454e-06,
+      "loss": 4.1332,
+      "step": 35580
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3155341340075853e-06,
+      "loss": 3.3543,
+      "step": 35581
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3154867256637168e-06,
+      "loss": 2.1994,
+      "step": 35582
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3154393173198485e-06,
+      "loss": 1.3827,
+      "step": 35583
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3153919089759798e-06,
+      "loss": 1.1809,
+      "step": 35584
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3153445006321112e-06,
+      "loss": 3.5384,
+      "step": 35585
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3152970922882427e-06,
+      "loss": 1.4542,
+      "step": 35586
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3152496839443742e-06,
+      "loss": 1.8725,
+      "step": 35587
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3152022756005057e-06,
+      "loss": 2.3821,
+      "step": 35588
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3151548672566372e-06,
+      "loss": 1.1839,
+      "step": 35589
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3151074589127687e-06,
+      "loss": 1.5756,
+      "step": 35590
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3150600505689002e-06,
+      "loss": 1.0333,
+      "step": 35591
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3150126422250317e-06,
+      "loss": 2.8216,
+      "step": 35592
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3149652338811632e-06,
+      "loss": 1.5118,
+      "step": 35593
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3149178255372947e-06,
+      "loss": 1.9743,
+      "step": 35594
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.314870417193426e-06,
+      "loss": 1.2554,
+      "step": 35595
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3148230088495577e-06,
+      "loss": 1.5166,
+      "step": 35596
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.314775600505689e-06,
+      "loss": 1.7068,
+      "step": 35597
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3147281921618204e-06,
+      "loss": 3.1553,
+      "step": 35598
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3146807838179521e-06,
+      "loss": 3.1871,
+      "step": 35599
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3146333754740834e-06,
+      "loss": 1.5007,
+      "step": 35600
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.314585967130215e-06,
+      "loss": 1.9759,
+      "step": 35601
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3145385587863464e-06,
+      "loss": 1.6025,
+      "step": 35602
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3144911504424779e-06,
+      "loss": 0.7501,
+      "step": 35603
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3144437420986094e-06,
+      "loss": 1.4218,
+      "step": 35604
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3143963337547409e-06,
+      "loss": 1.2338,
+      "step": 35605
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3143489254108724e-06,
+      "loss": 2.3519,
+      "step": 35606
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3143015170670039e-06,
+      "loss": 0.8845,
+      "step": 35607
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3142541087231351e-06,
+      "loss": 2.0806,
+      "step": 35608
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3142067003792668e-06,
+      "loss": 1.688,
+      "step": 35609
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3141592920353983e-06,
+      "loss": 2.1383,
+      "step": 35610
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3141118836915296e-06,
+      "loss": 1.7774,
+      "step": 35611
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3140644753476613e-06,
+      "loss": 2.1876,
+      "step": 35612
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3140170670037928e-06,
+      "loss": 3.1739,
+      "step": 35613
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.313969658659924e-06,
+      "loss": 2.6879,
+      "step": 35614
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3139222503160556e-06,
+      "loss": 1.7858,
+      "step": 35615
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3138748419721873e-06,
+      "loss": 1.9664,
+      "step": 35616
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3138274336283186e-06,
+      "loss": 2.4504,
+      "step": 35617
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.31378002528445e-06,
+      "loss": 3.0343,
+      "step": 35618
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3137326169405816e-06,
+      "loss": 2.2164,
+      "step": 35619
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.313685208596713e-06,
+      "loss": 3.3036,
+      "step": 35620
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3136378002528445e-06,
+      "loss": 3.1086,
+      "step": 35621
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.313590391908976e-06,
+      "loss": 1.7519,
+      "step": 35622
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 1.3135429835651075e-06,
+      "loss": 2.3151,
+      "step": 35623
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.313495575221239e-06,
+      "loss": 3.6485,
+      "step": 35624
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3134481668773705e-06,
+      "loss": 4.1511,
+      "step": 35625
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.313400758533502e-06,
+      "loss": 2.3628,
+      "step": 35626
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3133533501896335e-06,
+      "loss": 2.2342,
+      "step": 35627
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3133059418457648e-06,
+      "loss": 1.1778,
+      "step": 35628
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3132585335018965e-06,
+      "loss": 2.6569,
+      "step": 35629
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3132111251580278e-06,
+      "loss": 2.4255,
+      "step": 35630
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3131637168141592e-06,
+      "loss": 0.8897,
+      "step": 35631
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.313116308470291e-06,
+      "loss": 2.3849,
+      "step": 35632
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3130689001264222e-06,
+      "loss": 1.5601,
+      "step": 35633
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3130214917825537e-06,
+      "loss": 1.7172,
+      "step": 35634
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3129740834386852e-06,
+      "loss": 1.1549,
+      "step": 35635
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3129266750948167e-06,
+      "loss": 2.3657,
+      "step": 35636
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3128792667509482e-06,
+      "loss": 1.5462,
+      "step": 35637
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3128318584070797e-06,
+      "loss": 2.7204,
+      "step": 35638
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3127844500632112e-06,
+      "loss": 1.6169,
+      "step": 35639
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3127370417193427e-06,
+      "loss": 1.4018,
+      "step": 35640
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.312689633375474e-06,
+      "loss": 2.0633,
+      "step": 35641
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3126422250316057e-06,
+      "loss": 3.1332,
+      "step": 35642
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3125948166877372e-06,
+      "loss": 1.7312,
+      "step": 35643
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3125474083438684e-06,
+      "loss": 1.3674,
+      "step": 35644
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3125000000000001e-06,
+      "loss": 3.1286,
+      "step": 35645
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3124525916561316e-06,
+      "loss": 3.6376,
+      "step": 35646
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.312405183312263e-06,
+      "loss": 3.7246,
+      "step": 35647
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3123577749683944e-06,
+      "loss": 1.9981,
+      "step": 35648
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3123103666245261e-06,
+      "loss": 1.5373,
+      "step": 35649
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3122629582806574e-06,
+      "loss": 2.0605,
+      "step": 35650
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3122155499367889e-06,
+      "loss": 2.9615,
+      "step": 35651
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3121681415929204e-06,
+      "loss": 3.5891,
+      "step": 35652
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3121207332490519e-06,
+      "loss": 2.4319,
+      "step": 35653
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3120733249051834e-06,
+      "loss": 2.8619,
+      "step": 35654
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3120259165613148e-06,
+      "loss": 1.9843,
+      "step": 35655
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3119785082174463e-06,
+      "loss": 1.6158,
+      "step": 35656
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3119310998735778e-06,
+      "loss": 2.6,
+      "step": 35657
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3118836915297091e-06,
+      "loss": 2.2193,
+      "step": 35658
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3118362831858408e-06,
+      "loss": 3.3271,
+      "step": 35659
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3117888748419723e-06,
+      "loss": 3.0992,
+      "step": 35660
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3117414664981036e-06,
+      "loss": 1.6221,
+      "step": 35661
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3116940581542353e-06,
+      "loss": 1.7956,
+      "step": 35662
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3116466498103666e-06,
+      "loss": 1.8121,
+      "step": 35663
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.311599241466498e-06,
+      "loss": 2.4738,
+      "step": 35664
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3115518331226296e-06,
+      "loss": 1.364,
+      "step": 35665
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.311504424778761e-06,
+      "loss": 3.0728,
+      "step": 35666
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3114570164348925e-06,
+      "loss": 1.9428,
+      "step": 35667
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.311409608091024e-06,
+      "loss": 1.8519,
+      "step": 35668
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3113621997471555e-06,
+      "loss": 2.5746,
+      "step": 35669
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.311314791403287e-06,
+      "loss": 2.4219,
+      "step": 35670
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3112673830594185e-06,
+      "loss": 3.2191,
+      "step": 35671
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.31121997471555e-06,
+      "loss": 2.6821,
+      "step": 35672
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3111725663716815e-06,
+      "loss": 2.9388,
+      "step": 35673
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3111251580278128e-06,
+      "loss": 1.6843,
+      "step": 35674
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3110777496839445e-06,
+      "loss": 1.5118,
+      "step": 35675
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.311030341340076e-06,
+      "loss": 1.0519,
+      "step": 35676
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3109829329962073e-06,
+      "loss": 1.5408,
+      "step": 35677
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3109355246523387e-06,
+      "loss": 2.4974,
+      "step": 35678
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3108881163084704e-06,
+      "loss": 2.252,
+      "step": 35679
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3108407079646017e-06,
+      "loss": 2.8293,
+      "step": 35680
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3107932996207332e-06,
+      "loss": 3.6433,
+      "step": 35681
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.310745891276865e-06,
+      "loss": 1.702,
+      "step": 35682
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3106984829329962e-06,
+      "loss": 2.6618,
+      "step": 35683
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3106510745891277e-06,
+      "loss": 3.6095,
+      "step": 35684
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3106036662452592e-06,
+      "loss": 1.9897,
+      "step": 35685
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 1.3105562579013907e-06,
+      "loss": 2.5623,
+      "step": 35686
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3105088495575222e-06,
+      "loss": 3.033,
+      "step": 35687
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3104614412136537e-06,
+      "loss": 1.8777,
+      "step": 35688
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3104140328697852e-06,
+      "loss": 1.7204,
+      "step": 35689
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3103666245259167e-06,
+      "loss": 1.3809,
+      "step": 35690
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.310319216182048e-06,
+      "loss": 1.6071,
+      "step": 35691
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3102718078381796e-06,
+      "loss": 1.9059,
+      "step": 35692
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3102243994943111e-06,
+      "loss": 2.7084,
+      "step": 35693
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3101769911504424e-06,
+      "loss": 2.1919,
+      "step": 35694
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3101295828065741e-06,
+      "loss": 2.2501,
+      "step": 35695
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3100821744627054e-06,
+      "loss": 2.3256,
+      "step": 35696
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3100347661188369e-06,
+      "loss": 2.1656,
+      "step": 35697
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3099873577749684e-06,
+      "loss": 2.6864,
+      "step": 35698
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3099399494310999e-06,
+      "loss": 2.7113,
+      "step": 35699
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3098925410872314e-06,
+      "loss": 2.191,
+      "step": 35700
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3098451327433629e-06,
+      "loss": 2.23,
+      "step": 35701
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3097977243994943e-06,
+      "loss": 3.0844,
+      "step": 35702
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3097503160556258e-06,
+      "loss": 3.0853,
+      "step": 35703
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3097029077117573e-06,
+      "loss": 2.0404,
+      "step": 35704
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3096554993678888e-06,
+      "loss": 2.2784,
+      "step": 35705
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3096080910240203e-06,
+      "loss": 1.178,
+      "step": 35706
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3095606826801516e-06,
+      "loss": 1.6745,
+      "step": 35707
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3095132743362833e-06,
+      "loss": 1.1596,
+      "step": 35708
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3094658659924148e-06,
+      "loss": 1.8292,
+      "step": 35709
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.309418457648546e-06,
+      "loss": 1.6367,
+      "step": 35710
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3093710493046776e-06,
+      "loss": 2.273,
+      "step": 35711
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3093236409608093e-06,
+      "loss": 2.5701,
+      "step": 35712
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3092762326169405e-06,
+      "loss": 2.0023,
+      "step": 35713
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.309228824273072e-06,
+      "loss": 2.6864,
+      "step": 35714
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3091814159292037e-06,
+      "loss": 3.6008,
+      "step": 35715
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.309134007585335e-06,
+      "loss": 2.2612,
+      "step": 35716
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3090865992414665e-06,
+      "loss": 3.8646,
+      "step": 35717
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.309039190897598e-06,
+      "loss": 3.2334,
+      "step": 35718
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3089917825537295e-06,
+      "loss": 1.7719,
+      "step": 35719
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.308944374209861e-06,
+      "loss": 1.529,
+      "step": 35720
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3088969658659925e-06,
+      "loss": 1.5314,
+      "step": 35721
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.308849557522124e-06,
+      "loss": 2.1225,
+      "step": 35722
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3088021491782555e-06,
+      "loss": 1.6737,
+      "step": 35723
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3087547408343868e-06,
+      "loss": 1.1442,
+      "step": 35724
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3087073324905185e-06,
+      "loss": 2.0755,
+      "step": 35725
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.30865992414665e-06,
+      "loss": 2.6713,
+      "step": 35726
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3086125158027812e-06,
+      "loss": 2.1376,
+      "step": 35727
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.308565107458913e-06,
+      "loss": 1.1388,
+      "step": 35728
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3085176991150442e-06,
+      "loss": 2.5516,
+      "step": 35729
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3084702907711757e-06,
+      "loss": 1.6667,
+      "step": 35730
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3084228824273072e-06,
+      "loss": 3.1181,
+      "step": 35731
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3083754740834387e-06,
+      "loss": 1.6322,
+      "step": 35732
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3083280657395702e-06,
+      "loss": 2.2809,
+      "step": 35733
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3082806573957017e-06,
+      "loss": 1.4225,
+      "step": 35734
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3082332490518332e-06,
+      "loss": 2.4708,
+      "step": 35735
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3081858407079647e-06,
+      "loss": 2.2065,
+      "step": 35736
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3081384323640961e-06,
+      "loss": 2.1598,
+      "step": 35737
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3080910240202276e-06,
+      "loss": 2.7025,
+      "step": 35738
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3080436156763591e-06,
+      "loss": 1.8626,
+      "step": 35739
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3079962073324904e-06,
+      "loss": 2.1219,
+      "step": 35740
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3079487989886221e-06,
+      "loss": 2.5524,
+      "step": 35741
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3079013906447536e-06,
+      "loss": 1.4617,
+      "step": 35742
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3078539823008849e-06,
+      "loss": 2.0225,
+      "step": 35743
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3078065739570164e-06,
+      "loss": 2.3182,
+      "step": 35744
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.307759165613148e-06,
+      "loss": 3.1092,
+      "step": 35745
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3077117572692794e-06,
+      "loss": 1.6032,
+      "step": 35746
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3076643489254109e-06,
+      "loss": 2.2642,
+      "step": 35747
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3076169405815426e-06,
+      "loss": 1.6606,
+      "step": 35748
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 1.3075695322376738e-06,
+      "loss": 3.091,
+      "step": 35749
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3075221238938053e-06,
+      "loss": 2.0194,
+      "step": 35750
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3074747155499368e-06,
+      "loss": 1.7551,
+      "step": 35751
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3074273072060683e-06,
+      "loss": 1.0653,
+      "step": 35752
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3073798988621998e-06,
+      "loss": 0.8964,
+      "step": 35753
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3073324905183313e-06,
+      "loss": 1.1216,
+      "step": 35754
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3072850821744628e-06,
+      "loss": 1.4882,
+      "step": 35755
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3072376738305943e-06,
+      "loss": 2.3805,
+      "step": 35756
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3071902654867256e-06,
+      "loss": 2.0483,
+      "step": 35757
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3071428571428573e-06,
+      "loss": 3.2119,
+      "step": 35758
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3070954487989888e-06,
+      "loss": 1.5259,
+      "step": 35759
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.30704804045512e-06,
+      "loss": 2.8594,
+      "step": 35760
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3070006321112517e-06,
+      "loss": 2.3079,
+      "step": 35761
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306953223767383e-06,
+      "loss": 2.0322,
+      "step": 35762
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3069058154235145e-06,
+      "loss": 3.8273,
+      "step": 35763
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306858407079646e-06,
+      "loss": 2.9144,
+      "step": 35764
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3068109987357775e-06,
+      "loss": 2.6455,
+      "step": 35765
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306763590391909e-06,
+      "loss": 2.3764,
+      "step": 35766
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3067161820480405e-06,
+      "loss": 2.8186,
+      "step": 35767
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306668773704172e-06,
+      "loss": 2.3768,
+      "step": 35768
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3066213653603035e-06,
+      "loss": 2.0544,
+      "step": 35769
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306573957016435e-06,
+      "loss": 1.1031,
+      "step": 35770
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3065265486725665e-06,
+      "loss": 1.903,
+      "step": 35771
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306479140328698e-06,
+      "loss": 2.3347,
+      "step": 35772
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3064317319848292e-06,
+      "loss": 2.1392,
+      "step": 35773
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306384323640961e-06,
+      "loss": 2.6253,
+      "step": 35774
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3063369152970924e-06,
+      "loss": 3.0923,
+      "step": 35775
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3062895069532237e-06,
+      "loss": 1.5199,
+      "step": 35776
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3062420986093552e-06,
+      "loss": 3.5427,
+      "step": 35777
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.306194690265487e-06,
+      "loss": 3.3867,
+      "step": 35778
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3061472819216182e-06,
+      "loss": 2.4059,
+      "step": 35779
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3060998735777497e-06,
+      "loss": 3.0779,
+      "step": 35780
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3060524652338814e-06,
+      "loss": 1.8446,
+      "step": 35781
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3060050568900127e-06,
+      "loss": 1.9907,
+      "step": 35782
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3059576485461442e-06,
+      "loss": 2.1228,
+      "step": 35783
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3059102402022756e-06,
+      "loss": 1.6773,
+      "step": 35784
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3058628318584071e-06,
+      "loss": 2.0551,
+      "step": 35785
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3058154235145386e-06,
+      "loss": 2.6314,
+      "step": 35786
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.30576801517067e-06,
+      "loss": 0.6497,
+      "step": 35787
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3057206068268016e-06,
+      "loss": 2.1176,
+      "step": 35788
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3056731984829331e-06,
+      "loss": 2.1984,
+      "step": 35789
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3056257901390644e-06,
+      "loss": 2.0266,
+      "step": 35790
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.305578381795196e-06,
+      "loss": 0.6855,
+      "step": 35791
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3055309734513276e-06,
+      "loss": 2.8294,
+      "step": 35792
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3054835651074589e-06,
+      "loss": 2.4782,
+      "step": 35793
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3054361567635904e-06,
+      "loss": 2.01,
+      "step": 35794
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3053887484197218e-06,
+      "loss": 3.2302,
+      "step": 35795
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3053413400758533e-06,
+      "loss": 2.7838,
+      "step": 35796
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3052939317319848e-06,
+      "loss": 1.8723,
+      "step": 35797
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3052465233881163e-06,
+      "loss": 3.6073,
+      "step": 35798
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3051991150442478e-06,
+      "loss": 1.305,
+      "step": 35799
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3051517067003793e-06,
+      "loss": 2.1451,
+      "step": 35800
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3051042983565108e-06,
+      "loss": 3.3486,
+      "step": 35801
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3050568900126423e-06,
+      "loss": 2.8985,
+      "step": 35802
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3050094816687738e-06,
+      "loss": 2.4234,
+      "step": 35803
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3049620733249053e-06,
+      "loss": 2.1081,
+      "step": 35804
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3049146649810368e-06,
+      "loss": 1.1859,
+      "step": 35805
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.304867256637168e-06,
+      "loss": 2.2411,
+      "step": 35806
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3048198482932995e-06,
+      "loss": 1.8182,
+      "step": 35807
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3047724399494312e-06,
+      "loss": 2.4899,
+      "step": 35808
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3047250316055625e-06,
+      "loss": 3.3296,
+      "step": 35809
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.304677623261694e-06,
+      "loss": 2.3655,
+      "step": 35810
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3046302149178257e-06,
+      "loss": 2.7468,
+      "step": 35811
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.304582806573957e-06,
+      "loss": 3.2856,
+      "step": 35812
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 1.3045353982300885e-06,
+      "loss": 1.2965,
+      "step": 35813
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.30448798988622e-06,
+      "loss": 2.0625,
+      "step": 35814
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3044405815423515e-06,
+      "loss": 2.7889,
+      "step": 35815
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.304393173198483e-06,
+      "loss": 1.3876,
+      "step": 35816
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3043457648546145e-06,
+      "loss": 2.7075,
+      "step": 35817
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.304298356510746e-06,
+      "loss": 2.8891,
+      "step": 35818
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3042509481668774e-06,
+      "loss": 1.4745,
+      "step": 35819
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3042035398230087e-06,
+      "loss": 2.8401,
+      "step": 35820
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3041561314791404e-06,
+      "loss": 1.6019,
+      "step": 35821
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.304108723135272e-06,
+      "loss": 1.514,
+      "step": 35822
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3040613147914032e-06,
+      "loss": 1.8063,
+      "step": 35823
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.304013906447535e-06,
+      "loss": 2.2454,
+      "step": 35824
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3039664981036664e-06,
+      "loss": 0.9605,
+      "step": 35825
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3039190897597977e-06,
+      "loss": 0.9754,
+      "step": 35826
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3038716814159292e-06,
+      "loss": 1.846,
+      "step": 35827
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3038242730720607e-06,
+      "loss": 1.6583,
+      "step": 35828
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3037768647281922e-06,
+      "loss": 2.4106,
+      "step": 35829
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3037294563843237e-06,
+      "loss": 2.4748,
+      "step": 35830
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3036820480404551e-06,
+      "loss": 1.9969,
+      "step": 35831
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3036346396965866e-06,
+      "loss": 1.2804,
+      "step": 35832
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3035872313527181e-06,
+      "loss": 1.9231,
+      "step": 35833
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3035398230088496e-06,
+      "loss": 2.0095,
+      "step": 35834
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3034924146649811e-06,
+      "loss": 1.7262,
+      "step": 35835
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3034450063211126e-06,
+      "loss": 2.0983,
+      "step": 35836
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.303397597977244e-06,
+      "loss": 2.345,
+      "step": 35837
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3033501896333756e-06,
+      "loss": 1.6388,
+      "step": 35838
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3033027812895069e-06,
+      "loss": 1.0486,
+      "step": 35839
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3032553729456384e-06,
+      "loss": 2.758,
+      "step": 35840
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.30320796460177e-06,
+      "loss": 1.9857,
+      "step": 35841
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3031605562579013e-06,
+      "loss": 1.6036,
+      "step": 35842
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3031131479140328e-06,
+      "loss": 1.9247,
+      "step": 35843
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3030657395701645e-06,
+      "loss": 2.5247,
+      "step": 35844
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3030183312262958e-06,
+      "loss": 1.1803,
+      "step": 35845
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3029709228824273e-06,
+      "loss": 3.0227,
+      "step": 35846
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3029235145385588e-06,
+      "loss": 1.5868,
+      "step": 35847
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3028761061946903e-06,
+      "loss": 2.907,
+      "step": 35848
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3028286978508218e-06,
+      "loss": 1.5911,
+      "step": 35849
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3027812895069533e-06,
+      "loss": 1.7227,
+      "step": 35850
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3027338811630848e-06,
+      "loss": 1.711,
+      "step": 35851
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3026864728192163e-06,
+      "loss": 2.4384,
+      "step": 35852
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3026390644753475e-06,
+      "loss": 0.9838,
+      "step": 35853
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3025916561314793e-06,
+      "loss": 2.0288,
+      "step": 35854
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3025442477876107e-06,
+      "loss": 2.3176,
+      "step": 35855
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.302496839443742e-06,
+      "loss": 1.2392,
+      "step": 35856
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3024494310998737e-06,
+      "loss": 2.1401,
+      "step": 35857
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3024020227560052e-06,
+      "loss": 2.5684,
+      "step": 35858
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3023546144121365e-06,
+      "loss": 1.1523,
+      "step": 35859
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.302307206068268e-06,
+      "loss": 2.3757,
+      "step": 35860
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3022597977243995e-06,
+      "loss": 2.2082,
+      "step": 35861
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.302212389380531e-06,
+      "loss": 2.0863,
+      "step": 35862
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3021649810366625e-06,
+      "loss": 1.3284,
+      "step": 35863
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.302117572692794e-06,
+      "loss": 3.3827,
+      "step": 35864
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3020701643489255e-06,
+      "loss": 1.2189,
+      "step": 35865
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.302022756005057e-06,
+      "loss": 1.755,
+      "step": 35866
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3019753476611884e-06,
+      "loss": 3.0356,
+      "step": 35867
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.30192793931732e-06,
+      "loss": 3.5177,
+      "step": 35868
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3018805309734512e-06,
+      "loss": 1.11,
+      "step": 35869
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.301833122629583e-06,
+      "loss": 2.1177,
+      "step": 35870
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3017857142857144e-06,
+      "loss": 1.975,
+      "step": 35871
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3017383059418457e-06,
+      "loss": 2.1644,
+      "step": 35872
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3016908975979772e-06,
+      "loss": 1.7289,
+      "step": 35873
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3016434892541089e-06,
+      "loss": 1.8287,
+      "step": 35874
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3015960809102402e-06,
+      "loss": 1.2516,
+      "step": 35875
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 1.3015486725663717e-06,
+      "loss": 1.3536,
+      "step": 35876
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3015012642225034e-06,
+      "loss": 0.8943,
+      "step": 35877
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3014538558786346e-06,
+      "loss": 1.9844,
+      "step": 35878
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3014064475347661e-06,
+      "loss": 2.1602,
+      "step": 35879
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3013590391908976e-06,
+      "loss": 2.5731,
+      "step": 35880
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3013116308470291e-06,
+      "loss": 2.2365,
+      "step": 35881
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3012642225031606e-06,
+      "loss": 2.1952,
+      "step": 35882
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.301216814159292e-06,
+      "loss": 1.954,
+      "step": 35883
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3011694058154236e-06,
+      "loss": 2.4347,
+      "step": 35884
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.301121997471555e-06,
+      "loss": 1.4949,
+      "step": 35885
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3010745891276864e-06,
+      "loss": 1.8392,
+      "step": 35886
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.301027180783818e-06,
+      "loss": 2.5139,
+      "step": 35887
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3009797724399496e-06,
+      "loss": 1.5838,
+      "step": 35888
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3009323640960808e-06,
+      "loss": 2.6228,
+      "step": 35889
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3008849557522125e-06,
+      "loss": 0.8603,
+      "step": 35890
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3008375474083438e-06,
+      "loss": 2.9781,
+      "step": 35891
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3007901390644753e-06,
+      "loss": 2.411,
+      "step": 35892
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3007427307206068e-06,
+      "loss": 1.8009,
+      "step": 35893
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3006953223767383e-06,
+      "loss": 2.2586,
+      "step": 35894
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3006479140328698e-06,
+      "loss": 2.0583,
+      "step": 35895
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3006005056890013e-06,
+      "loss": 1.6393,
+      "step": 35896
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3005530973451328e-06,
+      "loss": 2.1069,
+      "step": 35897
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3005056890012643e-06,
+      "loss": 1.7789,
+      "step": 35898
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3004582806573958e-06,
+      "loss": 2.8414,
+      "step": 35899
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3004108723135273e-06,
+      "loss": 1.6869,
+      "step": 35900
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3003634639696587e-06,
+      "loss": 2.4706,
+      "step": 35901
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.30031605562579e-06,
+      "loss": 2.1207,
+      "step": 35902
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3002686472819217e-06,
+      "loss": 2.1002,
+      "step": 35903
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3002212389380532e-06,
+      "loss": 1.5316,
+      "step": 35904
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3001738305941845e-06,
+      "loss": 2.3235,
+      "step": 35905
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.300126422250316e-06,
+      "loss": 2.7705,
+      "step": 35906
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.3000790139064477e-06,
+      "loss": 1.5813,
+      "step": 35907
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.300031605562579e-06,
+      "loss": 1.5581,
+      "step": 35908
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2999841972187105e-06,
+      "loss": 1.584,
+      "step": 35909
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2999367888748422e-06,
+      "loss": 2.5394,
+      "step": 35910
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2998893805309735e-06,
+      "loss": 1.849,
+      "step": 35911
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.299841972187105e-06,
+      "loss": 2.3467,
+      "step": 35912
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2997945638432364e-06,
+      "loss": 3.4304,
+      "step": 35913
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.299747155499368e-06,
+      "loss": 3.3491,
+      "step": 35914
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2996997471554994e-06,
+      "loss": 1.7515,
+      "step": 35915
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.299652338811631e-06,
+      "loss": 2.2448,
+      "step": 35916
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2996049304677624e-06,
+      "loss": 1.3995,
+      "step": 35917
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.299557522123894e-06,
+      "loss": 1.32,
+      "step": 35918
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2995101137800252e-06,
+      "loss": 1.8282,
+      "step": 35919
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2994627054361569e-06,
+      "loss": 1.1845,
+      "step": 35920
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2994152970922884e-06,
+      "loss": 1.6268,
+      "step": 35921
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2993678887484197e-06,
+      "loss": 2.9006,
+      "step": 35922
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2993204804045514e-06,
+      "loss": 2.6875,
+      "step": 35923
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2992730720606826e-06,
+      "loss": 2.6916,
+      "step": 35924
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2992256637168141e-06,
+      "loss": 2.0376,
+      "step": 35925
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2991782553729456e-06,
+      "loss": 4.018,
+      "step": 35926
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2991308470290771e-06,
+      "loss": 2.6758,
+      "step": 35927
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2990834386852086e-06,
+      "loss": 2.6306,
+      "step": 35928
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.29903603034134e-06,
+      "loss": 3.1154,
+      "step": 35929
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2989886219974716e-06,
+      "loss": 2.9614,
+      "step": 35930
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.298941213653603e-06,
+      "loss": 2.3838,
+      "step": 35931
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2988938053097346e-06,
+      "loss": 2.3306,
+      "step": 35932
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.298846396965866e-06,
+      "loss": 2.228,
+      "step": 35933
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2987989886219976e-06,
+      "loss": 2.6157,
+      "step": 35934
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2987515802781288e-06,
+      "loss": 2.4768,
+      "step": 35935
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2987041719342603e-06,
+      "loss": 2.4089,
+      "step": 35936
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.298656763590392e-06,
+      "loss": 2.7625,
+      "step": 35937
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2986093552465233e-06,
+      "loss": 1.1014,
+      "step": 35938
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 1.2985619469026548e-06,
+      "loss": 1.3196,
+      "step": 35939
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2985145385587865e-06,
+      "loss": 1.2926,
+      "step": 35940
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2984671302149178e-06,
+      "loss": 2.6697,
+      "step": 35941
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2984197218710493e-06,
+      "loss": 1.8221,
+      "step": 35942
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2983723135271808e-06,
+      "loss": 3.0495,
+      "step": 35943
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2983249051833123e-06,
+      "loss": 2.2919,
+      "step": 35944
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2982774968394438e-06,
+      "loss": 1.848,
+      "step": 35945
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2982300884955753e-06,
+      "loss": 1.6958,
+      "step": 35946
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2981826801517068e-06,
+      "loss": 1.1391,
+      "step": 35947
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2981352718078382e-06,
+      "loss": 2.5672,
+      "step": 35948
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2980878634639695e-06,
+      "loss": 1.8392,
+      "step": 35949
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2980404551201012e-06,
+      "loss": 1.6798,
+      "step": 35950
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2979930467762327e-06,
+      "loss": 1.447,
+      "step": 35951
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.297945638432364e-06,
+      "loss": 1.0788,
+      "step": 35952
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2978982300884957e-06,
+      "loss": 1.2254,
+      "step": 35953
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2978508217446272e-06,
+      "loss": 1.3193,
+      "step": 35954
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2978034134007585e-06,
+      "loss": 0.8033,
+      "step": 35955
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.29775600505689e-06,
+      "loss": 3.1804,
+      "step": 35956
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2977085967130215e-06,
+      "loss": 2.1527,
+      "step": 35957
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.297661188369153e-06,
+      "loss": 2.2288,
+      "step": 35958
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2976137800252844e-06,
+      "loss": 2.421,
+      "step": 35959
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.297566371681416e-06,
+      "loss": 2.4819,
+      "step": 35960
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2975189633375474e-06,
+      "loss": 2.6405,
+      "step": 35961
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.297471554993679e-06,
+      "loss": 1.4221,
+      "step": 35962
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2974241466498104e-06,
+      "loss": 3.073,
+      "step": 35963
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.297376738305942e-06,
+      "loss": 1.4143,
+      "step": 35964
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2973293299620734e-06,
+      "loss": 2.4493,
+      "step": 35965
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2972819216182049e-06,
+      "loss": 2.0471,
+      "step": 35966
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2972345132743364e-06,
+      "loss": 1.9886,
+      "step": 35967
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2971871049304677e-06,
+      "loss": 1.3538,
+      "step": 35968
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2971396965865992e-06,
+      "loss": 2.1293,
+      "step": 35969
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2970922882427309e-06,
+      "loss": 2.2653,
+      "step": 35970
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2970448798988621e-06,
+      "loss": 1.8522,
+      "step": 35971
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2969974715549936e-06,
+      "loss": 2.5344,
+      "step": 35972
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2969500632111253e-06,
+      "loss": 2.7647,
+      "step": 35973
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2969026548672566e-06,
+      "loss": 2.3948,
+      "step": 35974
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2968552465233881e-06,
+      "loss": 2.0017,
+      "step": 35975
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2968078381795196e-06,
+      "loss": 1.4605,
+      "step": 35976
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.296760429835651e-06,
+      "loss": 2.5538,
+      "step": 35977
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2967130214917826e-06,
+      "loss": 2.945,
+      "step": 35978
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.296665613147914e-06,
+      "loss": 2.0458,
+      "step": 35979
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2966182048040456e-06,
+      "loss": 1.9365,
+      "step": 35980
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.296570796460177e-06,
+      "loss": 2.0568,
+      "step": 35981
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2965233881163083e-06,
+      "loss": 2.2415,
+      "step": 35982
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.29647597977244e-06,
+      "loss": 3.6639,
+      "step": 35983
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2964285714285715e-06,
+      "loss": 2.4263,
+      "step": 35984
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2963811630847028e-06,
+      "loss": 1.7199,
+      "step": 35985
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2963337547408345e-06,
+      "loss": 2.6005,
+      "step": 35986
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.296286346396966e-06,
+      "loss": 1.4906,
+      "step": 35987
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2962389380530973e-06,
+      "loss": 3.5608,
+      "step": 35988
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2961915297092288e-06,
+      "loss": 2.7038,
+      "step": 35989
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2961441213653603e-06,
+      "loss": 0.9023,
+      "step": 35990
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2960967130214918e-06,
+      "loss": 1.8108,
+      "step": 35991
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2960493046776233e-06,
+      "loss": 1.6072,
+      "step": 35992
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2960018963337548e-06,
+      "loss": 1.483,
+      "step": 35993
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2959544879898863e-06,
+      "loss": 2.3217,
+      "step": 35994
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2959070796460177e-06,
+      "loss": 1.8323,
+      "step": 35995
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2958596713021492e-06,
+      "loss": 1.6291,
+      "step": 35996
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2958122629582807e-06,
+      "loss": 1.2363,
+      "step": 35997
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2957648546144122e-06,
+      "loss": 2.6151,
+      "step": 35998
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2957174462705437e-06,
+      "loss": 2.9303,
+      "step": 35999
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2956700379266752e-06,
+      "loss": 1.2235,
+      "step": 36000
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2956226295828065e-06,
+      "loss": 2.1763,
+      "step": 36001
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.295575221238938e-06,
+      "loss": 1.115,
+      "step": 36002
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 1.2955278128950697e-06,
+      "loss": 1.159,
+      "step": 36003
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.295480404551201e-06,
+      "loss": 1.9851,
+      "step": 36004
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2954329962073325e-06,
+      "loss": 2.3045,
+      "step": 36005
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2953855878634642e-06,
+      "loss": 1.1552,
+      "step": 36006
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2953381795195954e-06,
+      "loss": 1.6121,
+      "step": 36007
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.295290771175727e-06,
+      "loss": 3.2382,
+      "step": 36008
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2952433628318584e-06,
+      "loss": 2.4751,
+      "step": 36009
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.29519595448799e-06,
+      "loss": 2.156,
+      "step": 36010
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2951485461441214e-06,
+      "loss": 2.2207,
+      "step": 36011
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.295101137800253e-06,
+      "loss": 3.4011,
+      "step": 36012
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2950537294563844e-06,
+      "loss": 1.917,
+      "step": 36013
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2950063211125159e-06,
+      "loss": 3.2476,
+      "step": 36014
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2949589127686472e-06,
+      "loss": 2.5305,
+      "step": 36015
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2949115044247789e-06,
+      "loss": 1.6478,
+      "step": 36016
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2948640960809104e-06,
+      "loss": 2.094,
+      "step": 36017
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2948166877370416e-06,
+      "loss": 3.2413,
+      "step": 36018
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2947692793931733e-06,
+      "loss": 0.8518,
+      "step": 36019
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2947218710493048e-06,
+      "loss": 3.404,
+      "step": 36020
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2946744627054361e-06,
+      "loss": 2.5554,
+      "step": 36021
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2946270543615676e-06,
+      "loss": 3.0727,
+      "step": 36022
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.294579646017699e-06,
+      "loss": 2.482,
+      "step": 36023
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2945322376738306e-06,
+      "loss": 1.3633,
+      "step": 36024
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.294484829329962e-06,
+      "loss": 2.058,
+      "step": 36025
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2944374209860936e-06,
+      "loss": 2.937,
+      "step": 36026
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.294390012642225e-06,
+      "loss": 3.6087,
+      "step": 36027
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2943426042983566e-06,
+      "loss": 1.6073,
+      "step": 36028
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.294295195954488e-06,
+      "loss": 1.4161,
+      "step": 36029
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2942477876106195e-06,
+      "loss": 2.2294,
+      "step": 36030
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.294200379266751e-06,
+      "loss": 2.5195,
+      "step": 36031
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2941529709228825e-06,
+      "loss": 2.2321,
+      "step": 36032
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.294105562579014e-06,
+      "loss": 2.5511,
+      "step": 36033
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2940581542351453e-06,
+      "loss": 1.5737,
+      "step": 36034
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2940107458912768e-06,
+      "loss": 2.0831,
+      "step": 36035
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2939633375474085e-06,
+      "loss": 2.6218,
+      "step": 36036
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2939159292035398e-06,
+      "loss": 1.0505,
+      "step": 36037
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2938685208596713e-06,
+      "loss": 2.8422,
+      "step": 36038
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.293821112515803e-06,
+      "loss": 1.5614,
+      "step": 36039
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2937737041719343e-06,
+      "loss": 2.4553,
+      "step": 36040
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2937262958280657e-06,
+      "loss": 2.2075,
+      "step": 36041
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2936788874841972e-06,
+      "loss": 2.3913,
+      "step": 36042
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2936314791403287e-06,
+      "loss": 1.356,
+      "step": 36043
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2935840707964602e-06,
+      "loss": 3.1369,
+      "step": 36044
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2935366624525917e-06,
+      "loss": 1.7181,
+      "step": 36045
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2934892541087232e-06,
+      "loss": 3.1997,
+      "step": 36046
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2934418457648547e-06,
+      "loss": 3.3775,
+      "step": 36047
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.293394437420986e-06,
+      "loss": 1.7226,
+      "step": 36048
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2933470290771177e-06,
+      "loss": 1.887,
+      "step": 36049
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2932996207332492e-06,
+      "loss": 1.6533,
+      "step": 36050
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2932522123893805e-06,
+      "loss": 2.7018,
+      "step": 36051
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2932048040455122e-06,
+      "loss": 1.9438,
+      "step": 36052
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2931573957016437e-06,
+      "loss": 1.6382,
+      "step": 36053
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.293109987357775e-06,
+      "loss": 1.4379,
+      "step": 36054
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2930625790139064e-06,
+      "loss": 0.7976,
+      "step": 36055
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.293015170670038e-06,
+      "loss": 3.0942,
+      "step": 36056
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2929677623261694e-06,
+      "loss": 2.5915,
+      "step": 36057
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.292920353982301e-06,
+      "loss": 1.082,
+      "step": 36058
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2928729456384324e-06,
+      "loss": 1.0516,
+      "step": 36059
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2928255372945639e-06,
+      "loss": 2.2873,
+      "step": 36060
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2927781289506954e-06,
+      "loss": 1.037,
+      "step": 36061
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2927307206068269e-06,
+      "loss": 2.0441,
+      "step": 36062
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2926833122629584e-06,
+      "loss": 3.0156,
+      "step": 36063
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2926359039190899e-06,
+      "loss": 1.2969,
+      "step": 36064
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2925884955752211e-06,
+      "loss": 1.0723,
+      "step": 36065
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 1.2925410872313528e-06,
+      "loss": 3.0265,
+      "step": 36066
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2924936788874841e-06,
+      "loss": 2.0304,
+      "step": 36067
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2924462705436156e-06,
+      "loss": 2.4938,
+      "step": 36068
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2923988621997473e-06,
+      "loss": 3.7618,
+      "step": 36069
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2923514538558786e-06,
+      "loss": 0.9853,
+      "step": 36070
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.29230404551201e-06,
+      "loss": 2.2429,
+      "step": 36071
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2922566371681416e-06,
+      "loss": 1.7001,
+      "step": 36072
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.292209228824273e-06,
+      "loss": 3.2082,
+      "step": 36073
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2921618204804046e-06,
+      "loss": 2.1169,
+      "step": 36074
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.292114412136536e-06,
+      "loss": 2.9907,
+      "step": 36075
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2920670037926675e-06,
+      "loss": 2.6814,
+      "step": 36076
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.292019595448799e-06,
+      "loss": 3.2185,
+      "step": 36077
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2919721871049303e-06,
+      "loss": 2.4237,
+      "step": 36078
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.291924778761062e-06,
+      "loss": 2.4532,
+      "step": 36079
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2918773704171935e-06,
+      "loss": 3.163,
+      "step": 36080
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2918299620733248e-06,
+      "loss": 2.6591,
+      "step": 36081
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2917825537294565e-06,
+      "loss": 2.3106,
+      "step": 36082
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.291735145385588e-06,
+      "loss": 1.0771,
+      "step": 36083
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2916877370417193e-06,
+      "loss": 3.5736,
+      "step": 36084
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2916403286978508e-06,
+      "loss": 1.626,
+      "step": 36085
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2915929203539825e-06,
+      "loss": 2.8041,
+      "step": 36086
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2915455120101138e-06,
+      "loss": 1.8691,
+      "step": 36087
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2914981036662452e-06,
+      "loss": 1.8257,
+      "step": 36088
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2914506953223767e-06,
+      "loss": 0.678,
+      "step": 36089
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2914032869785082e-06,
+      "loss": 2.2536,
+      "step": 36090
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2913558786346397e-06,
+      "loss": 2.8786,
+      "step": 36091
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2913084702907712e-06,
+      "loss": 1.857,
+      "step": 36092
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2912610619469027e-06,
+      "loss": 2.9186,
+      "step": 36093
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2912136536030342e-06,
+      "loss": 3.1226,
+      "step": 36094
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2911662452591657e-06,
+      "loss": 2.6184,
+      "step": 36095
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2911188369152972e-06,
+      "loss": 2.787,
+      "step": 36096
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2910714285714287e-06,
+      "loss": 0.9388,
+      "step": 36097
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.29102402022756e-06,
+      "loss": 2.563,
+      "step": 36098
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2909766118836917e-06,
+      "loss": 3.6335,
+      "step": 36099
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.290929203539823e-06,
+      "loss": 1.3701,
+      "step": 36100
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2908817951959544e-06,
+      "loss": 2.5213,
+      "step": 36101
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2908343868520861e-06,
+      "loss": 1.5186,
+      "step": 36102
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2907869785082174e-06,
+      "loss": 1.9481,
+      "step": 36103
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.290739570164349e-06,
+      "loss": 0.6343,
+      "step": 36104
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2906921618204804e-06,
+      "loss": 1.3729,
+      "step": 36105
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2906447534766119e-06,
+      "loss": 3.5966,
+      "step": 36106
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2905973451327434e-06,
+      "loss": 1.9464,
+      "step": 36107
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2905499367888749e-06,
+      "loss": 1.4985,
+      "step": 36108
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2905025284450064e-06,
+      "loss": 1.5764,
+      "step": 36109
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2904551201011379e-06,
+      "loss": 3.1718,
+      "step": 36110
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2904077117572691e-06,
+      "loss": 2.8307,
+      "step": 36111
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2903603034134008e-06,
+      "loss": 3.0491,
+      "step": 36112
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2903128950695323e-06,
+      "loss": 3.4325,
+      "step": 36113
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2902654867256636e-06,
+      "loss": 1.6479,
+      "step": 36114
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2902180783817953e-06,
+      "loss": 2.8706,
+      "step": 36115
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2901706700379268e-06,
+      "loss": 2.1288,
+      "step": 36116
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.290123261694058e-06,
+      "loss": 2.1612,
+      "step": 36117
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2900758533501896e-06,
+      "loss": 3.3936,
+      "step": 36118
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2900284450063213e-06,
+      "loss": 0.9606,
+      "step": 36119
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2899810366624526e-06,
+      "loss": 2.2922,
+      "step": 36120
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.289933628318584e-06,
+      "loss": 1.5685,
+      "step": 36121
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2898862199747156e-06,
+      "loss": 2.6541,
+      "step": 36122
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.289838811630847e-06,
+      "loss": 2.1912,
+      "step": 36123
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2897914032869785e-06,
+      "loss": 1.5944,
+      "step": 36124
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.28974399494311e-06,
+      "loss": 1.5202,
+      "step": 36125
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2896965865992415e-06,
+      "loss": 3.3754,
+      "step": 36126
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.289649178255373e-06,
+      "loss": 0.553,
+      "step": 36127
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.2896017699115045e-06,
+      "loss": 2.1535,
+      "step": 36128
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 1.289554361567636e-06,
+      "loss": 2.6923,
+      "step": 36129
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2895069532237675e-06,
+      "loss": 1.8907,
+      "step": 36130
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2894595448798988e-06,
+      "loss": 1.5388,
+      "step": 36131
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2894121365360305e-06,
+      "loss": 2.3311,
+      "step": 36132
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2893647281921618e-06,
+      "loss": 2.0034,
+      "step": 36133
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2893173198482932e-06,
+      "loss": 2.1742,
+      "step": 36134
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.289269911504425e-06,
+      "loss": 1.8045,
+      "step": 36135
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2892225031605562e-06,
+      "loss": 2.9328,
+      "step": 36136
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2891750948166877e-06,
+      "loss": 2.8214,
+      "step": 36137
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2891276864728192e-06,
+      "loss": 0.4708,
+      "step": 36138
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2890802781289507e-06,
+      "loss": 2.8629,
+      "step": 36139
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2890328697850822e-06,
+      "loss": 2.2159,
+      "step": 36140
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2889854614412137e-06,
+      "loss": 1.8605,
+      "step": 36141
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2889380530973452e-06,
+      "loss": 2.1889,
+      "step": 36142
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2888906447534767e-06,
+      "loss": 2.2575,
+      "step": 36143
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.288843236409608e-06,
+      "loss": 1.29,
+      "step": 36144
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2887958280657397e-06,
+      "loss": 3.2409,
+      "step": 36145
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2887484197218712e-06,
+      "loss": 2.8633,
+      "step": 36146
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2887010113780024e-06,
+      "loss": 2.1165,
+      "step": 36147
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2886536030341341e-06,
+      "loss": 2.043,
+      "step": 36148
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2886061946902656e-06,
+      "loss": 2.2071,
+      "step": 36149
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.288558786346397e-06,
+      "loss": 4.9657,
+      "step": 36150
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2885113780025284e-06,
+      "loss": 3.1405,
+      "step": 36151
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2884639696586601e-06,
+      "loss": 2.8369,
+      "step": 36152
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2884165613147914e-06,
+      "loss": 1.4198,
+      "step": 36153
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2883691529709229e-06,
+      "loss": 1.5145,
+      "step": 36154
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2883217446270544e-06,
+      "loss": 1.6302,
+      "step": 36155
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2882743362831859e-06,
+      "loss": 2.6033,
+      "step": 36156
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2882269279393174e-06,
+      "loss": 2.8237,
+      "step": 36157
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2881795195954488e-06,
+      "loss": 2.7333,
+      "step": 36158
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2881321112515803e-06,
+      "loss": 1.9181,
+      "step": 36159
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2880847029077118e-06,
+      "loss": 2.8399,
+      "step": 36160
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2880372945638433e-06,
+      "loss": 3.0719,
+      "step": 36161
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2879898862199748e-06,
+      "loss": 1.9168,
+      "step": 36162
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2879424778761063e-06,
+      "loss": 1.6398,
+      "step": 36163
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2878950695322376e-06,
+      "loss": 1.7466,
+      "step": 36164
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2878476611883693e-06,
+      "loss": 1.9088,
+      "step": 36165
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2878002528445006e-06,
+      "loss": 2.2444,
+      "step": 36166
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.287752844500632e-06,
+      "loss": 2.8379,
+      "step": 36167
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2877054361567638e-06,
+      "loss": 3.0129,
+      "step": 36168
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.287658027812895e-06,
+      "loss": 1.4567,
+      "step": 36169
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2876106194690265e-06,
+      "loss": 1.9891,
+      "step": 36170
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.287563211125158e-06,
+      "loss": 1.9656,
+      "step": 36171
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2875158027812895e-06,
+      "loss": 1.5326,
+      "step": 36172
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.287468394437421e-06,
+      "loss": 2.0014,
+      "step": 36173
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2874209860935525e-06,
+      "loss": 2.39,
+      "step": 36174
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.287373577749684e-06,
+      "loss": 2.256,
+      "step": 36175
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2873261694058155e-06,
+      "loss": 1.6368,
+      "step": 36176
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2872787610619468e-06,
+      "loss": 2.9546,
+      "step": 36177
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2872313527180785e-06,
+      "loss": 3.1252,
+      "step": 36178
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.28718394437421e-06,
+      "loss": 2.3369,
+      "step": 36179
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2871365360303413e-06,
+      "loss": 2.0156,
+      "step": 36180
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.287089127686473e-06,
+      "loss": 3.0145,
+      "step": 36181
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2870417193426044e-06,
+      "loss": 1.9664,
+      "step": 36182
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2869943109987357e-06,
+      "loss": 3.2728,
+      "step": 36183
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2869469026548672e-06,
+      "loss": 2.2143,
+      "step": 36184
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.286899494310999e-06,
+      "loss": 1.2119,
+      "step": 36185
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2868520859671302e-06,
+      "loss": 2.4454,
+      "step": 36186
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2868046776232617e-06,
+      "loss": 3.3855,
+      "step": 36187
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2867572692793932e-06,
+      "loss": 0.9551,
+      "step": 36188
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2867098609355247e-06,
+      "loss": 1.5291,
+      "step": 36189
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2866624525916562e-06,
+      "loss": 2.8948,
+      "step": 36190
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2866150442477877e-06,
+      "loss": 0.5728,
+      "step": 36191
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2865676359039192e-06,
+      "loss": 1.8645,
+      "step": 36192
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 1.2865202275600507e-06,
+      "loss": 0.9902,
+      "step": 36193
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2864728192161821e-06,
+      "loss": 1.5608,
+      "step": 36194
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2864254108723136e-06,
+      "loss": 2.7388,
+      "step": 36195
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2863780025284451e-06,
+      "loss": 2.3988,
+      "step": 36196
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2863305941845764e-06,
+      "loss": 2.9708,
+      "step": 36197
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2862831858407081e-06,
+      "loss": 3.0979,
+      "step": 36198
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2862357774968394e-06,
+      "loss": 1.6188,
+      "step": 36199
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2861883691529709e-06,
+      "loss": 2.5778,
+      "step": 36200
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2861409608091026e-06,
+      "loss": 1.6864,
+      "step": 36201
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2860935524652339e-06,
+      "loss": 1.3767,
+      "step": 36202
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2860461441213654e-06,
+      "loss": 2.8832,
+      "step": 36203
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2859987357774969e-06,
+      "loss": 2.8238,
+      "step": 36204
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2859513274336283e-06,
+      "loss": 2.9063,
+      "step": 36205
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2859039190897598e-06,
+      "loss": 2.8542,
+      "step": 36206
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2858565107458913e-06,
+      "loss": 3.5944,
+      "step": 36207
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2858091024020228e-06,
+      "loss": 0.6845,
+      "step": 36208
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2857616940581543e-06,
+      "loss": 2.6498,
+      "step": 36209
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2857142857142856e-06,
+      "loss": 1.3888,
+      "step": 36210
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2856668773704173e-06,
+      "loss": 4.411,
+      "step": 36211
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2856194690265488e-06,
+      "loss": 2.7115,
+      "step": 36212
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.28557206068268e-06,
+      "loss": 1.3414,
+      "step": 36213
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2855246523388116e-06,
+      "loss": 2.4261,
+      "step": 36214
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2854772439949433e-06,
+      "loss": 1.8019,
+      "step": 36215
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2854298356510745e-06,
+      "loss": 2.151,
+      "step": 36216
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.285382427307206e-06,
+      "loss": 2.7852,
+      "step": 36217
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2853350189633377e-06,
+      "loss": 1.3014,
+      "step": 36218
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.285287610619469e-06,
+      "loss": 0.9478,
+      "step": 36219
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2852402022756005e-06,
+      "loss": 1.5659,
+      "step": 36220
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.285192793931732e-06,
+      "loss": 2.344,
+      "step": 36221
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2851453855878635e-06,
+      "loss": 3.6915,
+      "step": 36222
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.285097977243995e-06,
+      "loss": 1.1426,
+      "step": 36223
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2850505689001265e-06,
+      "loss": 2.3145,
+      "step": 36224
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.285003160556258e-06,
+      "loss": 2.2117,
+      "step": 36225
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2849557522123895e-06,
+      "loss": 2.4215,
+      "step": 36226
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2849083438685208e-06,
+      "loss": 1.3714,
+      "step": 36227
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2848609355246525e-06,
+      "loss": 3.0447,
+      "step": 36228
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.284813527180784e-06,
+      "loss": 2.8527,
+      "step": 36229
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2847661188369152e-06,
+      "loss": 2.5337,
+      "step": 36230
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.284718710493047e-06,
+      "loss": 1.8837,
+      "step": 36231
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2846713021491782e-06,
+      "loss": 1.3034,
+      "step": 36232
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2846238938053097e-06,
+      "loss": 4.2163,
+      "step": 36233
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2845764854614412e-06,
+      "loss": 4.2943,
+      "step": 36234
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2845290771175727e-06,
+      "loss": 1.8966,
+      "step": 36235
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2844816687737042e-06,
+      "loss": 1.7091,
+      "step": 36236
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2844342604298357e-06,
+      "loss": 1.9171,
+      "step": 36237
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2843868520859672e-06,
+      "loss": 1.8638,
+      "step": 36238
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2843394437420987e-06,
+      "loss": 2.8574,
+      "step": 36239
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2842920353982301e-06,
+      "loss": 1.8682,
+      "step": 36240
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2842446270543616e-06,
+      "loss": 0.869,
+      "step": 36241
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2841972187104931e-06,
+      "loss": 1.209,
+      "step": 36242
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2841498103666244e-06,
+      "loss": 1.0873,
+      "step": 36243
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2841024020227561e-06,
+      "loss": 0.9197,
+      "step": 36244
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2840549936788876e-06,
+      "loss": 2.2857,
+      "step": 36245
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2840075853350189e-06,
+      "loss": 2.1356,
+      "step": 36246
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2839601769911504e-06,
+      "loss": 1.6307,
+      "step": 36247
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.283912768647282e-06,
+      "loss": 2.629,
+      "step": 36248
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2838653603034134e-06,
+      "loss": 2.7936,
+      "step": 36249
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2838179519595449e-06,
+      "loss": 3.1803,
+      "step": 36250
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2837705436156766e-06,
+      "loss": 1.9209,
+      "step": 36251
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2837231352718078e-06,
+      "loss": 1.0158,
+      "step": 36252
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2836757269279393e-06,
+      "loss": 2.8297,
+      "step": 36253
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2836283185840708e-06,
+      "loss": 1.9227,
+      "step": 36254
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2835809102402023e-06,
+      "loss": 2.4386,
+      "step": 36255
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 1.2835335018963338e-06,
+      "loss": 2.5459,
+      "step": 36256
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2834860935524653e-06,
+      "loss": 2.7121,
+      "step": 36257
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2834386852085968e-06,
+      "loss": 1.9426,
+      "step": 36258
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2833912768647283e-06,
+      "loss": 2.5241,
+      "step": 36259
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2833438685208596e-06,
+      "loss": 1.8592,
+      "step": 36260
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2832964601769913e-06,
+      "loss": 2.5598,
+      "step": 36261
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2832490518331228e-06,
+      "loss": 2.1925,
+      "step": 36262
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.283201643489254e-06,
+      "loss": 2.1882,
+      "step": 36263
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2831542351453857e-06,
+      "loss": 2.8923,
+      "step": 36264
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.283106826801517e-06,
+      "loss": 3.003,
+      "step": 36265
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2830594184576485e-06,
+      "loss": 0.5645,
+      "step": 36266
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.28301201011378e-06,
+      "loss": 1.663,
+      "step": 36267
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2829646017699115e-06,
+      "loss": 1.2316,
+      "step": 36268
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.282917193426043e-06,
+      "loss": 1.6818,
+      "step": 36269
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2828697850821745e-06,
+      "loss": 3.4992,
+      "step": 36270
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.282822376738306e-06,
+      "loss": 2.1446,
+      "step": 36271
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2827749683944375e-06,
+      "loss": 1.3248,
+      "step": 36272
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.282727560050569e-06,
+      "loss": 2.8274,
+      "step": 36273
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2826801517067005e-06,
+      "loss": 2.0553,
+      "step": 36274
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.282632743362832e-06,
+      "loss": 1.303,
+      "step": 36275
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2825853350189632e-06,
+      "loss": 2.0387,
+      "step": 36276
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.282537926675095e-06,
+      "loss": 4.0085,
+      "step": 36277
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2824905183312264e-06,
+      "loss": 1.7058,
+      "step": 36278
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2824431099873577e-06,
+      "loss": 2.1098,
+      "step": 36279
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2823957016434892e-06,
+      "loss": 1.2756,
+      "step": 36280
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.282348293299621e-06,
+      "loss": 0.8317,
+      "step": 36281
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2823008849557522e-06,
+      "loss": 1.4811,
+      "step": 36282
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2822534766118837e-06,
+      "loss": 2.2708,
+      "step": 36283
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2822060682680154e-06,
+      "loss": 3.5749,
+      "step": 36284
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2821586599241467e-06,
+      "loss": 2.4526,
+      "step": 36285
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2821112515802782e-06,
+      "loss": 1.6241,
+      "step": 36286
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2820638432364096e-06,
+      "loss": 1.6435,
+      "step": 36287
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2820164348925411e-06,
+      "loss": 1.2486,
+      "step": 36288
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2819690265486726e-06,
+      "loss": 1.7641,
+      "step": 36289
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2819216182048041e-06,
+      "loss": 2.5363,
+      "step": 36290
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2818742098609356e-06,
+      "loss": 1.4659,
+      "step": 36291
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2818268015170671e-06,
+      "loss": 1.714,
+      "step": 36292
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2817793931731984e-06,
+      "loss": 3.3571,
+      "step": 36293
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.28173198482933e-06,
+      "loss": 2.2167,
+      "step": 36294
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2816845764854616e-06,
+      "loss": 2.4565,
+      "step": 36295
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2816371681415929e-06,
+      "loss": 2.8728,
+      "step": 36296
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2815897597977246e-06,
+      "loss": 2.6707,
+      "step": 36297
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2815423514538558e-06,
+      "loss": 2.2621,
+      "step": 36298
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2814949431099873e-06,
+      "loss": 2.9249,
+      "step": 36299
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2814475347661188e-06,
+      "loss": 2.1915,
+      "step": 36300
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2814001264222503e-06,
+      "loss": 2.4918,
+      "step": 36301
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2813527180783818e-06,
+      "loss": 2.6817,
+      "step": 36302
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2813053097345133e-06,
+      "loss": 2.7206,
+      "step": 36303
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2812579013906448e-06,
+      "loss": 2.6492,
+      "step": 36304
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2812104930467763e-06,
+      "loss": 3.4379,
+      "step": 36305
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2811630847029076e-06,
+      "loss": 1.2682,
+      "step": 36306
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2811156763590393e-06,
+      "loss": 1.3448,
+      "step": 36307
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2810682680151708e-06,
+      "loss": 3.5978,
+      "step": 36308
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.281020859671302e-06,
+      "loss": 2.4013,
+      "step": 36309
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2809734513274338e-06,
+      "loss": 1.5673,
+      "step": 36310
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2809260429835652e-06,
+      "loss": 1.881,
+      "step": 36311
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2808786346396965e-06,
+      "loss": 3.0575,
+      "step": 36312
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.280831226295828e-06,
+      "loss": 1.28,
+      "step": 36313
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2807838179519597e-06,
+      "loss": 2.3135,
+      "step": 36314
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.280736409608091e-06,
+      "loss": 2.9061,
+      "step": 36315
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2806890012642225e-06,
+      "loss": 2.7546,
+      "step": 36316
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.280641592920354e-06,
+      "loss": 1.8749,
+      "step": 36317
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.2805941845764855e-06,
+      "loss": 1.3267,
+      "step": 36318
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 1.280546776232617e-06,
+      "loss": 0.9224,
+      "step": 36319
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2804993678887485e-06,
+      "loss": 2.3287,
+      "step": 36320
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.28045195954488e-06,
+      "loss": 3.158,
+      "step": 36321
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2804045512010114e-06,
+      "loss": 1.083,
+      "step": 36322
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.280357142857143e-06,
+      "loss": 2.1215,
+      "step": 36323
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2803097345132744e-06,
+      "loss": 1.883,
+      "step": 36324
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.280262326169406e-06,
+      "loss": 1.4333,
+      "step": 36325
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2802149178255372e-06,
+      "loss": 5.0685,
+      "step": 36326
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.280167509481669e-06,
+      "loss": 3.1022,
+      "step": 36327
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2801201011378002e-06,
+      "loss": 1.9836,
+      "step": 36328
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2800726927939317e-06,
+      "loss": 1.9053,
+      "step": 36329
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2800252844500634e-06,
+      "loss": 3.0186,
+      "step": 36330
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2799778761061947e-06,
+      "loss": 1.2987,
+      "step": 36331
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2799304677623262e-06,
+      "loss": 2.2314,
+      "step": 36332
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2798830594184577e-06,
+      "loss": 2.7167,
+      "step": 36333
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2798356510745891e-06,
+      "loss": 1.8458,
+      "step": 36334
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2797882427307206e-06,
+      "loss": 1.4453,
+      "step": 36335
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2797408343868521e-06,
+      "loss": 1.1979,
+      "step": 36336
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2796934260429836e-06,
+      "loss": 2.853,
+      "step": 36337
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2796460176991151e-06,
+      "loss": 1.7481,
+      "step": 36338
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2795986093552464e-06,
+      "loss": 2.0612,
+      "step": 36339
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.279551201011378e-06,
+      "loss": 2.1979,
+      "step": 36340
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2795037926675096e-06,
+      "loss": 1.2073,
+      "step": 36341
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2794563843236409e-06,
+      "loss": 3.7394,
+      "step": 36342
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2794089759797724e-06,
+      "loss": 2.8107,
+      "step": 36343
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.279361567635904e-06,
+      "loss": 2.2388,
+      "step": 36344
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2793141592920353e-06,
+      "loss": 2.4818,
+      "step": 36345
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2792667509481668e-06,
+      "loss": 0.4423,
+      "step": 36346
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2792193426042985e-06,
+      "loss": 2.3121,
+      "step": 36347
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2791719342604298e-06,
+      "loss": 2.0808,
+      "step": 36348
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2791245259165613e-06,
+      "loss": 2.2083,
+      "step": 36349
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2790771175726928e-06,
+      "loss": 1.7834,
+      "step": 36350
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2790297092288243e-06,
+      "loss": 4.5122,
+      "step": 36351
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2789823008849558e-06,
+      "loss": 3.0636,
+      "step": 36352
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2789348925410873e-06,
+      "loss": 0.3735,
+      "step": 36353
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2788874841972188e-06,
+      "loss": 3.2054,
+      "step": 36354
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2788400758533503e-06,
+      "loss": 1.8318,
+      "step": 36355
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2787926675094815e-06,
+      "loss": 2.5809,
+      "step": 36356
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2787452591656133e-06,
+      "loss": 2.8018,
+      "step": 36357
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2786978508217447e-06,
+      "loss": 1.6572,
+      "step": 36358
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278650442477876e-06,
+      "loss": 0.8693,
+      "step": 36359
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2786030341340077e-06,
+      "loss": 1.2309,
+      "step": 36360
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278555625790139e-06,
+      "loss": 0.8781,
+      "step": 36361
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2785082174462705e-06,
+      "loss": 2.4033,
+      "step": 36362
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278460809102402e-06,
+      "loss": 1.8829,
+      "step": 36363
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2784134007585335e-06,
+      "loss": 2.5786,
+      "step": 36364
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278365992414665e-06,
+      "loss": 3.0421,
+      "step": 36365
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2783185840707965e-06,
+      "loss": 3.5099,
+      "step": 36366
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278271175726928e-06,
+      "loss": 3.0131,
+      "step": 36367
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2782237673830595e-06,
+      "loss": 1.8229,
+      "step": 36368
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278176359039191e-06,
+      "loss": 2.5189,
+      "step": 36369
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2781289506953224e-06,
+      "loss": 1.8297,
+      "step": 36370
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.278081542351454e-06,
+      "loss": 2.7157,
+      "step": 36371
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2780341340075852e-06,
+      "loss": 2.7391,
+      "step": 36372
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.277986725663717e-06,
+      "loss": 2.9478,
+      "step": 36373
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2779393173198484e-06,
+      "loss": 1.3427,
+      "step": 36374
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2778919089759797e-06,
+      "loss": 2.9129,
+      "step": 36375
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2778445006321112e-06,
+      "loss": 2.1233,
+      "step": 36376
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2777970922882429e-06,
+      "loss": 2.3028,
+      "step": 36377
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2777496839443742e-06,
+      "loss": 2.4566,
+      "step": 36378
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2777022756005057e-06,
+      "loss": 1.851,
+      "step": 36379
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2776548672566374e-06,
+      "loss": 1.9031,
+      "step": 36380
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2776074589127686e-06,
+      "loss": 1.8385,
+      "step": 36381
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2775600505689001e-06,
+      "loss": 2.9258,
+      "step": 36382
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 1.2775126422250316e-06,
+      "loss": 3.1027,
+      "step": 36383
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2774652338811631e-06,
+      "loss": 2.0925,
+      "step": 36384
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2774178255372946e-06,
+      "loss": 1.5619,
+      "step": 36385
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.277370417193426e-06,
+      "loss": 1.7086,
+      "step": 36386
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2773230088495576e-06,
+      "loss": 2.6272,
+      "step": 36387
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.277275600505689e-06,
+      "loss": 1.6178,
+      "step": 36388
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2772281921618204e-06,
+      "loss": 2.3358,
+      "step": 36389
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.277180783817952e-06,
+      "loss": 1.7176,
+      "step": 36390
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2771333754740836e-06,
+      "loss": 1.2952,
+      "step": 36391
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2770859671302148e-06,
+      "loss": 1.8576,
+      "step": 36392
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2770385587863465e-06,
+      "loss": 1.8083,
+      "step": 36393
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2769911504424778e-06,
+      "loss": 1.7014,
+      "step": 36394
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2769437420986093e-06,
+      "loss": 1.5127,
+      "step": 36395
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2768963337547408e-06,
+      "loss": 2.1885,
+      "step": 36396
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2768489254108723e-06,
+      "loss": 3.0433,
+      "step": 36397
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2768015170670038e-06,
+      "loss": 3.0207,
+      "step": 36398
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2767541087231353e-06,
+      "loss": 1.9148,
+      "step": 36399
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2767067003792668e-06,
+      "loss": 1.7472,
+      "step": 36400
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2766592920353983e-06,
+      "loss": 1.006,
+      "step": 36401
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2766118836915298e-06,
+      "loss": 1.7714,
+      "step": 36402
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2765644753476613e-06,
+      "loss": 2.394,
+      "step": 36403
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2765170670037927e-06,
+      "loss": 1.7697,
+      "step": 36404
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.276469658659924e-06,
+      "loss": 1.1952,
+      "step": 36405
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2764222503160557e-06,
+      "loss": 2.5014,
+      "step": 36406
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2763748419721872e-06,
+      "loss": 2.7255,
+      "step": 36407
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2763274336283185e-06,
+      "loss": 4.7215,
+      "step": 36408
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.27628002528445e-06,
+      "loss": 0.921,
+      "step": 36409
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2762326169405817e-06,
+      "loss": 2.2843,
+      "step": 36410
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.276185208596713e-06,
+      "loss": 3.3763,
+      "step": 36411
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2761378002528445e-06,
+      "loss": 1.8908,
+      "step": 36412
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2760903919089762e-06,
+      "loss": 1.9539,
+      "step": 36413
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2760429835651075e-06,
+      "loss": 3.2129,
+      "step": 36414
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.275995575221239e-06,
+      "loss": 2.8669,
+      "step": 36415
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2759481668773704e-06,
+      "loss": 2.9634,
+      "step": 36416
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.275900758533502e-06,
+      "loss": 4.1331,
+      "step": 36417
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2758533501896334e-06,
+      "loss": 3.4315,
+      "step": 36418
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.275805941845765e-06,
+      "loss": 1.3865,
+      "step": 36419
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2757585335018964e-06,
+      "loss": 1.7062,
+      "step": 36420
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.275711125158028e-06,
+      "loss": 1.0215,
+      "step": 36421
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2756637168141592e-06,
+      "loss": 1.9698,
+      "step": 36422
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2756163084702909e-06,
+      "loss": 2.361,
+      "step": 36423
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2755689001264224e-06,
+      "loss": 2.3937,
+      "step": 36424
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2755214917825537e-06,
+      "loss": 2.5173,
+      "step": 36425
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2754740834386854e-06,
+      "loss": 1.6201,
+      "step": 36426
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2754266750948166e-06,
+      "loss": 1.4101,
+      "step": 36427
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2753792667509481e-06,
+      "loss": 1.4183,
+      "step": 36428
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2753318584070796e-06,
+      "loss": 1.0688,
+      "step": 36429
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2752844500632111e-06,
+      "loss": 2.288,
+      "step": 36430
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2752370417193426e-06,
+      "loss": 2.2406,
+      "step": 36431
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2751896333754741e-06,
+      "loss": 3.0675,
+      "step": 36432
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2751422250316056e-06,
+      "loss": 2.2065,
+      "step": 36433
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.275094816687737e-06,
+      "loss": 1.8991,
+      "step": 36434
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2750474083438686e-06,
+      "loss": 1.9676,
+      "step": 36435
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.275e-06,
+      "loss": 0.6059,
+      "step": 36436
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2749525916561316e-06,
+      "loss": 0.6729,
+      "step": 36437
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2749051833122628e-06,
+      "loss": 3.5022,
+      "step": 36438
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2748577749683946e-06,
+      "loss": 1.9551,
+      "step": 36439
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.274810366624526e-06,
+      "loss": 1.2561,
+      "step": 36440
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2747629582806573e-06,
+      "loss": 2.9836,
+      "step": 36441
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2747155499367888e-06,
+      "loss": 1.4358,
+      "step": 36442
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2746681415929205e-06,
+      "loss": 3.3633,
+      "step": 36443
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2746207332490518e-06,
+      "loss": 2.0741,
+      "step": 36444
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.2745733249051833e-06,
+      "loss": 1.7414,
+      "step": 36445
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 1.274525916561315e-06,
+      "loss": 2.2688,
+      "step": 36446
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2744785082174463e-06,
+      "loss": 0.8575,
+      "step": 36447
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2744310998735778e-06,
+      "loss": 2.1711,
+      "step": 36448
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2743836915297093e-06,
+      "loss": 1.3416,
+      "step": 36449
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2743362831858408e-06,
+      "loss": 2.6991,
+      "step": 36450
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2742888748419722e-06,
+      "loss": 1.1286,
+      "step": 36451
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2742414664981037e-06,
+      "loss": 1.6309,
+      "step": 36452
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2741940581542352e-06,
+      "loss": 1.5782,
+      "step": 36453
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2741466498103667e-06,
+      "loss": 1.3083,
+      "step": 36454
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.274099241466498e-06,
+      "loss": 2.4093,
+      "step": 36455
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2740518331226297e-06,
+      "loss": 2.0909,
+      "step": 36456
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2740044247787612e-06,
+      "loss": 1.4641,
+      "step": 36457
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2739570164348925e-06,
+      "loss": 0.8154,
+      "step": 36458
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2739096080910242e-06,
+      "loss": 0.9116,
+      "step": 36459
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2738621997471555e-06,
+      "loss": 3.5057,
+      "step": 36460
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.273814791403287e-06,
+      "loss": 1.7404,
+      "step": 36461
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2737673830594184e-06,
+      "loss": 1.5678,
+      "step": 36462
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.27371997471555e-06,
+      "loss": 1.9583,
+      "step": 36463
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2736725663716814e-06,
+      "loss": 2.2689,
+      "step": 36464
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.273625158027813e-06,
+      "loss": 3.2901,
+      "step": 36465
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2735777496839444e-06,
+      "loss": 1.985,
+      "step": 36466
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.273530341340076e-06,
+      "loss": 1.7376,
+      "step": 36467
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2734829329962074e-06,
+      "loss": 2.3076,
+      "step": 36468
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2734355246523389e-06,
+      "loss": 1.9519,
+      "step": 36469
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2733881163084704e-06,
+      "loss": 1.0247,
+      "step": 36470
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2733407079646017e-06,
+      "loss": 1.9376,
+      "step": 36471
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2732932996207334e-06,
+      "loss": 3.237,
+      "step": 36472
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2732458912768649e-06,
+      "loss": 3.4703,
+      "step": 36473
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2731984829329961e-06,
+      "loss": 2.6812,
+      "step": 36474
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2731510745891276e-06,
+      "loss": 3.2416,
+      "step": 36475
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2731036662452593e-06,
+      "loss": 2.3312,
+      "step": 36476
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2730562579013906e-06,
+      "loss": 1.0587,
+      "step": 36477
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2730088495575221e-06,
+      "loss": 2.6811,
+      "step": 36478
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2729614412136538e-06,
+      "loss": 2.2514,
+      "step": 36479
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.272914032869785e-06,
+      "loss": 0.6818,
+      "step": 36480
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2728666245259166e-06,
+      "loss": 2.8685,
+      "step": 36481
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.272819216182048e-06,
+      "loss": 1.5176,
+      "step": 36482
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2727718078381796e-06,
+      "loss": 2.8344,
+      "step": 36483
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.272724399494311e-06,
+      "loss": 2.2029,
+      "step": 36484
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2726769911504423e-06,
+      "loss": 1.8369,
+      "step": 36485
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.272629582806574e-06,
+      "loss": 1.7312,
+      "step": 36486
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2725821744627055e-06,
+      "loss": 2.9651,
+      "step": 36487
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2725347661188368e-06,
+      "loss": 1.5742,
+      "step": 36488
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2724873577749685e-06,
+      "loss": 1.995,
+      "step": 36489
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2724399494311e-06,
+      "loss": 1.8359,
+      "step": 36490
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2723925410872313e-06,
+      "loss": 3.0939,
+      "step": 36491
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2723451327433628e-06,
+      "loss": 4.2525,
+      "step": 36492
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2722977243994943e-06,
+      "loss": 2.3103,
+      "step": 36493
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2722503160556258e-06,
+      "loss": 1.3845,
+      "step": 36494
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2722029077117573e-06,
+      "loss": 1.8774,
+      "step": 36495
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2721554993678888e-06,
+      "loss": 0.7991,
+      "step": 36496
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2721080910240203e-06,
+      "loss": 2.1774,
+      "step": 36497
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2720606826801517e-06,
+      "loss": 2.3624,
+      "step": 36498
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2720132743362832e-06,
+      "loss": 2.0535,
+      "step": 36499
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2719658659924147e-06,
+      "loss": 1.9833,
+      "step": 36500
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2719184576485462e-06,
+      "loss": 2.4806,
+      "step": 36501
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2718710493046777e-06,
+      "loss": 3.8663,
+      "step": 36502
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2718236409608092e-06,
+      "loss": 1.4187,
+      "step": 36503
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2717762326169405e-06,
+      "loss": 1.0532,
+      "step": 36504
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.271728824273072e-06,
+      "loss": 4.1574,
+      "step": 36505
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2716814159292037e-06,
+      "loss": 2.5935,
+      "step": 36506
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.271634007585335e-06,
+      "loss": 2.7455,
+      "step": 36507
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2715865992414665e-06,
+      "loss": 1.5522,
+      "step": 36508
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 1.2715391908975982e-06,
+      "loss": 1.8326,
+      "step": 36509
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2714917825537294e-06,
+      "loss": 2.6676,
+      "step": 36510
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.271444374209861e-06,
+      "loss": 2.7142,
+      "step": 36511
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2713969658659924e-06,
+      "loss": 2.062,
+      "step": 36512
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.271349557522124e-06,
+      "loss": 1.6834,
+      "step": 36513
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2713021491782554e-06,
+      "loss": 2.5299,
+      "step": 36514
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.271254740834387e-06,
+      "loss": 1.7874,
+      "step": 36515
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2712073324905184e-06,
+      "loss": 2.1094,
+      "step": 36516
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2711599241466499e-06,
+      "loss": 1.071,
+      "step": 36517
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2711125158027812e-06,
+      "loss": 3.0075,
+      "step": 36518
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2710651074589129e-06,
+      "loss": 1.8273,
+      "step": 36519
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2710176991150444e-06,
+      "loss": 2.0916,
+      "step": 36520
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2709702907711756e-06,
+      "loss": 2.8592,
+      "step": 36521
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2709228824273073e-06,
+      "loss": 3.0784,
+      "step": 36522
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2708754740834388e-06,
+      "loss": 2.105,
+      "step": 36523
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2708280657395701e-06,
+      "loss": 2.0667,
+      "step": 36524
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2707806573957016e-06,
+      "loss": 3.055,
+      "step": 36525
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.270733249051833e-06,
+      "loss": 2.195,
+      "step": 36526
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2706858407079646e-06,
+      "loss": 2.3863,
+      "step": 36527
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.270638432364096e-06,
+      "loss": 3.7013,
+      "step": 36528
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2705910240202276e-06,
+      "loss": 3.0426,
+      "step": 36529
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.270543615676359e-06,
+      "loss": 2.8674,
+      "step": 36530
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2704962073324906e-06,
+      "loss": 2.5327,
+      "step": 36531
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.270448798988622e-06,
+      "loss": 1.4816,
+      "step": 36532
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2704013906447535e-06,
+      "loss": 3.3756,
+      "step": 36533
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.270353982300885e-06,
+      "loss": 1.8403,
+      "step": 36534
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2703065739570165e-06,
+      "loss": 1.8983,
+      "step": 36535
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.270259165613148e-06,
+      "loss": 1.2521,
+      "step": 36536
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2702117572692793e-06,
+      "loss": 2.6322,
+      "step": 36537
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2701643489254108e-06,
+      "loss": 2.3004,
+      "step": 36538
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2701169405815425e-06,
+      "loss": 3.4237,
+      "step": 36539
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2700695322376738e-06,
+      "loss": 2.6489,
+      "step": 36540
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2700221238938053e-06,
+      "loss": 1.8427,
+      "step": 36541
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.269974715549937e-06,
+      "loss": 2.9485,
+      "step": 36542
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2699273072060683e-06,
+      "loss": 1.4963,
+      "step": 36543
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2698798988621997e-06,
+      "loss": 1.7695,
+      "step": 36544
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2698324905183312e-06,
+      "loss": 2.9512,
+      "step": 36545
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2697850821744627e-06,
+      "loss": 0.8229,
+      "step": 36546
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2697376738305942e-06,
+      "loss": 3.2021,
+      "step": 36547
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2696902654867257e-06,
+      "loss": 2.3339,
+      "step": 36548
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2696428571428572e-06,
+      "loss": 0.8512,
+      "step": 36549
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2695954487989887e-06,
+      "loss": 2.2188,
+      "step": 36550
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.26954804045512e-06,
+      "loss": 2.4083,
+      "step": 36551
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2695006321112517e-06,
+      "loss": 2.1346,
+      "step": 36552
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2694532237673832e-06,
+      "loss": 2.5688,
+      "step": 36553
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2694058154235145e-06,
+      "loss": 2.1796,
+      "step": 36554
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2693584070796462e-06,
+      "loss": 1.3396,
+      "step": 36555
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2693109987357777e-06,
+      "loss": 1.0893,
+      "step": 36556
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.269263590391909e-06,
+      "loss": 1.5841,
+      "step": 36557
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2692161820480404e-06,
+      "loss": 2.386,
+      "step": 36558
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.269168773704172e-06,
+      "loss": 1.3997,
+      "step": 36559
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2691213653603034e-06,
+      "loss": 2.0124,
+      "step": 36560
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.269073957016435e-06,
+      "loss": 3.4739,
+      "step": 36561
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2690265486725664e-06,
+      "loss": 1.1898,
+      "step": 36562
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2689791403286979e-06,
+      "loss": 1.6909,
+      "step": 36563
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2689317319848294e-06,
+      "loss": 3.1156,
+      "step": 36564
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2688843236409609e-06,
+      "loss": 2.5986,
+      "step": 36565
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2688369152970924e-06,
+      "loss": 2.5802,
+      "step": 36566
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2687895069532239e-06,
+      "loss": 2.4955,
+      "step": 36567
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2687420986093553e-06,
+      "loss": 4.3611,
+      "step": 36568
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2686946902654868e-06,
+      "loss": 2.6475,
+      "step": 36569
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2686472819216181e-06,
+      "loss": 3.3348,
+      "step": 36570
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2685998735777496e-06,
+      "loss": 3.1502,
+      "step": 36571
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2685524652338813e-06,
+      "loss": 2.2294,
+      "step": 36572
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 1.2685050568900126e-06,
+      "loss": 1.3726,
+      "step": 36573
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.268457648546144e-06,
+      "loss": 1.9184,
+      "step": 36574
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2684102402022758e-06,
+      "loss": 1.9355,
+      "step": 36575
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.268362831858407e-06,
+      "loss": 2.7385,
+      "step": 36576
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2683154235145386e-06,
+      "loss": 0.3465,
+      "step": 36577
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.26826801517067e-06,
+      "loss": 2.7532,
+      "step": 36578
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2682206068268016e-06,
+      "loss": 2.132,
+      "step": 36579
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.268173198482933e-06,
+      "loss": 3.2733,
+      "step": 36580
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2681257901390645e-06,
+      "loss": 2.7504,
+      "step": 36581
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.268078381795196e-06,
+      "loss": 2.0728,
+      "step": 36582
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2680309734513275e-06,
+      "loss": 3.3888,
+      "step": 36583
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2679835651074588e-06,
+      "loss": 1.9176,
+      "step": 36584
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2679361567635905e-06,
+      "loss": 2.427,
+      "step": 36585
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.267888748419722e-06,
+      "loss": 2.0378,
+      "step": 36586
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2678413400758533e-06,
+      "loss": 2.2069,
+      "step": 36587
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.267793931731985e-06,
+      "loss": 2.7151,
+      "step": 36588
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2677465233881165e-06,
+      "loss": 1.7929,
+      "step": 36589
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2676991150442478e-06,
+      "loss": 2.0362,
+      "step": 36590
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2676517067003792e-06,
+      "loss": 0.9027,
+      "step": 36591
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2676042983565107e-06,
+      "loss": 3.7637,
+      "step": 36592
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2675568900126422e-06,
+      "loss": 2.8211,
+      "step": 36593
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2675094816687737e-06,
+      "loss": 1.7037,
+      "step": 36594
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2674620733249052e-06,
+      "loss": 1.7778,
+      "step": 36595
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2674146649810367e-06,
+      "loss": 1.7273,
+      "step": 36596
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2673672566371682e-06,
+      "loss": 1.3814,
+      "step": 36597
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2673198482932997e-06,
+      "loss": 1.7881,
+      "step": 36598
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2672724399494312e-06,
+      "loss": 0.8204,
+      "step": 36599
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2672250316055627e-06,
+      "loss": 0.8464,
+      "step": 36600
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2671776232616942e-06,
+      "loss": 1.7184,
+      "step": 36601
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2671302149178257e-06,
+      "loss": 4.2255,
+      "step": 36602
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.267082806573957e-06,
+      "loss": 2.2994,
+      "step": 36603
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2670353982300884e-06,
+      "loss": 3.0837,
+      "step": 36604
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2669879898862201e-06,
+      "loss": 1.2003,
+      "step": 36605
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2669405815423514e-06,
+      "loss": 1.9615,
+      "step": 36606
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.266893173198483e-06,
+      "loss": 3.7544,
+      "step": 36607
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2668457648546146e-06,
+      "loss": 2.5245,
+      "step": 36608
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2667983565107459e-06,
+      "loss": 2.3085,
+      "step": 36609
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2667509481668774e-06,
+      "loss": 1.4905,
+      "step": 36610
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2667035398230089e-06,
+      "loss": 1.4171,
+      "step": 36611
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2666561314791404e-06,
+      "loss": 1.6257,
+      "step": 36612
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2666087231352719e-06,
+      "loss": 1.2753,
+      "step": 36613
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2665613147914031e-06,
+      "loss": 1.7636,
+      "step": 36614
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2665139064475348e-06,
+      "loss": 2.9845,
+      "step": 36615
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2664664981036663e-06,
+      "loss": 1.9221,
+      "step": 36616
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2664190897597976e-06,
+      "loss": 3.3233,
+      "step": 36617
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2663716814159293e-06,
+      "loss": 2.8734,
+      "step": 36618
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2663242730720608e-06,
+      "loss": 3.804,
+      "step": 36619
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.266276864728192e-06,
+      "loss": 1.7427,
+      "step": 36620
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2662294563843236e-06,
+      "loss": 1.7603,
+      "step": 36621
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2661820480404553e-06,
+      "loss": 1.3982,
+      "step": 36622
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2661346396965866e-06,
+      "loss": 1.6042,
+      "step": 36623
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.266087231352718e-06,
+      "loss": 2.1154,
+      "step": 36624
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2660398230088496e-06,
+      "loss": 2.6155,
+      "step": 36625
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.265992414664981e-06,
+      "loss": 3.5352,
+      "step": 36626
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2659450063211125e-06,
+      "loss": 1.1865,
+      "step": 36627
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.265897597977244e-06,
+      "loss": 2.5723,
+      "step": 36628
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2658501896333755e-06,
+      "loss": 3.9856,
+      "step": 36629
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.265802781289507e-06,
+      "loss": 1.6332,
+      "step": 36630
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2657553729456385e-06,
+      "loss": 2.7932,
+      "step": 36631
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.26570796460177e-06,
+      "loss": 1.7748,
+      "step": 36632
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2656605562579015e-06,
+      "loss": 1.826,
+      "step": 36633
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2656131479140328e-06,
+      "loss": 1.6662,
+      "step": 36634
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2655657395701645e-06,
+      "loss": 2.3845,
+      "step": 36635
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 1.2655183312262958e-06,
+      "loss": 1.7654,
+      "step": 36636
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2654709228824272e-06,
+      "loss": 3.3966,
+      "step": 36637
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.265423514538559e-06,
+      "loss": 2.3786,
+      "step": 36638
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2653761061946902e-06,
+      "loss": 3.1509,
+      "step": 36639
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2653286978508217e-06,
+      "loss": 1.806,
+      "step": 36640
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2652812895069532e-06,
+      "loss": 1.901,
+      "step": 36641
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2652338811630847e-06,
+      "loss": 1.6911,
+      "step": 36642
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2651864728192162e-06,
+      "loss": 1.5256,
+      "step": 36643
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2651390644753477e-06,
+      "loss": 1.4995,
+      "step": 36644
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2650916561314792e-06,
+      "loss": 2.1661,
+      "step": 36645
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2650442477876107e-06,
+      "loss": 0.4806,
+      "step": 36646
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.264996839443742e-06,
+      "loss": 2.6082,
+      "step": 36647
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2649494310998737e-06,
+      "loss": 3.1705,
+      "step": 36648
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2649020227560052e-06,
+      "loss": 2.9252,
+      "step": 36649
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2648546144121364e-06,
+      "loss": 1.9896,
+      "step": 36650
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2648072060682681e-06,
+      "loss": 2.8324,
+      "step": 36651
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2647597977243996e-06,
+      "loss": 2.075,
+      "step": 36652
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.264712389380531e-06,
+      "loss": 2.8409,
+      "step": 36653
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2646649810366624e-06,
+      "loss": 2.3092,
+      "step": 36654
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2646175726927941e-06,
+      "loss": 3.5967,
+      "step": 36655
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2645701643489254e-06,
+      "loss": 3.0327,
+      "step": 36656
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2645227560050569e-06,
+      "loss": 1.6779,
+      "step": 36657
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2644753476611884e-06,
+      "loss": 1.1011,
+      "step": 36658
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2644279393173199e-06,
+      "loss": 2.2657,
+      "step": 36659
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2643805309734514e-06,
+      "loss": 3.2441,
+      "step": 36660
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2643331226295828e-06,
+      "loss": 2.6727,
+      "step": 36661
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2642857142857143e-06,
+      "loss": 1.212,
+      "step": 36662
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2642383059418458e-06,
+      "loss": 3.1209,
+      "step": 36663
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2641908975979773e-06,
+      "loss": 1.3548,
+      "step": 36664
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2641434892541088e-06,
+      "loss": 2.0024,
+      "step": 36665
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2640960809102403e-06,
+      "loss": 2.8201,
+      "step": 36666
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2640486725663716e-06,
+      "loss": 1.6685,
+      "step": 36667
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2640012642225033e-06,
+      "loss": 3.3408,
+      "step": 36668
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2639538558786346e-06,
+      "loss": 2.9095,
+      "step": 36669
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263906447534766e-06,
+      "loss": 2.2204,
+      "step": 36670
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2638590391908978e-06,
+      "loss": 3.1024,
+      "step": 36671
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263811630847029e-06,
+      "loss": 1.6597,
+      "step": 36672
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2637642225031605e-06,
+      "loss": 1.7281,
+      "step": 36673
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263716814159292e-06,
+      "loss": 2.6415,
+      "step": 36674
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2636694058154235e-06,
+      "loss": 2.5306,
+      "step": 36675
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263621997471555e-06,
+      "loss": 2.7536,
+      "step": 36676
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2635745891276865e-06,
+      "loss": 1.1267,
+      "step": 36677
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263527180783818e-06,
+      "loss": 1.8081,
+      "step": 36678
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2634797724399495e-06,
+      "loss": 1.5901,
+      "step": 36679
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2634323640960808e-06,
+      "loss": 2.1133,
+      "step": 36680
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2633849557522125e-06,
+      "loss": 2.0913,
+      "step": 36681
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263337547408344e-06,
+      "loss": 1.0499,
+      "step": 36682
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2632901390644753e-06,
+      "loss": 2.0866,
+      "step": 36683
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263242730720607e-06,
+      "loss": 2.9281,
+      "step": 36684
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2631953223767385e-06,
+      "loss": 1.6346,
+      "step": 36685
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2631479140328697e-06,
+      "loss": 1.6781,
+      "step": 36686
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2631005056890012e-06,
+      "loss": 1.9174,
+      "step": 36687
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.263053097345133e-06,
+      "loss": 3.1099,
+      "step": 36688
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2630056890012642e-06,
+      "loss": 1.2783,
+      "step": 36689
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2629582806573957e-06,
+      "loss": 1.2353,
+      "step": 36690
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2629108723135272e-06,
+      "loss": 3.5169,
+      "step": 36691
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2628634639696587e-06,
+      "loss": 2.4071,
+      "step": 36692
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2628160556257902e-06,
+      "loss": 2.3266,
+      "step": 36693
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2627686472819217e-06,
+      "loss": 2.584,
+      "step": 36694
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2627212389380532e-06,
+      "loss": 1.7971,
+      "step": 36695
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2626738305941847e-06,
+      "loss": 1.8625,
+      "step": 36696
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2626264222503161e-06,
+      "loss": 2.3301,
+      "step": 36697
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2625790139064476e-06,
+      "loss": 1.7166,
+      "step": 36698
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 1.2625316055625791e-06,
+      "loss": 2.6081,
+      "step": 36699
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2624841972187104e-06,
+      "loss": 2.4904,
+      "step": 36700
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2624367888748421e-06,
+      "loss": 2.6234,
+      "step": 36701
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2623893805309734e-06,
+      "loss": 1.6859,
+      "step": 36702
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2623419721871049e-06,
+      "loss": 2.124,
+      "step": 36703
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2622945638432366e-06,
+      "loss": 5.082,
+      "step": 36704
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2622471554993679e-06,
+      "loss": 2.6988,
+      "step": 36705
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2621997471554994e-06,
+      "loss": 1.3959,
+      "step": 36706
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2621523388116309e-06,
+      "loss": 2.1254,
+      "step": 36707
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2621049304677623e-06,
+      "loss": 0.7462,
+      "step": 36708
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2620575221238938e-06,
+      "loss": 1.21,
+      "step": 36709
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2620101137800253e-06,
+      "loss": 2.1215,
+      "step": 36710
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2619627054361568e-06,
+      "loss": 0.3426,
+      "step": 36711
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2619152970922883e-06,
+      "loss": 3.3257,
+      "step": 36712
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2618678887484196e-06,
+      "loss": 1.6506,
+      "step": 36713
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2618204804045513e-06,
+      "loss": 1.4926,
+      "step": 36714
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2617730720606828e-06,
+      "loss": 2.749,
+      "step": 36715
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.261725663716814e-06,
+      "loss": 1.3563,
+      "step": 36716
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2616782553729458e-06,
+      "loss": 1.8699,
+      "step": 36717
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2616308470290773e-06,
+      "loss": 1.7488,
+      "step": 36718
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2615834386852085e-06,
+      "loss": 2.6414,
+      "step": 36719
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.26153603034134e-06,
+      "loss": 1.4448,
+      "step": 36720
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2614886219974717e-06,
+      "loss": 1.5281,
+      "step": 36721
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.261441213653603e-06,
+      "loss": 3.1227,
+      "step": 36722
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2613938053097345e-06,
+      "loss": 1.2963,
+      "step": 36723
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.261346396965866e-06,
+      "loss": 3.037,
+      "step": 36724
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2612989886219975e-06,
+      "loss": 1.5082,
+      "step": 36725
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.261251580278129e-06,
+      "loss": 1.286,
+      "step": 36726
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2612041719342605e-06,
+      "loss": 2.2313,
+      "step": 36727
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.261156763590392e-06,
+      "loss": 1.7451,
+      "step": 36728
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2611093552465235e-06,
+      "loss": 1.6233,
+      "step": 36729
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.261061946902655e-06,
+      "loss": 3.1773,
+      "step": 36730
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2610145385587865e-06,
+      "loss": 2.1246,
+      "step": 36731
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.260967130214918e-06,
+      "loss": 2.0483,
+      "step": 36732
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2609197218710492e-06,
+      "loss": 1.0261,
+      "step": 36733
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.260872313527181e-06,
+      "loss": 1.6611,
+      "step": 36734
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2608249051833122e-06,
+      "loss": 1.0717,
+      "step": 36735
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2607774968394437e-06,
+      "loss": 2.8395,
+      "step": 36736
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2607300884955754e-06,
+      "loss": 1.2893,
+      "step": 36737
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2606826801517067e-06,
+      "loss": 1.2017,
+      "step": 36738
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2606352718078382e-06,
+      "loss": 2.6835,
+      "step": 36739
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2605878634639697e-06,
+      "loss": 1.4779,
+      "step": 36740
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2605404551201012e-06,
+      "loss": 2.5673,
+      "step": 36741
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2604930467762327e-06,
+      "loss": 3.0802,
+      "step": 36742
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2604456384323641e-06,
+      "loss": 1.7691,
+      "step": 36743
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2603982300884956e-06,
+      "loss": 2.577,
+      "step": 36744
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2603508217446271e-06,
+      "loss": 0.4542,
+      "step": 36745
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2603034134007584e-06,
+      "loss": 2.0417,
+      "step": 36746
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2602560050568901e-06,
+      "loss": 1.3895,
+      "step": 36747
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2602085967130216e-06,
+      "loss": 2.5672,
+      "step": 36748
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2601611883691529e-06,
+      "loss": 1.2483,
+      "step": 36749
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2601137800252846e-06,
+      "loss": 0.3393,
+      "step": 36750
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.260066371681416e-06,
+      "loss": 3.1352,
+      "step": 36751
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2600189633375474e-06,
+      "loss": 1.3403,
+      "step": 36752
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2599715549936789e-06,
+      "loss": 0.6437,
+      "step": 36753
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2599241466498104e-06,
+      "loss": 2.9142,
+      "step": 36754
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2598767383059418e-06,
+      "loss": 1.8734,
+      "step": 36755
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2598293299620733e-06,
+      "loss": 2.6599,
+      "step": 36756
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2597819216182048e-06,
+      "loss": 2.5808,
+      "step": 36757
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2597345132743363e-06,
+      "loss": 1.5241,
+      "step": 36758
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2596871049304678e-06,
+      "loss": 3.4775,
+      "step": 36759
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2596396965865993e-06,
+      "loss": 2.7791,
+      "step": 36760
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2595922882427308e-06,
+      "loss": 2.6528,
+      "step": 36761
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2595448798988623e-06,
+      "loss": 2.4315,
+      "step": 36762
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 1.2594974715549936e-06,
+      "loss": 2.5604,
+      "step": 36763
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2594500632111253e-06,
+      "loss": 1.832,
+      "step": 36764
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2594026548672566e-06,
+      "loss": 2.6794,
+      "step": 36765
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.259355246523388e-06,
+      "loss": 2.337,
+      "step": 36766
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2593078381795197e-06,
+      "loss": 1.8133,
+      "step": 36767
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.259260429835651e-06,
+      "loss": 1.807,
+      "step": 36768
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2592130214917825e-06,
+      "loss": 2.517,
+      "step": 36769
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.259165613147914e-06,
+      "loss": 2.2522,
+      "step": 36770
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2591182048040455e-06,
+      "loss": 1.8998,
+      "step": 36771
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.259070796460177e-06,
+      "loss": 2.4266,
+      "step": 36772
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2590233881163085e-06,
+      "loss": 1.4121,
+      "step": 36773
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.25897597977244e-06,
+      "loss": 1.5913,
+      "step": 36774
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2589285714285715e-06,
+      "loss": 1.8245,
+      "step": 36775
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2588811630847028e-06,
+      "loss": 1.5615,
+      "step": 36776
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2588337547408345e-06,
+      "loss": 1.9929,
+      "step": 36777
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.258786346396966e-06,
+      "loss": 2.1608,
+      "step": 36778
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2587389380530972e-06,
+      "loss": 0.9507,
+      "step": 36779
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.258691529709229e-06,
+      "loss": 1.8692,
+      "step": 36780
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2586441213653604e-06,
+      "loss": 2.2582,
+      "step": 36781
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2585967130214917e-06,
+      "loss": 3.7571,
+      "step": 36782
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2585493046776232e-06,
+      "loss": 1.6962,
+      "step": 36783
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.258501896333755e-06,
+      "loss": 2.0349,
+      "step": 36784
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2584544879898862e-06,
+      "loss": 2.6206,
+      "step": 36785
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2584070796460177e-06,
+      "loss": 1.8624,
+      "step": 36786
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2583596713021492e-06,
+      "loss": 2.1873,
+      "step": 36787
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2583122629582807e-06,
+      "loss": 2.481,
+      "step": 36788
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2582648546144122e-06,
+      "loss": 2.269,
+      "step": 36789
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2582174462705436e-06,
+      "loss": 1.7306,
+      "step": 36790
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2581700379266751e-06,
+      "loss": 3.2109,
+      "step": 36791
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2581226295828066e-06,
+      "loss": 1.5952,
+      "step": 36792
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2580752212389381e-06,
+      "loss": 2.6865,
+      "step": 36793
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2580278128950696e-06,
+      "loss": 0.6385,
+      "step": 36794
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2579804045512011e-06,
+      "loss": 2.081,
+      "step": 36795
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2579329962073324e-06,
+      "loss": 1.3701,
+      "step": 36796
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.257885587863464e-06,
+      "loss": 2.4589,
+      "step": 36797
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2578381795195954e-06,
+      "loss": 2.498,
+      "step": 36798
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2577907711757269e-06,
+      "loss": 1.7376,
+      "step": 36799
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2577433628318586e-06,
+      "loss": 0.9355,
+      "step": 36800
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2576959544879898e-06,
+      "loss": 1.0652,
+      "step": 36801
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2576485461441213e-06,
+      "loss": 2.0106,
+      "step": 36802
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2576011378002528e-06,
+      "loss": 0.853,
+      "step": 36803
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2575537294563843e-06,
+      "loss": 2.4535,
+      "step": 36804
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2575063211125158e-06,
+      "loss": 2.5227,
+      "step": 36805
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2574589127686473e-06,
+      "loss": 2.6582,
+      "step": 36806
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2574115044247788e-06,
+      "loss": 1.7259,
+      "step": 36807
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2573640960809103e-06,
+      "loss": 2.1989,
+      "step": 36808
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2573166877370416e-06,
+      "loss": 2.0527,
+      "step": 36809
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2572692793931733e-06,
+      "loss": 2.6483,
+      "step": 36810
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2572218710493048e-06,
+      "loss": 3.1559,
+      "step": 36811
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.257174462705436e-06,
+      "loss": 2.6147,
+      "step": 36812
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2571270543615678e-06,
+      "loss": 2.4791,
+      "step": 36813
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2570796460176992e-06,
+      "loss": 1.9161,
+      "step": 36814
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2570322376738305e-06,
+      "loss": 1.944,
+      "step": 36815
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.256984829329962e-06,
+      "loss": 2.0618,
+      "step": 36816
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2569374209860937e-06,
+      "loss": 3.189,
+      "step": 36817
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.256890012642225e-06,
+      "loss": 0.6128,
+      "step": 36818
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2568426042983565e-06,
+      "loss": 1.2033,
+      "step": 36819
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.256795195954488e-06,
+      "loss": 1.7571,
+      "step": 36820
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2567477876106195e-06,
+      "loss": 1.8306,
+      "step": 36821
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.256700379266751e-06,
+      "loss": 2.7003,
+      "step": 36822
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2566529709228825e-06,
+      "loss": 2.7846,
+      "step": 36823
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.256605562579014e-06,
+      "loss": 3.4358,
+      "step": 36824
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.2565581542351454e-06,
+      "loss": 2.39,
+      "step": 36825
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 1.256510745891277e-06,
+      "loss": 1.9397,
+      "step": 36826
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2564633375474084e-06,
+      "loss": 2.5711,
+      "step": 36827
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.25641592920354e-06,
+      "loss": 1.2964,
+      "step": 36828
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2563685208596712e-06,
+      "loss": 1.3995,
+      "step": 36829
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.256321112515803e-06,
+      "loss": 1.7829,
+      "step": 36830
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2562737041719342e-06,
+      "loss": 2.6078,
+      "step": 36831
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2562262958280657e-06,
+      "loss": 3.0951,
+      "step": 36832
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2561788874841974e-06,
+      "loss": 3.7466,
+      "step": 36833
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2561314791403287e-06,
+      "loss": 3.19,
+      "step": 36834
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2560840707964602e-06,
+      "loss": 1.3085,
+      "step": 36835
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2560366624525917e-06,
+      "loss": 2.812,
+      "step": 36836
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2559892541087231e-06,
+      "loss": 1.2738,
+      "step": 36837
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2559418457648546e-06,
+      "loss": 1.5974,
+      "step": 36838
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2558944374209861e-06,
+      "loss": 1.5925,
+      "step": 36839
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2558470290771176e-06,
+      "loss": 2.8172,
+      "step": 36840
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2557996207332491e-06,
+      "loss": 1.8497,
+      "step": 36841
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2557522123893804e-06,
+      "loss": 2.7677,
+      "step": 36842
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.255704804045512e-06,
+      "loss": 2.1828,
+      "step": 36843
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2556573957016436e-06,
+      "loss": 2.7735,
+      "step": 36844
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2556099873577749e-06,
+      "loss": 0.9292,
+      "step": 36845
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2555625790139066e-06,
+      "loss": 2.6697,
+      "step": 36846
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.255515170670038e-06,
+      "loss": 2.1498,
+      "step": 36847
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2554677623261693e-06,
+      "loss": 1.686,
+      "step": 36848
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2554203539823008e-06,
+      "loss": 2.0584,
+      "step": 36849
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2553729456384325e-06,
+      "loss": 1.107,
+      "step": 36850
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2553255372945638e-06,
+      "loss": 1.2776,
+      "step": 36851
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2552781289506953e-06,
+      "loss": 2.6728,
+      "step": 36852
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2552307206068268e-06,
+      "loss": 1.9468,
+      "step": 36853
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2551833122629583e-06,
+      "loss": 3.4731,
+      "step": 36854
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2551359039190898e-06,
+      "loss": 2.0125,
+      "step": 36855
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2550884955752213e-06,
+      "loss": 1.9854,
+      "step": 36856
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2550410872313528e-06,
+      "loss": 2.3571,
+      "step": 36857
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2549936788874843e-06,
+      "loss": 1.062,
+      "step": 36858
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2549462705436158e-06,
+      "loss": 2.8496,
+      "step": 36859
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2548988621997473e-06,
+      "loss": 2.225,
+      "step": 36860
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2548514538558787e-06,
+      "loss": 2.0348,
+      "step": 36861
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.25480404551201e-06,
+      "loss": 2.7578,
+      "step": 36862
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2547566371681417e-06,
+      "loss": 2.5314,
+      "step": 36863
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.254709228824273e-06,
+      "loss": 1.5574,
+      "step": 36864
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2546618204804045e-06,
+      "loss": 0.9034,
+      "step": 36865
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2546144121365362e-06,
+      "loss": 2.8525,
+      "step": 36866
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2545670037926675e-06,
+      "loss": 1.486,
+      "step": 36867
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.254519595448799e-06,
+      "loss": 2.1497,
+      "step": 36868
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2544721871049305e-06,
+      "loss": 1.3202,
+      "step": 36869
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.254424778761062e-06,
+      "loss": 1.3875,
+      "step": 36870
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2543773704171935e-06,
+      "loss": 1.2066,
+      "step": 36871
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.254329962073325e-06,
+      "loss": 2.9998,
+      "step": 36872
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2542825537294564e-06,
+      "loss": 1.5736,
+      "step": 36873
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.254235145385588e-06,
+      "loss": 1.6186,
+      "step": 36874
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2541877370417192e-06,
+      "loss": 2.4802,
+      "step": 36875
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.254140328697851e-06,
+      "loss": 2.4154,
+      "step": 36876
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2540929203539824e-06,
+      "loss": 2.4046,
+      "step": 36877
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2540455120101137e-06,
+      "loss": 3.2846,
+      "step": 36878
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2539981036662454e-06,
+      "loss": 2.4371,
+      "step": 36879
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2539506953223769e-06,
+      "loss": 3.4086,
+      "step": 36880
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2539032869785082e-06,
+      "loss": 1.9113,
+      "step": 36881
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2538558786346397e-06,
+      "loss": 1.7624,
+      "step": 36882
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2538084702907714e-06,
+      "loss": 2.0177,
+      "step": 36883
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2537610619469026e-06,
+      "loss": 2.3355,
+      "step": 36884
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2537136536030341e-06,
+      "loss": 2.8414,
+      "step": 36885
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2536662452591656e-06,
+      "loss": 1.6142,
+      "step": 36886
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2536188369152971e-06,
+      "loss": 1.5633,
+      "step": 36887
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.2535714285714286e-06,
+      "loss": 1.0784,
+      "step": 36888
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 1.25352402022756e-06,
+      "loss": 2.1863,
+      "step": 36889
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2534766118836916e-06,
+      "loss": 3.5927,
+      "step": 36890
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.253429203539823e-06,
+      "loss": 2.0386,
+      "step": 36891
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2533817951959544e-06,
+      "loss": 2.2111,
+      "step": 36892
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.253334386852086e-06,
+      "loss": 0.8657,
+      "step": 36893
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2532869785082176e-06,
+      "loss": 3.0553,
+      "step": 36894
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2532395701643488e-06,
+      "loss": 1.5945,
+      "step": 36895
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2531921618204805e-06,
+      "loss": 1.3211,
+      "step": 36896
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2531447534766118e-06,
+      "loss": 2.1167,
+      "step": 36897
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2530973451327433e-06,
+      "loss": 3.4205,
+      "step": 36898
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2530499367888748e-06,
+      "loss": 1.7701,
+      "step": 36899
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2530025284450063e-06,
+      "loss": 1.9212,
+      "step": 36900
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2529551201011378e-06,
+      "loss": 1.791,
+      "step": 36901
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2529077117572693e-06,
+      "loss": 2.0077,
+      "step": 36902
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2528603034134008e-06,
+      "loss": 1.4354,
+      "step": 36903
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2528128950695323e-06,
+      "loss": 0.6065,
+      "step": 36904
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2527654867256638e-06,
+      "loss": 2.6271,
+      "step": 36905
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2527180783817953e-06,
+      "loss": 1.5656,
+      "step": 36906
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2526706700379267e-06,
+      "loss": 2.5288,
+      "step": 36907
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.252623261694058e-06,
+      "loss": 2.1459,
+      "step": 36908
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2525758533501897e-06,
+      "loss": 1.1624,
+      "step": 36909
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2525284450063212e-06,
+      "loss": 2.9976,
+      "step": 36910
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2524810366624525e-06,
+      "loss": 1.9606,
+      "step": 36911
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.252433628318584e-06,
+      "loss": 2.5691,
+      "step": 36912
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2523862199747157e-06,
+      "loss": 3.9093,
+      "step": 36913
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.252338811630847e-06,
+      "loss": 2.1563,
+      "step": 36914
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2522914032869785e-06,
+      "loss": 2.962,
+      "step": 36915
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2522439949431102e-06,
+      "loss": 2.5777,
+      "step": 36916
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2521965865992415e-06,
+      "loss": 3.2423,
+      "step": 36917
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.252149178255373e-06,
+      "loss": 2.0111,
+      "step": 36918
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2521017699115044e-06,
+      "loss": 0.7691,
+      "step": 36919
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.252054361567636e-06,
+      "loss": 1.6884,
+      "step": 36920
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2520069532237674e-06,
+      "loss": 1.5952,
+      "step": 36921
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.251959544879899e-06,
+      "loss": 0.3276,
+      "step": 36922
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2519121365360304e-06,
+      "loss": 3.7656,
+      "step": 36923
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.251864728192162e-06,
+      "loss": 2.004,
+      "step": 36924
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2518173198482932e-06,
+      "loss": 2.655,
+      "step": 36925
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2517699115044249e-06,
+      "loss": 1.2097,
+      "step": 36926
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2517225031605564e-06,
+      "loss": 3.2257,
+      "step": 36927
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2516750948166877e-06,
+      "loss": 3.1964,
+      "step": 36928
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2516276864728194e-06,
+      "loss": 1.4555,
+      "step": 36929
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2515802781289506e-06,
+      "loss": 2.2421,
+      "step": 36930
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2515328697850821e-06,
+      "loss": 3.8323,
+      "step": 36931
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2514854614412136e-06,
+      "loss": 1.262,
+      "step": 36932
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2514380530973451e-06,
+      "loss": 2.5577,
+      "step": 36933
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2513906447534766e-06,
+      "loss": 2.2083,
+      "step": 36934
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2513432364096081e-06,
+      "loss": 1.8593,
+      "step": 36935
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2512958280657396e-06,
+      "loss": 1.6747,
+      "step": 36936
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.251248419721871e-06,
+      "loss": 1.2033,
+      "step": 36937
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2512010113780026e-06,
+      "loss": 2.1193,
+      "step": 36938
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.251153603034134e-06,
+      "loss": 2.4048,
+      "step": 36939
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2511061946902656e-06,
+      "loss": 0.9294,
+      "step": 36940
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2510587863463968e-06,
+      "loss": 2.0597,
+      "step": 36941
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2510113780025286e-06,
+      "loss": 3.9725,
+      "step": 36942
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.25096396965866e-06,
+      "loss": 2.2888,
+      "step": 36943
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2509165613147913e-06,
+      "loss": 2.5229,
+      "step": 36944
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2508691529709228e-06,
+      "loss": 2.4479,
+      "step": 36945
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2508217446270545e-06,
+      "loss": 2.6454,
+      "step": 36946
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2507743362831858e-06,
+      "loss": 3.024,
+      "step": 36947
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2507269279393173e-06,
+      "loss": 1.7399,
+      "step": 36948
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.250679519595449e-06,
+      "loss": 2.1779,
+      "step": 36949
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2506321112515803e-06,
+      "loss": 3.0564,
+      "step": 36950
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2505847029077118e-06,
+      "loss": 3.3313,
+      "step": 36951
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2505372945638433e-06,
+      "loss": 2.1558,
+      "step": 36952
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 1.2504898862199748e-06,
+      "loss": 2.1006,
+      "step": 36953
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2504424778761062e-06,
+      "loss": 2.6609,
+      "step": 36954
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2503950695322377e-06,
+      "loss": 2.6728,
+      "step": 36955
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2503476611883692e-06,
+      "loss": 1.4053,
+      "step": 36956
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2503002528445007e-06,
+      "loss": 2.9158,
+      "step": 36957
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.250252844500632e-06,
+      "loss": 3.088,
+      "step": 36958
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2502054361567637e-06,
+      "loss": 1.7356,
+      "step": 36959
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2501580278128952e-06,
+      "loss": 2.5289,
+      "step": 36960
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2501106194690265e-06,
+      "loss": 2.7127,
+      "step": 36961
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2500632111251582e-06,
+      "loss": 2.4252,
+      "step": 36962
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2500158027812895e-06,
+      "loss": 3.0698,
+      "step": 36963
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.249968394437421e-06,
+      "loss": 1.6979,
+      "step": 36964
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2499209860935524e-06,
+      "loss": 3.1738,
+      "step": 36965
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.249873577749684e-06,
+      "loss": 2.025,
+      "step": 36966
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2498261694058154e-06,
+      "loss": 1.4776,
+      "step": 36967
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.249778761061947e-06,
+      "loss": 2.6733,
+      "step": 36968
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2497313527180784e-06,
+      "loss": 0.2204,
+      "step": 36969
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.24968394437421e-06,
+      "loss": 1.5452,
+      "step": 36970
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2496365360303414e-06,
+      "loss": 3.2162,
+      "step": 36971
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2495891276864729e-06,
+      "loss": 1.2307,
+      "step": 36972
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2495417193426044e-06,
+      "loss": 1.8247,
+      "step": 36973
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2494943109987357e-06,
+      "loss": 2.5885,
+      "step": 36974
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2494469026548674e-06,
+      "loss": 2.3219,
+      "step": 36975
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2493994943109989e-06,
+      "loss": 1.4186,
+      "step": 36976
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2493520859671301e-06,
+      "loss": 3.6658,
+      "step": 36977
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2493046776232616e-06,
+      "loss": 2.3361,
+      "step": 36978
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2492572692793933e-06,
+      "loss": 3.343,
+      "step": 36979
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2492098609355246e-06,
+      "loss": 4.1683,
+      "step": 36980
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2491624525916561e-06,
+      "loss": 2.2715,
+      "step": 36981
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2491150442477878e-06,
+      "loss": 1.7392,
+      "step": 36982
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.249067635903919e-06,
+      "loss": 2.2139,
+      "step": 36983
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2490202275600506e-06,
+      "loss": 2.2833,
+      "step": 36984
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.248972819216182e-06,
+      "loss": 2.463,
+      "step": 36985
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2489254108723136e-06,
+      "loss": 2.674,
+      "step": 36986
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.248878002528445e-06,
+      "loss": 2.0857,
+      "step": 36987
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2488305941845766e-06,
+      "loss": 1.1868,
+      "step": 36988
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.248783185840708e-06,
+      "loss": 2.4924,
+      "step": 36989
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2487357774968395e-06,
+      "loss": 2.0971,
+      "step": 36990
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2486883691529708e-06,
+      "loss": 2.8941,
+      "step": 36991
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2486409608091025e-06,
+      "loss": 3.7753,
+      "step": 36992
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.248593552465234e-06,
+      "loss": 2.2817,
+      "step": 36993
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2485461441213653e-06,
+      "loss": 1.8317,
+      "step": 36994
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.248498735777497e-06,
+      "loss": 1.5486,
+      "step": 36995
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2484513274336283e-06,
+      "loss": 1.762,
+      "step": 36996
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2484039190897598e-06,
+      "loss": 1.7281,
+      "step": 36997
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2483565107458913e-06,
+      "loss": 2.9751,
+      "step": 36998
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2483091024020228e-06,
+      "loss": 2.4103,
+      "step": 36999
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2482616940581543e-06,
+      "loss": 0.9398,
+      "step": 37000
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2482142857142857e-06,
+      "loss": 3.0443,
+      "step": 37001
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2481668773704172e-06,
+      "loss": 1.9215,
+      "step": 37002
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2481194690265487e-06,
+      "loss": 2.6587,
+      "step": 37003
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2480720606826802e-06,
+      "loss": 1.3128,
+      "step": 37004
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2480246523388117e-06,
+      "loss": 2.4525,
+      "step": 37005
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2479772439949432e-06,
+      "loss": 1.6106,
+      "step": 37006
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2479298356510745e-06,
+      "loss": 2.6114,
+      "step": 37007
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2478824273072062e-06,
+      "loss": 3.0827,
+      "step": 37008
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2478350189633377e-06,
+      "loss": 1.9664,
+      "step": 37009
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.247787610619469e-06,
+      "loss": 0.96,
+      "step": 37010
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2477402022756005e-06,
+      "loss": 1.6918,
+      "step": 37011
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2476927939317322e-06,
+      "loss": 2.9551,
+      "step": 37012
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2476453855878634e-06,
+      "loss": 1.7603,
+      "step": 37013
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.247597977243995e-06,
+      "loss": 1.7906,
+      "step": 37014
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.2475505689001266e-06,
+      "loss": 2.6366,
+      "step": 37015
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 1.247503160556258e-06,
+      "loss": 2.8146,
+      "step": 37016
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2474557522123894e-06,
+      "loss": 2.7562,
+      "step": 37017
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.247408343868521e-06,
+      "loss": 1.7067,
+      "step": 37018
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2473609355246524e-06,
+      "loss": 2.0722,
+      "step": 37019
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2473135271807839e-06,
+      "loss": 2.1313,
+      "step": 37020
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2472661188369154e-06,
+      "loss": 2.7923,
+      "step": 37021
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2472187104930469e-06,
+      "loss": 1.9719,
+      "step": 37022
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2471713021491784e-06,
+      "loss": 3.0064,
+      "step": 37023
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2471238938053096e-06,
+      "loss": 2.9744,
+      "step": 37024
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2470764854614413e-06,
+      "loss": 2.8503,
+      "step": 37025
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2470290771175728e-06,
+      "loss": 3.3449,
+      "step": 37026
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2469816687737041e-06,
+      "loss": 2.1708,
+      "step": 37027
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2469342604298358e-06,
+      "loss": 2.7213,
+      "step": 37028
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.246886852085967e-06,
+      "loss": 1.0015,
+      "step": 37029
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2468394437420986e-06,
+      "loss": 2.7676,
+      "step": 37030
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.24679203539823e-06,
+      "loss": 1.6175,
+      "step": 37031
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2467446270543616e-06,
+      "loss": 4.4025,
+      "step": 37032
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.246697218710493e-06,
+      "loss": 3.3036,
+      "step": 37033
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2466498103666246e-06,
+      "loss": 1.2282,
+      "step": 37034
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.246602402022756e-06,
+      "loss": 3.0641,
+      "step": 37035
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2465549936788875e-06,
+      "loss": 3.3151,
+      "step": 37036
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.246507585335019e-06,
+      "loss": 1.5581,
+      "step": 37037
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2464601769911505e-06,
+      "loss": 1.1725,
+      "step": 37038
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.246412768647282e-06,
+      "loss": 1.0646,
+      "step": 37039
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2463653603034133e-06,
+      "loss": 1.047,
+      "step": 37040
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2463179519595448e-06,
+      "loss": 2.6027,
+      "step": 37041
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2462705436156765e-06,
+      "loss": 2.7118,
+      "step": 37042
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2462231352718078e-06,
+      "loss": 1.1438,
+      "step": 37043
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2461757269279393e-06,
+      "loss": 2.7905,
+      "step": 37044
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.246128318584071e-06,
+      "loss": 2.5313,
+      "step": 37045
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2460809102402023e-06,
+      "loss": 2.7806,
+      "step": 37046
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2460335018963337e-06,
+      "loss": 1.5524,
+      "step": 37047
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2459860935524652e-06,
+      "loss": 2.3795,
+      "step": 37048
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2459386852085967e-06,
+      "loss": 1.4589,
+      "step": 37049
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2458912768647282e-06,
+      "loss": 1.6762,
+      "step": 37050
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2458438685208597e-06,
+      "loss": 2.8649,
+      "step": 37051
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2457964601769912e-06,
+      "loss": 2.1194,
+      "step": 37052
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2457490518331227e-06,
+      "loss": 1.3906,
+      "step": 37053
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.245701643489254e-06,
+      "loss": 2.4707,
+      "step": 37054
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2456542351453857e-06,
+      "loss": 2.6357,
+      "step": 37055
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2456068268015172e-06,
+      "loss": 1.8606,
+      "step": 37056
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2455594184576485e-06,
+      "loss": 1.6925,
+      "step": 37057
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2455120101137802e-06,
+      "loss": 2.8629,
+      "step": 37058
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2454646017699117e-06,
+      "loss": 1.9002,
+      "step": 37059
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.245417193426043e-06,
+      "loss": 4.8707,
+      "step": 37060
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2453697850821744e-06,
+      "loss": 1.4062,
+      "step": 37061
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.245322376738306e-06,
+      "loss": 1.391,
+      "step": 37062
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2452749683944374e-06,
+      "loss": 2.2458,
+      "step": 37063
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.245227560050569e-06,
+      "loss": 2.8206,
+      "step": 37064
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2451801517067004e-06,
+      "loss": 0.7889,
+      "step": 37065
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2451327433628319e-06,
+      "loss": 3.4035,
+      "step": 37066
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2450853350189634e-06,
+      "loss": 2.0406,
+      "step": 37067
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2450379266750949e-06,
+      "loss": 3.1815,
+      "step": 37068
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2449905183312264e-06,
+      "loss": 1.6316,
+      "step": 37069
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2449431099873579e-06,
+      "loss": 2.1962,
+      "step": 37070
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2448957016434893e-06,
+      "loss": 2.8438,
+      "step": 37071
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2448482932996208e-06,
+      "loss": 1.7434,
+      "step": 37072
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2448008849557521e-06,
+      "loss": 2.2729,
+      "step": 37073
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2447534766118836e-06,
+      "loss": 2.4864,
+      "step": 37074
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2447060682680153e-06,
+      "loss": 1.95,
+      "step": 37075
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2446586599241466e-06,
+      "loss": 1.0322,
+      "step": 37076
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.244611251580278e-06,
+      "loss": 1.515,
+      "step": 37077
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.2445638432364098e-06,
+      "loss": 1.9574,
+      "step": 37078
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 1.244516434892541e-06,
+      "loss": 3.618,
+      "step": 37079
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2444690265486726e-06,
+      "loss": 1.8355,
+      "step": 37080
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.244421618204804e-06,
+      "loss": 2.7163,
+      "step": 37081
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2443742098609356e-06,
+      "loss": 2.4836,
+      "step": 37082
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.244326801517067e-06,
+      "loss": 2.3097,
+      "step": 37083
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2442793931731985e-06,
+      "loss": 1.8977,
+      "step": 37084
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.24423198482933e-06,
+      "loss": 2.2633,
+      "step": 37085
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2441845764854615e-06,
+      "loss": 1.8983,
+      "step": 37086
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2441371681415928e-06,
+      "loss": 2.8386,
+      "step": 37087
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2440897597977245e-06,
+      "loss": 2.0729,
+      "step": 37088
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.244042351453856e-06,
+      "loss": 2.9948,
+      "step": 37089
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2439949431099873e-06,
+      "loss": 1.5411,
+      "step": 37090
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.243947534766119e-06,
+      "loss": 1.9952,
+      "step": 37091
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2439001264222505e-06,
+      "loss": 2.7191,
+      "step": 37092
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2438527180783818e-06,
+      "loss": 1.1798,
+      "step": 37093
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2438053097345132e-06,
+      "loss": 2.6267,
+      "step": 37094
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2437579013906447e-06,
+      "loss": 3.2421,
+      "step": 37095
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2437104930467762e-06,
+      "loss": 2.0436,
+      "step": 37096
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2436630847029077e-06,
+      "loss": 2.574,
+      "step": 37097
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2436156763590392e-06,
+      "loss": 1.5733,
+      "step": 37098
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2435682680151707e-06,
+      "loss": 3.2814,
+      "step": 37099
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2435208596713022e-06,
+      "loss": 1.5136,
+      "step": 37100
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2434734513274337e-06,
+      "loss": 3.581,
+      "step": 37101
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2434260429835652e-06,
+      "loss": 1.6272,
+      "step": 37102
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2433786346396967e-06,
+      "loss": 3.0083,
+      "step": 37103
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2433312262958282e-06,
+      "loss": 2.6651,
+      "step": 37104
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2432838179519597e-06,
+      "loss": 2.0154,
+      "step": 37105
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.243236409608091e-06,
+      "loss": 2.61,
+      "step": 37106
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2431890012642224e-06,
+      "loss": 1.3276,
+      "step": 37107
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2431415929203541e-06,
+      "loss": 2.785,
+      "step": 37108
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2430941845764854e-06,
+      "loss": 1.1743,
+      "step": 37109
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.243046776232617e-06,
+      "loss": 2.1281,
+      "step": 37110
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2429993678887486e-06,
+      "loss": 1.1133,
+      "step": 37111
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2429519595448799e-06,
+      "loss": 1.6142,
+      "step": 37112
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2429045512010114e-06,
+      "loss": 2.594,
+      "step": 37113
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2428571428571429e-06,
+      "loss": 1.8562,
+      "step": 37114
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2428097345132744e-06,
+      "loss": 2.4876,
+      "step": 37115
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2427623261694059e-06,
+      "loss": 1.7264,
+      "step": 37116
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2427149178255374e-06,
+      "loss": 2.8875,
+      "step": 37117
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2426675094816688e-06,
+      "loss": 2.3834,
+      "step": 37118
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2426201011378003e-06,
+      "loss": 0.94,
+      "step": 37119
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2425726927939316e-06,
+      "loss": 2.5622,
+      "step": 37120
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2425252844500633e-06,
+      "loss": 1.1811,
+      "step": 37121
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2424778761061948e-06,
+      "loss": 1.4017,
+      "step": 37122
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.242430467762326e-06,
+      "loss": 2.6413,
+      "step": 37123
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2423830594184578e-06,
+      "loss": 2.4426,
+      "step": 37124
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2423356510745893e-06,
+      "loss": 2.1063,
+      "step": 37125
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2422882427307206e-06,
+      "loss": 1.322,
+      "step": 37126
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.242240834386852e-06,
+      "loss": 0.9403,
+      "step": 37127
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2421934260429836e-06,
+      "loss": 3.6792,
+      "step": 37128
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.242146017699115e-06,
+      "loss": 1.8158,
+      "step": 37129
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2420986093552465e-06,
+      "loss": 1.0898,
+      "step": 37130
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.242051201011378e-06,
+      "loss": 1.5268,
+      "step": 37131
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2420037926675095e-06,
+      "loss": 2.7108,
+      "step": 37132
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.241956384323641e-06,
+      "loss": 1.7439,
+      "step": 37133
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2419089759797725e-06,
+      "loss": 4.0956,
+      "step": 37134
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.241861567635904e-06,
+      "loss": 1.4612,
+      "step": 37135
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2418141592920355e-06,
+      "loss": 2.3791,
+      "step": 37136
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.241766750948167e-06,
+      "loss": 4.5746,
+      "step": 37137
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2417193426042985e-06,
+      "loss": 1.096,
+      "step": 37138
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2416719342604298e-06,
+      "loss": 3.0585,
+      "step": 37139
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2416245259165612e-06,
+      "loss": 0.5773,
+      "step": 37140
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.241577117572693e-06,
+      "loss": 0.396,
+      "step": 37141
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2415297092288242e-06,
+      "loss": 2.8709,
+      "step": 37142
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 1.2414823008849557e-06,
+      "loss": 2.4478,
+      "step": 37143
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2414348925410874e-06,
+      "loss": 2.3942,
+      "step": 37144
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2413874841972187e-06,
+      "loss": 2.9831,
+      "step": 37145
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2413400758533502e-06,
+      "loss": 1.9619,
+      "step": 37146
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2412926675094817e-06,
+      "loss": 0.6998,
+      "step": 37147
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2412452591656132e-06,
+      "loss": 3.0341,
+      "step": 37148
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2411978508217447e-06,
+      "loss": 1.3569,
+      "step": 37149
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2411504424778762e-06,
+      "loss": 1.41,
+      "step": 37150
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2411030341340077e-06,
+      "loss": 1.4982,
+      "step": 37151
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2410556257901392e-06,
+      "loss": 1.8637,
+      "step": 37152
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2410082174462704e-06,
+      "loss": 1.3586,
+      "step": 37153
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2409608091024021e-06,
+      "loss": 2.6212,
+      "step": 37154
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2409134007585336e-06,
+      "loss": 1.0132,
+      "step": 37155
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.240865992414665e-06,
+      "loss": 1.5639,
+      "step": 37156
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2408185840707966e-06,
+      "loss": 2.3618,
+      "step": 37157
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2407711757269281e-06,
+      "loss": 1.211,
+      "step": 37158
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2407237673830594e-06,
+      "loss": 2.6189,
+      "step": 37159
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2406763590391909e-06,
+      "loss": 1.9072,
+      "step": 37160
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2406289506953224e-06,
+      "loss": 2.5445,
+      "step": 37161
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2405815423514539e-06,
+      "loss": 0.3973,
+      "step": 37162
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2405341340075854e-06,
+      "loss": 1.49,
+      "step": 37163
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2404867256637169e-06,
+      "loss": 1.9568,
+      "step": 37164
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2404393173198483e-06,
+      "loss": 3.0102,
+      "step": 37165
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2403919089759798e-06,
+      "loss": 2.8949,
+      "step": 37166
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2403445006321113e-06,
+      "loss": 0.9684,
+      "step": 37167
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2402970922882428e-06,
+      "loss": 3.2039,
+      "step": 37168
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2402496839443743e-06,
+      "loss": 1.9502,
+      "step": 37169
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2402022756005056e-06,
+      "loss": 1.207,
+      "step": 37170
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2401548672566373e-06,
+      "loss": 2.7201,
+      "step": 37171
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2401074589127686e-06,
+      "loss": 2.7445,
+      "step": 37172
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2400600505689e-06,
+      "loss": 2.3196,
+      "step": 37173
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2400126422250318e-06,
+      "loss": 2.0149,
+      "step": 37174
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.239965233881163e-06,
+      "loss": 3.1538,
+      "step": 37175
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2399178255372945e-06,
+      "loss": 2.3231,
+      "step": 37176
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.239870417193426e-06,
+      "loss": 1.4624,
+      "step": 37177
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2398230088495575e-06,
+      "loss": 2.3463,
+      "step": 37178
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.239775600505689e-06,
+      "loss": 1.2338,
+      "step": 37179
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2397281921618205e-06,
+      "loss": 0.6254,
+      "step": 37180
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.239680783817952e-06,
+      "loss": 2.2261,
+      "step": 37181
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2396333754740835e-06,
+      "loss": 1.295,
+      "step": 37182
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2395859671302148e-06,
+      "loss": 2.2638,
+      "step": 37183
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2395385587863465e-06,
+      "loss": 3.0891,
+      "step": 37184
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.239491150442478e-06,
+      "loss": 0.9304,
+      "step": 37185
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2394437420986093e-06,
+      "loss": 1.9609,
+      "step": 37186
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.239396333754741e-06,
+      "loss": 2.7709,
+      "step": 37187
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2393489254108725e-06,
+      "loss": 1.8386,
+      "step": 37188
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2393015170670037e-06,
+      "loss": 2.1724,
+      "step": 37189
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2392541087231352e-06,
+      "loss": 2.307,
+      "step": 37190
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2392067003792667e-06,
+      "loss": 1.2384,
+      "step": 37191
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2391592920353982e-06,
+      "loss": 3.2427,
+      "step": 37192
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2391118836915297e-06,
+      "loss": 1.7245,
+      "step": 37193
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2390644753476612e-06,
+      "loss": 2.4123,
+      "step": 37194
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2390170670037927e-06,
+      "loss": 1.6134,
+      "step": 37195
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2389696586599242e-06,
+      "loss": 0.8066,
+      "step": 37196
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2389222503160557e-06,
+      "loss": 1.8068,
+      "step": 37197
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2388748419721872e-06,
+      "loss": 2.6123,
+      "step": 37198
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2388274336283187e-06,
+      "loss": 2.3433,
+      "step": 37199
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2387800252844501e-06,
+      "loss": 2.2447,
+      "step": 37200
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2387326169405816e-06,
+      "loss": 2.1294,
+      "step": 37201
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.238685208596713e-06,
+      "loss": 3.3766,
+      "step": 37202
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2386378002528444e-06,
+      "loss": 3.7559,
+      "step": 37203
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2385903919089761e-06,
+      "loss": 1.4195,
+      "step": 37204
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2385429835651074e-06,
+      "loss": 2.5026,
+      "step": 37205
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 1.2384955752212389e-06,
+      "loss": 2.6992,
+      "step": 37206
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2384481668773706e-06,
+      "loss": 1.845,
+      "step": 37207
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2384007585335019e-06,
+      "loss": 2.5912,
+      "step": 37208
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2383533501896334e-06,
+      "loss": 0.977,
+      "step": 37209
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2383059418457649e-06,
+      "loss": 2.2469,
+      "step": 37210
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2382585335018963e-06,
+      "loss": 2.618,
+      "step": 37211
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2382111251580278e-06,
+      "loss": 3.0984,
+      "step": 37212
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2381637168141593e-06,
+      "loss": 2.8851,
+      "step": 37213
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2381163084702908e-06,
+      "loss": 1.4935,
+      "step": 37214
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2380689001264223e-06,
+      "loss": 2.6266,
+      "step": 37215
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2380214917825536e-06,
+      "loss": 2.6361,
+      "step": 37216
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2379740834386853e-06,
+      "loss": 1.7385,
+      "step": 37217
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2379266750948168e-06,
+      "loss": 3.1622,
+      "step": 37218
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237879266750948e-06,
+      "loss": 2.5498,
+      "step": 37219
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2378318584070798e-06,
+      "loss": 1.7586,
+      "step": 37220
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2377844500632113e-06,
+      "loss": 2.4498,
+      "step": 37221
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2377370417193425e-06,
+      "loss": 1.1782,
+      "step": 37222
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237689633375474e-06,
+      "loss": 2.765,
+      "step": 37223
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2376422250316055e-06,
+      "loss": 2.2323,
+      "step": 37224
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237594816687737e-06,
+      "loss": 2.3166,
+      "step": 37225
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2375474083438685e-06,
+      "loss": 3.4713,
+      "step": 37226
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2375e-06,
+      "loss": 1.3293,
+      "step": 37227
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2374525916561315e-06,
+      "loss": 1.2341,
+      "step": 37228
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237405183312263e-06,
+      "loss": 1.5517,
+      "step": 37229
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2373577749683945e-06,
+      "loss": 1.7549,
+      "step": 37230
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237310366624526e-06,
+      "loss": 2.9102,
+      "step": 37231
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2372629582806575e-06,
+      "loss": 2.7756,
+      "step": 37232
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237215549936789e-06,
+      "loss": 1.6515,
+      "step": 37233
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2371681415929205e-06,
+      "loss": 2.6749,
+      "step": 37234
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2371207332490517e-06,
+      "loss": 2.0793,
+      "step": 37235
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2370733249051832e-06,
+      "loss": 3.6308,
+      "step": 37236
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.237025916561315e-06,
+      "loss": 1.6704,
+      "step": 37237
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2369785082174462e-06,
+      "loss": 3.1156,
+      "step": 37238
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2369310998735777e-06,
+      "loss": 2.7968,
+      "step": 37239
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2368836915297094e-06,
+      "loss": 1.7204,
+      "step": 37240
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2368362831858407e-06,
+      "loss": 2.9132,
+      "step": 37241
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2367888748419722e-06,
+      "loss": 1.6498,
+      "step": 37242
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2367414664981037e-06,
+      "loss": 3.6172,
+      "step": 37243
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2366940581542352e-06,
+      "loss": 2.1171,
+      "step": 37244
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2366466498103667e-06,
+      "loss": 2.2014,
+      "step": 37245
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2365992414664981e-06,
+      "loss": 1.4059,
+      "step": 37246
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2365518331226296e-06,
+      "loss": 2.2401,
+      "step": 37247
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2365044247787611e-06,
+      "loss": 3.0414,
+      "step": 37248
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2364570164348924e-06,
+      "loss": 3.2684,
+      "step": 37249
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2364096080910241e-06,
+      "loss": 2.9254,
+      "step": 37250
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2363621997471556e-06,
+      "loss": 3.6944,
+      "step": 37251
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2363147914032869e-06,
+      "loss": 1.713,
+      "step": 37252
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2362673830594186e-06,
+      "loss": 1.7791,
+      "step": 37253
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.23621997471555e-06,
+      "loss": 1.2901,
+      "step": 37254
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2361725663716814e-06,
+      "loss": 1.7584,
+      "step": 37255
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2361251580278129e-06,
+      "loss": 2.6092,
+      "step": 37256
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2360777496839444e-06,
+      "loss": 0.6998,
+      "step": 37257
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2360303413400758e-06,
+      "loss": 2.6129,
+      "step": 37258
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2359829329962073e-06,
+      "loss": 1.9949,
+      "step": 37259
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2359355246523388e-06,
+      "loss": 2.7477,
+      "step": 37260
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2358881163084703e-06,
+      "loss": 0.6531,
+      "step": 37261
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2358407079646018e-06,
+      "loss": 2.2738,
+      "step": 37262
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2357932996207333e-06,
+      "loss": 3.411,
+      "step": 37263
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2357458912768648e-06,
+      "loss": 0.9771,
+      "step": 37264
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2356984829329963e-06,
+      "loss": 1.6983,
+      "step": 37265
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2356510745891278e-06,
+      "loss": 1.7241,
+      "step": 37266
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2356036662452593e-06,
+      "loss": 2.3072,
+      "step": 37267
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.2355562579013906e-06,
+      "loss": 2.4677,
+      "step": 37268
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 1.235508849557522e-06,
+      "loss": 2.2543,
+      "step": 37269
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2354614412136538e-06,
+      "loss": 2.6363,
+      "step": 37270
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.235414032869785e-06,
+      "loss": 1.5593,
+      "step": 37271
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2353666245259165e-06,
+      "loss": 3.2244,
+      "step": 37272
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2353192161820482e-06,
+      "loss": 2.7076,
+      "step": 37273
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2352718078381795e-06,
+      "loss": 1.5535,
+      "step": 37274
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.235224399494311e-06,
+      "loss": 1.2869,
+      "step": 37275
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2351769911504425e-06,
+      "loss": 2.8486,
+      "step": 37276
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.235129582806574e-06,
+      "loss": 2.4536,
+      "step": 37277
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2350821744627055e-06,
+      "loss": 1.6214,
+      "step": 37278
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.235034766118837e-06,
+      "loss": 1.55,
+      "step": 37279
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2349873577749685e-06,
+      "loss": 3.6685,
+      "step": 37280
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2349399494311e-06,
+      "loss": 3.0296,
+      "step": 37281
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2348925410872312e-06,
+      "loss": 1.7925,
+      "step": 37282
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.234845132743363e-06,
+      "loss": 3.285,
+      "step": 37283
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2347977243994944e-06,
+      "loss": 2.9019,
+      "step": 37284
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2347503160556257e-06,
+      "loss": 0.9377,
+      "step": 37285
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2347029077117574e-06,
+      "loss": 1.7011,
+      "step": 37286
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.234655499367889e-06,
+      "loss": 1.5796,
+      "step": 37287
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2346080910240202e-06,
+      "loss": 1.1082,
+      "step": 37288
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2345606826801517e-06,
+      "loss": 2.9987,
+      "step": 37289
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2345132743362832e-06,
+      "loss": 0.7266,
+      "step": 37290
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2344658659924147e-06,
+      "loss": 1.9727,
+      "step": 37291
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2344184576485462e-06,
+      "loss": 3.0589,
+      "step": 37292
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2343710493046776e-06,
+      "loss": 2.4037,
+      "step": 37293
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2343236409608091e-06,
+      "loss": 1.8574,
+      "step": 37294
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2342762326169406e-06,
+      "loss": 3.5792,
+      "step": 37295
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2342288242730721e-06,
+      "loss": 2.5724,
+      "step": 37296
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2341814159292036e-06,
+      "loss": 3.6762,
+      "step": 37297
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2341340075853351e-06,
+      "loss": 1.8003,
+      "step": 37298
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2340865992414666e-06,
+      "loss": 0.9927,
+      "step": 37299
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.234039190897598e-06,
+      "loss": 1.2285,
+      "step": 37300
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2339917825537294e-06,
+      "loss": 3.2236,
+      "step": 37301
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2339443742098609e-06,
+      "loss": 1.8976,
+      "step": 37302
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2338969658659926e-06,
+      "loss": 2.4741,
+      "step": 37303
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2338495575221238e-06,
+      "loss": 2.4428,
+      "step": 37304
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2338021491782553e-06,
+      "loss": 3.2629,
+      "step": 37305
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.233754740834387e-06,
+      "loss": 0.7589,
+      "step": 37306
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2337073324905183e-06,
+      "loss": 1.1232,
+      "step": 37307
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2336599241466498e-06,
+      "loss": 2.9169,
+      "step": 37308
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2336125158027813e-06,
+      "loss": 2.0979,
+      "step": 37309
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2335651074589128e-06,
+      "loss": 1.9628,
+      "step": 37310
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2335176991150443e-06,
+      "loss": 3.1303,
+      "step": 37311
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2334702907711756e-06,
+      "loss": 2.3099,
+      "step": 37312
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2334228824273073e-06,
+      "loss": 3.4864,
+      "step": 37313
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2333754740834388e-06,
+      "loss": 2.3783,
+      "step": 37314
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.23332806573957e-06,
+      "loss": 2.3314,
+      "step": 37315
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2332806573957018e-06,
+      "loss": 2.892,
+      "step": 37316
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2332332490518332e-06,
+      "loss": 4.0403,
+      "step": 37317
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2331858407079645e-06,
+      "loss": 1.8228,
+      "step": 37318
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.233138432364096e-06,
+      "loss": 1.2552,
+      "step": 37319
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2330910240202277e-06,
+      "loss": 1.9894,
+      "step": 37320
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.233043615676359e-06,
+      "loss": 2.2779,
+      "step": 37321
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2329962073324905e-06,
+      "loss": 1.98,
+      "step": 37322
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.232948798988622e-06,
+      "loss": 2.2504,
+      "step": 37323
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2329013906447535e-06,
+      "loss": 2.6095,
+      "step": 37324
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.232853982300885e-06,
+      "loss": 1.1035,
+      "step": 37325
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2328065739570165e-06,
+      "loss": 2.7021,
+      "step": 37326
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.232759165613148e-06,
+      "loss": 3.4695,
+      "step": 37327
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2327117572692794e-06,
+      "loss": 1.5842,
+      "step": 37328
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.232664348925411e-06,
+      "loss": 1.4975,
+      "step": 37329
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2326169405815424e-06,
+      "loss": 0.8227,
+      "step": 37330
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.232569532237674e-06,
+      "loss": 1.7306,
+      "step": 37331
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.2325221238938052e-06,
+      "loss": 1.8994,
+      "step": 37332
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 1.232474715549937e-06,
+      "loss": 2.2176,
+      "step": 37333
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2324273072060682e-06,
+      "loss": 3.586,
+      "step": 37334
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2323798988621997e-06,
+      "loss": 2.0732,
+      "step": 37335
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2323324905183314e-06,
+      "loss": 2.6718,
+      "step": 37336
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2322850821744627e-06,
+      "loss": 2.8449,
+      "step": 37337
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2322376738305942e-06,
+      "loss": 1.8037,
+      "step": 37338
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2321902654867257e-06,
+      "loss": 0.5771,
+      "step": 37339
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2321428571428571e-06,
+      "loss": 1.1279,
+      "step": 37340
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2320954487989886e-06,
+      "loss": 1.09,
+      "step": 37341
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2320480404551201e-06,
+      "loss": 1.8887,
+      "step": 37342
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2320006321112516e-06,
+      "loss": 2.839,
+      "step": 37343
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2319532237673831e-06,
+      "loss": 1.1454,
+      "step": 37344
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2319058154235144e-06,
+      "loss": 2.9383,
+      "step": 37345
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.231858407079646e-06,
+      "loss": 1.7807,
+      "step": 37346
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2318109987357776e-06,
+      "loss": 1.8622,
+      "step": 37347
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2317635903919089e-06,
+      "loss": 5.6751,
+      "step": 37348
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2317161820480406e-06,
+      "loss": 3.2393,
+      "step": 37349
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.231668773704172e-06,
+      "loss": 3.28,
+      "step": 37350
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2316213653603033e-06,
+      "loss": 3.0033,
+      "step": 37351
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2315739570164348e-06,
+      "loss": 2.2081,
+      "step": 37352
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2315265486725665e-06,
+      "loss": 3.0053,
+      "step": 37353
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2314791403286978e-06,
+      "loss": 3.0056,
+      "step": 37354
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2314317319848293e-06,
+      "loss": 1.5048,
+      "step": 37355
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2313843236409608e-06,
+      "loss": 2.1479,
+      "step": 37356
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2313369152970923e-06,
+      "loss": 1.8797,
+      "step": 37357
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2312895069532238e-06,
+      "loss": 1.2142,
+      "step": 37358
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2312420986093553e-06,
+      "loss": 1.9953,
+      "step": 37359
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2311946902654868e-06,
+      "loss": 2.1691,
+      "step": 37360
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2311472819216183e-06,
+      "loss": 2.2378,
+      "step": 37361
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2310998735777498e-06,
+      "loss": 2.8607,
+      "step": 37362
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2310524652338813e-06,
+      "loss": 2.3952,
+      "step": 37363
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2310050568900127e-06,
+      "loss": 2.2115,
+      "step": 37364
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230957648546144e-06,
+      "loss": 0.5354,
+      "step": 37365
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2309102402022757e-06,
+      "loss": 2.0694,
+      "step": 37366
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230862831858407e-06,
+      "loss": 1.0808,
+      "step": 37367
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2308154235145385e-06,
+      "loss": 3.2148,
+      "step": 37368
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2307680151706702e-06,
+      "loss": 1.4662,
+      "step": 37369
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2307206068268015e-06,
+      "loss": 2.4778,
+      "step": 37370
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230673198482933e-06,
+      "loss": 1.343,
+      "step": 37371
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2306257901390645e-06,
+      "loss": 1.9882,
+      "step": 37372
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230578381795196e-06,
+      "loss": 2.2389,
+      "step": 37373
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2305309734513275e-06,
+      "loss": 2.6154,
+      "step": 37374
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230483565107459e-06,
+      "loss": 2.0076,
+      "step": 37375
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2304361567635904e-06,
+      "loss": 2.0372,
+      "step": 37376
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230388748419722e-06,
+      "loss": 0.7963,
+      "step": 37377
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2303413400758532e-06,
+      "loss": 2.0902,
+      "step": 37378
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.230293931731985e-06,
+      "loss": 1.301,
+      "step": 37379
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2302465233881164e-06,
+      "loss": 1.2816,
+      "step": 37380
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2301991150442477e-06,
+      "loss": 2.568,
+      "step": 37381
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2301517067003794e-06,
+      "loss": 2.0399,
+      "step": 37382
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2301042983565109e-06,
+      "loss": 1.8846,
+      "step": 37383
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2300568900126422e-06,
+      "loss": 1.239,
+      "step": 37384
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2300094816687737e-06,
+      "loss": 2.0139,
+      "step": 37385
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2299620733249054e-06,
+      "loss": 1.5824,
+      "step": 37386
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2299146649810366e-06,
+      "loss": 2.3819,
+      "step": 37387
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2298672566371681e-06,
+      "loss": 1.0806,
+      "step": 37388
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2298198482932996e-06,
+      "loss": 2.8979,
+      "step": 37389
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2297724399494311e-06,
+      "loss": 2.3925,
+      "step": 37390
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2297250316055626e-06,
+      "loss": 2.5175,
+      "step": 37391
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.229677623261694e-06,
+      "loss": 2.9985,
+      "step": 37392
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2296302149178256e-06,
+      "loss": 1.4092,
+      "step": 37393
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.229582806573957e-06,
+      "loss": 1.9493,
+      "step": 37394
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.2295353982300886e-06,
+      "loss": 0.8327,
+      "step": 37395
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 1.22948798988622e-06,
+      "loss": 2.5336,
+      "step": 37396
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2294405815423516e-06,
+      "loss": 2.1485,
+      "step": 37397
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2293931731984828e-06,
+      "loss": 5.483,
+      "step": 37398
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2293457648546145e-06,
+      "loss": 2.874,
+      "step": 37399
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2292983565107458e-06,
+      "loss": 1.3569,
+      "step": 37400
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2292509481668773e-06,
+      "loss": 1.1373,
+      "step": 37401
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.229203539823009e-06,
+      "loss": 2.9916,
+      "step": 37402
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2291561314791403e-06,
+      "loss": 2.8773,
+      "step": 37403
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2291087231352718e-06,
+      "loss": 1.9286,
+      "step": 37404
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2290613147914033e-06,
+      "loss": 3.2134,
+      "step": 37405
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2290139064475348e-06,
+      "loss": 2.3495,
+      "step": 37406
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2289664981036663e-06,
+      "loss": 3.1412,
+      "step": 37407
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2289190897597978e-06,
+      "loss": 1.8602,
+      "step": 37408
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2288716814159293e-06,
+      "loss": 3.7968,
+      "step": 37409
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2288242730720607e-06,
+      "loss": 2.1778,
+      "step": 37410
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.228776864728192e-06,
+      "loss": 2.8213,
+      "step": 37411
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2287294563843237e-06,
+      "loss": 2.6134,
+      "step": 37412
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2286820480404552e-06,
+      "loss": 0.8074,
+      "step": 37413
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2286346396965865e-06,
+      "loss": 1.344,
+      "step": 37414
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2285872313527182e-06,
+      "loss": 2.2449,
+      "step": 37415
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2285398230088497e-06,
+      "loss": 2.298,
+      "step": 37416
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.228492414664981e-06,
+      "loss": 1.8785,
+      "step": 37417
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2284450063211125e-06,
+      "loss": 3.0982,
+      "step": 37418
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2283975979772442e-06,
+      "loss": 2.0076,
+      "step": 37419
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2283501896333755e-06,
+      "loss": 2.4362,
+      "step": 37420
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.228302781289507e-06,
+      "loss": 1.3698,
+      "step": 37421
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2282553729456384e-06,
+      "loss": 1.7007,
+      "step": 37422
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.22820796460177e-06,
+      "loss": 2.4766,
+      "step": 37423
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2281605562579014e-06,
+      "loss": 1.654,
+      "step": 37424
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.228113147914033e-06,
+      "loss": 3.3337,
+      "step": 37425
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2280657395701644e-06,
+      "loss": 0.9434,
+      "step": 37426
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.228018331226296e-06,
+      "loss": 2.2657,
+      "step": 37427
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2279709228824274e-06,
+      "loss": 2.8303,
+      "step": 37428
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2279235145385589e-06,
+      "loss": 1.3721,
+      "step": 37429
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2278761061946904e-06,
+      "loss": 1.8336,
+      "step": 37430
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2278286978508217e-06,
+      "loss": 2.0493,
+      "step": 37431
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2277812895069534e-06,
+      "loss": 1.6171,
+      "step": 37432
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2277338811630846e-06,
+      "loss": 0.5551,
+      "step": 37433
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2276864728192161e-06,
+      "loss": 2.7621,
+      "step": 37434
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2276390644753478e-06,
+      "loss": 1.8446,
+      "step": 37435
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2275916561314791e-06,
+      "loss": 2.1323,
+      "step": 37436
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2275442477876106e-06,
+      "loss": 4.0793,
+      "step": 37437
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2274968394437421e-06,
+      "loss": 2.2745,
+      "step": 37438
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2274494310998736e-06,
+      "loss": 2.1704,
+      "step": 37439
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.227402022756005e-06,
+      "loss": 2.4339,
+      "step": 37440
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2273546144121366e-06,
+      "loss": 1.4593,
+      "step": 37441
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.227307206068268e-06,
+      "loss": 1.5134,
+      "step": 37442
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2272597977243996e-06,
+      "loss": 2.3036,
+      "step": 37443
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2272123893805308e-06,
+      "loss": 3.1753,
+      "step": 37444
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2271649810366626e-06,
+      "loss": 3.405,
+      "step": 37445
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.227117572692794e-06,
+      "loss": 1.6528,
+      "step": 37446
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2270701643489253e-06,
+      "loss": 3.4466,
+      "step": 37447
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.227022756005057e-06,
+      "loss": 2.848,
+      "step": 37448
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2269753476611885e-06,
+      "loss": 1.6335,
+      "step": 37449
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2269279393173198e-06,
+      "loss": 1.776,
+      "step": 37450
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2268805309734513e-06,
+      "loss": 1.95,
+      "step": 37451
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.226833122629583e-06,
+      "loss": 1.5885,
+      "step": 37452
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2267857142857143e-06,
+      "loss": 1.3401,
+      "step": 37453
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2267383059418458e-06,
+      "loss": 1.299,
+      "step": 37454
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2266908975979773e-06,
+      "loss": 1.5858,
+      "step": 37455
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2266434892541088e-06,
+      "loss": 2.2075,
+      "step": 37456
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2265960809102402e-06,
+      "loss": 2.2713,
+      "step": 37457
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2265486725663717e-06,
+      "loss": 1.1397,
+      "step": 37458
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 1.2265012642225032e-06,
+      "loss": 3.415,
+      "step": 37459
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2264538558786347e-06,
+      "loss": 1.8006,
+      "step": 37460
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.226406447534766e-06,
+      "loss": 0.9996,
+      "step": 37461
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2263590391908977e-06,
+      "loss": 2.3006,
+      "step": 37462
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2263116308470292e-06,
+      "loss": 0.524,
+      "step": 37463
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2262642225031605e-06,
+      "loss": 0.9908,
+      "step": 37464
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2262168141592922e-06,
+      "loss": 1.1338,
+      "step": 37465
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2261694058154235e-06,
+      "loss": 1.6789,
+      "step": 37466
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.226121997471555e-06,
+      "loss": 0.511,
+      "step": 37467
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2260745891276864e-06,
+      "loss": 1.9891,
+      "step": 37468
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.226027180783818e-06,
+      "loss": 2.2025,
+      "step": 37469
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2259797724399494e-06,
+      "loss": 1.9544,
+      "step": 37470
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.225932364096081e-06,
+      "loss": 3.6549,
+      "step": 37471
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2258849557522124e-06,
+      "loss": 2.7218,
+      "step": 37472
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.225837547408344e-06,
+      "loss": 1.7247,
+      "step": 37473
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2257901390644754e-06,
+      "loss": 2.827,
+      "step": 37474
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2257427307206069e-06,
+      "loss": 2.0253,
+      "step": 37475
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2256953223767384e-06,
+      "loss": 2.4762,
+      "step": 37476
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2256479140328697e-06,
+      "loss": 2.1841,
+      "step": 37477
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2256005056890014e-06,
+      "loss": 3.2042,
+      "step": 37478
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2255530973451329e-06,
+      "loss": 0.9175,
+      "step": 37479
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2255056890012641e-06,
+      "loss": 2.5723,
+      "step": 37480
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2254582806573956e-06,
+      "loss": 1.801,
+      "step": 37481
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2254108723135273e-06,
+      "loss": 2.8422,
+      "step": 37482
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2253634639696586e-06,
+      "loss": 3.4562,
+      "step": 37483
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2253160556257901e-06,
+      "loss": 2.036,
+      "step": 37484
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2252686472819218e-06,
+      "loss": 1.4386,
+      "step": 37485
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.225221238938053e-06,
+      "loss": 1.6703,
+      "step": 37486
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2251738305941846e-06,
+      "loss": 1.6,
+      "step": 37487
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.225126422250316e-06,
+      "loss": 2.8386,
+      "step": 37488
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2250790139064476e-06,
+      "loss": 1.8408,
+      "step": 37489
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.225031605562579e-06,
+      "loss": 1.0752,
+      "step": 37490
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2249841972187106e-06,
+      "loss": 3.1368,
+      "step": 37491
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.224936788874842e-06,
+      "loss": 2.9914,
+      "step": 37492
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2248893805309735e-06,
+      "loss": 2.8033,
+      "step": 37493
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2248419721871048e-06,
+      "loss": 2.3713,
+      "step": 37494
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2247945638432365e-06,
+      "loss": 1.5876,
+      "step": 37495
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.224747155499368e-06,
+      "loss": 2.9706,
+      "step": 37496
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2246997471554993e-06,
+      "loss": 2.0028,
+      "step": 37497
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.224652338811631e-06,
+      "loss": 2.0866,
+      "step": 37498
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2246049304677623e-06,
+      "loss": 1.5934,
+      "step": 37499
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2245575221238938e-06,
+      "loss": 2.8632,
+      "step": 37500
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2245101137800253e-06,
+      "loss": 2.3419,
+      "step": 37501
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2244627054361568e-06,
+      "loss": 2.2038,
+      "step": 37502
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2244152970922883e-06,
+      "loss": 1.6533,
+      "step": 37503
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2243678887484197e-06,
+      "loss": 1.6348,
+      "step": 37504
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2243204804045512e-06,
+      "loss": 2.6476,
+      "step": 37505
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2242730720606827e-06,
+      "loss": 3.7319,
+      "step": 37506
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2242256637168142e-06,
+      "loss": 1.6187,
+      "step": 37507
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2241782553729457e-06,
+      "loss": 1.3801,
+      "step": 37508
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2241308470290772e-06,
+      "loss": 3.4806,
+      "step": 37509
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2240834386852085e-06,
+      "loss": 1.2319,
+      "step": 37510
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2240360303413402e-06,
+      "loss": 2.208,
+      "step": 37511
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2239886219974717e-06,
+      "loss": 2.264,
+      "step": 37512
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.223941213653603e-06,
+      "loss": 3.1858,
+      "step": 37513
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2238938053097345e-06,
+      "loss": 1.8432,
+      "step": 37514
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2238463969658662e-06,
+      "loss": 1.9227,
+      "step": 37515
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2237989886219974e-06,
+      "loss": 2.8179,
+      "step": 37516
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.223751580278129e-06,
+      "loss": 2.1932,
+      "step": 37517
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2237041719342606e-06,
+      "loss": 2.0101,
+      "step": 37518
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.223656763590392e-06,
+      "loss": 1.7607,
+      "step": 37519
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2236093552465234e-06,
+      "loss": 1.8313,
+      "step": 37520
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.223561946902655e-06,
+      "loss": 1.3351,
+      "step": 37521
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2235145385587864e-06,
+      "loss": 1.8862,
+      "step": 37522
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 1.2234671302149179e-06,
+      "loss": 2.3934,
+      "step": 37523
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2234197218710494e-06,
+      "loss": 3.2332,
+      "step": 37524
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2233723135271809e-06,
+      "loss": 2.4507,
+      "step": 37525
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2233249051833124e-06,
+      "loss": 2.5909,
+      "step": 37526
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2232774968394436e-06,
+      "loss": 1.791,
+      "step": 37527
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2232300884955753e-06,
+      "loss": 1.9545,
+      "step": 37528
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2231826801517068e-06,
+      "loss": 2.0931,
+      "step": 37529
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2231352718078381e-06,
+      "loss": 3.1736,
+      "step": 37530
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2230878634639698e-06,
+      "loss": 2.2814,
+      "step": 37531
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.223040455120101e-06,
+      "loss": 2.7795,
+      "step": 37532
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2229930467762326e-06,
+      "loss": 1.8631,
+      "step": 37533
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.222945638432364e-06,
+      "loss": 2.128,
+      "step": 37534
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2228982300884956e-06,
+      "loss": 2.3435,
+      "step": 37535
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.222850821744627e-06,
+      "loss": 2.1022,
+      "step": 37536
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2228034134007586e-06,
+      "loss": 2.9619,
+      "step": 37537
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.22275600505689e-06,
+      "loss": 1.5251,
+      "step": 37538
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2227085967130215e-06,
+      "loss": 3.1535,
+      "step": 37539
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.222661188369153e-06,
+      "loss": 2.5902,
+      "step": 37540
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2226137800252845e-06,
+      "loss": 1.4088,
+      "step": 37541
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.222566371681416e-06,
+      "loss": 1.2598,
+      "step": 37542
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2225189633375473e-06,
+      "loss": 2.2052,
+      "step": 37543
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.222471554993679e-06,
+      "loss": 2.3621,
+      "step": 37544
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2224241466498105e-06,
+      "loss": 3.0678,
+      "step": 37545
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2223767383059418e-06,
+      "loss": 1.2955,
+      "step": 37546
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2223293299620733e-06,
+      "loss": 1.398,
+      "step": 37547
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.222281921618205e-06,
+      "loss": 1.56,
+      "step": 37548
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2222345132743363e-06,
+      "loss": 1.7126,
+      "step": 37549
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2221871049304677e-06,
+      "loss": 0.8583,
+      "step": 37550
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2221396965865995e-06,
+      "loss": 1.718,
+      "step": 37551
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2220922882427307e-06,
+      "loss": 2.4646,
+      "step": 37552
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2220448798988622e-06,
+      "loss": 0.7826,
+      "step": 37553
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2219974715549937e-06,
+      "loss": 2.3931,
+      "step": 37554
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2219500632111252e-06,
+      "loss": 2.7158,
+      "step": 37555
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2219026548672567e-06,
+      "loss": 2.047,
+      "step": 37556
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2218552465233882e-06,
+      "loss": 2.363,
+      "step": 37557
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2218078381795197e-06,
+      "loss": 2.5685,
+      "step": 37558
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2217604298356512e-06,
+      "loss": 2.6405,
+      "step": 37559
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2217130214917825e-06,
+      "loss": 1.9531,
+      "step": 37560
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2216656131479142e-06,
+      "loss": 2.1764,
+      "step": 37561
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2216182048040457e-06,
+      "loss": 2.823,
+      "step": 37562
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.221570796460177e-06,
+      "loss": 2.8453,
+      "step": 37563
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2215233881163086e-06,
+      "loss": 2.5196,
+      "step": 37564
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.22147597977244e-06,
+      "loss": 2.5251,
+      "step": 37565
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2214285714285714e-06,
+      "loss": 2.2936,
+      "step": 37566
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.221381163084703e-06,
+      "loss": 3.3186,
+      "step": 37567
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2213337547408344e-06,
+      "loss": 1.5796,
+      "step": 37568
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2212863463969659e-06,
+      "loss": 2.9606,
+      "step": 37569
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2212389380530974e-06,
+      "loss": 2.133,
+      "step": 37570
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2211915297092289e-06,
+      "loss": 2.3343,
+      "step": 37571
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2211441213653604e-06,
+      "loss": 2.1382,
+      "step": 37572
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2210967130214919e-06,
+      "loss": 2.0782,
+      "step": 37573
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2210493046776233e-06,
+      "loss": 2.0588,
+      "step": 37574
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2210018963337548e-06,
+      "loss": 2.0693,
+      "step": 37575
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2209544879898861e-06,
+      "loss": 2.1398,
+      "step": 37576
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2209070796460178e-06,
+      "loss": 3.3152,
+      "step": 37577
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2208596713021493e-06,
+      "loss": 2.7377,
+      "step": 37578
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2208122629582806e-06,
+      "loss": 1.1625,
+      "step": 37579
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.220764854614412e-06,
+      "loss": 1.2166,
+      "step": 37580
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2207174462705438e-06,
+      "loss": 2.4325,
+      "step": 37581
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.220670037926675e-06,
+      "loss": 1.1382,
+      "step": 37582
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2206226295828066e-06,
+      "loss": 2.91,
+      "step": 37583
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2205752212389383e-06,
+      "loss": 4.4177,
+      "step": 37584
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.2205278128950696e-06,
+      "loss": 3.6457,
+      "step": 37585
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 1.220480404551201e-06,
+      "loss": 1.4812,
+      "step": 37586
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2204329962073325e-06,
+      "loss": 3.1312,
+      "step": 37587
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.220385587863464e-06,
+      "loss": 1.4154,
+      "step": 37588
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2203381795195955e-06,
+      "loss": 2.1051,
+      "step": 37589
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2202907711757268e-06,
+      "loss": 3.6086,
+      "step": 37590
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2202433628318585e-06,
+      "loss": 3.0293,
+      "step": 37591
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.22019595448799e-06,
+      "loss": 1.2087,
+      "step": 37592
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2201485461441213e-06,
+      "loss": 1.3951,
+      "step": 37593
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.220101137800253e-06,
+      "loss": 2.8683,
+      "step": 37594
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2200537294563845e-06,
+      "loss": 2.0799,
+      "step": 37595
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2200063211125158e-06,
+      "loss": 2.0105,
+      "step": 37596
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2199589127686472e-06,
+      "loss": 0.6967,
+      "step": 37597
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2199115044247787e-06,
+      "loss": 2.0404,
+      "step": 37598
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2198640960809102e-06,
+      "loss": 1.354,
+      "step": 37599
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2198166877370417e-06,
+      "loss": 3.334,
+      "step": 37600
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2197692793931732e-06,
+      "loss": 1.6283,
+      "step": 37601
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2197218710493047e-06,
+      "loss": 2.6925,
+      "step": 37602
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2196744627054362e-06,
+      "loss": 1.3064,
+      "step": 37603
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2196270543615677e-06,
+      "loss": 4.1461,
+      "step": 37604
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2195796460176992e-06,
+      "loss": 2.7209,
+      "step": 37605
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2195322376738307e-06,
+      "loss": 2.5261,
+      "step": 37606
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2194848293299622e-06,
+      "loss": 2.5109,
+      "step": 37607
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2194374209860937e-06,
+      "loss": 1.0453,
+      "step": 37608
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.219390012642225e-06,
+      "loss": 1.0631,
+      "step": 37609
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2193426042983564e-06,
+      "loss": 1.5167,
+      "step": 37610
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2192951959544881e-06,
+      "loss": 1.3824,
+      "step": 37611
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2192477876106194e-06,
+      "loss": 1.6343,
+      "step": 37612
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.219200379266751e-06,
+      "loss": 2.1589,
+      "step": 37613
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2191529709228826e-06,
+      "loss": 1.4634,
+      "step": 37614
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2191055625790139e-06,
+      "loss": 2.0941,
+      "step": 37615
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2190581542351454e-06,
+      "loss": 2.052,
+      "step": 37616
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2190107458912769e-06,
+      "loss": 1.7161,
+      "step": 37617
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2189633375474084e-06,
+      "loss": 1.6206,
+      "step": 37618
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2189159292035399e-06,
+      "loss": 2.1115,
+      "step": 37619
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2188685208596714e-06,
+      "loss": 1.2797,
+      "step": 37620
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2188211125158028e-06,
+      "loss": 2.0148,
+      "step": 37621
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2187737041719343e-06,
+      "loss": 2.6332,
+      "step": 37622
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2187262958280656e-06,
+      "loss": 2.1377,
+      "step": 37623
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2186788874841973e-06,
+      "loss": 1.9623,
+      "step": 37624
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2186314791403288e-06,
+      "loss": 1.132,
+      "step": 37625
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.21858407079646e-06,
+      "loss": 2.9917,
+      "step": 37626
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2185366624525918e-06,
+      "loss": 3.6696,
+      "step": 37627
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.218489254108723e-06,
+      "loss": 2.5675,
+      "step": 37628
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2184418457648546e-06,
+      "loss": 2.5868,
+      "step": 37629
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.218394437420986e-06,
+      "loss": 1.1302,
+      "step": 37630
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2183470290771176e-06,
+      "loss": 2.2417,
+      "step": 37631
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.218299620733249e-06,
+      "loss": 1.4408,
+      "step": 37632
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2182522123893805e-06,
+      "loss": 1.4814,
+      "step": 37633
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.218204804045512e-06,
+      "loss": 3.9644,
+      "step": 37634
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2181573957016435e-06,
+      "loss": 1.6434,
+      "step": 37635
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.218109987357775e-06,
+      "loss": 2.1683,
+      "step": 37636
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2180625790139065e-06,
+      "loss": 1.5573,
+      "step": 37637
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.218015170670038e-06,
+      "loss": 1.8324,
+      "step": 37638
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2179677623261693e-06,
+      "loss": 2.8205,
+      "step": 37639
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.217920353982301e-06,
+      "loss": 4.0098,
+      "step": 37640
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2178729456384325e-06,
+      "loss": 2.3758,
+      "step": 37641
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2178255372945638e-06,
+      "loss": 1.7523,
+      "step": 37642
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2177781289506952e-06,
+      "loss": 2.7727,
+      "step": 37643
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.217730720606827e-06,
+      "loss": 1.9563,
+      "step": 37644
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2176833122629582e-06,
+      "loss": 2.2511,
+      "step": 37645
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2176359039190897e-06,
+      "loss": 1.9944,
+      "step": 37646
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2175884955752214e-06,
+      "loss": 0.9474,
+      "step": 37647
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2175410872313527e-06,
+      "loss": 2.5029,
+      "step": 37648
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 1.2174936788874842e-06,
+      "loss": 1.9609,
+      "step": 37649
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2174462705436157e-06,
+      "loss": 1.9705,
+      "step": 37650
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2173988621997472e-06,
+      "loss": 2.3397,
+      "step": 37651
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2173514538558787e-06,
+      "loss": 2.635,
+      "step": 37652
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2173040455120102e-06,
+      "loss": 2.5302,
+      "step": 37653
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2172566371681417e-06,
+      "loss": 2.3762,
+      "step": 37654
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2172092288242732e-06,
+      "loss": 1.8456,
+      "step": 37655
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2171618204804044e-06,
+      "loss": 2.9179,
+      "step": 37656
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2171144121365361e-06,
+      "loss": 1.793,
+      "step": 37657
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2170670037926676e-06,
+      "loss": 1.7301,
+      "step": 37658
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.217019595448799e-06,
+      "loss": 1.7496,
+      "step": 37659
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2169721871049306e-06,
+      "loss": 2.4389,
+      "step": 37660
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.216924778761062e-06,
+      "loss": 0.7663,
+      "step": 37661
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2168773704171934e-06,
+      "loss": 3.561,
+      "step": 37662
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2168299620733249e-06,
+      "loss": 1.8644,
+      "step": 37663
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2167825537294564e-06,
+      "loss": 4.5462,
+      "step": 37664
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2167351453855879e-06,
+      "loss": 1.889,
+      "step": 37665
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2166877370417194e-06,
+      "loss": 1.8952,
+      "step": 37666
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2166403286978509e-06,
+      "loss": 1.549,
+      "step": 37667
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2165929203539823e-06,
+      "loss": 2.5975,
+      "step": 37668
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2165455120101138e-06,
+      "loss": 2.7966,
+      "step": 37669
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2164981036662453e-06,
+      "loss": 2.3014,
+      "step": 37670
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2164506953223768e-06,
+      "loss": 2.0695,
+      "step": 37671
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.216403286978508e-06,
+      "loss": 2.9538,
+      "step": 37672
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2163558786346398e-06,
+      "loss": 1.5031,
+      "step": 37673
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2163084702907713e-06,
+      "loss": 1.7098,
+      "step": 37674
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2162610619469026e-06,
+      "loss": 2.6614,
+      "step": 37675
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.216213653603034e-06,
+      "loss": 2.3394,
+      "step": 37676
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2161662452591658e-06,
+      "loss": 2.6315,
+      "step": 37677
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.216118836915297e-06,
+      "loss": 1.2831,
+      "step": 37678
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2160714285714285e-06,
+      "loss": 1.8812,
+      "step": 37679
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2160240202275602e-06,
+      "loss": 2.7168,
+      "step": 37680
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2159766118836915e-06,
+      "loss": 2.569,
+      "step": 37681
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.215929203539823e-06,
+      "loss": 2.7516,
+      "step": 37682
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2158817951959545e-06,
+      "loss": 3.0608,
+      "step": 37683
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.215834386852086e-06,
+      "loss": 2.0402,
+      "step": 37684
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2157869785082175e-06,
+      "loss": 1.3844,
+      "step": 37685
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.215739570164349e-06,
+      "loss": 2.5628,
+      "step": 37686
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2156921618204805e-06,
+      "loss": 2.2847,
+      "step": 37687
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.215644753476612e-06,
+      "loss": 2.3785,
+      "step": 37688
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2155973451327433e-06,
+      "loss": 2.1075,
+      "step": 37689
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.215549936788875e-06,
+      "loss": 2.1046,
+      "step": 37690
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2155025284450065e-06,
+      "loss": 3.205,
+      "step": 37691
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2154551201011377e-06,
+      "loss": 2.2498,
+      "step": 37692
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2154077117572694e-06,
+      "loss": 1.9506,
+      "step": 37693
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2153603034134007e-06,
+      "loss": 2.3656,
+      "step": 37694
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2153128950695322e-06,
+      "loss": 0.6698,
+      "step": 37695
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2152654867256637e-06,
+      "loss": 2.1667,
+      "step": 37696
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2152180783817952e-06,
+      "loss": 2.8275,
+      "step": 37697
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2151706700379267e-06,
+      "loss": 1.4201,
+      "step": 37698
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2151232616940582e-06,
+      "loss": 2.7616,
+      "step": 37699
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2150758533501897e-06,
+      "loss": 2.1619,
+      "step": 37700
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2150284450063212e-06,
+      "loss": 1.6579,
+      "step": 37701
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2149810366624527e-06,
+      "loss": 2.9301,
+      "step": 37702
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2149336283185841e-06,
+      "loss": 0.5137,
+      "step": 37703
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2148862199747156e-06,
+      "loss": 1.4963,
+      "step": 37704
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.214838811630847e-06,
+      "loss": 2.9,
+      "step": 37705
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2147914032869786e-06,
+      "loss": 2.8302,
+      "step": 37706
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2147439949431101e-06,
+      "loss": 1.7707,
+      "step": 37707
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2146965865992414e-06,
+      "loss": 1.1233,
+      "step": 37708
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2146491782553729e-06,
+      "loss": 2.4698,
+      "step": 37709
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2146017699115046e-06,
+      "loss": 0.6459,
+      "step": 37710
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2145543615676359e-06,
+      "loss": 1.7795,
+      "step": 37711
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.2145069532237674e-06,
+      "loss": 2.188,
+      "step": 37712
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 1.214459544879899e-06,
+      "loss": 2.0386,
+      "step": 37713
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2144121365360303e-06,
+      "loss": 2.1959,
+      "step": 37714
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2143647281921618e-06,
+      "loss": 2.02,
+      "step": 37715
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2143173198482933e-06,
+      "loss": 0.7799,
+      "step": 37716
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2142699115044248e-06,
+      "loss": 3.209,
+      "step": 37717
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2142225031605563e-06,
+      "loss": 1.4802,
+      "step": 37718
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2141750948166878e-06,
+      "loss": 1.8036,
+      "step": 37719
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2141276864728193e-06,
+      "loss": 1.7164,
+      "step": 37720
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2140802781289508e-06,
+      "loss": 0.2746,
+      "step": 37721
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.214032869785082e-06,
+      "loss": 1.3536,
+      "step": 37722
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2139854614412138e-06,
+      "loss": 2.4158,
+      "step": 37723
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2139380530973453e-06,
+      "loss": 1.8252,
+      "step": 37724
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2138906447534765e-06,
+      "loss": 2.1364,
+      "step": 37725
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2138432364096083e-06,
+      "loss": 3.1009,
+      "step": 37726
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2137958280657395e-06,
+      "loss": 1.3628,
+      "step": 37727
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.213748419721871e-06,
+      "loss": 1.3693,
+      "step": 37728
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2137010113780025e-06,
+      "loss": 2.2049,
+      "step": 37729
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.213653603034134e-06,
+      "loss": 1.7032,
+      "step": 37730
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2136061946902655e-06,
+      "loss": 1.7633,
+      "step": 37731
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.213558786346397e-06,
+      "loss": 2.3332,
+      "step": 37732
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2135113780025285e-06,
+      "loss": 1.0197,
+      "step": 37733
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.21346396965866e-06,
+      "loss": 1.6216,
+      "step": 37734
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2134165613147915e-06,
+      "loss": 3.4233,
+      "step": 37735
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.213369152970923e-06,
+      "loss": 2.2099,
+      "step": 37736
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2133217446270545e-06,
+      "loss": 2.3686,
+      "step": 37737
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2132743362831857e-06,
+      "loss": 2.1744,
+      "step": 37738
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2132269279393172e-06,
+      "loss": 2.3098,
+      "step": 37739
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.213179519595449e-06,
+      "loss": 1.7774,
+      "step": 37740
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2131321112515802e-06,
+      "loss": 1.5347,
+      "step": 37741
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2130847029077117e-06,
+      "loss": 1.4498,
+      "step": 37742
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2130372945638434e-06,
+      "loss": 3.3186,
+      "step": 37743
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2129898862199747e-06,
+      "loss": 2.4232,
+      "step": 37744
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2129424778761062e-06,
+      "loss": 4.2095,
+      "step": 37745
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2128950695322377e-06,
+      "loss": 2.2181,
+      "step": 37746
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2128476611883692e-06,
+      "loss": 1.7657,
+      "step": 37747
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2128002528445007e-06,
+      "loss": 2.6384,
+      "step": 37748
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2127528445006322e-06,
+      "loss": 1.0519,
+      "step": 37749
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2127054361567636e-06,
+      "loss": 1.5962,
+      "step": 37750
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2126580278128951e-06,
+      "loss": 1.509,
+      "step": 37751
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2126106194690264e-06,
+      "loss": 3.0648,
+      "step": 37752
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2125632111251581e-06,
+      "loss": 1.5747,
+      "step": 37753
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2125158027812896e-06,
+      "loss": 1.7095,
+      "step": 37754
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2124683944374209e-06,
+      "loss": 2.2758,
+      "step": 37755
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2124209860935526e-06,
+      "loss": 1.6415,
+      "step": 37756
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.212373577749684e-06,
+      "loss": 1.73,
+      "step": 37757
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2123261694058154e-06,
+      "loss": 2.3288,
+      "step": 37758
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2122787610619469e-06,
+      "loss": 1.1046,
+      "step": 37759
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2122313527180784e-06,
+      "loss": 2.1724,
+      "step": 37760
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2121839443742098e-06,
+      "loss": 2.8835,
+      "step": 37761
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2121365360303413e-06,
+      "loss": 1.6553,
+      "step": 37762
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2120891276864728e-06,
+      "loss": 1.7373,
+      "step": 37763
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2120417193426043e-06,
+      "loss": 2.6264,
+      "step": 37764
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2119943109987358e-06,
+      "loss": 2.7451,
+      "step": 37765
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2119469026548673e-06,
+      "loss": 3.5274,
+      "step": 37766
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2118994943109988e-06,
+      "loss": 1.1943,
+      "step": 37767
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2118520859671303e-06,
+      "loss": 3.6142,
+      "step": 37768
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2118046776232618e-06,
+      "loss": 1.7646,
+      "step": 37769
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2117572692793933e-06,
+      "loss": 1.7929,
+      "step": 37770
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2117098609355246e-06,
+      "loss": 2.228,
+      "step": 37771
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.211662452591656e-06,
+      "loss": 1.485,
+      "step": 37772
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2116150442477878e-06,
+      "loss": 1.3201,
+      "step": 37773
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.211567635903919e-06,
+      "loss": 3.0416,
+      "step": 37774
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2115202275600505e-06,
+      "loss": 1.9926,
+      "step": 37775
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 1.2114728192161822e-06,
+      "loss": 3.0323,
+      "step": 37776
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2114254108723135e-06,
+      "loss": 2.0355,
+      "step": 37777
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.211378002528445e-06,
+      "loss": 1.892,
+      "step": 37778
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2113305941845765e-06,
+      "loss": 1.527,
+      "step": 37779
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.211283185840708e-06,
+      "loss": 1.802,
+      "step": 37780
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2112357774968395e-06,
+      "loss": 1.1616,
+      "step": 37781
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.211188369152971e-06,
+      "loss": 3.0692,
+      "step": 37782
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2111409608091025e-06,
+      "loss": 3.029,
+      "step": 37783
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.211093552465234e-06,
+      "loss": 2.1139,
+      "step": 37784
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2110461441213652e-06,
+      "loss": 2.0821,
+      "step": 37785
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.210998735777497e-06,
+      "loss": 2.9536,
+      "step": 37786
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2109513274336284e-06,
+      "loss": 2.1808,
+      "step": 37787
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2109039190897597e-06,
+      "loss": 1.6421,
+      "step": 37788
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2108565107458914e-06,
+      "loss": 0.9304,
+      "step": 37789
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.210809102402023e-06,
+      "loss": 1.6244,
+      "step": 37790
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2107616940581542e-06,
+      "loss": 2.3176,
+      "step": 37791
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2107142857142857e-06,
+      "loss": 2.93,
+      "step": 37792
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2106668773704172e-06,
+      "loss": 3.3087,
+      "step": 37793
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2106194690265487e-06,
+      "loss": 1.7658,
+      "step": 37794
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2105720606826802e-06,
+      "loss": 2.7114,
+      "step": 37795
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2105246523388116e-06,
+      "loss": 1.4463,
+      "step": 37796
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2104772439949431e-06,
+      "loss": 2.5635,
+      "step": 37797
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2104298356510746e-06,
+      "loss": 1.3817,
+      "step": 37798
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2103824273072061e-06,
+      "loss": 3.2032,
+      "step": 37799
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2103350189633376e-06,
+      "loss": 1.5423,
+      "step": 37800
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2102876106194691e-06,
+      "loss": 2.0324,
+      "step": 37801
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2102402022756006e-06,
+      "loss": 2.5807,
+      "step": 37802
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.210192793931732e-06,
+      "loss": 1.5367,
+      "step": 37803
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2101453855878634e-06,
+      "loss": 2.198,
+      "step": 37804
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2100979772439949e-06,
+      "loss": 2.921,
+      "step": 37805
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2100505689001266e-06,
+      "loss": 1.8256,
+      "step": 37806
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2100031605562578e-06,
+      "loss": 2.3696,
+      "step": 37807
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2099557522123893e-06,
+      "loss": 1.8907,
+      "step": 37808
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.209908343868521e-06,
+      "loss": 2.0299,
+      "step": 37809
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2098609355246523e-06,
+      "loss": 3.6486,
+      "step": 37810
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2098135271807838e-06,
+      "loss": 2.4366,
+      "step": 37811
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2097661188369153e-06,
+      "loss": 2.4641,
+      "step": 37812
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2097187104930468e-06,
+      "loss": 2.7184,
+      "step": 37813
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2096713021491783e-06,
+      "loss": 2.3012,
+      "step": 37814
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2096238938053098e-06,
+      "loss": 2.0627,
+      "step": 37815
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2095764854614413e-06,
+      "loss": 2.1745,
+      "step": 37816
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2095290771175728e-06,
+      "loss": 3.024,
+      "step": 37817
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.209481668773704e-06,
+      "loss": 2.3173,
+      "step": 37818
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2094342604298358e-06,
+      "loss": 2.4777,
+      "step": 37819
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2093868520859672e-06,
+      "loss": 2.0548,
+      "step": 37820
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2093394437420985e-06,
+      "loss": 2.264,
+      "step": 37821
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2092920353982302e-06,
+      "loss": 3.1241,
+      "step": 37822
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2092446270543617e-06,
+      "loss": 1.0771,
+      "step": 37823
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.209197218710493e-06,
+      "loss": 0.8989,
+      "step": 37824
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2091498103666245e-06,
+      "loss": 3.3418,
+      "step": 37825
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.209102402022756e-06,
+      "loss": 2.9019,
+      "step": 37826
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2090549936788875e-06,
+      "loss": 1.0184,
+      "step": 37827
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.209007585335019e-06,
+      "loss": 3.2299,
+      "step": 37828
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2089601769911505e-06,
+      "loss": 2.2599,
+      "step": 37829
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.208912768647282e-06,
+      "loss": 1.191,
+      "step": 37830
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2088653603034134e-06,
+      "loss": 0.9771,
+      "step": 37831
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.208817951959545e-06,
+      "loss": 1.8287,
+      "step": 37832
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2087705436156764e-06,
+      "loss": 3.1657,
+      "step": 37833
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.208723135271808e-06,
+      "loss": 1.8861,
+      "step": 37834
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2086757269279394e-06,
+      "loss": 1.7925,
+      "step": 37835
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.208628318584071e-06,
+      "loss": 1.0049,
+      "step": 37836
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2085809102402022e-06,
+      "loss": 0.3252,
+      "step": 37837
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2085335018963337e-06,
+      "loss": 0.5749,
+      "step": 37838
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 1.2084860935524654e-06,
+      "loss": 2.4495,
+      "step": 37839
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2084386852085967e-06,
+      "loss": 2.6155,
+      "step": 37840
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2083912768647282e-06,
+      "loss": 1.7227,
+      "step": 37841
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2083438685208599e-06,
+      "loss": 2.6541,
+      "step": 37842
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2082964601769911e-06,
+      "loss": 2.4264,
+      "step": 37843
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2082490518331226e-06,
+      "loss": 2.8048,
+      "step": 37844
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2082016434892541e-06,
+      "loss": 0.3997,
+      "step": 37845
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2081542351453856e-06,
+      "loss": 2.5722,
+      "step": 37846
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2081068268015171e-06,
+      "loss": 1.9385,
+      "step": 37847
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2080594184576486e-06,
+      "loss": 2.6526,
+      "step": 37848
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.20801201011378e-06,
+      "loss": 2.9594,
+      "step": 37849
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2079646017699116e-06,
+      "loss": 2.3338,
+      "step": 37850
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2079171934260429e-06,
+      "loss": 3.0932,
+      "step": 37851
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2078697850821746e-06,
+      "loss": 2.3852,
+      "step": 37852
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.207822376738306e-06,
+      "loss": 0.462,
+      "step": 37853
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2077749683944373e-06,
+      "loss": 2.3546,
+      "step": 37854
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.207727560050569e-06,
+      "loss": 2.3796,
+      "step": 37855
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2076801517067005e-06,
+      "loss": 1.2298,
+      "step": 37856
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2076327433628318e-06,
+      "loss": 2.7555,
+      "step": 37857
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2075853350189633e-06,
+      "loss": 1.9235,
+      "step": 37858
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2075379266750948e-06,
+      "loss": 2.4258,
+      "step": 37859
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2074905183312263e-06,
+      "loss": 2.5255,
+      "step": 37860
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2074431099873578e-06,
+      "loss": 1.338,
+      "step": 37861
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2073957016434893e-06,
+      "loss": 1.4687,
+      "step": 37862
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2073482932996208e-06,
+      "loss": 1.565,
+      "step": 37863
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2073008849557523e-06,
+      "loss": 1.5613,
+      "step": 37864
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2072534766118838e-06,
+      "loss": 2.5499,
+      "step": 37865
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2072060682680153e-06,
+      "loss": 2.9033,
+      "step": 37866
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2071586599241467e-06,
+      "loss": 1.6454,
+      "step": 37867
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.207111251580278e-06,
+      "loss": 2.7873,
+      "step": 37868
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2070638432364097e-06,
+      "loss": 2.2048,
+      "step": 37869
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.207016434892541e-06,
+      "loss": 1.0832,
+      "step": 37870
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2069690265486725e-06,
+      "loss": 3.9001,
+      "step": 37871
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2069216182048042e-06,
+      "loss": 1.9639,
+      "step": 37872
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2068742098609355e-06,
+      "loss": 2.5282,
+      "step": 37873
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.206826801517067e-06,
+      "loss": 2.3684,
+      "step": 37874
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2067793931731985e-06,
+      "loss": 1.8125,
+      "step": 37875
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.20673198482933e-06,
+      "loss": 2.6187,
+      "step": 37876
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2066845764854615e-06,
+      "loss": 2.7832,
+      "step": 37877
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.206637168141593e-06,
+      "loss": 3.8929,
+      "step": 37878
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2065897597977244e-06,
+      "loss": 3.2768,
+      "step": 37879
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.206542351453856e-06,
+      "loss": 1.9126,
+      "step": 37880
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2064949431099872e-06,
+      "loss": 1.3513,
+      "step": 37881
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.206447534766119e-06,
+      "loss": 2.7263,
+      "step": 37882
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2064001264222504e-06,
+      "loss": 2.4429,
+      "step": 37883
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2063527180783817e-06,
+      "loss": 1.0805,
+      "step": 37884
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2063053097345134e-06,
+      "loss": 2.5367,
+      "step": 37885
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2062579013906449e-06,
+      "loss": 2.2372,
+      "step": 37886
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2062104930467762e-06,
+      "loss": 2.2655,
+      "step": 37887
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2061630847029077e-06,
+      "loss": 1.5777,
+      "step": 37888
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2061156763590394e-06,
+      "loss": 2.0625,
+      "step": 37889
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2060682680151706e-06,
+      "loss": 1.4387,
+      "step": 37890
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2060208596713021e-06,
+      "loss": 2.7836,
+      "step": 37891
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2059734513274336e-06,
+      "loss": 2.3758,
+      "step": 37892
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2059260429835651e-06,
+      "loss": 2.595,
+      "step": 37893
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2058786346396966e-06,
+      "loss": 2.3321,
+      "step": 37894
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.205831226295828e-06,
+      "loss": 2.7306,
+      "step": 37895
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2057838179519596e-06,
+      "loss": 1.7639,
+      "step": 37896
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.205736409608091e-06,
+      "loss": 1.7118,
+      "step": 37897
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2056890012642226e-06,
+      "loss": 1.1404,
+      "step": 37898
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.205641592920354e-06,
+      "loss": 2.4246,
+      "step": 37899
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2055941845764856e-06,
+      "loss": 2.1009,
+      "step": 37900
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2055467762326168e-06,
+      "loss": 3.0709,
+      "step": 37901
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2054993678887485e-06,
+      "loss": 1.4379,
+      "step": 37902
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 1.2054519595448798e-06,
+      "loss": 2.4287,
+      "step": 37903
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2054045512010113e-06,
+      "loss": 2.3879,
+      "step": 37904
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.205357142857143e-06,
+      "loss": 2.2376,
+      "step": 37905
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2053097345132743e-06,
+      "loss": 3.9509,
+      "step": 37906
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2052623261694058e-06,
+      "loss": 1.7674,
+      "step": 37907
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2052149178255373e-06,
+      "loss": 3.0555,
+      "step": 37908
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2051675094816688e-06,
+      "loss": 1.4845,
+      "step": 37909
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2051201011378003e-06,
+      "loss": 2.1158,
+      "step": 37910
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2050726927939318e-06,
+      "loss": 3.131,
+      "step": 37911
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2050252844500633e-06,
+      "loss": 2.6483,
+      "step": 37912
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2049778761061947e-06,
+      "loss": 1.1931,
+      "step": 37913
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.204930467762326e-06,
+      "loss": 1.2732,
+      "step": 37914
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2048830594184577e-06,
+      "loss": 1.8604,
+      "step": 37915
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2048356510745892e-06,
+      "loss": 2.7227,
+      "step": 37916
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2047882427307205e-06,
+      "loss": 0.6157,
+      "step": 37917
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2047408343868522e-06,
+      "loss": 2.5255,
+      "step": 37918
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2046934260429837e-06,
+      "loss": 1.6177,
+      "step": 37919
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.204646017699115e-06,
+      "loss": 3.7939,
+      "step": 37920
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2045986093552465e-06,
+      "loss": 2.0278,
+      "step": 37921
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2045512010113782e-06,
+      "loss": 2.4237,
+      "step": 37922
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2045037926675095e-06,
+      "loss": 2.9653,
+      "step": 37923
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.204456384323641e-06,
+      "loss": 3.133,
+      "step": 37924
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2044089759797724e-06,
+      "loss": 1.2652,
+      "step": 37925
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.204361567635904e-06,
+      "loss": 2.1191,
+      "step": 37926
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2043141592920354e-06,
+      "loss": 2.1076,
+      "step": 37927
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.204266750948167e-06,
+      "loss": 1.8851,
+      "step": 37928
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2042193426042984e-06,
+      "loss": 2.0321,
+      "step": 37929
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.20417193426043e-06,
+      "loss": 1.2357,
+      "step": 37930
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2041245259165614e-06,
+      "loss": 2.3189,
+      "step": 37931
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2040771175726929e-06,
+      "loss": 2.4472,
+      "step": 37932
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2040297092288244e-06,
+      "loss": 1.6988,
+      "step": 37933
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2039823008849557e-06,
+      "loss": 1.9741,
+      "step": 37934
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2039348925410874e-06,
+      "loss": 2.4296,
+      "step": 37935
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2038874841972186e-06,
+      "loss": 0.6441,
+      "step": 37936
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2038400758533501e-06,
+      "loss": 1.9321,
+      "step": 37937
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2037926675094818e-06,
+      "loss": 2.4961,
+      "step": 37938
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2037452591656131e-06,
+      "loss": 2.496,
+      "step": 37939
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2036978508217446e-06,
+      "loss": 2.2667,
+      "step": 37940
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2036504424778761e-06,
+      "loss": 1.912,
+      "step": 37941
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2036030341340076e-06,
+      "loss": 2.524,
+      "step": 37942
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.203555625790139e-06,
+      "loss": 1.6014,
+      "step": 37943
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2035082174462706e-06,
+      "loss": 1.2847,
+      "step": 37944
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.203460809102402e-06,
+      "loss": 1.9241,
+      "step": 37945
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2034134007585336e-06,
+      "loss": 2.8767,
+      "step": 37946
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2033659924146648e-06,
+      "loss": 1.3042,
+      "step": 37947
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2033185840707966e-06,
+      "loss": 1.7029,
+      "step": 37948
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.203271175726928e-06,
+      "loss": 2.1734,
+      "step": 37949
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2032237673830593e-06,
+      "loss": 1.4124,
+      "step": 37950
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.203176359039191e-06,
+      "loss": 0.6202,
+      "step": 37951
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2031289506953225e-06,
+      "loss": 2.9802,
+      "step": 37952
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2030815423514538e-06,
+      "loss": 2.4353,
+      "step": 37953
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2030341340075853e-06,
+      "loss": 1.4143,
+      "step": 37954
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.202986725663717e-06,
+      "loss": 1.2711,
+      "step": 37955
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2029393173198483e-06,
+      "loss": 1.2742,
+      "step": 37956
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2028919089759798e-06,
+      "loss": 1.5806,
+      "step": 37957
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2028445006321113e-06,
+      "loss": 3.4684,
+      "step": 37958
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2027970922882428e-06,
+      "loss": 0.7802,
+      "step": 37959
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2027496839443742e-06,
+      "loss": 2.3317,
+      "step": 37960
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2027022756005057e-06,
+      "loss": 2.428,
+      "step": 37961
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2026548672566372e-06,
+      "loss": 2.1204,
+      "step": 37962
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2026074589127687e-06,
+      "loss": 2.5082,
+      "step": 37963
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2025600505689002e-06,
+      "loss": 3.4602,
+      "step": 37964
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2025126422250317e-06,
+      "loss": 1.8632,
+      "step": 37965
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 1.2024652338811632e-06,
+      "loss": 3.0565,
+      "step": 37966
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2024178255372945e-06,
+      "loss": 3.4724,
+      "step": 37967
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2023704171934262e-06,
+      "loss": 2.2351,
+      "step": 37968
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2023230088495575e-06,
+      "loss": 1.8729,
+      "step": 37969
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.202275600505689e-06,
+      "loss": 0.7808,
+      "step": 37970
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2022281921618207e-06,
+      "loss": 2.6996,
+      "step": 37971
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.202180783817952e-06,
+      "loss": 1.9088,
+      "step": 37972
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2021333754740834e-06,
+      "loss": 1.0803,
+      "step": 37973
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.202085967130215e-06,
+      "loss": 1.0317,
+      "step": 37974
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2020385587863464e-06,
+      "loss": 2.8696,
+      "step": 37975
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.201991150442478e-06,
+      "loss": 2.9706,
+      "step": 37976
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2019437420986094e-06,
+      "loss": 2.1692,
+      "step": 37977
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2018963337547409e-06,
+      "loss": 1.9209,
+      "step": 37978
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2018489254108724e-06,
+      "loss": 1.5729,
+      "step": 37979
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2018015170670037e-06,
+      "loss": 2.1115,
+      "step": 37980
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2017541087231354e-06,
+      "loss": 2.0765,
+      "step": 37981
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2017067003792669e-06,
+      "loss": 2.045,
+      "step": 37982
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2016592920353981e-06,
+      "loss": 1.5277,
+      "step": 37983
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2016118836915298e-06,
+      "loss": 1.369,
+      "step": 37984
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2015644753476613e-06,
+      "loss": 1.1985,
+      "step": 37985
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2015170670037926e-06,
+      "loss": 5.9135,
+      "step": 37986
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2014696586599241e-06,
+      "loss": 2.0768,
+      "step": 37987
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2014222503160558e-06,
+      "loss": 1.6058,
+      "step": 37988
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.201374841972187e-06,
+      "loss": 1.7597,
+      "step": 37989
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2013274336283186e-06,
+      "loss": 1.0315,
+      "step": 37990
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.20128002528445e-06,
+      "loss": 1.8904,
+      "step": 37991
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2012326169405816e-06,
+      "loss": 1.7848,
+      "step": 37992
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.201185208596713e-06,
+      "loss": 1.6868,
+      "step": 37993
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2011378002528446e-06,
+      "loss": 1.6563,
+      "step": 37994
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.201090391908976e-06,
+      "loss": 1.5103,
+      "step": 37995
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2010429835651075e-06,
+      "loss": 3.1342,
+      "step": 37996
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.200995575221239e-06,
+      "loss": 1.4195,
+      "step": 37997
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2009481668773705e-06,
+      "loss": 2.3939,
+      "step": 37998
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.5335701598579041,
+      "eval_loss": 2.050567626953125,
+      "eval_runtime": 158.8365,
+      "eval_samples_per_second": 17.723,
+      "eval_steps_per_second": 4.432,
+      "step": 37998
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.200900758533502e-06,
+      "loss": 1.0384,
+      "step": 37999
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2008533501896333e-06,
+      "loss": 1.5866,
+      "step": 38000
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.200805941845765e-06,
+      "loss": 1.4666,
+      "step": 38001
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2007585335018963e-06,
+      "loss": 1.8245,
+      "step": 38002
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2007111251580278e-06,
+      "loss": 1.751,
+      "step": 38003
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2006637168141595e-06,
+      "loss": 3.0463,
+      "step": 38004
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2006163084702908e-06,
+      "loss": 1.6242,
+      "step": 38005
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2005689001264223e-06,
+      "loss": 1.8085,
+      "step": 38006
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2005214917825537e-06,
+      "loss": 3.1297,
+      "step": 38007
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2004740834386852e-06,
+      "loss": 2.5078,
+      "step": 38008
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2004266750948167e-06,
+      "loss": 1.012,
+      "step": 38009
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2003792667509482e-06,
+      "loss": 2.943,
+      "step": 38010
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2003318584070797e-06,
+      "loss": 2.1336,
+      "step": 38011
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2002844500632112e-06,
+      "loss": 2.6985,
+      "step": 38012
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2002370417193425e-06,
+      "loss": 2.385,
+      "step": 38013
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2001896333754742e-06,
+      "loss": 2.6248,
+      "step": 38014
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2001422250316057e-06,
+      "loss": 2.4081,
+      "step": 38015
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.200094816687737e-06,
+      "loss": 1.8759,
+      "step": 38016
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2000474083438685e-06,
+      "loss": 1.9622,
+      "step": 38017
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 2.1763,
+      "step": 38018
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1999525916561314e-06,
+      "loss": 1.5113,
+      "step": 38019
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.199905183312263e-06,
+      "loss": 2.3155,
+      "step": 38020
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1998577749683946e-06,
+      "loss": 2.2025,
+      "step": 38021
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.199810366624526e-06,
+      "loss": 0.8609,
+      "step": 38022
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1997629582806574e-06,
+      "loss": 1.1253,
+      "step": 38023
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.199715549936789e-06,
+      "loss": 2.1377,
+      "step": 38024
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1996681415929204e-06,
+      "loss": 2.6372,
+      "step": 38025
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1996207332490519e-06,
+      "loss": 1.0468,
+      "step": 38026
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1995733249051834e-06,
+      "loss": 2.3668,
+      "step": 38027
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1995259165613149e-06,
+      "loss": 2.1758,
+      "step": 38028
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 1.1994785082174464e-06,
+      "loss": 3.9343,
+      "step": 38029
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1994310998735776e-06,
+      "loss": 0.8889,
+      "step": 38030
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1993836915297093e-06,
+      "loss": 4.6556,
+      "step": 38031
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1993362831858408e-06,
+      "loss": 2.0211,
+      "step": 38032
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1992888748419721e-06,
+      "loss": 2.443,
+      "step": 38033
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1992414664981038e-06,
+      "loss": 1.6706,
+      "step": 38034
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.199194058154235e-06,
+      "loss": 2.4322,
+      "step": 38035
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1991466498103666e-06,
+      "loss": 3.0478,
+      "step": 38036
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.199099241466498e-06,
+      "loss": 3.3403,
+      "step": 38037
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1990518331226296e-06,
+      "loss": 3.5792,
+      "step": 38038
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.199004424778761e-06,
+      "loss": 1.0101,
+      "step": 38039
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1989570164348926e-06,
+      "loss": 4.0734,
+      "step": 38040
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.198909608091024e-06,
+      "loss": 1.7978,
+      "step": 38041
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1988621997471555e-06,
+      "loss": 2.9891,
+      "step": 38042
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.198814791403287e-06,
+      "loss": 2.7136,
+      "step": 38043
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1987673830594185e-06,
+      "loss": 2.4093,
+      "step": 38044
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.19871997471555e-06,
+      "loss": 2.4369,
+      "step": 38045
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1986725663716813e-06,
+      "loss": 2.8863,
+      "step": 38046
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.198625158027813e-06,
+      "loss": 1.2968,
+      "step": 38047
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1985777496839445e-06,
+      "loss": 0.8389,
+      "step": 38048
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1985303413400758e-06,
+      "loss": 2.35,
+      "step": 38049
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1984829329962073e-06,
+      "loss": 2.6233,
+      "step": 38050
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.198435524652339e-06,
+      "loss": 1.6247,
+      "step": 38051
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1983881163084703e-06,
+      "loss": 3.1008,
+      "step": 38052
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1983407079646017e-06,
+      "loss": 2.9746,
+      "step": 38053
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1982932996207335e-06,
+      "loss": 2.7854,
+      "step": 38054
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1982458912768647e-06,
+      "loss": 2.3963,
+      "step": 38055
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1981984829329962e-06,
+      "loss": 3.8769,
+      "step": 38056
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1981510745891277e-06,
+      "loss": 2.9567,
+      "step": 38057
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1981036662452592e-06,
+      "loss": 1.4638,
+      "step": 38058
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1980562579013907e-06,
+      "loss": 3.0744,
+      "step": 38059
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1980088495575222e-06,
+      "loss": 1.6774,
+      "step": 38060
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1979614412136537e-06,
+      "loss": 2.5777,
+      "step": 38061
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1979140328697852e-06,
+      "loss": 2.936,
+      "step": 38062
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1978666245259165e-06,
+      "loss": 2.3481,
+      "step": 38063
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1978192161820482e-06,
+      "loss": 1.5573,
+      "step": 38064
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1977718078381794e-06,
+      "loss": 1.1921,
+      "step": 38065
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.197724399494311e-06,
+      "loss": 3.6371,
+      "step": 38066
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1976769911504426e-06,
+      "loss": 1.4119,
+      "step": 38067
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.197629582806574e-06,
+      "loss": 1.7401,
+      "step": 38068
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1975821744627054e-06,
+      "loss": 1.9955,
+      "step": 38069
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.197534766118837e-06,
+      "loss": 2.4433,
+      "step": 38070
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1974873577749684e-06,
+      "loss": 2.5068,
+      "step": 38071
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1974399494310999e-06,
+      "loss": 1.4918,
+      "step": 38072
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1973925410872314e-06,
+      "loss": 1.4746,
+      "step": 38073
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1973451327433629e-06,
+      "loss": 2.0362,
+      "step": 38074
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1972977243994944e-06,
+      "loss": 2.0874,
+      "step": 38075
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1972503160556256e-06,
+      "loss": 1.6289,
+      "step": 38076
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1972029077117573e-06,
+      "loss": 0.9343,
+      "step": 38077
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1971554993678888e-06,
+      "loss": 1.8883,
+      "step": 38078
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1971080910240201e-06,
+      "loss": 1.394,
+      "step": 38079
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1970606826801518e-06,
+      "loss": 1.9712,
+      "step": 38080
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1970132743362833e-06,
+      "loss": 2.6266,
+      "step": 38081
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1969658659924146e-06,
+      "loss": 1.4542,
+      "step": 38082
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.196918457648546e-06,
+      "loss": 2.6612,
+      "step": 38083
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1968710493046778e-06,
+      "loss": 0.8425,
+      "step": 38084
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.196823640960809e-06,
+      "loss": 2.5576,
+      "step": 38085
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1967762326169406e-06,
+      "loss": 1.7492,
+      "step": 38086
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.196728824273072e-06,
+      "loss": 2.2495,
+      "step": 38087
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1966814159292036e-06,
+      "loss": 1.1864,
+      "step": 38088
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.196634007585335e-06,
+      "loss": 1.6883,
+      "step": 38089
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1965865992414665e-06,
+      "loss": 2.3731,
+      "step": 38090
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.196539190897598e-06,
+      "loss": 1.7794,
+      "step": 38091
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 1.1964917825537295e-06,
+      "loss": 2.4368,
+      "step": 38092
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.196444374209861e-06,
+      "loss": 1.9625,
+      "step": 38093
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1963969658659925e-06,
+      "loss": 1.3388,
+      "step": 38094
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.196349557522124e-06,
+      "loss": 1.9626,
+      "step": 38095
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1963021491782553e-06,
+      "loss": 1.0871,
+      "step": 38096
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.196254740834387e-06,
+      "loss": 1.9093,
+      "step": 38097
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1962073324905183e-06,
+      "loss": 1.7265,
+      "step": 38098
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1961599241466498e-06,
+      "loss": 1.8183,
+      "step": 38099
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1961125158027815e-06,
+      "loss": 3.7111,
+      "step": 38100
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1960651074589127e-06,
+      "loss": 2.9828,
+      "step": 38101
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1960176991150442e-06,
+      "loss": 1.8968,
+      "step": 38102
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1959702907711757e-06,
+      "loss": 2.0883,
+      "step": 38103
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1959228824273072e-06,
+      "loss": 2.0316,
+      "step": 38104
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1958754740834387e-06,
+      "loss": 1.6412,
+      "step": 38105
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1958280657395702e-06,
+      "loss": 3.7482,
+      "step": 38106
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1957806573957017e-06,
+      "loss": 2.9177,
+      "step": 38107
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1957332490518332e-06,
+      "loss": 2.5023,
+      "step": 38108
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1956858407079645e-06,
+      "loss": 1.7721,
+      "step": 38109
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1956384323640962e-06,
+      "loss": 2.8269,
+      "step": 38110
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1955910240202277e-06,
+      "loss": 1.4375,
+      "step": 38111
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.195543615676359e-06,
+      "loss": 2.6268,
+      "step": 38112
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1954962073324906e-06,
+      "loss": 2.2491,
+      "step": 38113
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1954487989886221e-06,
+      "loss": 0.9708,
+      "step": 38114
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1954013906447534e-06,
+      "loss": 2.5312,
+      "step": 38115
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.195353982300885e-06,
+      "loss": 2.3415,
+      "step": 38116
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1953065739570166e-06,
+      "loss": 1.707,
+      "step": 38117
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1952591656131479e-06,
+      "loss": 2.2365,
+      "step": 38118
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1952117572692794e-06,
+      "loss": 2.4875,
+      "step": 38119
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1951643489254109e-06,
+      "loss": 1.4674,
+      "step": 38120
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1951169405815424e-06,
+      "loss": 0.6611,
+      "step": 38121
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1950695322376739e-06,
+      "loss": 1.9026,
+      "step": 38122
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1950221238938054e-06,
+      "loss": 2.7545,
+      "step": 38123
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1949747155499368e-06,
+      "loss": 1.2083,
+      "step": 38124
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1949273072060683e-06,
+      "loss": 2.695,
+      "step": 38125
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1948798988621998e-06,
+      "loss": 1.5909,
+      "step": 38126
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1948324905183313e-06,
+      "loss": 2.5831,
+      "step": 38127
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1947850821744628e-06,
+      "loss": 2.4049,
+      "step": 38128
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194737673830594e-06,
+      "loss": 2.0298,
+      "step": 38129
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1946902654867258e-06,
+      "loss": 1.9348,
+      "step": 38130
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194642857142857e-06,
+      "loss": 3.5212,
+      "step": 38131
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1945954487989886e-06,
+      "loss": 3.5148,
+      "step": 38132
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1945480404551203e-06,
+      "loss": 2.7582,
+      "step": 38133
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1945006321112516e-06,
+      "loss": 2.8114,
+      "step": 38134
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194453223767383e-06,
+      "loss": 1.6386,
+      "step": 38135
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1944058154235145e-06,
+      "loss": 1.2714,
+      "step": 38136
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194358407079646e-06,
+      "loss": 1.6444,
+      "step": 38137
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1943109987357775e-06,
+      "loss": 3.8307,
+      "step": 38138
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194263590391909e-06,
+      "loss": 1.6898,
+      "step": 38139
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1942161820480405e-06,
+      "loss": 1.7352,
+      "step": 38140
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194168773704172e-06,
+      "loss": 2.2176,
+      "step": 38141
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1941213653603033e-06,
+      "loss": 1.7526,
+      "step": 38142
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.194073957016435e-06,
+      "loss": 2.3007,
+      "step": 38143
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1940265486725665e-06,
+      "loss": 2.2758,
+      "step": 38144
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1939791403286978e-06,
+      "loss": 1.8567,
+      "step": 38145
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1939317319848293e-06,
+      "loss": 1.8512,
+      "step": 38146
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.193884323640961e-06,
+      "loss": 2.2966,
+      "step": 38147
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1938369152970922e-06,
+      "loss": 2.0029,
+      "step": 38148
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1937895069532237e-06,
+      "loss": 1.5823,
+      "step": 38149
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1937420986093554e-06,
+      "loss": 1.3826,
+      "step": 38150
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1936946902654867e-06,
+      "loss": 1.8301,
+      "step": 38151
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1936472819216182e-06,
+      "loss": 2.8365,
+      "step": 38152
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1935998735777497e-06,
+      "loss": 1.5156,
+      "step": 38153
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1935524652338812e-06,
+      "loss": 1.7197,
+      "step": 38154
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1935050568900127e-06,
+      "loss": 2.3277,
+      "step": 38155
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 1.1934576485461442e-06,
+      "loss": 2.8371,
+      "step": 38156
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1934102402022757e-06,
+      "loss": 2.1261,
+      "step": 38157
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1933628318584072e-06,
+      "loss": 1.7991,
+      "step": 38158
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1933154235145384e-06,
+      "loss": 1.8004,
+      "step": 38159
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1932680151706701e-06,
+      "loss": 2.0,
+      "step": 38160
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1932206068268016e-06,
+      "loss": 1.6389,
+      "step": 38161
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.193173198482933e-06,
+      "loss": 1.6325,
+      "step": 38162
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1931257901390646e-06,
+      "loss": 2.3299,
+      "step": 38163
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.193078381795196e-06,
+      "loss": 2.2689,
+      "step": 38164
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1930309734513274e-06,
+      "loss": 2.7662,
+      "step": 38165
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1929835651074589e-06,
+      "loss": 2.7092,
+      "step": 38166
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1929361567635904e-06,
+      "loss": 3.0193,
+      "step": 38167
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1928887484197219e-06,
+      "loss": 2.0286,
+      "step": 38168
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1928413400758534e-06,
+      "loss": 1.1725,
+      "step": 38169
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1927939317319849e-06,
+      "loss": 2.8595,
+      "step": 38170
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1927465233881163e-06,
+      "loss": 2.5291,
+      "step": 38171
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1926991150442478e-06,
+      "loss": 2.8067,
+      "step": 38172
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1926517067003793e-06,
+      "loss": 1.6914,
+      "step": 38173
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1926042983565108e-06,
+      "loss": 2.8836,
+      "step": 38174
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.192556890012642e-06,
+      "loss": 3.4771,
+      "step": 38175
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1925094816687738e-06,
+      "loss": 2.5128,
+      "step": 38176
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1924620733249053e-06,
+      "loss": 2.5798,
+      "step": 38177
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1924146649810366e-06,
+      "loss": 2.4712,
+      "step": 38178
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.192367256637168e-06,
+      "loss": 2.975,
+      "step": 38179
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1923198482932998e-06,
+      "loss": 0.6384,
+      "step": 38180
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.192272439949431e-06,
+      "loss": 1.7143,
+      "step": 38181
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1922250316055625e-06,
+      "loss": 1.2228,
+      "step": 38182
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1921776232616942e-06,
+      "loss": 1.4978,
+      "step": 38183
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1921302149178255e-06,
+      "loss": 1.9191,
+      "step": 38184
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.192082806573957e-06,
+      "loss": 2.9308,
+      "step": 38185
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1920353982300885e-06,
+      "loss": 2.9925,
+      "step": 38186
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.19198798988622e-06,
+      "loss": 1.4757,
+      "step": 38187
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1919405815423515e-06,
+      "loss": 2.5646,
+      "step": 38188
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.191893173198483e-06,
+      "loss": 2.3445,
+      "step": 38189
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1918457648546145e-06,
+      "loss": 1.685,
+      "step": 38190
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.191798356510746e-06,
+      "loss": 2.964,
+      "step": 38191
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1917509481668773e-06,
+      "loss": 1.1184,
+      "step": 38192
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.191703539823009e-06,
+      "loss": 1.0827,
+      "step": 38193
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1916561314791405e-06,
+      "loss": 2.8627,
+      "step": 38194
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1916087231352717e-06,
+      "loss": 2.1984,
+      "step": 38195
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1915613147914034e-06,
+      "loss": 1.9147,
+      "step": 38196
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1915139064475347e-06,
+      "loss": 1.1056,
+      "step": 38197
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1914664981036662e-06,
+      "loss": 1.8262,
+      "step": 38198
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1914190897597977e-06,
+      "loss": 1.6233,
+      "step": 38199
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1913716814159292e-06,
+      "loss": 2.7887,
+      "step": 38200
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1913242730720607e-06,
+      "loss": 3.6263,
+      "step": 38201
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1912768647281922e-06,
+      "loss": 1.637,
+      "step": 38202
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1912294563843237e-06,
+      "loss": 3.085,
+      "step": 38203
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1911820480404552e-06,
+      "loss": 1.4692,
+      "step": 38204
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1911346396965867e-06,
+      "loss": 1.5561,
+      "step": 38205
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1910872313527181e-06,
+      "loss": 1.337,
+      "step": 38206
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1910398230088496e-06,
+      "loss": 1.6803,
+      "step": 38207
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.190992414664981e-06,
+      "loss": 2.2352,
+      "step": 38208
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1909450063211126e-06,
+      "loss": 2.3225,
+      "step": 38209
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1908975979772441e-06,
+      "loss": 2.096,
+      "step": 38210
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1908501896333754e-06,
+      "loss": 2.0833,
+      "step": 38211
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1908027812895069e-06,
+      "loss": 1.4153,
+      "step": 38212
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1907553729456386e-06,
+      "loss": 3.75,
+      "step": 38213
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1907079646017699e-06,
+      "loss": 1.1936,
+      "step": 38214
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1906605562579014e-06,
+      "loss": 3.061,
+      "step": 38215
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.190613147914033e-06,
+      "loss": 1.271,
+      "step": 38216
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1905657395701643e-06,
+      "loss": 1.7307,
+      "step": 38217
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1905183312262958e-06,
+      "loss": 2.0067,
+      "step": 38218
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 1.1904709228824273e-06,
+      "loss": 2.4446,
+      "step": 38219
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1904235145385588e-06,
+      "loss": 1.4542,
+      "step": 38220
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1903761061946903e-06,
+      "loss": 1.91,
+      "step": 38221
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1903286978508218e-06,
+      "loss": 1.7989,
+      "step": 38222
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1902812895069533e-06,
+      "loss": 2.8414,
+      "step": 38223
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1902338811630848e-06,
+      "loss": 3.4507,
+      "step": 38224
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.190186472819216e-06,
+      "loss": 2.3131,
+      "step": 38225
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1901390644753478e-06,
+      "loss": 3.0067,
+      "step": 38226
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1900916561314793e-06,
+      "loss": 1.5836,
+      "step": 38227
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1900442477876105e-06,
+      "loss": 2.7943,
+      "step": 38228
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1899968394437423e-06,
+      "loss": 3.9084,
+      "step": 38229
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1899494310998735e-06,
+      "loss": 2.1946,
+      "step": 38230
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.189902022756005e-06,
+      "loss": 2.4151,
+      "step": 38231
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1898546144121365e-06,
+      "loss": 2.249,
+      "step": 38232
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.189807206068268e-06,
+      "loss": 1.7372,
+      "step": 38233
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1897597977243995e-06,
+      "loss": 0.4323,
+      "step": 38234
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.189712389380531e-06,
+      "loss": 2.6448,
+      "step": 38235
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1896649810366625e-06,
+      "loss": 2.9489,
+      "step": 38236
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.189617572692794e-06,
+      "loss": 3.8783,
+      "step": 38237
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1895701643489255e-06,
+      "loss": 1.7483,
+      "step": 38238
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.189522756005057e-06,
+      "loss": 2.8442,
+      "step": 38239
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1894753476611885e-06,
+      "loss": 1.7996,
+      "step": 38240
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1894279393173197e-06,
+      "loss": 2.9711,
+      "step": 38241
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1893805309734514e-06,
+      "loss": 2.5802,
+      "step": 38242
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.189333122629583e-06,
+      "loss": 1.628,
+      "step": 38243
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1892857142857142e-06,
+      "loss": 3.1793,
+      "step": 38244
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1892383059418457e-06,
+      "loss": 1.6194,
+      "step": 38245
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1891908975979774e-06,
+      "loss": 2.2724,
+      "step": 38246
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1891434892541087e-06,
+      "loss": 2.478,
+      "step": 38247
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1890960809102402e-06,
+      "loss": 1.4847,
+      "step": 38248
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1890486725663719e-06,
+      "loss": 3.4165,
+      "step": 38249
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1890012642225032e-06,
+      "loss": 2.8519,
+      "step": 38250
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1889538558786347e-06,
+      "loss": 3.1581,
+      "step": 38251
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1889064475347662e-06,
+      "loss": 1.3851,
+      "step": 38252
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1888590391908976e-06,
+      "loss": 2.9384,
+      "step": 38253
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1888116308470291e-06,
+      "loss": 3.3776,
+      "step": 38254
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1887642225031606e-06,
+      "loss": 1.5765,
+      "step": 38255
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1887168141592921e-06,
+      "loss": 2.9213,
+      "step": 38256
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1886694058154236e-06,
+      "loss": 1.5019,
+      "step": 38257
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1886219974715549e-06,
+      "loss": 2.8187,
+      "step": 38258
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1885745891276866e-06,
+      "loss": 0.6656,
+      "step": 38259
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.188527180783818e-06,
+      "loss": 2.3023,
+      "step": 38260
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1884797724399494e-06,
+      "loss": 1.2979,
+      "step": 38261
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.188432364096081e-06,
+      "loss": 0.2631,
+      "step": 38262
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1883849557522124e-06,
+      "loss": 3.349,
+      "step": 38263
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1883375474083438e-06,
+      "loss": 2.4399,
+      "step": 38264
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1882901390644753e-06,
+      "loss": 1.6837,
+      "step": 38265
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1882427307206068e-06,
+      "loss": 2.3879,
+      "step": 38266
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1881953223767383e-06,
+      "loss": 1.5235,
+      "step": 38267
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1881479140328698e-06,
+      "loss": 2.6885,
+      "step": 38268
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1881005056890013e-06,
+      "loss": 1.801,
+      "step": 38269
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1880530973451328e-06,
+      "loss": 1.7319,
+      "step": 38270
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1880056890012643e-06,
+      "loss": 2.5849,
+      "step": 38271
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1879582806573958e-06,
+      "loss": 2.0329,
+      "step": 38272
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1879108723135273e-06,
+      "loss": 2.4818,
+      "step": 38273
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1878634639696586e-06,
+      "loss": 2.462,
+      "step": 38274
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1878160556257903e-06,
+      "loss": 2.6675,
+      "step": 38275
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1877686472819218e-06,
+      "loss": 1.4778,
+      "step": 38276
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.187721238938053e-06,
+      "loss": 0.9551,
+      "step": 38277
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1876738305941845e-06,
+      "loss": 2.4229,
+      "step": 38278
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1876264222503162e-06,
+      "loss": 1.4189,
+      "step": 38279
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1875790139064475e-06,
+      "loss": 2.0438,
+      "step": 38280
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.187531605562579e-06,
+      "loss": 1.8275,
+      "step": 38281
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 1.1874841972187107e-06,
+      "loss": 2.1861,
+      "step": 38282
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.187436788874842e-06,
+      "loss": 2.8346,
+      "step": 38283
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1873893805309735e-06,
+      "loss": 2.0674,
+      "step": 38284
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.187341972187105e-06,
+      "loss": 3.8328,
+      "step": 38285
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1872945638432365e-06,
+      "loss": 2.2521,
+      "step": 38286
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.187247155499368e-06,
+      "loss": 2.4951,
+      "step": 38287
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1871997471554992e-06,
+      "loss": 1.932,
+      "step": 38288
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.187152338811631e-06,
+      "loss": 2.4004,
+      "step": 38289
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1871049304677624e-06,
+      "loss": 1.9877,
+      "step": 38290
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1870575221238937e-06,
+      "loss": 1.3202,
+      "step": 38291
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1870101137800254e-06,
+      "loss": 1.8577,
+      "step": 38292
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.186962705436157e-06,
+      "loss": 2.8208,
+      "step": 38293
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1869152970922882e-06,
+      "loss": 2.123,
+      "step": 38294
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1868678887484197e-06,
+      "loss": 1.3519,
+      "step": 38295
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1868204804045512e-06,
+      "loss": 2.2909,
+      "step": 38296
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1867730720606827e-06,
+      "loss": 3.8968,
+      "step": 38297
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1867256637168142e-06,
+      "loss": 2.5992,
+      "step": 38298
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1866782553729456e-06,
+      "loss": 1.4101,
+      "step": 38299
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1866308470290771e-06,
+      "loss": 1.6081,
+      "step": 38300
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1865834386852086e-06,
+      "loss": 0.4469,
+      "step": 38301
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1865360303413401e-06,
+      "loss": 2.1608,
+      "step": 38302
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1864886219974716e-06,
+      "loss": 1.5569,
+      "step": 38303
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1864412136536031e-06,
+      "loss": 2.9376,
+      "step": 38304
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1863938053097346e-06,
+      "loss": 1.6242,
+      "step": 38305
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.186346396965866e-06,
+      "loss": 0.864,
+      "step": 38306
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1862989886219974e-06,
+      "loss": 2.5476,
+      "step": 38307
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1862515802781289e-06,
+      "loss": 2.0069,
+      "step": 38308
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1862041719342606e-06,
+      "loss": 3.4755,
+      "step": 38309
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1861567635903918e-06,
+      "loss": 2.9665,
+      "step": 38310
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1861093552465233e-06,
+      "loss": 1.3908,
+      "step": 38311
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.186061946902655e-06,
+      "loss": 2.3369,
+      "step": 38312
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1860145385587863e-06,
+      "loss": 1.5825,
+      "step": 38313
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1859671302149178e-06,
+      "loss": 1.8253,
+      "step": 38314
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1859197218710493e-06,
+      "loss": 1.8266,
+      "step": 38315
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1858723135271808e-06,
+      "loss": 2.3516,
+      "step": 38316
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1858249051833123e-06,
+      "loss": 2.5133,
+      "step": 38317
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1857774968394438e-06,
+      "loss": 3.3578,
+      "step": 38318
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1857300884955753e-06,
+      "loss": 1.5147,
+      "step": 38319
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1856826801517068e-06,
+      "loss": 1.2828,
+      "step": 38320
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.185635271807838e-06,
+      "loss": 1.2246,
+      "step": 38321
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1855878634639698e-06,
+      "loss": 3.6025,
+      "step": 38322
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1855404551201012e-06,
+      "loss": 1.37,
+      "step": 38323
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1854930467762325e-06,
+      "loss": 0.9749,
+      "step": 38324
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1854456384323642e-06,
+      "loss": 3.2079,
+      "step": 38325
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1853982300884957e-06,
+      "loss": 1.9146,
+      "step": 38326
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.185350821744627e-06,
+      "loss": 1.9586,
+      "step": 38327
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1853034134007585e-06,
+      "loss": 1.7118,
+      "step": 38328
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.18525600505689e-06,
+      "loss": 3.5502,
+      "step": 38329
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1852085967130215e-06,
+      "loss": 2.9471,
+      "step": 38330
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.185161188369153e-06,
+      "loss": 1.6693,
+      "step": 38331
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1851137800252845e-06,
+      "loss": 2.2227,
+      "step": 38332
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.185066371681416e-06,
+      "loss": 1.8716,
+      "step": 38333
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1850189633375475e-06,
+      "loss": 0.8883,
+      "step": 38334
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.184971554993679e-06,
+      "loss": 1.4151,
+      "step": 38335
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1849241466498104e-06,
+      "loss": 3.2505,
+      "step": 38336
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.184876738305942e-06,
+      "loss": 3.4481,
+      "step": 38337
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1848293299620734e-06,
+      "loss": 2.3484,
+      "step": 38338
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.184781921618205e-06,
+      "loss": 1.434,
+      "step": 38339
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1847345132743362e-06,
+      "loss": 1.6006,
+      "step": 38340
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1846871049304677e-06,
+      "loss": 1.4337,
+      "step": 38341
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1846396965865994e-06,
+      "loss": 2.8383,
+      "step": 38342
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1845922882427307e-06,
+      "loss": 1.8905,
+      "step": 38343
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1845448798988622e-06,
+      "loss": 2.2609,
+      "step": 38344
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1844974715549939e-06,
+      "loss": 2.0346,
+      "step": 38345
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 1.1844500632111251e-06,
+      "loss": 2.5426,
+      "step": 38346
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1844026548672566e-06,
+      "loss": 2.6251,
+      "step": 38347
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1843552465233881e-06,
+      "loss": 2.4974,
+      "step": 38348
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1843078381795196e-06,
+      "loss": 1.5065,
+      "step": 38349
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1842604298356511e-06,
+      "loss": 3.8281,
+      "step": 38350
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1842130214917826e-06,
+      "loss": 2.8615,
+      "step": 38351
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.184165613147914e-06,
+      "loss": 2.2228,
+      "step": 38352
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1841182048040456e-06,
+      "loss": 2.604,
+      "step": 38353
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1840707964601769e-06,
+      "loss": 2.6377,
+      "step": 38354
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1840233881163086e-06,
+      "loss": 1.1476,
+      "step": 38355
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.18397597977244e-06,
+      "loss": 1.639,
+      "step": 38356
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1839285714285713e-06,
+      "loss": 1.9268,
+      "step": 38357
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.183881163084703e-06,
+      "loss": 1.3627,
+      "step": 38358
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1838337547408345e-06,
+      "loss": 3.0497,
+      "step": 38359
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1837863463969658e-06,
+      "loss": 1.7471,
+      "step": 38360
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1837389380530973e-06,
+      "loss": 1.7416,
+      "step": 38361
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1836915297092288e-06,
+      "loss": 1.4691,
+      "step": 38362
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1836441213653603e-06,
+      "loss": 0.8986,
+      "step": 38363
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1835967130214918e-06,
+      "loss": 2.1206,
+      "step": 38364
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1835493046776233e-06,
+      "loss": 2.7245,
+      "step": 38365
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1835018963337548e-06,
+      "loss": 1.9951,
+      "step": 38366
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1834544879898863e-06,
+      "loss": 0.9463,
+      "step": 38367
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1834070796460178e-06,
+      "loss": 3.1989,
+      "step": 38368
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1833596713021493e-06,
+      "loss": 1.2571,
+      "step": 38369
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1833122629582807e-06,
+      "loss": 2.8817,
+      "step": 38370
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1832648546144122e-06,
+      "loss": 1.695,
+      "step": 38371
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1832174462705437e-06,
+      "loss": 1.4891,
+      "step": 38372
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.183170037926675e-06,
+      "loss": 1.0092,
+      "step": 38373
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1831226295828065e-06,
+      "loss": 2.3788,
+      "step": 38374
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1830752212389382e-06,
+      "loss": 2.3108,
+      "step": 38375
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1830278128950695e-06,
+      "loss": 2.7249,
+      "step": 38376
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.182980404551201e-06,
+      "loss": 1.9731,
+      "step": 38377
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1829329962073327e-06,
+      "loss": 2.458,
+      "step": 38378
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.182885587863464e-06,
+      "loss": 1.3469,
+      "step": 38379
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1828381795195955e-06,
+      "loss": 1.9823,
+      "step": 38380
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.182790771175727e-06,
+      "loss": 2.5971,
+      "step": 38381
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1827433628318584e-06,
+      "loss": 2.1779,
+      "step": 38382
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.18269595448799e-06,
+      "loss": 1.9086,
+      "step": 38383
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1826485461441214e-06,
+      "loss": 2.4884,
+      "step": 38384
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.182601137800253e-06,
+      "loss": 2.4105,
+      "step": 38385
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1825537294563844e-06,
+      "loss": 1.8194,
+      "step": 38386
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1825063211125157e-06,
+      "loss": 1.3505,
+      "step": 38387
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1824589127686474e-06,
+      "loss": 1.8078,
+      "step": 38388
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1824115044247789e-06,
+      "loss": 1.2891,
+      "step": 38389
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1823640960809102e-06,
+      "loss": 2.8978,
+      "step": 38390
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1823166877370419e-06,
+      "loss": 2.7176,
+      "step": 38391
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1822692793931734e-06,
+      "loss": 1.2966,
+      "step": 38392
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1822218710493046e-06,
+      "loss": 1.9108,
+      "step": 38393
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1821744627054361e-06,
+      "loss": 2.9107,
+      "step": 38394
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1821270543615676e-06,
+      "loss": 2.2256,
+      "step": 38395
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1820796460176991e-06,
+      "loss": 1.0876,
+      "step": 38396
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1820322376738306e-06,
+      "loss": 1.7742,
+      "step": 38397
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.181984829329962e-06,
+      "loss": 1.6044,
+      "step": 38398
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1819374209860936e-06,
+      "loss": 1.3851,
+      "step": 38399
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.181890012642225e-06,
+      "loss": 1.4715,
+      "step": 38400
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1818426042983566e-06,
+      "loss": 1.2227,
+      "step": 38401
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.181795195954488e-06,
+      "loss": 2.4504,
+      "step": 38402
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1817477876106196e-06,
+      "loss": 3.223,
+      "step": 38403
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.181700379266751e-06,
+      "loss": 1.5777,
+      "step": 38404
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1816529709228825e-06,
+      "loss": 1.7858,
+      "step": 38405
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1816055625790138e-06,
+      "loss": 2.706,
+      "step": 38406
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1815581542351453e-06,
+      "loss": 1.6326,
+      "step": 38407
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.181510745891277e-06,
+      "loss": 3.6196,
+      "step": 38408
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 1.1814633375474083e-06,
+      "loss": 2.3128,
+      "step": 38409
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1814159292035398e-06,
+      "loss": 1.2879,
+      "step": 38410
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1813685208596715e-06,
+      "loss": 1.0375,
+      "step": 38411
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1813211125158028e-06,
+      "loss": 2.3691,
+      "step": 38412
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1812737041719343e-06,
+      "loss": 1.6211,
+      "step": 38413
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1812262958280658e-06,
+      "loss": 2.3223,
+      "step": 38414
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1811788874841973e-06,
+      "loss": 1.8339,
+      "step": 38415
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1811314791403287e-06,
+      "loss": 3.127,
+      "step": 38416
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.18108407079646e-06,
+      "loss": 1.4805,
+      "step": 38417
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1810366624525917e-06,
+      "loss": 2.1759,
+      "step": 38418
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1809892541087232e-06,
+      "loss": 1.7436,
+      "step": 38419
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1809418457648545e-06,
+      "loss": 1.5979,
+      "step": 38420
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1808944374209862e-06,
+      "loss": 2.2971,
+      "step": 38421
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1808470290771177e-06,
+      "loss": 3.5815,
+      "step": 38422
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.180799620733249e-06,
+      "loss": 2.5784,
+      "step": 38423
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1807522123893805e-06,
+      "loss": 2.8382,
+      "step": 38424
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1807048040455122e-06,
+      "loss": 1.3194,
+      "step": 38425
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1806573957016435e-06,
+      "loss": 2.8176,
+      "step": 38426
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.180609987357775e-06,
+      "loss": 2.8962,
+      "step": 38427
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1805625790139064e-06,
+      "loss": 1.8426,
+      "step": 38428
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.180515170670038e-06,
+      "loss": 2.0175,
+      "step": 38429
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1804677623261694e-06,
+      "loss": 1.8035,
+      "step": 38430
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.180420353982301e-06,
+      "loss": 1.574,
+      "step": 38431
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1803729456384324e-06,
+      "loss": 3.5245,
+      "step": 38432
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.180325537294564e-06,
+      "loss": 1.7059,
+      "step": 38433
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1802781289506954e-06,
+      "loss": 0.455,
+      "step": 38434
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1802307206068269e-06,
+      "loss": 2.3565,
+      "step": 38435
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1801833122629584e-06,
+      "loss": 1.8,
+      "step": 38436
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1801359039190897e-06,
+      "loss": 3.1378,
+      "step": 38437
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1800884955752214e-06,
+      "loss": 2.3418,
+      "step": 38438
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1800410872313526e-06,
+      "loss": 1.5365,
+      "step": 38439
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1799936788874841e-06,
+      "loss": 2.9107,
+      "step": 38440
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1799462705436158e-06,
+      "loss": 3.3724,
+      "step": 38441
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1798988621997471e-06,
+      "loss": 1.4054,
+      "step": 38442
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1798514538558786e-06,
+      "loss": 2.4739,
+      "step": 38443
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1798040455120101e-06,
+      "loss": 1.783,
+      "step": 38444
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1797566371681416e-06,
+      "loss": 2.6397,
+      "step": 38445
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.179709228824273e-06,
+      "loss": 2.9049,
+      "step": 38446
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1796618204804046e-06,
+      "loss": 2.1472,
+      "step": 38447
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.179614412136536e-06,
+      "loss": 1.9826,
+      "step": 38448
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1795670037926676e-06,
+      "loss": 3.8752,
+      "step": 38449
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1795195954487988e-06,
+      "loss": 2.7378,
+      "step": 38450
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1794721871049306e-06,
+      "loss": 1.8021,
+      "step": 38451
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.179424778761062e-06,
+      "loss": 1.7233,
+      "step": 38452
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1793773704171933e-06,
+      "loss": 2.984,
+      "step": 38453
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.179329962073325e-06,
+      "loss": 1.9829,
+      "step": 38454
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1792825537294565e-06,
+      "loss": 2.2492,
+      "step": 38455
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1792351453855878e-06,
+      "loss": 2.2201,
+      "step": 38456
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1791877370417193e-06,
+      "loss": 2.3132,
+      "step": 38457
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.179140328697851e-06,
+      "loss": 2.3237,
+      "step": 38458
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1790929203539823e-06,
+      "loss": 2.8521,
+      "step": 38459
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1790455120101138e-06,
+      "loss": 1.8853,
+      "step": 38460
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1789981036662453e-06,
+      "loss": 2.7596,
+      "step": 38461
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1789506953223768e-06,
+      "loss": 1.8895,
+      "step": 38462
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1789032869785082e-06,
+      "loss": 1.9924,
+      "step": 38463
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1788558786346397e-06,
+      "loss": 2.0512,
+      "step": 38464
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1788084702907712e-06,
+      "loss": 2.4187,
+      "step": 38465
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1787610619469027e-06,
+      "loss": 1.8207,
+      "step": 38466
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1787136536030342e-06,
+      "loss": 1.0581,
+      "step": 38467
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1786662452591657e-06,
+      "loss": 0.5467,
+      "step": 38468
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1786188369152972e-06,
+      "loss": 1.8353,
+      "step": 38469
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1785714285714285e-06,
+      "loss": 2.6386,
+      "step": 38470
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1785240202275602e-06,
+      "loss": 1.9353,
+      "step": 38471
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 1.1784766118836915e-06,
+      "loss": 2.5031,
+      "step": 38472
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.178429203539823e-06,
+      "loss": 0.5836,
+      "step": 38473
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1783817951959547e-06,
+      "loss": 2.9332,
+      "step": 38474
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.178334386852086e-06,
+      "loss": 2.3597,
+      "step": 38475
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1782869785082174e-06,
+      "loss": 2.3125,
+      "step": 38476
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.178239570164349e-06,
+      "loss": 0.7093,
+      "step": 38477
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1781921618204804e-06,
+      "loss": 1.8006,
+      "step": 38478
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.178144753476612e-06,
+      "loss": 2.1944,
+      "step": 38479
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1780973451327434e-06,
+      "loss": 0.3301,
+      "step": 38480
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.178049936788875e-06,
+      "loss": 3.648,
+      "step": 38481
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1780025284450064e-06,
+      "loss": 1.95,
+      "step": 38482
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1779551201011377e-06,
+      "loss": 1.9634,
+      "step": 38483
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1779077117572694e-06,
+      "loss": 3.2458,
+      "step": 38484
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1778603034134009e-06,
+      "loss": 1.7917,
+      "step": 38485
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1778128950695321e-06,
+      "loss": 1.7651,
+      "step": 38486
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1777654867256638e-06,
+      "loss": 1.5711,
+      "step": 38487
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1777180783817953e-06,
+      "loss": 3.4561,
+      "step": 38488
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1776706700379266e-06,
+      "loss": 1.659,
+      "step": 38489
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1776232616940581e-06,
+      "loss": 2.0173,
+      "step": 38490
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1775758533501898e-06,
+      "loss": 3.0272,
+      "step": 38491
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.177528445006321e-06,
+      "loss": 1.8013,
+      "step": 38492
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1774810366624526e-06,
+      "loss": 1.0066,
+      "step": 38493
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.177433628318584e-06,
+      "loss": 2.063,
+      "step": 38494
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1773862199747156e-06,
+      "loss": 1.9631,
+      "step": 38495
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.177338811630847e-06,
+      "loss": 1.3794,
+      "step": 38496
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1772914032869786e-06,
+      "loss": 2.5434,
+      "step": 38497
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.17724399494311e-06,
+      "loss": 2.7541,
+      "step": 38498
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1771965865992415e-06,
+      "loss": 2.9226,
+      "step": 38499
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.177149178255373e-06,
+      "loss": 1.2642,
+      "step": 38500
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1771017699115045e-06,
+      "loss": 2.7372,
+      "step": 38501
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1770543615676358e-06,
+      "loss": 2.7474,
+      "step": 38502
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1770069532237673e-06,
+      "loss": 1.785,
+      "step": 38503
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.176959544879899e-06,
+      "loss": 2.0352,
+      "step": 38504
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1769121365360303e-06,
+      "loss": 2.0361,
+      "step": 38505
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1768647281921618e-06,
+      "loss": 2.379,
+      "step": 38506
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1768173198482935e-06,
+      "loss": 2.7625,
+      "step": 38507
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1767699115044248e-06,
+      "loss": 2.3167,
+      "step": 38508
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1767225031605563e-06,
+      "loss": 2.039,
+      "step": 38509
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1766750948166877e-06,
+      "loss": 3.53,
+      "step": 38510
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1766276864728192e-06,
+      "loss": 3.1932,
+      "step": 38511
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1765802781289507e-06,
+      "loss": 2.2044,
+      "step": 38512
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1765328697850822e-06,
+      "loss": 1.2364,
+      "step": 38513
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1764854614412137e-06,
+      "loss": 1.4668,
+      "step": 38514
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1764380530973452e-06,
+      "loss": 2.6253,
+      "step": 38515
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1763906447534765e-06,
+      "loss": 1.2729,
+      "step": 38516
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1763432364096082e-06,
+      "loss": 3.1137,
+      "step": 38517
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1762958280657397e-06,
+      "loss": 1.1262,
+      "step": 38518
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.176248419721871e-06,
+      "loss": 2.1163,
+      "step": 38519
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1762010113780027e-06,
+      "loss": 2.5917,
+      "step": 38520
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1761536030341342e-06,
+      "loss": 1.9739,
+      "step": 38521
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1761061946902654e-06,
+      "loss": 1.4278,
+      "step": 38522
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.176058786346397e-06,
+      "loss": 1.3584,
+      "step": 38523
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1760113780025284e-06,
+      "loss": 2.326,
+      "step": 38524
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.17596396965866e-06,
+      "loss": 1.8085,
+      "step": 38525
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1759165613147914e-06,
+      "loss": 2.3168,
+      "step": 38526
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.175869152970923e-06,
+      "loss": 2.0246,
+      "step": 38527
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1758217446270544e-06,
+      "loss": 1.1892,
+      "step": 38528
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1757743362831859e-06,
+      "loss": 2.8017,
+      "step": 38529
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1757269279393174e-06,
+      "loss": 2.1757,
+      "step": 38530
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1756795195954489e-06,
+      "loss": 2.4592,
+      "step": 38531
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1756321112515804e-06,
+      "loss": 2.4634,
+      "step": 38532
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1755847029077119e-06,
+      "loss": 1.3708,
+      "step": 38533
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1755372945638433e-06,
+      "loss": 1.2438,
+      "step": 38534
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1754898862199746e-06,
+      "loss": 2.2707,
+      "step": 38535
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 1.1754424778761061e-06,
+      "loss": 2.3842,
+      "step": 38536
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1753950695322378e-06,
+      "loss": 3.1748,
+      "step": 38537
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.175347661188369e-06,
+      "loss": 1.4733,
+      "step": 38538
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1753002528445006e-06,
+      "loss": 1.6753,
+      "step": 38539
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1752528445006323e-06,
+      "loss": 1.2739,
+      "step": 38540
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1752054361567636e-06,
+      "loss": 2.2875,
+      "step": 38541
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.175158027812895e-06,
+      "loss": 2.5075,
+      "step": 38542
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1751106194690266e-06,
+      "loss": 1.5811,
+      "step": 38543
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.175063211125158e-06,
+      "loss": 2.8394,
+      "step": 38544
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1750158027812895e-06,
+      "loss": 2.844,
+      "step": 38545
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.174968394437421e-06,
+      "loss": 2.7154,
+      "step": 38546
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1749209860935525e-06,
+      "loss": 2.3956,
+      "step": 38547
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.174873577749684e-06,
+      "loss": 1.8438,
+      "step": 38548
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1748261694058153e-06,
+      "loss": 1.2231,
+      "step": 38549
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.174778761061947e-06,
+      "loss": 2.2985,
+      "step": 38550
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1747313527180785e-06,
+      "loss": 2.5889,
+      "step": 38551
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1746839443742098e-06,
+      "loss": 2.8365,
+      "step": 38552
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1746365360303415e-06,
+      "loss": 1.9648,
+      "step": 38553
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.174589127686473e-06,
+      "loss": 3.7006,
+      "step": 38554
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1745417193426043e-06,
+      "loss": 2.1541,
+      "step": 38555
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1744943109987357e-06,
+      "loss": 3.2926,
+      "step": 38556
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1744469026548672e-06,
+      "loss": 1.7073,
+      "step": 38557
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1743994943109987e-06,
+      "loss": 3.6174,
+      "step": 38558
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1743520859671302e-06,
+      "loss": 1.2369,
+      "step": 38559
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1743046776232617e-06,
+      "loss": 0.6125,
+      "step": 38560
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1742572692793932e-06,
+      "loss": 1.8322,
+      "step": 38561
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1742098609355247e-06,
+      "loss": 3.3808,
+      "step": 38562
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1741624525916562e-06,
+      "loss": 2.1372,
+      "step": 38563
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1741150442477877e-06,
+      "loss": 2.2207,
+      "step": 38564
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1740676359039192e-06,
+      "loss": 1.2751,
+      "step": 38565
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1740202275600505e-06,
+      "loss": 1.6962,
+      "step": 38566
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1739728192161822e-06,
+      "loss": 2.4888,
+      "step": 38567
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1739254108723134e-06,
+      "loss": 2.5474,
+      "step": 38568
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.173878002528445e-06,
+      "loss": 1.3398,
+      "step": 38569
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1738305941845766e-06,
+      "loss": 1.7707,
+      "step": 38570
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.173783185840708e-06,
+      "loss": 2.5193,
+      "step": 38571
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1737357774968394e-06,
+      "loss": 2.4576,
+      "step": 38572
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.173688369152971e-06,
+      "loss": 2.3186,
+      "step": 38573
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1736409608091024e-06,
+      "loss": 2.7627,
+      "step": 38574
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1735935524652339e-06,
+      "loss": 2.4649,
+      "step": 38575
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1735461441213654e-06,
+      "loss": 2.2681,
+      "step": 38576
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1734987357774969e-06,
+      "loss": 2.8264,
+      "step": 38577
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1734513274336284e-06,
+      "loss": 1.973,
+      "step": 38578
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1734039190897596e-06,
+      "loss": 2.2325,
+      "step": 38579
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1733565107458913e-06,
+      "loss": 2.8007,
+      "step": 38580
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1733091024020228e-06,
+      "loss": 1.156,
+      "step": 38581
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1732616940581541e-06,
+      "loss": 2.367,
+      "step": 38582
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1732142857142858e-06,
+      "loss": 2.6758,
+      "step": 38583
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1731668773704173e-06,
+      "loss": 2.7142,
+      "step": 38584
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1731194690265486e-06,
+      "loss": 2.8837,
+      "step": 38585
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.17307206068268e-06,
+      "loss": 1.8228,
+      "step": 38586
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1730246523388118e-06,
+      "loss": 2.3878,
+      "step": 38587
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.172977243994943e-06,
+      "loss": 0.6064,
+      "step": 38588
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1729298356510746e-06,
+      "loss": 1.8956,
+      "step": 38589
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.172882427307206e-06,
+      "loss": 0.9679,
+      "step": 38590
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1728350189633376e-06,
+      "loss": 0.5392,
+      "step": 38591
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.172787610619469e-06,
+      "loss": 1.8391,
+      "step": 38592
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1727402022756005e-06,
+      "loss": 2.1254,
+      "step": 38593
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.172692793931732e-06,
+      "loss": 2.7988,
+      "step": 38594
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1726453855878635e-06,
+      "loss": 1.4641,
+      "step": 38595
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.172597977243995e-06,
+      "loss": 1.5946,
+      "step": 38596
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1725505689001265e-06,
+      "loss": 1.4947,
+      "step": 38597
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.172503160556258e-06,
+      "loss": 1.2351,
+      "step": 38598
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 1.1724557522123893e-06,
+      "loss": 3.2193,
+      "step": 38599
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.172408343868521e-06,
+      "loss": 3.4191,
+      "step": 38600
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1723609355246523e-06,
+      "loss": 1.731,
+      "step": 38601
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1723135271807838e-06,
+      "loss": 2.1267,
+      "step": 38602
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1722661188369155e-06,
+      "loss": 2.7776,
+      "step": 38603
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1722187104930467e-06,
+      "loss": 3.2577,
+      "step": 38604
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1721713021491782e-06,
+      "loss": 1.6212,
+      "step": 38605
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1721238938053097e-06,
+      "loss": 2.0388,
+      "step": 38606
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1720764854614412e-06,
+      "loss": 3.0731,
+      "step": 38607
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1720290771175727e-06,
+      "loss": 2.739,
+      "step": 38608
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1719816687737042e-06,
+      "loss": 2.8805,
+      "step": 38609
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1719342604298357e-06,
+      "loss": 2.1077,
+      "step": 38610
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1718868520859672e-06,
+      "loss": 2.4789,
+      "step": 38611
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1718394437420985e-06,
+      "loss": 1.2708,
+      "step": 38612
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1717920353982302e-06,
+      "loss": 3.0779,
+      "step": 38613
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1717446270543617e-06,
+      "loss": 1.7382,
+      "step": 38614
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.171697218710493e-06,
+      "loss": 2.4614,
+      "step": 38615
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1716498103666246e-06,
+      "loss": 3.4661,
+      "step": 38616
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1716024020227561e-06,
+      "loss": 2.9011,
+      "step": 38617
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1715549936788874e-06,
+      "loss": 3.6802,
+      "step": 38618
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.171507585335019e-06,
+      "loss": 1.0111,
+      "step": 38619
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1714601769911506e-06,
+      "loss": 3.0719,
+      "step": 38620
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1714127686472819e-06,
+      "loss": 1.5273,
+      "step": 38621
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1713653603034134e-06,
+      "loss": 3.2916,
+      "step": 38622
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1713179519595449e-06,
+      "loss": 1.6535,
+      "step": 38623
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1712705436156764e-06,
+      "loss": 1.9902,
+      "step": 38624
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1712231352718079e-06,
+      "loss": 0.6989,
+      "step": 38625
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1711757269279394e-06,
+      "loss": 2.1802,
+      "step": 38626
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1711283185840708e-06,
+      "loss": 1.1445,
+      "step": 38627
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1710809102402023e-06,
+      "loss": 2.4962,
+      "step": 38628
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1710335018963338e-06,
+      "loss": 2.1572,
+      "step": 38629
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1709860935524653e-06,
+      "loss": 1.7942,
+      "step": 38630
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1709386852085968e-06,
+      "loss": 2.5483,
+      "step": 38631
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170891276864728e-06,
+      "loss": 1.6181,
+      "step": 38632
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1708438685208598e-06,
+      "loss": 3.0475,
+      "step": 38633
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170796460176991e-06,
+      "loss": 2.1232,
+      "step": 38634
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1707490518331226e-06,
+      "loss": 2.2606,
+      "step": 38635
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1707016434892543e-06,
+      "loss": 2.1911,
+      "step": 38636
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1706542351453856e-06,
+      "loss": 0.7801,
+      "step": 38637
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170606826801517e-06,
+      "loss": 2.4352,
+      "step": 38638
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1705594184576485e-06,
+      "loss": 0.9984,
+      "step": 38639
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.17051201011378e-06,
+      "loss": 1.7346,
+      "step": 38640
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1704646017699115e-06,
+      "loss": 1.9549,
+      "step": 38641
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170417193426043e-06,
+      "loss": 2.0244,
+      "step": 38642
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1703697850821745e-06,
+      "loss": 1.0886,
+      "step": 38643
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170322376738306e-06,
+      "loss": 1.6352,
+      "step": 38644
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1702749683944373e-06,
+      "loss": 2.3462,
+      "step": 38645
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170227560050569e-06,
+      "loss": 3.0721,
+      "step": 38646
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1701801517067005e-06,
+      "loss": 2.7418,
+      "step": 38647
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1701327433628318e-06,
+      "loss": 1.6656,
+      "step": 38648
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1700853350189635e-06,
+      "loss": 2.0584,
+      "step": 38649
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.170037926675095e-06,
+      "loss": 3.2178,
+      "step": 38650
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1699905183312262e-06,
+      "loss": 3.2507,
+      "step": 38651
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1699431099873577e-06,
+      "loss": 1.0632,
+      "step": 38652
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1698957016434894e-06,
+      "loss": 0.8231,
+      "step": 38653
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1698482932996207e-06,
+      "loss": 1.4381,
+      "step": 38654
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1698008849557522e-06,
+      "loss": 3.1166,
+      "step": 38655
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1697534766118837e-06,
+      "loss": 1.0394,
+      "step": 38656
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1697060682680152e-06,
+      "loss": 3.3885,
+      "step": 38657
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1696586599241467e-06,
+      "loss": 2.5858,
+      "step": 38658
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1696112515802782e-06,
+      "loss": 2.5555,
+      "step": 38659
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1695638432364097e-06,
+      "loss": 2.5813,
+      "step": 38660
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1695164348925412e-06,
+      "loss": 1.8672,
+      "step": 38661
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 1.1694690265486726e-06,
+      "loss": 0.2736,
+      "step": 38662
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1694216182048041e-06,
+      "loss": 0.8998,
+      "step": 38663
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1693742098609356e-06,
+      "loss": 3.2343,
+      "step": 38664
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.169326801517067e-06,
+      "loss": 1.7519,
+      "step": 38665
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1692793931731986e-06,
+      "loss": 1.3258,
+      "step": 38666
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.16923198482933e-06,
+      "loss": 0.9682,
+      "step": 38667
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1691845764854614e-06,
+      "loss": 2.1273,
+      "step": 38668
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.169137168141593e-06,
+      "loss": 2.8428,
+      "step": 38669
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1690897597977244e-06,
+      "loss": 2.3045,
+      "step": 38670
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1690423514538559e-06,
+      "loss": 2.3228,
+      "step": 38671
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1689949431099874e-06,
+      "loss": 1.8339,
+      "step": 38672
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1689475347661189e-06,
+      "loss": 2.9889,
+      "step": 38673
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1689001264222503e-06,
+      "loss": 3.1208,
+      "step": 38674
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1688527180783818e-06,
+      "loss": 2.0114,
+      "step": 38675
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1688053097345133e-06,
+      "loss": 3.7913,
+      "step": 38676
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1687579013906448e-06,
+      "loss": 2.4414,
+      "step": 38677
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.168710493046776e-06,
+      "loss": 2.1357,
+      "step": 38678
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1686630847029078e-06,
+      "loss": 1.3916,
+      "step": 38679
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1686156763590393e-06,
+      "loss": 1.9531,
+      "step": 38680
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1685682680151706e-06,
+      "loss": 2.7557,
+      "step": 38681
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1685208596713023e-06,
+      "loss": 3.1071,
+      "step": 38682
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1684734513274338e-06,
+      "loss": 1.5017,
+      "step": 38683
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.168426042983565e-06,
+      "loss": 1.8441,
+      "step": 38684
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1683786346396965e-06,
+      "loss": 0.9763,
+      "step": 38685
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1683312262958282e-06,
+      "loss": 1.3432,
+      "step": 38686
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1682838179519595e-06,
+      "loss": 3.489,
+      "step": 38687
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.168236409608091e-06,
+      "loss": 1.3089,
+      "step": 38688
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1681890012642225e-06,
+      "loss": 2.3504,
+      "step": 38689
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.168141592920354e-06,
+      "loss": 1.8184,
+      "step": 38690
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1680941845764855e-06,
+      "loss": 2.2441,
+      "step": 38691
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.168046776232617e-06,
+      "loss": 1.8747,
+      "step": 38692
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1679993678887485e-06,
+      "loss": 2.2786,
+      "step": 38693
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.16795195954488e-06,
+      "loss": 2.3504,
+      "step": 38694
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1679045512010113e-06,
+      "loss": 1.2691,
+      "step": 38695
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.167857142857143e-06,
+      "loss": 1.5001,
+      "step": 38696
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1678097345132745e-06,
+      "loss": 2.6645,
+      "step": 38697
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1677623261694057e-06,
+      "loss": 1.614,
+      "step": 38698
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1677149178255374e-06,
+      "loss": 3.3559,
+      "step": 38699
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1676675094816687e-06,
+      "loss": 1.8229,
+      "step": 38700
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1676201011378002e-06,
+      "loss": 1.9315,
+      "step": 38701
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1675726927939317e-06,
+      "loss": 1.5779,
+      "step": 38702
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1675252844500632e-06,
+      "loss": 2.3727,
+      "step": 38703
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1674778761061947e-06,
+      "loss": 1.616,
+      "step": 38704
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1674304677623262e-06,
+      "loss": 1.4741,
+      "step": 38705
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1673830594184577e-06,
+      "loss": 2.7561,
+      "step": 38706
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1673356510745892e-06,
+      "loss": 2.012,
+      "step": 38707
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1672882427307207e-06,
+      "loss": 0.957,
+      "step": 38708
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1672408343868521e-06,
+      "loss": 1.751,
+      "step": 38709
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1671934260429836e-06,
+      "loss": 1.7107,
+      "step": 38710
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.167146017699115e-06,
+      "loss": 1.6607,
+      "step": 38711
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1670986093552466e-06,
+      "loss": 2.1212,
+      "step": 38712
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1670512010113781e-06,
+      "loss": 1.3435,
+      "step": 38713
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1670037926675094e-06,
+      "loss": 1.372,
+      "step": 38714
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1669563843236409e-06,
+      "loss": 2.5767,
+      "step": 38715
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1669089759797726e-06,
+      "loss": 1.2984,
+      "step": 38716
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1668615676359039e-06,
+      "loss": 3.0377,
+      "step": 38717
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1668141592920354e-06,
+      "loss": 2.1695,
+      "step": 38718
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.166766750948167e-06,
+      "loss": 0.7955,
+      "step": 38719
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1667193426042983e-06,
+      "loss": 1.8249,
+      "step": 38720
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1666719342604298e-06,
+      "loss": 2.1824,
+      "step": 38721
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1666245259165613e-06,
+      "loss": 1.6097,
+      "step": 38722
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1665771175726928e-06,
+      "loss": 2.5703,
+      "step": 38723
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1665297092288243e-06,
+      "loss": 1.3091,
+      "step": 38724
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1664823008849558e-06,
+      "loss": 2.6589,
+      "step": 38725
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 1.1664348925410873e-06,
+      "loss": 2.0118,
+      "step": 38726
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1663874841972188e-06,
+      "loss": 2.3017,
+      "step": 38727
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.16634007585335e-06,
+      "loss": 2.454,
+      "step": 38728
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1662926675094818e-06,
+      "loss": 3.078,
+      "step": 38729
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1662452591656133e-06,
+      "loss": 2.8738,
+      "step": 38730
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1661978508217446e-06,
+      "loss": 1.6735,
+      "step": 38731
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1661504424778763e-06,
+      "loss": 2.7397,
+      "step": 38732
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1661030341340075e-06,
+      "loss": 1.7874,
+      "step": 38733
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.166055625790139e-06,
+      "loss": 2.8836,
+      "step": 38734
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1660082174462705e-06,
+      "loss": 1.1213,
+      "step": 38735
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.165960809102402e-06,
+      "loss": 2.4024,
+      "step": 38736
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1659134007585335e-06,
+      "loss": 2.0709,
+      "step": 38737
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.165865992414665e-06,
+      "loss": 2.3409,
+      "step": 38738
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1658185840707965e-06,
+      "loss": 1.4844,
+      "step": 38739
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.165771175726928e-06,
+      "loss": 2.1205,
+      "step": 38740
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1657237673830595e-06,
+      "loss": 1.1181,
+      "step": 38741
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.165676359039191e-06,
+      "loss": 2.3157,
+      "step": 38742
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1656289506953225e-06,
+      "loss": 3.9612,
+      "step": 38743
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1655815423514537e-06,
+      "loss": 3.1329,
+      "step": 38744
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1655341340075854e-06,
+      "loss": 2.403,
+      "step": 38745
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.165486725663717e-06,
+      "loss": 1.2152,
+      "step": 38746
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1654393173198482e-06,
+      "loss": 2.2066,
+      "step": 38747
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1653919089759797e-06,
+      "loss": 0.9926,
+      "step": 38748
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1653445006321114e-06,
+      "loss": 1.2789,
+      "step": 38749
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1652970922882427e-06,
+      "loss": 1.9092,
+      "step": 38750
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1652496839443742e-06,
+      "loss": 1.1146,
+      "step": 38751
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1652022756005059e-06,
+      "loss": 1.5058,
+      "step": 38752
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1651548672566372e-06,
+      "loss": 2.4435,
+      "step": 38753
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1651074589127687e-06,
+      "loss": 2.2562,
+      "step": 38754
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1650600505689002e-06,
+      "loss": 1.6629,
+      "step": 38755
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1650126422250316e-06,
+      "loss": 2.2761,
+      "step": 38756
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1649652338811631e-06,
+      "loss": 1.6108,
+      "step": 38757
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1649178255372946e-06,
+      "loss": 2.553,
+      "step": 38758
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1648704171934261e-06,
+      "loss": 2.1739,
+      "step": 38759
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1648230088495576e-06,
+      "loss": 1.2062,
+      "step": 38760
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1647756005056889e-06,
+      "loss": 2.6337,
+      "step": 38761
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1647281921618206e-06,
+      "loss": 2.3595,
+      "step": 38762
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.164680783817952e-06,
+      "loss": 2.421,
+      "step": 38763
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1646333754740834e-06,
+      "loss": 1.9654,
+      "step": 38764
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.164585967130215e-06,
+      "loss": 1.7456,
+      "step": 38765
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1645385587863464e-06,
+      "loss": 2.1311,
+      "step": 38766
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1644911504424778e-06,
+      "loss": 0.979,
+      "step": 38767
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1644437420986093e-06,
+      "loss": 2.6547,
+      "step": 38768
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1643963337547408e-06,
+      "loss": 2.3965,
+      "step": 38769
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1643489254108723e-06,
+      "loss": 3.0649,
+      "step": 38770
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1643015170670038e-06,
+      "loss": 1.9307,
+      "step": 38771
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1642541087231353e-06,
+      "loss": 0.9953,
+      "step": 38772
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1642067003792668e-06,
+      "loss": 1.333,
+      "step": 38773
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1641592920353983e-06,
+      "loss": 1.3788,
+      "step": 38774
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1641118836915298e-06,
+      "loss": 2.7272,
+      "step": 38775
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1640644753476613e-06,
+      "loss": 2.8707,
+      "step": 38776
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1640170670037926e-06,
+      "loss": 2.836,
+      "step": 38777
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1639696586599243e-06,
+      "loss": 1.7618,
+      "step": 38778
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1639222503160558e-06,
+      "loss": 2.155,
+      "step": 38779
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.163874841972187e-06,
+      "loss": 1.1759,
+      "step": 38780
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1638274336283185e-06,
+      "loss": 1.3733,
+      "step": 38781
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1637800252844502e-06,
+      "loss": 1.7594,
+      "step": 38782
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1637326169405815e-06,
+      "loss": 2.5645,
+      "step": 38783
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.163685208596713e-06,
+      "loss": 1.9032,
+      "step": 38784
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1636378002528447e-06,
+      "loss": 1.0242,
+      "step": 38785
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.163590391908976e-06,
+      "loss": 1.3008,
+      "step": 38786
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1635429835651075e-06,
+      "loss": 3.0954,
+      "step": 38787
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.163495575221239e-06,
+      "loss": 1.6682,
+      "step": 38788
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 1.1634481668773705e-06,
+      "loss": 2.5318,
+      "step": 38789
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.163400758533502e-06,
+      "loss": 2.5936,
+      "step": 38790
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1633533501896334e-06,
+      "loss": 1.6455,
+      "step": 38791
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.163305941845765e-06,
+      "loss": 3.3584,
+      "step": 38792
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1632585335018964e-06,
+      "loss": 2.4277,
+      "step": 38793
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1632111251580277e-06,
+      "loss": 2.1239,
+      "step": 38794
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1631637168141594e-06,
+      "loss": 1.2454,
+      "step": 38795
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.163116308470291e-06,
+      "loss": 1.9135,
+      "step": 38796
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1630689001264222e-06,
+      "loss": 1.3853,
+      "step": 38797
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1630214917825539e-06,
+      "loss": 2.9624,
+      "step": 38798
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1629740834386852e-06,
+      "loss": 2.0025,
+      "step": 38799
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1629266750948167e-06,
+      "loss": 2.5981,
+      "step": 38800
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1628792667509482e-06,
+      "loss": 2.3809,
+      "step": 38801
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1628318584070796e-06,
+      "loss": 2.6892,
+      "step": 38802
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1627844500632111e-06,
+      "loss": 1.6391,
+      "step": 38803
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1627370417193426e-06,
+      "loss": 1.6069,
+      "step": 38804
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1626896333754741e-06,
+      "loss": 0.6531,
+      "step": 38805
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1626422250316056e-06,
+      "loss": 2.2276,
+      "step": 38806
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1625948166877371e-06,
+      "loss": 1.3245,
+      "step": 38807
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1625474083438686e-06,
+      "loss": 3.6269,
+      "step": 38808
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1625e-06,
+      "loss": 2.6141,
+      "step": 38809
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1624525916561314e-06,
+      "loss": 1.7738,
+      "step": 38810
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.162405183312263e-06,
+      "loss": 1.8334,
+      "step": 38811
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1623577749683946e-06,
+      "loss": 3.3766,
+      "step": 38812
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1623103666245258e-06,
+      "loss": 2.3723,
+      "step": 38813
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1622629582806573e-06,
+      "loss": 2.3926,
+      "step": 38814
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.162215549936789e-06,
+      "loss": 1.5441,
+      "step": 38815
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1621681415929203e-06,
+      "loss": 1.4924,
+      "step": 38816
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1621207332490518e-06,
+      "loss": 1.2621,
+      "step": 38817
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1620733249051835e-06,
+      "loss": 0.9363,
+      "step": 38818
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1620259165613148e-06,
+      "loss": 3.3409,
+      "step": 38819
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1619785082174463e-06,
+      "loss": 1.4503,
+      "step": 38820
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1619310998735778e-06,
+      "loss": 2.6486,
+      "step": 38821
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1618836915297093e-06,
+      "loss": 1.8106,
+      "step": 38822
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1618362831858408e-06,
+      "loss": 1.829,
+      "step": 38823
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1617888748419723e-06,
+      "loss": 3.0074,
+      "step": 38824
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1617414664981038e-06,
+      "loss": 2.8424,
+      "step": 38825
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1616940581542352e-06,
+      "loss": 0.6617,
+      "step": 38826
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1616466498103665e-06,
+      "loss": 2.0192,
+      "step": 38827
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1615992414664982e-06,
+      "loss": 2.2963,
+      "step": 38828
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1615518331226297e-06,
+      "loss": 1.4053,
+      "step": 38829
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.161504424778761e-06,
+      "loss": 1.5026,
+      "step": 38830
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1614570164348927e-06,
+      "loss": 3.8532,
+      "step": 38831
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.161409608091024e-06,
+      "loss": 3.1047,
+      "step": 38832
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1613621997471555e-06,
+      "loss": 1.4039,
+      "step": 38833
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.161314791403287e-06,
+      "loss": 1.7094,
+      "step": 38834
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1612673830594185e-06,
+      "loss": 2.5427,
+      "step": 38835
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.16121997471555e-06,
+      "loss": 2.8758,
+      "step": 38836
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1611725663716815e-06,
+      "loss": 1.7232,
+      "step": 38837
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.161125158027813e-06,
+      "loss": 1.0551,
+      "step": 38838
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1610777496839444e-06,
+      "loss": 2.7053,
+      "step": 38839
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.161030341340076e-06,
+      "loss": 2.0456,
+      "step": 38840
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1609829329962074e-06,
+      "loss": 2.3571,
+      "step": 38841
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.160935524652339e-06,
+      "loss": 3.2922,
+      "step": 38842
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1608881163084702e-06,
+      "loss": 1.5156,
+      "step": 38843
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1608407079646017e-06,
+      "loss": 1.0151,
+      "step": 38844
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1607932996207334e-06,
+      "loss": 1.2771,
+      "step": 38845
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1607458912768647e-06,
+      "loss": 2.3388,
+      "step": 38846
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1606984829329962e-06,
+      "loss": 2.4608,
+      "step": 38847
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1606510745891279e-06,
+      "loss": 2.2918,
+      "step": 38848
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1606036662452591e-06,
+      "loss": 3.2997,
+      "step": 38849
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1605562579013906e-06,
+      "loss": 2.7679,
+      "step": 38850
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1605088495575221e-06,
+      "loss": 2.7576,
+      "step": 38851
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 1.1604614412136536e-06,
+      "loss": 1.8287,
+      "step": 38852
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1604140328697851e-06,
+      "loss": 0.8426,
+      "step": 38853
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1603666245259166e-06,
+      "loss": 2.9667,
+      "step": 38854
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.160319216182048e-06,
+      "loss": 1.5068,
+      "step": 38855
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1602718078381796e-06,
+      "loss": 2.3394,
+      "step": 38856
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1602243994943109e-06,
+      "loss": 0.9976,
+      "step": 38857
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1601769911504426e-06,
+      "loss": 1.2255,
+      "step": 38858
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.160129582806574e-06,
+      "loss": 1.7674,
+      "step": 38859
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1600821744627053e-06,
+      "loss": 0.9057,
+      "step": 38860
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.160034766118837e-06,
+      "loss": 1.7531,
+      "step": 38861
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1599873577749685e-06,
+      "loss": 2.2158,
+      "step": 38862
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1599399494310998e-06,
+      "loss": 1.1023,
+      "step": 38863
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1598925410872313e-06,
+      "loss": 0.822,
+      "step": 38864
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1598451327433628e-06,
+      "loss": 2.0532,
+      "step": 38865
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1597977243994943e-06,
+      "loss": 1.6511,
+      "step": 38866
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1597503160556258e-06,
+      "loss": 2.4397,
+      "step": 38867
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1597029077117573e-06,
+      "loss": 3.2828,
+      "step": 38868
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1596554993678888e-06,
+      "loss": 1.6928,
+      "step": 38869
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1596080910240203e-06,
+      "loss": 2.2893,
+      "step": 38870
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1595606826801518e-06,
+      "loss": 1.9934,
+      "step": 38871
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1595132743362833e-06,
+      "loss": 1.8763,
+      "step": 38872
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1594658659924147e-06,
+      "loss": 1.4506,
+      "step": 38873
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1594184576485462e-06,
+      "loss": 0.6875,
+      "step": 38874
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1593710493046777e-06,
+      "loss": 1.0968,
+      "step": 38875
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.159323640960809e-06,
+      "loss": 2.3514,
+      "step": 38876
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1592762326169405e-06,
+      "loss": 0.7108,
+      "step": 38877
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1592288242730722e-06,
+      "loss": 2.0673,
+      "step": 38878
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1591814159292035e-06,
+      "loss": 3.5797,
+      "step": 38879
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.159134007585335e-06,
+      "loss": 1.431,
+      "step": 38880
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1590865992414667e-06,
+      "loss": 1.4884,
+      "step": 38881
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.159039190897598e-06,
+      "loss": 1.9156,
+      "step": 38882
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1589917825537295e-06,
+      "loss": 2.5552,
+      "step": 38883
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.158944374209861e-06,
+      "loss": 2.0217,
+      "step": 38884
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1588969658659924e-06,
+      "loss": 2.171,
+      "step": 38885
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.158849557522124e-06,
+      "loss": 1.0286,
+      "step": 38886
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1588021491782554e-06,
+      "loss": 2.5822,
+      "step": 38887
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.158754740834387e-06,
+      "loss": 1.7006,
+      "step": 38888
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1587073324905184e-06,
+      "loss": 1.3597,
+      "step": 38889
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1586599241466497e-06,
+      "loss": 1.6467,
+      "step": 38890
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1586125158027814e-06,
+      "loss": 3.5639,
+      "step": 38891
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1585651074589129e-06,
+      "loss": 3.5231,
+      "step": 38892
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1585176991150442e-06,
+      "loss": 2.1942,
+      "step": 38893
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1584702907711759e-06,
+      "loss": 2.6899,
+      "step": 38894
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1584228824273074e-06,
+      "loss": 1.8415,
+      "step": 38895
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1583754740834386e-06,
+      "loss": 1.6107,
+      "step": 38896
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1583280657395701e-06,
+      "loss": 2.9998,
+      "step": 38897
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1582806573957016e-06,
+      "loss": 1.7402,
+      "step": 38898
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1582332490518331e-06,
+      "loss": 1.7345,
+      "step": 38899
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1581858407079646e-06,
+      "loss": 2.397,
+      "step": 38900
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.158138432364096e-06,
+      "loss": 1.0796,
+      "step": 38901
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1580910240202276e-06,
+      "loss": 1.7514,
+      "step": 38902
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.158043615676359e-06,
+      "loss": 3.1661,
+      "step": 38903
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1579962073324906e-06,
+      "loss": 3.1072,
+      "step": 38904
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.157948798988622e-06,
+      "loss": 2.6814,
+      "step": 38905
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1579013906447536e-06,
+      "loss": 1.6492,
+      "step": 38906
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.157853982300885e-06,
+      "loss": 1.8322,
+      "step": 38907
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1578065739570165e-06,
+      "loss": 2.1754,
+      "step": 38908
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1577591656131478e-06,
+      "loss": 2.5263,
+      "step": 38909
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1577117572692793e-06,
+      "loss": 1.5756,
+      "step": 38910
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.157664348925411e-06,
+      "loss": 2.2093,
+      "step": 38911
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1576169405815423e-06,
+      "loss": 2.3351,
+      "step": 38912
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1575695322376738e-06,
+      "loss": 1.7829,
+      "step": 38913
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1575221238938055e-06,
+      "loss": 1.1809,
+      "step": 38914
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1574747155499368e-06,
+      "loss": 4.0592,
+      "step": 38915
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 1.1574273072060683e-06,
+      "loss": 2.0032,
+      "step": 38916
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1573798988621998e-06,
+      "loss": 2.2706,
+      "step": 38917
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1573324905183313e-06,
+      "loss": 2.2458,
+      "step": 38918
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1572850821744628e-06,
+      "loss": 2.0107,
+      "step": 38919
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1572376738305942e-06,
+      "loss": 2.6203,
+      "step": 38920
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1571902654867257e-06,
+      "loss": 3.2777,
+      "step": 38921
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1571428571428572e-06,
+      "loss": 2.1451,
+      "step": 38922
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1570954487989885e-06,
+      "loss": 1.8603,
+      "step": 38923
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1570480404551202e-06,
+      "loss": 2.2041,
+      "step": 38924
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1570006321112517e-06,
+      "loss": 2.5188,
+      "step": 38925
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.156953223767383e-06,
+      "loss": 2.6433,
+      "step": 38926
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1569058154235147e-06,
+      "loss": 2.5146,
+      "step": 38927
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1568584070796462e-06,
+      "loss": 2.0996,
+      "step": 38928
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1568109987357775e-06,
+      "loss": 2.3727,
+      "step": 38929
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.156763590391909e-06,
+      "loss": 4.3533,
+      "step": 38930
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1567161820480404e-06,
+      "loss": 1.7154,
+      "step": 38931
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.156668773704172e-06,
+      "loss": 2.5196,
+      "step": 38932
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1566213653603034e-06,
+      "loss": 2.3209,
+      "step": 38933
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.156573957016435e-06,
+      "loss": 3.1407,
+      "step": 38934
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1565265486725664e-06,
+      "loss": 3.334,
+      "step": 38935
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.156479140328698e-06,
+      "loss": 3.9014,
+      "step": 38936
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1564317319848294e-06,
+      "loss": 2.4325,
+      "step": 38937
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1563843236409609e-06,
+      "loss": 2.6857,
+      "step": 38938
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1563369152970924e-06,
+      "loss": 1.9401,
+      "step": 38939
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1562895069532239e-06,
+      "loss": 2.0028,
+      "step": 38940
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1562420986093554e-06,
+      "loss": 2.1284,
+      "step": 38941
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1561946902654866e-06,
+      "loss": 2.8819,
+      "step": 38942
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1561472819216181e-06,
+      "loss": 3.1335,
+      "step": 38943
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1560998735777498e-06,
+      "loss": 1.4715,
+      "step": 38944
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1560524652338811e-06,
+      "loss": 1.2102,
+      "step": 38945
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1560050568900126e-06,
+      "loss": 1.8073,
+      "step": 38946
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1559576485461443e-06,
+      "loss": 2.012,
+      "step": 38947
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1559102402022756e-06,
+      "loss": 2.5482,
+      "step": 38948
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.155862831858407e-06,
+      "loss": 2.4543,
+      "step": 38949
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1558154235145386e-06,
+      "loss": 2.1389,
+      "step": 38950
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.15576801517067e-06,
+      "loss": 1.8548,
+      "step": 38951
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1557206068268016e-06,
+      "loss": 1.1345,
+      "step": 38952
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.155673198482933e-06,
+      "loss": 2.9865,
+      "step": 38953
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1556257901390646e-06,
+      "loss": 0.7065,
+      "step": 38954
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.155578381795196e-06,
+      "loss": 1.203,
+      "step": 38955
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1555309734513273e-06,
+      "loss": 0.5945,
+      "step": 38956
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.155483565107459e-06,
+      "loss": 2.0809,
+      "step": 38957
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1554361567635905e-06,
+      "loss": 3.1629,
+      "step": 38958
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1553887484197218e-06,
+      "loss": 2.6779,
+      "step": 38959
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1553413400758535e-06,
+      "loss": 2.0916,
+      "step": 38960
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1552939317319848e-06,
+      "loss": 1.3719,
+      "step": 38961
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1552465233881163e-06,
+      "loss": 1.8576,
+      "step": 38962
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1551991150442478e-06,
+      "loss": 3.0165,
+      "step": 38963
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1551517067003793e-06,
+      "loss": 2.528,
+      "step": 38964
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1551042983565108e-06,
+      "loss": 1.1519,
+      "step": 38965
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1550568900126422e-06,
+      "loss": 2.1918,
+      "step": 38966
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1550094816687737e-06,
+      "loss": 4.0452,
+      "step": 38967
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1549620733249052e-06,
+      "loss": 2.971,
+      "step": 38968
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1549146649810367e-06,
+      "loss": 1.5388,
+      "step": 38969
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1548672566371682e-06,
+      "loss": 2.548,
+      "step": 38970
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1548198482932997e-06,
+      "loss": 2.2857,
+      "step": 38971
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.154772439949431e-06,
+      "loss": 2.0909,
+      "step": 38972
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1547250316055627e-06,
+      "loss": 2.4595,
+      "step": 38973
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1546776232616942e-06,
+      "loss": 1.4484,
+      "step": 38974
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1546302149178255e-06,
+      "loss": 2.4876,
+      "step": 38975
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.154582806573957e-06,
+      "loss": 2.5021,
+      "step": 38976
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1545353982300887e-06,
+      "loss": 3.5043,
+      "step": 38977
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.15448798988622e-06,
+      "loss": 3.2005,
+      "step": 38978
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 1.1544405815423514e-06,
+      "loss": 1.8503,
+      "step": 38979
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1543931731984831e-06,
+      "loss": 2.9939,
+      "step": 38980
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1543457648546144e-06,
+      "loss": 1.9819,
+      "step": 38981
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.154298356510746e-06,
+      "loss": 3.2668,
+      "step": 38982
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1542509481668774e-06,
+      "loss": 1.6375,
+      "step": 38983
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.154203539823009e-06,
+      "loss": 1.3431,
+      "step": 38984
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1541561314791404e-06,
+      "loss": 2.2421,
+      "step": 38985
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1541087231352717e-06,
+      "loss": 2.5123,
+      "step": 38986
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1540613147914034e-06,
+      "loss": 2.1162,
+      "step": 38987
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1540139064475349e-06,
+      "loss": 0.8085,
+      "step": 38988
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1539664981036661e-06,
+      "loss": 2.56,
+      "step": 38989
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1539190897597978e-06,
+      "loss": 2.8318,
+      "step": 38990
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1538716814159293e-06,
+      "loss": 2.2041,
+      "step": 38991
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1538242730720606e-06,
+      "loss": 0.9301,
+      "step": 38992
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1537768647281921e-06,
+      "loss": 1.4806,
+      "step": 38993
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1537294563843236e-06,
+      "loss": 0.707,
+      "step": 38994
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.153682048040455e-06,
+      "loss": 1.8884,
+      "step": 38995
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1536346396965866e-06,
+      "loss": 1.2687,
+      "step": 38996
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.153587231352718e-06,
+      "loss": 2.7548,
+      "step": 38997
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1535398230088496e-06,
+      "loss": 2.8501,
+      "step": 38998
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.153492414664981e-06,
+      "loss": 2.128,
+      "step": 38999
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1534450063211126e-06,
+      "loss": 2.4041,
+      "step": 39000
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.153397597977244e-06,
+      "loss": 0.7521,
+      "step": 39001
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1533501896333755e-06,
+      "loss": 1.3117,
+      "step": 39002
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.153302781289507e-06,
+      "loss": 1.3709,
+      "step": 39003
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1532553729456385e-06,
+      "loss": 2.8962,
+      "step": 39004
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1532079646017698e-06,
+      "loss": 2.8237,
+      "step": 39005
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1531605562579013e-06,
+      "loss": 2.3572,
+      "step": 39006
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.153113147914033e-06,
+      "loss": 3.2087,
+      "step": 39007
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1530657395701643e-06,
+      "loss": 2.7876,
+      "step": 39008
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1530183312262958e-06,
+      "loss": 1.6378,
+      "step": 39009
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1529709228824275e-06,
+      "loss": 2.5067,
+      "step": 39010
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1529235145385588e-06,
+      "loss": 1.8774,
+      "step": 39011
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1528761061946903e-06,
+      "loss": 2.7453,
+      "step": 39012
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1528286978508217e-06,
+      "loss": 1.9952,
+      "step": 39013
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1527812895069532e-06,
+      "loss": 2.0275,
+      "step": 39014
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1527338811630847e-06,
+      "loss": 2.41,
+      "step": 39015
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1526864728192162e-06,
+      "loss": 2.0983,
+      "step": 39016
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1526390644753477e-06,
+      "loss": 2.727,
+      "step": 39017
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1525916561314792e-06,
+      "loss": 1.5455,
+      "step": 39018
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1525442477876105e-06,
+      "loss": 2.5189,
+      "step": 39019
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1524968394437422e-06,
+      "loss": 1.6764,
+      "step": 39020
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1524494310998737e-06,
+      "loss": 1.6592,
+      "step": 39021
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.152402022756005e-06,
+      "loss": 2.3956,
+      "step": 39022
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1523546144121367e-06,
+      "loss": 2.1506,
+      "step": 39023
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1523072060682682e-06,
+      "loss": 2.9428,
+      "step": 39024
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1522597977243994e-06,
+      "loss": 0.8818,
+      "step": 39025
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.152212389380531e-06,
+      "loss": 1.8684,
+      "step": 39026
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1521649810366624e-06,
+      "loss": 1.9055,
+      "step": 39027
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.152117572692794e-06,
+      "loss": 2.5356,
+      "step": 39028
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1520701643489254e-06,
+      "loss": 2.3619,
+      "step": 39029
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.152022756005057e-06,
+      "loss": 4.6325,
+      "step": 39030
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1519753476611884e-06,
+      "loss": 2.3189,
+      "step": 39031
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1519279393173199e-06,
+      "loss": 2.2005,
+      "step": 39032
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1518805309734514e-06,
+      "loss": 3.3219,
+      "step": 39033
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1518331226295829e-06,
+      "loss": 2.4847,
+      "step": 39034
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1517857142857144e-06,
+      "loss": 2.2035,
+      "step": 39035
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1517383059418459e-06,
+      "loss": 2.5961,
+      "step": 39036
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1516908975979773e-06,
+      "loss": 1.9134,
+      "step": 39037
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1516434892541086e-06,
+      "loss": 1.2394,
+      "step": 39038
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1515960809102401e-06,
+      "loss": 1.592,
+      "step": 39039
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1515486725663718e-06,
+      "loss": 0.5948,
+      "step": 39040
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.151501264222503e-06,
+      "loss": 1.747,
+      "step": 39041
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 1.1514538558786346e-06,
+      "loss": 3.1815,
+      "step": 39042
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1514064475347663e-06,
+      "loss": 1.9774,
+      "step": 39043
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1513590391908976e-06,
+      "loss": 2.5106,
+      "step": 39044
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.151311630847029e-06,
+      "loss": 2.7594,
+      "step": 39045
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1512642225031606e-06,
+      "loss": 0.3556,
+      "step": 39046
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.151216814159292e-06,
+      "loss": 3.666,
+      "step": 39047
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1511694058154235e-06,
+      "loss": 2.1974,
+      "step": 39048
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.151121997471555e-06,
+      "loss": 1.832,
+      "step": 39049
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1510745891276865e-06,
+      "loss": 3.464,
+      "step": 39050
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.151027180783818e-06,
+      "loss": 0.9323,
+      "step": 39051
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1509797724399493e-06,
+      "loss": 1.1988,
+      "step": 39052
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.150932364096081e-06,
+      "loss": 2.421,
+      "step": 39053
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1508849557522125e-06,
+      "loss": 1.4965,
+      "step": 39054
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1508375474083438e-06,
+      "loss": 1.1187,
+      "step": 39055
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1507901390644755e-06,
+      "loss": 1.555,
+      "step": 39056
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.150742730720607e-06,
+      "loss": 1.0948,
+      "step": 39057
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1506953223767383e-06,
+      "loss": 2.0743,
+      "step": 39058
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1506479140328697e-06,
+      "loss": 1.5727,
+      "step": 39059
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1506005056890012e-06,
+      "loss": 1.8095,
+      "step": 39060
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1505530973451327e-06,
+      "loss": 1.7664,
+      "step": 39061
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1505056890012642e-06,
+      "loss": 2.4484,
+      "step": 39062
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1504582806573957e-06,
+      "loss": 1.3133,
+      "step": 39063
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1504108723135272e-06,
+      "loss": 2.2308,
+      "step": 39064
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1503634639696587e-06,
+      "loss": 1.4242,
+      "step": 39065
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1503160556257902e-06,
+      "loss": 1.9354,
+      "step": 39066
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1502686472819217e-06,
+      "loss": 1.5096,
+      "step": 39067
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1502212389380532e-06,
+      "loss": 2.6084,
+      "step": 39068
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1501738305941847e-06,
+      "loss": 1.9733,
+      "step": 39069
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1501264222503162e-06,
+      "loss": 2.8208,
+      "step": 39070
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1500790139064474e-06,
+      "loss": 1.9679,
+      "step": 39071
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.150031605562579e-06,
+      "loss": 0.9142,
+      "step": 39072
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1499841972187106e-06,
+      "loss": 1.5737,
+      "step": 39073
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.149936788874842e-06,
+      "loss": 0.8138,
+      "step": 39074
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1498893805309734e-06,
+      "loss": 2.3378,
+      "step": 39075
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1498419721871051e-06,
+      "loss": 1.6118,
+      "step": 39076
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1497945638432364e-06,
+      "loss": 2.0997,
+      "step": 39077
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1497471554993679e-06,
+      "loss": 0.4507,
+      "step": 39078
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1496997471554994e-06,
+      "loss": 1.4005,
+      "step": 39079
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1496523388116309e-06,
+      "loss": 1.6368,
+      "step": 39080
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1496049304677624e-06,
+      "loss": 1.4491,
+      "step": 39081
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1495575221238939e-06,
+      "loss": 2.7503,
+      "step": 39082
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1495101137800253e-06,
+      "loss": 2.7182,
+      "step": 39083
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1494627054361568e-06,
+      "loss": 2.0083,
+      "step": 39084
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1494152970922881e-06,
+      "loss": 3.0746,
+      "step": 39085
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1493678887484198e-06,
+      "loss": 2.9195,
+      "step": 39086
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1493204804045513e-06,
+      "loss": 1.6914,
+      "step": 39087
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1492730720606826e-06,
+      "loss": 2.4854,
+      "step": 39088
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1492256637168143e-06,
+      "loss": 1.9457,
+      "step": 39089
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1491782553729458e-06,
+      "loss": 3.7787,
+      "step": 39090
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.149130847029077e-06,
+      "loss": 3.4075,
+      "step": 39091
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1490834386852086e-06,
+      "loss": 2.3029,
+      "step": 39092
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.14903603034134e-06,
+      "loss": 2.5797,
+      "step": 39093
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1489886219974716e-06,
+      "loss": 2.7581,
+      "step": 39094
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.148941213653603e-06,
+      "loss": 1.2203,
+      "step": 39095
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1488938053097345e-06,
+      "loss": 1.6381,
+      "step": 39096
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.148846396965866e-06,
+      "loss": 0.543,
+      "step": 39097
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1487989886219975e-06,
+      "loss": 1.4553,
+      "step": 39098
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.148751580278129e-06,
+      "loss": 2.4781,
+      "step": 39099
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1487041719342605e-06,
+      "loss": 0.3331,
+      "step": 39100
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.148656763590392e-06,
+      "loss": 2.1118,
+      "step": 39101
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1486093552465235e-06,
+      "loss": 1.2352,
+      "step": 39102
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.148561946902655e-06,
+      "loss": 0.639,
+      "step": 39103
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1485145385587863e-06,
+      "loss": 2.1206,
+      "step": 39104
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1484671302149178e-06,
+      "loss": 2.1878,
+      "step": 39105
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 1.1484197218710495e-06,
+      "loss": 1.918,
+      "step": 39106
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1483723135271807e-06,
+      "loss": 2.9409,
+      "step": 39107
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1483249051833122e-06,
+      "loss": 1.6337,
+      "step": 39108
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.148277496839444e-06,
+      "loss": 0.2741,
+      "step": 39109
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1482300884955752e-06,
+      "loss": 0.475,
+      "step": 39110
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1481826801517067e-06,
+      "loss": 2.0371,
+      "step": 39111
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1481352718078382e-06,
+      "loss": 3.5211,
+      "step": 39112
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1480878634639697e-06,
+      "loss": 2.4186,
+      "step": 39113
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1480404551201012e-06,
+      "loss": 2.2073,
+      "step": 39114
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1479930467762325e-06,
+      "loss": 2.1468,
+      "step": 39115
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1479456384323642e-06,
+      "loss": 1.8249,
+      "step": 39116
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1478982300884957e-06,
+      "loss": 1.0667,
+      "step": 39117
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.147850821744627e-06,
+      "loss": 4.3684,
+      "step": 39118
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1478034134007586e-06,
+      "loss": 2.1978,
+      "step": 39119
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1477560050568901e-06,
+      "loss": 1.192,
+      "step": 39120
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1477085967130214e-06,
+      "loss": 2.0852,
+      "step": 39121
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.147661188369153e-06,
+      "loss": 2.2408,
+      "step": 39122
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1476137800252846e-06,
+      "loss": 2.3657,
+      "step": 39123
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1475663716814159e-06,
+      "loss": 1.8142,
+      "step": 39124
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1475189633375474e-06,
+      "loss": 1.2155,
+      "step": 39125
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1474715549936789e-06,
+      "loss": 1.9069,
+      "step": 39126
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1474241466498104e-06,
+      "loss": 2.5003,
+      "step": 39127
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1473767383059419e-06,
+      "loss": 2.7607,
+      "step": 39128
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1473293299620734e-06,
+      "loss": 2.5889,
+      "step": 39129
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1472819216182048e-06,
+      "loss": 1.1639,
+      "step": 39130
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1472345132743363e-06,
+      "loss": 1.011,
+      "step": 39131
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1471871049304678e-06,
+      "loss": 1.621,
+      "step": 39132
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1471396965865993e-06,
+      "loss": 0.3244,
+      "step": 39133
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1470922882427308e-06,
+      "loss": 3.7701,
+      "step": 39134
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.147044879898862e-06,
+      "loss": 1.6778,
+      "step": 39135
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1469974715549938e-06,
+      "loss": 1.7355,
+      "step": 39136
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.146950063211125e-06,
+      "loss": 2.2194,
+      "step": 39137
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1469026548672566e-06,
+      "loss": 1.9499,
+      "step": 39138
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1468552465233883e-06,
+      "loss": 1.7346,
+      "step": 39139
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1468078381795196e-06,
+      "loss": 2.6581,
+      "step": 39140
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.146760429835651e-06,
+      "loss": 2.0923,
+      "step": 39141
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1467130214917825e-06,
+      "loss": 0.6076,
+      "step": 39142
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.146665613147914e-06,
+      "loss": 0.44,
+      "step": 39143
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1466182048040455e-06,
+      "loss": 1.4158,
+      "step": 39144
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.146570796460177e-06,
+      "loss": 1.5971,
+      "step": 39145
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1465233881163085e-06,
+      "loss": 2.3196,
+      "step": 39146
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.14647597977244e-06,
+      "loss": 1.5534,
+      "step": 39147
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1464285714285713e-06,
+      "loss": 2.5743,
+      "step": 39148
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.146381163084703e-06,
+      "loss": 1.7637,
+      "step": 39149
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1463337547408345e-06,
+      "loss": 2.627,
+      "step": 39150
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1462863463969658e-06,
+      "loss": 1.8612,
+      "step": 39151
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1462389380530975e-06,
+      "loss": 1.9377,
+      "step": 39152
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.146191529709229e-06,
+      "loss": 2.3326,
+      "step": 39153
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1461441213653602e-06,
+      "loss": 2.8861,
+      "step": 39154
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1460967130214917e-06,
+      "loss": 1.604,
+      "step": 39155
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1460493046776234e-06,
+      "loss": 0.7055,
+      "step": 39156
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1460018963337547e-06,
+      "loss": 2.1266,
+      "step": 39157
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1459544879898862e-06,
+      "loss": 1.2511,
+      "step": 39158
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1459070796460177e-06,
+      "loss": 2.1054,
+      "step": 39159
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1458596713021492e-06,
+      "loss": 3.2712,
+      "step": 39160
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1458122629582807e-06,
+      "loss": 2.1216,
+      "step": 39161
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1457648546144122e-06,
+      "loss": 2.2891,
+      "step": 39162
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1457174462705437e-06,
+      "loss": 1.2376,
+      "step": 39163
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1456700379266752e-06,
+      "loss": 1.4118,
+      "step": 39164
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1456226295828066e-06,
+      "loss": 2.1925,
+      "step": 39165
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1455752212389381e-06,
+      "loss": 1.8863,
+      "step": 39166
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1455278128950696e-06,
+      "loss": 1.3699,
+      "step": 39167
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.145480404551201e-06,
+      "loss": 3.1474,
+      "step": 39168
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 1.1454329962073326e-06,
+      "loss": 3.4964,
+      "step": 39169
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.145385587863464e-06,
+      "loss": 3.705,
+      "step": 39170
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1453381795195954e-06,
+      "loss": 2.08,
+      "step": 39171
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.145290771175727e-06,
+      "loss": 2.4578,
+      "step": 39172
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1452433628318584e-06,
+      "loss": 1.385,
+      "step": 39173
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1451959544879899e-06,
+      "loss": 2.0128,
+      "step": 39174
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1451485461441214e-06,
+      "loss": 1.6436,
+      "step": 39175
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1451011378002529e-06,
+      "loss": 1.6898,
+      "step": 39176
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1450537294563843e-06,
+      "loss": 3.5444,
+      "step": 39177
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1450063211125158e-06,
+      "loss": 1.9473,
+      "step": 39178
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1449589127686473e-06,
+      "loss": 2.6091,
+      "step": 39179
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1449115044247788e-06,
+      "loss": 1.7218,
+      "step": 39180
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.14486409608091e-06,
+      "loss": 2.3089,
+      "step": 39181
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1448166877370418e-06,
+      "loss": 2.1562,
+      "step": 39182
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1447692793931733e-06,
+      "loss": 3.1149,
+      "step": 39183
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1447218710493046e-06,
+      "loss": 2.1071,
+      "step": 39184
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1446744627054363e-06,
+      "loss": 2.5137,
+      "step": 39185
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1446270543615678e-06,
+      "loss": 2.0415,
+      "step": 39186
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.144579646017699e-06,
+      "loss": 2.162,
+      "step": 39187
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1445322376738305e-06,
+      "loss": 3.0352,
+      "step": 39188
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1444848293299622e-06,
+      "loss": 2.0592,
+      "step": 39189
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1444374209860935e-06,
+      "loss": 1.8346,
+      "step": 39190
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.144390012642225e-06,
+      "loss": 0.607,
+      "step": 39191
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1443426042983565e-06,
+      "loss": 3.2738,
+      "step": 39192
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.144295195954488e-06,
+      "loss": 2.8657,
+      "step": 39193
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1442477876106195e-06,
+      "loss": 2.3231,
+      "step": 39194
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.144200379266751e-06,
+      "loss": 3.1145,
+      "step": 39195
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1441529709228825e-06,
+      "loss": 2.461,
+      "step": 39196
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.144105562579014e-06,
+      "loss": 1.4074,
+      "step": 39197
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1440581542351455e-06,
+      "loss": 2.5585,
+      "step": 39198
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.144010745891277e-06,
+      "loss": 1.0564,
+      "step": 39199
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1439633375474085e-06,
+      "loss": 2.3196,
+      "step": 39200
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1439159292035397e-06,
+      "loss": 2.909,
+      "step": 39201
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1438685208596714e-06,
+      "loss": 1.6031,
+      "step": 39202
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1438211125158027e-06,
+      "loss": 1.8721,
+      "step": 39203
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1437737041719342e-06,
+      "loss": 0.7522,
+      "step": 39204
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.143726295828066e-06,
+      "loss": 1.1672,
+      "step": 39205
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1436788874841972e-06,
+      "loss": 0.9373,
+      "step": 39206
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1436314791403287e-06,
+      "loss": 1.941,
+      "step": 39207
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1435840707964602e-06,
+      "loss": 1.3136,
+      "step": 39208
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1435366624525917e-06,
+      "loss": 1.4566,
+      "step": 39209
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1434892541087232e-06,
+      "loss": 3.7062,
+      "step": 39210
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1434418457648547e-06,
+      "loss": 0.8431,
+      "step": 39211
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1433944374209861e-06,
+      "loss": 1.7146,
+      "step": 39212
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1433470290771176e-06,
+      "loss": 2.6301,
+      "step": 39213
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.143299620733249e-06,
+      "loss": 2.4515,
+      "step": 39214
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1432522123893806e-06,
+      "loss": 2.698,
+      "step": 39215
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1432048040455121e-06,
+      "loss": 1.3335,
+      "step": 39216
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1431573957016434e-06,
+      "loss": 2.0475,
+      "step": 39217
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.143109987357775e-06,
+      "loss": 2.395,
+      "step": 39218
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1430625790139066e-06,
+      "loss": 1.6291,
+      "step": 39219
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1430151706700379e-06,
+      "loss": 1.8018,
+      "step": 39220
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1429677623261694e-06,
+      "loss": 1.3181,
+      "step": 39221
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.142920353982301e-06,
+      "loss": 2.1539,
+      "step": 39222
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1428729456384323e-06,
+      "loss": 2.312,
+      "step": 39223
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1428255372945638e-06,
+      "loss": 2.2996,
+      "step": 39224
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1427781289506953e-06,
+      "loss": 3.688,
+      "step": 39225
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1427307206068268e-06,
+      "loss": 2.0109,
+      "step": 39226
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1426833122629583e-06,
+      "loss": 2.295,
+      "step": 39227
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1426359039190898e-06,
+      "loss": 1.4384,
+      "step": 39228
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1425884955752213e-06,
+      "loss": 2.5585,
+      "step": 39229
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1425410872313528e-06,
+      "loss": 2.3957,
+      "step": 39230
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1424936788874843e-06,
+      "loss": 3.0233,
+      "step": 39231
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 1.1424462705436158e-06,
+      "loss": 1.7917,
+      "step": 39232
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1423988621997473e-06,
+      "loss": 2.4447,
+      "step": 39233
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1423514538558786e-06,
+      "loss": 1.2214,
+      "step": 39234
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1423040455120103e-06,
+      "loss": 1.9623,
+      "step": 39235
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1422566371681415e-06,
+      "loss": 3.2524,
+      "step": 39236
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.142209228824273e-06,
+      "loss": 2.0683,
+      "step": 39237
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1421618204804047e-06,
+      "loss": 1.3783,
+      "step": 39238
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.142114412136536e-06,
+      "loss": 1.9644,
+      "step": 39239
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1420670037926675e-06,
+      "loss": 0.5211,
+      "step": 39240
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.142019595448799e-06,
+      "loss": 1.4022,
+      "step": 39241
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1419721871049305e-06,
+      "loss": 4.1432,
+      "step": 39242
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.141924778761062e-06,
+      "loss": 2.8072,
+      "step": 39243
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1418773704171935e-06,
+      "loss": 2.3239,
+      "step": 39244
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.141829962073325e-06,
+      "loss": 2.2566,
+      "step": 39245
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1417825537294565e-06,
+      "loss": 1.4245,
+      "step": 39246
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1417351453855877e-06,
+      "loss": 1.7398,
+      "step": 39247
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1416877370417194e-06,
+      "loss": 2.6514,
+      "step": 39248
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.141640328697851e-06,
+      "loss": 1.3943,
+      "step": 39249
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1415929203539822e-06,
+      "loss": 3.0648,
+      "step": 39250
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.141545512010114e-06,
+      "loss": 1.4384,
+      "step": 39251
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1414981036662454e-06,
+      "loss": 2.4061,
+      "step": 39252
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1414506953223767e-06,
+      "loss": 2.3237,
+      "step": 39253
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1414032869785082e-06,
+      "loss": 2.7764,
+      "step": 39254
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1413558786346399e-06,
+      "loss": 1.6051,
+      "step": 39255
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1413084702907712e-06,
+      "loss": 2.1568,
+      "step": 39256
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1412610619469027e-06,
+      "loss": 2.9996,
+      "step": 39257
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1412136536030342e-06,
+      "loss": 2.6799,
+      "step": 39258
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1411662452591656e-06,
+      "loss": 2.0658,
+      "step": 39259
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1411188369152971e-06,
+      "loss": 2.3933,
+      "step": 39260
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1410714285714286e-06,
+      "loss": 1.2002,
+      "step": 39261
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1410240202275601e-06,
+      "loss": 1.8366,
+      "step": 39262
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1409766118836916e-06,
+      "loss": 3.3039,
+      "step": 39263
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1409292035398229e-06,
+      "loss": 1.6275,
+      "step": 39264
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1408817951959546e-06,
+      "loss": 3.9671,
+      "step": 39265
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.140834386852086e-06,
+      "loss": 2.8757,
+      "step": 39266
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1407869785082174e-06,
+      "loss": 2.5047,
+      "step": 39267
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.140739570164349e-06,
+      "loss": 1.1597,
+      "step": 39268
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1406921618204804e-06,
+      "loss": 2.7261,
+      "step": 39269
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1406447534766118e-06,
+      "loss": 2.2519,
+      "step": 39270
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1405973451327433e-06,
+      "loss": 1.3446,
+      "step": 39271
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1405499367888748e-06,
+      "loss": 2.1466,
+      "step": 39272
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1405025284450063e-06,
+      "loss": 1.7214,
+      "step": 39273
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1404551201011378e-06,
+      "loss": 2.3417,
+      "step": 39274
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1404077117572693e-06,
+      "loss": 2.0652,
+      "step": 39275
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1403603034134008e-06,
+      "loss": 1.9035,
+      "step": 39276
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1403128950695323e-06,
+      "loss": 1.4536,
+      "step": 39277
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1402654867256638e-06,
+      "loss": 1.467,
+      "step": 39278
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1402180783817953e-06,
+      "loss": 2.2462,
+      "step": 39279
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1401706700379266e-06,
+      "loss": 1.5479,
+      "step": 39280
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1401232616940583e-06,
+      "loss": 2.361,
+      "step": 39281
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1400758533501898e-06,
+      "loss": 0.8343,
+      "step": 39282
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.140028445006321e-06,
+      "loss": 2.1849,
+      "step": 39283
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1399810366624525e-06,
+      "loss": 2.1628,
+      "step": 39284
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1399336283185842e-06,
+      "loss": 3.1031,
+      "step": 39285
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1398862199747155e-06,
+      "loss": 1.7849,
+      "step": 39286
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.139838811630847e-06,
+      "loss": 2.3462,
+      "step": 39287
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1397914032869787e-06,
+      "loss": 1.4154,
+      "step": 39288
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.13974399494311e-06,
+      "loss": 2.4076,
+      "step": 39289
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1396965865992415e-06,
+      "loss": 1.5718,
+      "step": 39290
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.139649178255373e-06,
+      "loss": 2.5861,
+      "step": 39291
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1396017699115045e-06,
+      "loss": 2.6705,
+      "step": 39292
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.139554361567636e-06,
+      "loss": 1.5356,
+      "step": 39293
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1395069532237674e-06,
+      "loss": 2.8618,
+      "step": 39294
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.139459544879899e-06,
+      "loss": 2.4001,
+      "step": 39295
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 1.1394121365360304e-06,
+      "loss": 1.1508,
+      "step": 39296
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1393647281921617e-06,
+      "loss": 2.4807,
+      "step": 39297
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1393173198482934e-06,
+      "loss": 2.9214,
+      "step": 39298
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.139269911504425e-06,
+      "loss": 1.1937,
+      "step": 39299
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1392225031605562e-06,
+      "loss": 2.1592,
+      "step": 39300
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1391750948166879e-06,
+      "loss": 2.7682,
+      "step": 39301
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1391276864728192e-06,
+      "loss": 1.6645,
+      "step": 39302
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1390802781289507e-06,
+      "loss": 1.7633,
+      "step": 39303
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1390328697850822e-06,
+      "loss": 2.6176,
+      "step": 39304
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1389854614412136e-06,
+      "loss": 0.9925,
+      "step": 39305
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1389380530973451e-06,
+      "loss": 1.9224,
+      "step": 39306
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1388906447534766e-06,
+      "loss": 3.3833,
+      "step": 39307
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1388432364096081e-06,
+      "loss": 1.7174,
+      "step": 39308
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1387958280657396e-06,
+      "loss": 2.008,
+      "step": 39309
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1387484197218711e-06,
+      "loss": 3.3417,
+      "step": 39310
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1387010113780026e-06,
+      "loss": 1.7672,
+      "step": 39311
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.138653603034134e-06,
+      "loss": 2.4591,
+      "step": 39312
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1386061946902654e-06,
+      "loss": 0.9212,
+      "step": 39313
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.138558786346397e-06,
+      "loss": 2.531,
+      "step": 39314
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1385113780025286e-06,
+      "loss": 2.1629,
+      "step": 39315
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1384639696586599e-06,
+      "loss": 3.1095,
+      "step": 39316
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1384165613147913e-06,
+      "loss": 2.3957,
+      "step": 39317
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.138369152970923e-06,
+      "loss": 3.6414,
+      "step": 39318
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1383217446270543e-06,
+      "loss": 2.3772,
+      "step": 39319
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1382743362831858e-06,
+      "loss": 2.5125,
+      "step": 39320
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1382269279393175e-06,
+      "loss": 2.0408,
+      "step": 39321
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1381795195954488e-06,
+      "loss": 1.7101,
+      "step": 39322
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1381321112515803e-06,
+      "loss": 1.4427,
+      "step": 39323
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1380847029077118e-06,
+      "loss": 0.7878,
+      "step": 39324
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1380372945638433e-06,
+      "loss": 2.7535,
+      "step": 39325
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1379898862199748e-06,
+      "loss": 1.1513,
+      "step": 39326
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1379424778761063e-06,
+      "loss": 2.9898,
+      "step": 39327
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1378950695322378e-06,
+      "loss": 1.8893,
+      "step": 39328
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1378476611883692e-06,
+      "loss": 2.9932,
+      "step": 39329
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1378002528445005e-06,
+      "loss": 3.6463,
+      "step": 39330
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1377528445006322e-06,
+      "loss": 0.5167,
+      "step": 39331
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1377054361567637e-06,
+      "loss": 1.3257,
+      "step": 39332
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.137658027812895e-06,
+      "loss": 2.8683,
+      "step": 39333
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1376106194690267e-06,
+      "loss": 2.32,
+      "step": 39334
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.137563211125158e-06,
+      "loss": 0.5234,
+      "step": 39335
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1375158027812895e-06,
+      "loss": 1.2745,
+      "step": 39336
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.137468394437421e-06,
+      "loss": 2.1658,
+      "step": 39337
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1374209860935525e-06,
+      "loss": 1.7416,
+      "step": 39338
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.137373577749684e-06,
+      "loss": 2.1405,
+      "step": 39339
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1373261694058155e-06,
+      "loss": 2.9464,
+      "step": 39340
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.137278761061947e-06,
+      "loss": 1.6415,
+      "step": 39341
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1372313527180784e-06,
+      "loss": 1.8185,
+      "step": 39342
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.13718394437421e-06,
+      "loss": 1.9189,
+      "step": 39343
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1371365360303414e-06,
+      "loss": 3.0818,
+      "step": 39344
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.137089127686473e-06,
+      "loss": 2.7653,
+      "step": 39345
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1370417193426042e-06,
+      "loss": 0.8804,
+      "step": 39346
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.136994310998736e-06,
+      "loss": 2.3385,
+      "step": 39347
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1369469026548674e-06,
+      "loss": 2.1824,
+      "step": 39348
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1368994943109987e-06,
+      "loss": 2.7295,
+      "step": 39349
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1368520859671302e-06,
+      "loss": 0.4457,
+      "step": 39350
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1368046776232619e-06,
+      "loss": 2.5204,
+      "step": 39351
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1367572692793931e-06,
+      "loss": 1.0806,
+      "step": 39352
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1367098609355246e-06,
+      "loss": 1.2129,
+      "step": 39353
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1366624525916563e-06,
+      "loss": 2.0366,
+      "step": 39354
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1366150442477876e-06,
+      "loss": 2.1715,
+      "step": 39355
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1365676359039191e-06,
+      "loss": 3.0909,
+      "step": 39356
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1365202275600506e-06,
+      "loss": 2.1992,
+      "step": 39357
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.136472819216182e-06,
+      "loss": 1.1495,
+      "step": 39358
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 1.1364254108723136e-06,
+      "loss": 1.1388,
+      "step": 39359
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.136378002528445e-06,
+      "loss": 1.9591,
+      "step": 39360
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1363305941845766e-06,
+      "loss": 2.6666,
+      "step": 39361
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.136283185840708e-06,
+      "loss": 3.6755,
+      "step": 39362
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1362357774968393e-06,
+      "loss": 3.325,
+      "step": 39363
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.136188369152971e-06,
+      "loss": 1.8916,
+      "step": 39364
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1361409608091025e-06,
+      "loss": 0.9791,
+      "step": 39365
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1360935524652338e-06,
+      "loss": 1.4426,
+      "step": 39366
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1360461441213655e-06,
+      "loss": 2.683,
+      "step": 39367
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1359987357774968e-06,
+      "loss": 3.2886,
+      "step": 39368
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1359513274336283e-06,
+      "loss": 1.0787,
+      "step": 39369
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1359039190897598e-06,
+      "loss": 2.0077,
+      "step": 39370
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1358565107458913e-06,
+      "loss": 1.8679,
+      "step": 39371
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1358091024020228e-06,
+      "loss": 2.3334,
+      "step": 39372
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1357616940581543e-06,
+      "loss": 1.7866,
+      "step": 39373
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1357142857142858e-06,
+      "loss": 1.915,
+      "step": 39374
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1356668773704173e-06,
+      "loss": 1.6673,
+      "step": 39375
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1356194690265487e-06,
+      "loss": 2.1282,
+      "step": 39376
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1355720606826802e-06,
+      "loss": 1.4842,
+      "step": 39377
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1355246523388117e-06,
+      "loss": 2.1932,
+      "step": 39378
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.135477243994943e-06,
+      "loss": 0.9699,
+      "step": 39379
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1354298356510747e-06,
+      "loss": 1.3178,
+      "step": 39380
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1353824273072062e-06,
+      "loss": 2.8797,
+      "step": 39381
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1353350189633375e-06,
+      "loss": 1.2796,
+      "step": 39382
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.135287610619469e-06,
+      "loss": 1.9793,
+      "step": 39383
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1352402022756007e-06,
+      "loss": 2.2228,
+      "step": 39384
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.135192793931732e-06,
+      "loss": 2.584,
+      "step": 39385
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1351453855878635e-06,
+      "loss": 1.9364,
+      "step": 39386
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.135097977243995e-06,
+      "loss": 1.3907,
+      "step": 39387
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1350505689001264e-06,
+      "loss": 3.4528,
+      "step": 39388
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.135003160556258e-06,
+      "loss": 1.8741,
+      "step": 39389
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1349557522123894e-06,
+      "loss": 1.9417,
+      "step": 39390
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.134908343868521e-06,
+      "loss": 1.3383,
+      "step": 39391
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1348609355246524e-06,
+      "loss": 3.097,
+      "step": 39392
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1348135271807837e-06,
+      "loss": 2.0418,
+      "step": 39393
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1347661188369154e-06,
+      "loss": 1.8329,
+      "step": 39394
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1347187104930469e-06,
+      "loss": 2.6094,
+      "step": 39395
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1346713021491782e-06,
+      "loss": 2.6221,
+      "step": 39396
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1346238938053099e-06,
+      "loss": 2.3822,
+      "step": 39397
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1345764854614411e-06,
+      "loss": 1.4324,
+      "step": 39398
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1345290771175726e-06,
+      "loss": 0.9273,
+      "step": 39399
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1344816687737041e-06,
+      "loss": 1.6188,
+      "step": 39400
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1344342604298356e-06,
+      "loss": 1.4957,
+      "step": 39401
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1343868520859671e-06,
+      "loss": 1.5699,
+      "step": 39402
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1343394437420986e-06,
+      "loss": 2.0919,
+      "step": 39403
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.13429203539823e-06,
+      "loss": 1.6687,
+      "step": 39404
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1342446270543616e-06,
+      "loss": 2.0138,
+      "step": 39405
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.134197218710493e-06,
+      "loss": 1.0899,
+      "step": 39406
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1341498103666246e-06,
+      "loss": 2.9211,
+      "step": 39407
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.134102402022756e-06,
+      "loss": 1.3162,
+      "step": 39408
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1340549936788874e-06,
+      "loss": 2.3388,
+      "step": 39409
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.134007585335019e-06,
+      "loss": 3.0039,
+      "step": 39410
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1339601769911505e-06,
+      "loss": 2.4275,
+      "step": 39411
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1339127686472818e-06,
+      "loss": 2.2494,
+      "step": 39412
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1338653603034133e-06,
+      "loss": 3.063,
+      "step": 39413
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.133817951959545e-06,
+      "loss": 2.8309,
+      "step": 39414
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1337705436156763e-06,
+      "loss": 0.7248,
+      "step": 39415
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1337231352718078e-06,
+      "loss": 2.0369,
+      "step": 39416
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1336757269279395e-06,
+      "loss": 1.5404,
+      "step": 39417
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1336283185840708e-06,
+      "loss": 2.137,
+      "step": 39418
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1335809102402023e-06,
+      "loss": 2.1422,
+      "step": 39419
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1335335018963338e-06,
+      "loss": 2.1112,
+      "step": 39420
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1334860935524653e-06,
+      "loss": 1.1356,
+      "step": 39421
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 1.1334386852085968e-06,
+      "loss": 1.4284,
+      "step": 39422
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1333912768647282e-06,
+      "loss": 3.5115,
+      "step": 39423
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1333438685208597e-06,
+      "loss": 2.2091,
+      "step": 39424
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1332964601769912e-06,
+      "loss": 2.3851,
+      "step": 39425
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1332490518331225e-06,
+      "loss": 2.0727,
+      "step": 39426
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1332016434892542e-06,
+      "loss": 2.1149,
+      "step": 39427
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1331542351453857e-06,
+      "loss": 1.6156,
+      "step": 39428
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.133106826801517e-06,
+      "loss": 2.1563,
+      "step": 39429
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1330594184576487e-06,
+      "loss": 1.3801,
+      "step": 39430
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.13301201011378e-06,
+      "loss": 1.3639,
+      "step": 39431
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1329646017699115e-06,
+      "loss": 0.4529,
+      "step": 39432
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.132917193426043e-06,
+      "loss": 2.0689,
+      "step": 39433
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1328697850821744e-06,
+      "loss": 2.6894,
+      "step": 39434
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.132822376738306e-06,
+      "loss": 1.7411,
+      "step": 39435
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1327749683944374e-06,
+      "loss": 2.6927,
+      "step": 39436
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.132727560050569e-06,
+      "loss": 3.0837,
+      "step": 39437
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1326801517067004e-06,
+      "loss": 0.9552,
+      "step": 39438
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.132632743362832e-06,
+      "loss": 2.4349,
+      "step": 39439
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1325853350189634e-06,
+      "loss": 1.7371,
+      "step": 39440
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1325379266750949e-06,
+      "loss": 2.5616,
+      "step": 39441
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1324905183312262e-06,
+      "loss": 1.3719,
+      "step": 39442
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1324431099873579e-06,
+      "loss": 1.7291,
+      "step": 39443
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1323957016434894e-06,
+      "loss": 1.8209,
+      "step": 39444
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1323482932996206e-06,
+      "loss": 2.5882,
+      "step": 39445
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1323008849557521e-06,
+      "loss": 2.8373,
+      "step": 39446
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1322534766118838e-06,
+      "loss": 2.3809,
+      "step": 39447
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1322060682680151e-06,
+      "loss": 1.5252,
+      "step": 39448
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1321586599241466e-06,
+      "loss": 3.4997,
+      "step": 39449
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1321112515802783e-06,
+      "loss": 2.5699,
+      "step": 39450
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1320638432364096e-06,
+      "loss": 4.2024,
+      "step": 39451
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.132016434892541e-06,
+      "loss": 2.4647,
+      "step": 39452
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1319690265486726e-06,
+      "loss": 1.4979,
+      "step": 39453
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.131921618204804e-06,
+      "loss": 3.4724,
+      "step": 39454
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1318742098609356e-06,
+      "loss": 1.6768,
+      "step": 39455
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.131826801517067e-06,
+      "loss": 1.2221,
+      "step": 39456
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1317793931731986e-06,
+      "loss": 2.0158,
+      "step": 39457
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.13173198482933e-06,
+      "loss": 1.8083,
+      "step": 39458
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1316845764854613e-06,
+      "loss": 1.4277,
+      "step": 39459
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.131637168141593e-06,
+      "loss": 3.4847,
+      "step": 39460
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1315897597977245e-06,
+      "loss": 2.5188,
+      "step": 39461
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1315423514538558e-06,
+      "loss": 1.6479,
+      "step": 39462
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1314949431099875e-06,
+      "loss": 1.3239,
+      "step": 39463
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1314475347661188e-06,
+      "loss": 3.3759,
+      "step": 39464
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1314001264222503e-06,
+      "loss": 3.1821,
+      "step": 39465
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1313527180783818e-06,
+      "loss": 1.934,
+      "step": 39466
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1313053097345133e-06,
+      "loss": 1.7841,
+      "step": 39467
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1312579013906448e-06,
+      "loss": 2.388,
+      "step": 39468
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1312104930467762e-06,
+      "loss": 2.3031,
+      "step": 39469
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1311630847029077e-06,
+      "loss": 2.6666,
+      "step": 39470
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1311156763590392e-06,
+      "loss": 2.6724,
+      "step": 39471
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1310682680151707e-06,
+      "loss": 0.9109,
+      "step": 39472
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1310208596713022e-06,
+      "loss": 1.6474,
+      "step": 39473
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1309734513274337e-06,
+      "loss": 3.2554,
+      "step": 39474
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.130926042983565e-06,
+      "loss": 1.9057,
+      "step": 39475
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1308786346396967e-06,
+      "loss": 1.3589,
+      "step": 39476
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1308312262958282e-06,
+      "loss": 2.0641,
+      "step": 39477
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1307838179519595e-06,
+      "loss": 1.9887,
+      "step": 39478
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.130736409608091e-06,
+      "loss": 3.3065,
+      "step": 39479
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1306890012642227e-06,
+      "loss": 1.3454,
+      "step": 39480
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.130641592920354e-06,
+      "loss": 1.7323,
+      "step": 39481
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1305941845764854e-06,
+      "loss": 3.2094,
+      "step": 39482
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1305467762326171e-06,
+      "loss": 2.493,
+      "step": 39483
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1304993678887484e-06,
+      "loss": 2.1306,
+      "step": 39484
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.13045195954488e-06,
+      "loss": 1.3513,
+      "step": 39485
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 1.1304045512010114e-06,
+      "loss": 2.0841,
+      "step": 39486
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.130357142857143e-06,
+      "loss": 0.71,
+      "step": 39487
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1303097345132744e-06,
+      "loss": 2.607,
+      "step": 39488
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1302623261694059e-06,
+      "loss": 3.1935,
+      "step": 39489
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1302149178255374e-06,
+      "loss": 2.7055,
+      "step": 39490
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1301675094816689e-06,
+      "loss": 1.558,
+      "step": 39491
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1301201011378001e-06,
+      "loss": 1.3203,
+      "step": 39492
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1300726927939318e-06,
+      "loss": 1.4232,
+      "step": 39493
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1300252844500633e-06,
+      "loss": 1.9066,
+      "step": 39494
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1299778761061946e-06,
+      "loss": 1.8311,
+      "step": 39495
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1299304677623263e-06,
+      "loss": 2.4924,
+      "step": 39496
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1298830594184576e-06,
+      "loss": 1.8159,
+      "step": 39497
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.129835651074589e-06,
+      "loss": 2.9448,
+      "step": 39498
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1297882427307206e-06,
+      "loss": 3.3012,
+      "step": 39499
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.129740834386852e-06,
+      "loss": 1.5278,
+      "step": 39500
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1296934260429836e-06,
+      "loss": 2.3889,
+      "step": 39501
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.129646017699115e-06,
+      "loss": 1.2488,
+      "step": 39502
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1295986093552466e-06,
+      "loss": 1.3455,
+      "step": 39503
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.129551201011378e-06,
+      "loss": 2.2822,
+      "step": 39504
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1295037926675095e-06,
+      "loss": 2.4094,
+      "step": 39505
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.129456384323641e-06,
+      "loss": 1.0244,
+      "step": 39506
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1294089759797725e-06,
+      "loss": 1.352,
+      "step": 39507
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1293615676359038e-06,
+      "loss": 3.6431,
+      "step": 39508
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1293141592920355e-06,
+      "loss": 2.3354,
+      "step": 39509
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.129266750948167e-06,
+      "loss": 2.4656,
+      "step": 39510
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1292193426042983e-06,
+      "loss": 2.3941,
+      "step": 39511
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1291719342604298e-06,
+      "loss": 1.2786,
+      "step": 39512
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1291245259165615e-06,
+      "loss": 2.2459,
+      "step": 39513
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1290771175726928e-06,
+      "loss": 3.2899,
+      "step": 39514
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1290297092288243e-06,
+      "loss": 2.5269,
+      "step": 39515
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.128982300884956e-06,
+      "loss": 1.4105,
+      "step": 39516
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1289348925410872e-06,
+      "loss": 1.5505,
+      "step": 39517
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1288874841972187e-06,
+      "loss": 2.4154,
+      "step": 39518
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1288400758533502e-06,
+      "loss": 2.9589,
+      "step": 39519
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1287926675094817e-06,
+      "loss": 1.733,
+      "step": 39520
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1287452591656132e-06,
+      "loss": 1.2104,
+      "step": 39521
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1286978508217447e-06,
+      "loss": 2.5958,
+      "step": 39522
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1286504424778762e-06,
+      "loss": 1.9964,
+      "step": 39523
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1286030341340077e-06,
+      "loss": 1.6091,
+      "step": 39524
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.128555625790139e-06,
+      "loss": 2.7919,
+      "step": 39525
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1285082174462707e-06,
+      "loss": 2.5539,
+      "step": 39526
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1284608091024022e-06,
+      "loss": 1.2427,
+      "step": 39527
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1284134007585334e-06,
+      "loss": 1.7612,
+      "step": 39528
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1283659924146651e-06,
+      "loss": 1.9332,
+      "step": 39529
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1283185840707964e-06,
+      "loss": 2.1988,
+      "step": 39530
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.128271175726928e-06,
+      "loss": 1.997,
+      "step": 39531
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1282237673830594e-06,
+      "loss": 1.5126,
+      "step": 39532
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.128176359039191e-06,
+      "loss": 2.1797,
+      "step": 39533
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1281289506953224e-06,
+      "loss": 2.4084,
+      "step": 39534
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1280815423514539e-06,
+      "loss": 2.8898,
+      "step": 39535
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1280341340075854e-06,
+      "loss": 2.39,
+      "step": 39536
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1279867256637169e-06,
+      "loss": 1.7723,
+      "step": 39537
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1279393173198484e-06,
+      "loss": 2.9633,
+      "step": 39538
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1278919089759799e-06,
+      "loss": 1.6924,
+      "step": 39539
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1278445006321113e-06,
+      "loss": 3.2975,
+      "step": 39540
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1277970922882426e-06,
+      "loss": 2.6435,
+      "step": 39541
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1277496839443741e-06,
+      "loss": 2.7175,
+      "step": 39542
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1277022756005058e-06,
+      "loss": 2.418,
+      "step": 39543
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.127654867256637e-06,
+      "loss": 2.6236,
+      "step": 39544
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1276074589127686e-06,
+      "loss": 1.77,
+      "step": 39545
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1275600505689003e-06,
+      "loss": 1.7458,
+      "step": 39546
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1275126422250316e-06,
+      "loss": 2.4467,
+      "step": 39547
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.127465233881163e-06,
+      "loss": 3.2475,
+      "step": 39548
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 1.1274178255372946e-06,
+      "loss": 2.0015,
+      "step": 39549
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.127370417193426e-06,
+      "loss": 2.5237,
+      "step": 39550
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1273230088495575e-06,
+      "loss": 0.7583,
+      "step": 39551
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.127275600505689e-06,
+      "loss": 1.9638,
+      "step": 39552
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1272281921618205e-06,
+      "loss": 2.6356,
+      "step": 39553
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.127180783817952e-06,
+      "loss": 1.2507,
+      "step": 39554
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1271333754740833e-06,
+      "loss": 1.3319,
+      "step": 39555
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.127085967130215e-06,
+      "loss": 1.3097,
+      "step": 39556
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1270385587863465e-06,
+      "loss": 1.3533,
+      "step": 39557
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1269911504424778e-06,
+      "loss": 2.1127,
+      "step": 39558
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1269437420986095e-06,
+      "loss": 2.6143,
+      "step": 39559
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.126896333754741e-06,
+      "loss": 2.249,
+      "step": 39560
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1268489254108723e-06,
+      "loss": 1.7338,
+      "step": 39561
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1268015170670037e-06,
+      "loss": 1.964,
+      "step": 39562
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1267541087231352e-06,
+      "loss": 2.1598,
+      "step": 39563
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1267067003792667e-06,
+      "loss": 1.3826,
+      "step": 39564
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1266592920353982e-06,
+      "loss": 1.9814,
+      "step": 39565
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1266118836915297e-06,
+      "loss": 2.5716,
+      "step": 39566
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1265644753476612e-06,
+      "loss": 2.3293,
+      "step": 39567
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1265170670037927e-06,
+      "loss": 1.2155,
+      "step": 39568
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1264696586599242e-06,
+      "loss": 0.6039,
+      "step": 39569
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1264222503160557e-06,
+      "loss": 2.8197,
+      "step": 39570
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1263748419721872e-06,
+      "loss": 3.4682,
+      "step": 39571
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1263274336283187e-06,
+      "loss": 1.8799,
+      "step": 39572
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1262800252844502e-06,
+      "loss": 1.9258,
+      "step": 39573
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1262326169405814e-06,
+      "loss": 1.2682,
+      "step": 39574
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.126185208596713e-06,
+      "loss": 0.7841,
+      "step": 39575
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1261378002528446e-06,
+      "loss": 2.1661,
+      "step": 39576
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.126090391908976e-06,
+      "loss": 0.9835,
+      "step": 39577
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1260429835651074e-06,
+      "loss": 1.4094,
+      "step": 39578
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1259955752212391e-06,
+      "loss": 2.4426,
+      "step": 39579
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1259481668773704e-06,
+      "loss": 1.9921,
+      "step": 39580
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1259007585335019e-06,
+      "loss": 1.3702,
+      "step": 39581
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1258533501896334e-06,
+      "loss": 1.5641,
+      "step": 39582
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1258059418457649e-06,
+      "loss": 2.1808,
+      "step": 39583
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1257585335018964e-06,
+      "loss": 2.8242,
+      "step": 39584
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1257111251580279e-06,
+      "loss": 2.0494,
+      "step": 39585
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1256637168141593e-06,
+      "loss": 3.1409,
+      "step": 39586
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1256163084702908e-06,
+      "loss": 4.0247,
+      "step": 39587
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1255689001264221e-06,
+      "loss": 0.483,
+      "step": 39588
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1255214917825538e-06,
+      "loss": 3.1931,
+      "step": 39589
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1254740834386853e-06,
+      "loss": 0.8633,
+      "step": 39590
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1254266750948166e-06,
+      "loss": 4.1322,
+      "step": 39591
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1253792667509483e-06,
+      "loss": 1.2523,
+      "step": 39592
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1253318584070798e-06,
+      "loss": 2.5398,
+      "step": 39593
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.125284450063211e-06,
+      "loss": 1.2392,
+      "step": 39594
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1252370417193426e-06,
+      "loss": 2.474,
+      "step": 39595
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.125189633375474e-06,
+      "loss": 2.7078,
+      "step": 39596
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1251422250316056e-06,
+      "loss": 1.2505,
+      "step": 39597
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.125094816687737e-06,
+      "loss": 2.3349,
+      "step": 39598
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1250474083438685e-06,
+      "loss": 2.2647,
+      "step": 39599
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.125e-06,
+      "loss": 3.3067,
+      "step": 39600
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1249525916561315e-06,
+      "loss": 2.5808,
+      "step": 39601
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.124905183312263e-06,
+      "loss": 1.6571,
+      "step": 39602
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1248577749683945e-06,
+      "loss": 1.9515,
+      "step": 39603
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.124810366624526e-06,
+      "loss": 2.8521,
+      "step": 39604
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1247629582806575e-06,
+      "loss": 3.1314,
+      "step": 39605
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.124715549936789e-06,
+      "loss": 1.8715,
+      "step": 39606
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1246681415929203e-06,
+      "loss": 2.1057,
+      "step": 39607
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1246207332490518e-06,
+      "loss": 2.4535,
+      "step": 39608
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1245733249051835e-06,
+      "loss": 1.8813,
+      "step": 39609
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1245259165613147e-06,
+      "loss": 2.1338,
+      "step": 39610
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.1244785082174462e-06,
+      "loss": 1.4322,
+      "step": 39611
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 1.124431099873578e-06,
+      "loss": 2.3079,
+      "step": 39612
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1243836915297092e-06,
+      "loss": 1.2969,
+      "step": 39613
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1243362831858407e-06,
+      "loss": 2.0694,
+      "step": 39614
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1242888748419722e-06,
+      "loss": 1.3263,
+      "step": 39615
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1242414664981037e-06,
+      "loss": 2.4627,
+      "step": 39616
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1241940581542352e-06,
+      "loss": 3.0093,
+      "step": 39617
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1241466498103667e-06,
+      "loss": 2.8031,
+      "step": 39618
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1240992414664982e-06,
+      "loss": 1.9562,
+      "step": 39619
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1240518331226297e-06,
+      "loss": 1.8553,
+      "step": 39620
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.124004424778761e-06,
+      "loss": 2.4739,
+      "step": 39621
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1239570164348926e-06,
+      "loss": 0.3061,
+      "step": 39622
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1239096080910241e-06,
+      "loss": 2.0062,
+      "step": 39623
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1238621997471554e-06,
+      "loss": 1.2033,
+      "step": 39624
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1238147914032871e-06,
+      "loss": 2.4639,
+      "step": 39625
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1237673830594186e-06,
+      "loss": 2.2665,
+      "step": 39626
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1237199747155499e-06,
+      "loss": 2.2858,
+      "step": 39627
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1236725663716814e-06,
+      "loss": 2.3589,
+      "step": 39628
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1236251580278129e-06,
+      "loss": 1.9443,
+      "step": 39629
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1235777496839444e-06,
+      "loss": 1.3786,
+      "step": 39630
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1235303413400759e-06,
+      "loss": 1.9032,
+      "step": 39631
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1234829329962074e-06,
+      "loss": 3.0473,
+      "step": 39632
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1234355246523388e-06,
+      "loss": 0.8664,
+      "step": 39633
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1233881163084703e-06,
+      "loss": 2.4297,
+      "step": 39634
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1233407079646018e-06,
+      "loss": 1.1853,
+      "step": 39635
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1232932996207333e-06,
+      "loss": 3.5615,
+      "step": 39636
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1232458912768648e-06,
+      "loss": 1.4797,
+      "step": 39637
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1231984829329963e-06,
+      "loss": 1.8489,
+      "step": 39638
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1231510745891278e-06,
+      "loss": 2.5846,
+      "step": 39639
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.123103666245259e-06,
+      "loss": 2.1314,
+      "step": 39640
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1230562579013906e-06,
+      "loss": 2.5161,
+      "step": 39641
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1230088495575223e-06,
+      "loss": 2.4606,
+      "step": 39642
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1229614412136536e-06,
+      "loss": 2.0815,
+      "step": 39643
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122914032869785e-06,
+      "loss": 3.008,
+      "step": 39644
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1228666245259168e-06,
+      "loss": 2.2045,
+      "step": 39645
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122819216182048e-06,
+      "loss": 1.9866,
+      "step": 39646
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1227718078381795e-06,
+      "loss": 0.8375,
+      "step": 39647
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122724399494311e-06,
+      "loss": 2.4089,
+      "step": 39648
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1226769911504425e-06,
+      "loss": 2.1746,
+      "step": 39649
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122629582806574e-06,
+      "loss": 1.7668,
+      "step": 39650
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1225821744627055e-06,
+      "loss": 2.4466,
+      "step": 39651
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122534766118837e-06,
+      "loss": 2.6221,
+      "step": 39652
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1224873577749685e-06,
+      "loss": 2.2148,
+      "step": 39653
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1224399494310998e-06,
+      "loss": 2.6033,
+      "step": 39654
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1223925410872315e-06,
+      "loss": 1.9403,
+      "step": 39655
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122345132743363e-06,
+      "loss": 2.1868,
+      "step": 39656
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1222977243994942e-06,
+      "loss": 3.1542,
+      "step": 39657
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.122250316055626e-06,
+      "loss": 0.6694,
+      "step": 39658
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1222029077117574e-06,
+      "loss": 2.8224,
+      "step": 39659
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1221554993678887e-06,
+      "loss": 2.1444,
+      "step": 39660
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1221080910240202e-06,
+      "loss": 2.2306,
+      "step": 39661
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1220606826801517e-06,
+      "loss": 2.1818,
+      "step": 39662
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1220132743362832e-06,
+      "loss": 2.124,
+      "step": 39663
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1219658659924147e-06,
+      "loss": 1.8073,
+      "step": 39664
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1219184576485462e-06,
+      "loss": 2.1076,
+      "step": 39665
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1218710493046777e-06,
+      "loss": 2.5625,
+      "step": 39666
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1218236409608092e-06,
+      "loss": 0.4446,
+      "step": 39667
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1217762326169406e-06,
+      "loss": 2.5821,
+      "step": 39668
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1217288242730721e-06,
+      "loss": 2.6079,
+      "step": 39669
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1216814159292036e-06,
+      "loss": 1.1115,
+      "step": 39670
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.121634007585335e-06,
+      "loss": 0.8059,
+      "step": 39671
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1215865992414666e-06,
+      "loss": 2.0241,
+      "step": 39672
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.121539190897598e-06,
+      "loss": 2.622,
+      "step": 39673
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1214917825537294e-06,
+      "loss": 2.5475,
+      "step": 39674
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.121444374209861e-06,
+      "loss": 4.0613,
+      "step": 39675
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 1.1213969658659924e-06,
+      "loss": 1.2245,
+      "step": 39676
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1213495575221239e-06,
+      "loss": 1.3471,
+      "step": 39677
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1213021491782554e-06,
+      "loss": 1.7096,
+      "step": 39678
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1212547408343869e-06,
+      "loss": 3.1057,
+      "step": 39679
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1212073324905183e-06,
+      "loss": 1.4448,
+      "step": 39680
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1211599241466498e-06,
+      "loss": 1.6151,
+      "step": 39681
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1211125158027813e-06,
+      "loss": 1.3233,
+      "step": 39682
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1210651074589128e-06,
+      "loss": 1.8282,
+      "step": 39683
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.121017699115044e-06,
+      "loss": 0.9416,
+      "step": 39684
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1209702907711758e-06,
+      "loss": 1.8803,
+      "step": 39685
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1209228824273073e-06,
+      "loss": 2.8395,
+      "step": 39686
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1208754740834386e-06,
+      "loss": 1.6569,
+      "step": 39687
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1208280657395703e-06,
+      "loss": 2.0025,
+      "step": 39688
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1207806573957018e-06,
+      "loss": 2.3652,
+      "step": 39689
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.120733249051833e-06,
+      "loss": 1.073,
+      "step": 39690
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1206858407079645e-06,
+      "loss": 0.3762,
+      "step": 39691
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1206384323640962e-06,
+      "loss": 2.3742,
+      "step": 39692
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1205910240202275e-06,
+      "loss": 1.5115,
+      "step": 39693
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.120543615676359e-06,
+      "loss": 3.024,
+      "step": 39694
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1204962073324905e-06,
+      "loss": 2.1971,
+      "step": 39695
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.120448798988622e-06,
+      "loss": 2.1555,
+      "step": 39696
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1204013906447535e-06,
+      "loss": 2.8883,
+      "step": 39697
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.120353982300885e-06,
+      "loss": 1.9333,
+      "step": 39698
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1203065739570165e-06,
+      "loss": 2.3012,
+      "step": 39699
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.120259165613148e-06,
+      "loss": 1.6889,
+      "step": 39700
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1202117572692795e-06,
+      "loss": 3.2981,
+      "step": 39701
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.120164348925411e-06,
+      "loss": 2.1378,
+      "step": 39702
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1201169405815425e-06,
+      "loss": 2.4991,
+      "step": 39703
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1200695322376737e-06,
+      "loss": 2.5916,
+      "step": 39704
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1200221238938054e-06,
+      "loss": 3.1521,
+      "step": 39705
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1199747155499367e-06,
+      "loss": 3.4547,
+      "step": 39706
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1199273072060682e-06,
+      "loss": 2.1712,
+      "step": 39707
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1198798988622e-06,
+      "loss": 1.2843,
+      "step": 39708
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1198324905183312e-06,
+      "loss": 1.2105,
+      "step": 39709
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1197850821744627e-06,
+      "loss": 1.4391,
+      "step": 39710
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1197376738305942e-06,
+      "loss": 2.3904,
+      "step": 39711
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1196902654867257e-06,
+      "loss": 0.9771,
+      "step": 39712
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1196428571428572e-06,
+      "loss": 1.8876,
+      "step": 39713
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1195954487989887e-06,
+      "loss": 2.6619,
+      "step": 39714
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1195480404551201e-06,
+      "loss": 3.4527,
+      "step": 39715
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1195006321112516e-06,
+      "loss": 1.5628,
+      "step": 39716
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.119453223767383e-06,
+      "loss": 1.5288,
+      "step": 39717
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1194058154235146e-06,
+      "loss": 1.5657,
+      "step": 39718
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1193584070796461e-06,
+      "loss": 1.8292,
+      "step": 39719
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1193109987357774e-06,
+      "loss": 2.8296,
+      "step": 39720
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.119263590391909e-06,
+      "loss": 2.1485,
+      "step": 39721
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1192161820480406e-06,
+      "loss": 2.7366,
+      "step": 39722
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1191687737041719e-06,
+      "loss": 1.3924,
+      "step": 39723
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1191213653603034e-06,
+      "loss": 1.5639,
+      "step": 39724
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.119073957016435e-06,
+      "loss": 1.552,
+      "step": 39725
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1190265486725663e-06,
+      "loss": 3.9726,
+      "step": 39726
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1189791403286978e-06,
+      "loss": 2.1969,
+      "step": 39727
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1189317319848293e-06,
+      "loss": 3.8162,
+      "step": 39728
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1188843236409608e-06,
+      "loss": 1.6734,
+      "step": 39729
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1188369152970923e-06,
+      "loss": 2.7046,
+      "step": 39730
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1187895069532238e-06,
+      "loss": 2.1203,
+      "step": 39731
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1187420986093553e-06,
+      "loss": 2.0657,
+      "step": 39732
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1186946902654868e-06,
+      "loss": 0.8242,
+      "step": 39733
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1186472819216183e-06,
+      "loss": 1.4577,
+      "step": 39734
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1185998735777498e-06,
+      "loss": 1.7256,
+      "step": 39735
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1185524652338813e-06,
+      "loss": 2.391,
+      "step": 39736
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1185050568900126e-06,
+      "loss": 2.2306,
+      "step": 39737
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1184576485461443e-06,
+      "loss": 3.5531,
+      "step": 39738
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 1.1184102402022755e-06,
+      "loss": 2.2836,
+      "step": 39739
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.118362831858407e-06,
+      "loss": 2.5105,
+      "step": 39740
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1183154235145387e-06,
+      "loss": 2.2068,
+      "step": 39741
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.11826801517067e-06,
+      "loss": 1.0889,
+      "step": 39742
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1182206068268015e-06,
+      "loss": 2.726,
+      "step": 39743
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.118173198482933e-06,
+      "loss": 2.749,
+      "step": 39744
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1181257901390645e-06,
+      "loss": 2.0497,
+      "step": 39745
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.118078381795196e-06,
+      "loss": 2.3666,
+      "step": 39746
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1180309734513275e-06,
+      "loss": 1.6758,
+      "step": 39747
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.117983565107459e-06,
+      "loss": 2.0852,
+      "step": 39748
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1179361567635905e-06,
+      "loss": 2.8246,
+      "step": 39749
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1178887484197217e-06,
+      "loss": 3.3453,
+      "step": 39750
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1178413400758534e-06,
+      "loss": 4.007,
+      "step": 39751
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.117793931731985e-06,
+      "loss": 2.11,
+      "step": 39752
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1177465233881162e-06,
+      "loss": 2.0521,
+      "step": 39753
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.117699115044248e-06,
+      "loss": 1.9144,
+      "step": 39754
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1176517067003794e-06,
+      "loss": 1.489,
+      "step": 39755
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1176042983565107e-06,
+      "loss": 2.0917,
+      "step": 39756
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1175568900126422e-06,
+      "loss": 2.1456,
+      "step": 39757
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1175094816687739e-06,
+      "loss": 2.9742,
+      "step": 39758
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1174620733249052e-06,
+      "loss": 2.4036,
+      "step": 39759
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1174146649810367e-06,
+      "loss": 2.1469,
+      "step": 39760
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1173672566371682e-06,
+      "loss": 2.2126,
+      "step": 39761
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1173198482932996e-06,
+      "loss": 1.1922,
+      "step": 39762
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1172724399494311e-06,
+      "loss": 1.8587,
+      "step": 39763
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1172250316055626e-06,
+      "loss": 0.8528,
+      "step": 39764
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1171776232616941e-06,
+      "loss": 2.1819,
+      "step": 39765
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1171302149178256e-06,
+      "loss": 3.9798,
+      "step": 39766
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.117082806573957e-06,
+      "loss": 1.015,
+      "step": 39767
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1170353982300886e-06,
+      "loss": 3.0049,
+      "step": 39768
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.11698798988622e-06,
+      "loss": 2.6435,
+      "step": 39769
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1169405815423514e-06,
+      "loss": 1.2464,
+      "step": 39770
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.116893173198483e-06,
+      "loss": 2.2993,
+      "step": 39771
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1168457648546144e-06,
+      "loss": 3.5205,
+      "step": 39772
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1167983565107458e-06,
+      "loss": 2.0186,
+      "step": 39773
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1167509481668775e-06,
+      "loss": 2.3096,
+      "step": 39774
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1167035398230088e-06,
+      "loss": 1.6731,
+      "step": 39775
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1166561314791403e-06,
+      "loss": 2.3901,
+      "step": 39776
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1166087231352718e-06,
+      "loss": 2.0175,
+      "step": 39777
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1165613147914033e-06,
+      "loss": 3.2227,
+      "step": 39778
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1165139064475348e-06,
+      "loss": 2.6271,
+      "step": 39779
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1164664981036663e-06,
+      "loss": 2.5265,
+      "step": 39780
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1164190897597978e-06,
+      "loss": 0.94,
+      "step": 39781
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1163716814159293e-06,
+      "loss": 0.8672,
+      "step": 39782
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1163242730720606e-06,
+      "loss": 3.0551,
+      "step": 39783
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1162768647281923e-06,
+      "loss": 1.5594,
+      "step": 39784
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1162294563843238e-06,
+      "loss": 2.0243,
+      "step": 39785
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.116182048040455e-06,
+      "loss": 1.8968,
+      "step": 39786
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1161346396965867e-06,
+      "loss": 1.7248,
+      "step": 39787
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1160872313527182e-06,
+      "loss": 1.4067,
+      "step": 39788
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1160398230088495e-06,
+      "loss": 1.5679,
+      "step": 39789
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.115992414664981e-06,
+      "loss": 2.9734,
+      "step": 39790
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1159450063211127e-06,
+      "loss": 1.8167,
+      "step": 39791
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.115897597977244e-06,
+      "loss": 0.9503,
+      "step": 39792
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1158501896333755e-06,
+      "loss": 2.0549,
+      "step": 39793
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.115802781289507e-06,
+      "loss": 2.4792,
+      "step": 39794
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1157553729456385e-06,
+      "loss": 2.0544,
+      "step": 39795
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.11570796460177e-06,
+      "loss": 3.0337,
+      "step": 39796
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1156605562579014e-06,
+      "loss": 1.9709,
+      "step": 39797
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.115613147914033e-06,
+      "loss": 1.0567,
+      "step": 39798
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1155657395701644e-06,
+      "loss": 1.1015,
+      "step": 39799
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.115518331226296e-06,
+      "loss": 3.0882,
+      "step": 39800
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.1154709228824274e-06,
+      "loss": 2.716,
+      "step": 39801
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 1.115423514538559e-06,
+      "loss": 2.9067,
+      "step": 39802
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1153761061946902e-06,
+      "loss": 1.8583,
+      "step": 39803
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1153286978508219e-06,
+      "loss": 1.4976,
+      "step": 39804
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1152812895069532e-06,
+      "loss": 2.5151,
+      "step": 39805
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1152338811630847e-06,
+      "loss": 1.2205,
+      "step": 39806
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1151864728192164e-06,
+      "loss": 3.8492,
+      "step": 39807
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1151390644753476e-06,
+      "loss": 1.5202,
+      "step": 39808
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1150916561314791e-06,
+      "loss": 2.3349,
+      "step": 39809
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1150442477876106e-06,
+      "loss": 1.7948,
+      "step": 39810
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1149968394437421e-06,
+      "loss": 1.0874,
+      "step": 39811
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1149494310998736e-06,
+      "loss": 1.8153,
+      "step": 39812
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1149020227560051e-06,
+      "loss": 2.5094,
+      "step": 39813
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1148546144121366e-06,
+      "loss": 2.8155,
+      "step": 39814
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.114807206068268e-06,
+      "loss": 3.2257,
+      "step": 39815
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1147597977243994e-06,
+      "loss": 0.9035,
+      "step": 39816
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.114712389380531e-06,
+      "loss": 2.4857,
+      "step": 39817
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1146649810366626e-06,
+      "loss": 1.6452,
+      "step": 39818
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1146175726927939e-06,
+      "loss": 1.9521,
+      "step": 39819
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1145701643489253e-06,
+      "loss": 2.9777,
+      "step": 39820
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.114522756005057e-06,
+      "loss": 1.4263,
+      "step": 39821
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1144753476611883e-06,
+      "loss": 1.9225,
+      "step": 39822
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1144279393173198e-06,
+      "loss": 1.8338,
+      "step": 39823
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1143805309734513e-06,
+      "loss": 1.6643,
+      "step": 39824
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1143331226295828e-06,
+      "loss": 1.0779,
+      "step": 39825
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1142857142857143e-06,
+      "loss": 3.2904,
+      "step": 39826
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1142383059418458e-06,
+      "loss": 1.3307,
+      "step": 39827
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1141908975979773e-06,
+      "loss": 4.1773,
+      "step": 39828
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1141434892541088e-06,
+      "loss": 0.7826,
+      "step": 39829
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1140960809102403e-06,
+      "loss": 1.1375,
+      "step": 39830
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1140486725663718e-06,
+      "loss": 1.4788,
+      "step": 39831
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1140012642225032e-06,
+      "loss": 1.0789,
+      "step": 39832
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1139538558786345e-06,
+      "loss": 2.8377,
+      "step": 39833
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1139064475347662e-06,
+      "loss": 1.09,
+      "step": 39834
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1138590391908975e-06,
+      "loss": 2.5065,
+      "step": 39835
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.113811630847029e-06,
+      "loss": 0.7349,
+      "step": 39836
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1137642225031607e-06,
+      "loss": 2.836,
+      "step": 39837
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.113716814159292e-06,
+      "loss": 2.7815,
+      "step": 39838
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1136694058154235e-06,
+      "loss": 2.2886,
+      "step": 39839
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.113621997471555e-06,
+      "loss": 2.4619,
+      "step": 39840
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1135745891276865e-06,
+      "loss": 2.2072,
+      "step": 39841
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.113527180783818e-06,
+      "loss": 3.4119,
+      "step": 39842
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1134797724399495e-06,
+      "loss": 1.9823,
+      "step": 39843
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.113432364096081e-06,
+      "loss": 2.5219,
+      "step": 39844
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1133849557522124e-06,
+      "loss": 2.8592,
+      "step": 39845
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1133375474083437e-06,
+      "loss": 1.7351,
+      "step": 39846
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1132901390644754e-06,
+      "loss": 1.249,
+      "step": 39847
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.113242730720607e-06,
+      "loss": 1.6391,
+      "step": 39848
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1131953223767382e-06,
+      "loss": 1.6729,
+      "step": 39849
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.11314791403287e-06,
+      "loss": 1.9341,
+      "step": 39850
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1131005056890014e-06,
+      "loss": 2.1641,
+      "step": 39851
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1130530973451327e-06,
+      "loss": 1.275,
+      "step": 39852
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1130056890012642e-06,
+      "loss": 2.5609,
+      "step": 39853
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1129582806573959e-06,
+      "loss": 1.2937,
+      "step": 39854
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1129108723135271e-06,
+      "loss": 2.5983,
+      "step": 39855
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1128634639696586e-06,
+      "loss": 3.4962,
+      "step": 39856
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1128160556257901e-06,
+      "loss": 1.5665,
+      "step": 39857
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1127686472819216e-06,
+      "loss": 2.3453,
+      "step": 39858
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1127212389380531e-06,
+      "loss": 1.6858,
+      "step": 39859
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1126738305941846e-06,
+      "loss": 1.4896,
+      "step": 39860
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.112626422250316e-06,
+      "loss": 2.6509,
+      "step": 39861
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1125790139064476e-06,
+      "loss": 2.3073,
+      "step": 39862
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.112531605562579e-06,
+      "loss": 2.3327,
+      "step": 39863
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1124841972187106e-06,
+      "loss": 2.3236,
+      "step": 39864
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.112436788874842e-06,
+      "loss": 2.0847,
+      "step": 39865
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 1.1123893805309733e-06,
+      "loss": 2.5189,
+      "step": 39866
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.112341972187105e-06,
+      "loss": 1.1154,
+      "step": 39867
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1122945638432363e-06,
+      "loss": 0.7209,
+      "step": 39868
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1122471554993678e-06,
+      "loss": 1.655,
+      "step": 39869
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1121997471554995e-06,
+      "loss": 3.0283,
+      "step": 39870
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1121523388116308e-06,
+      "loss": 2.512,
+      "step": 39871
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1121049304677623e-06,
+      "loss": 1.8693,
+      "step": 39872
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1120575221238938e-06,
+      "loss": 1.7874,
+      "step": 39873
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1120101137800253e-06,
+      "loss": 2.6598,
+      "step": 39874
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1119627054361568e-06,
+      "loss": 3.1922,
+      "step": 39875
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1119152970922883e-06,
+      "loss": 1.7759,
+      "step": 39876
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1118678887484198e-06,
+      "loss": 1.6951,
+      "step": 39877
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1118204804045513e-06,
+      "loss": 2.2078,
+      "step": 39878
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1117730720606825e-06,
+      "loss": 3.1052,
+      "step": 39879
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1117256637168142e-06,
+      "loss": 2.5679,
+      "step": 39880
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1116782553729457e-06,
+      "loss": 3.3458,
+      "step": 39881
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.111630847029077e-06,
+      "loss": 0.9867,
+      "step": 39882
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1115834386852087e-06,
+      "loss": 0.9314,
+      "step": 39883
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1115360303413402e-06,
+      "loss": 0.8438,
+      "step": 39884
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1114886219974715e-06,
+      "loss": 3.0706,
+      "step": 39885
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.111441213653603e-06,
+      "loss": 2.2431,
+      "step": 39886
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1113938053097347e-06,
+      "loss": 2.8734,
+      "step": 39887
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.111346396965866e-06,
+      "loss": 2.0878,
+      "step": 39888
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1112989886219975e-06,
+      "loss": 2.5216,
+      "step": 39889
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.111251580278129e-06,
+      "loss": 0.7393,
+      "step": 39890
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1112041719342604e-06,
+      "loss": 1.3117,
+      "step": 39891
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.111156763590392e-06,
+      "loss": 2.1399,
+      "step": 39892
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1111093552465234e-06,
+      "loss": 2.1637,
+      "step": 39893
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.111061946902655e-06,
+      "loss": 1.789,
+      "step": 39894
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1110145385587864e-06,
+      "loss": 0.8864,
+      "step": 39895
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.110967130214918e-06,
+      "loss": 1.8281,
+      "step": 39896
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1109197218710494e-06,
+      "loss": 0.7858,
+      "step": 39897
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1108723135271809e-06,
+      "loss": 1.1283,
+      "step": 39898
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1108249051833122e-06,
+      "loss": 2.7703,
+      "step": 39899
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1107774968394439e-06,
+      "loss": 2.4132,
+      "step": 39900
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1107300884955752e-06,
+      "loss": 2.6669,
+      "step": 39901
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1106826801517066e-06,
+      "loss": 1.9694,
+      "step": 39902
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1106352718078383e-06,
+      "loss": 2.7823,
+      "step": 39903
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1105878634639696e-06,
+      "loss": 0.9742,
+      "step": 39904
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1105404551201011e-06,
+      "loss": 3.3011,
+      "step": 39905
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1104930467762326e-06,
+      "loss": 1.2281,
+      "step": 39906
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.110445638432364e-06,
+      "loss": 2.7657,
+      "step": 39907
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1103982300884956e-06,
+      "loss": 2.7301,
+      "step": 39908
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.110350821744627e-06,
+      "loss": 0.9284,
+      "step": 39909
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1103034134007586e-06,
+      "loss": 2.8724,
+      "step": 39910
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.11025600505689e-06,
+      "loss": 0.6172,
+      "step": 39911
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1102085967130214e-06,
+      "loss": 3.9279,
+      "step": 39912
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.110161188369153e-06,
+      "loss": 0.7526,
+      "step": 39913
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1101137800252845e-06,
+      "loss": 1.7865,
+      "step": 39914
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1100663716814158e-06,
+      "loss": 1.9777,
+      "step": 39915
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1100189633375475e-06,
+      "loss": 1.3071,
+      "step": 39916
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.109971554993679e-06,
+      "loss": 2.0544,
+      "step": 39917
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1099241466498103e-06,
+      "loss": 3.3045,
+      "step": 39918
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1098767383059418e-06,
+      "loss": 3.4678,
+      "step": 39919
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1098293299620735e-06,
+      "loss": 3.8753,
+      "step": 39920
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1097819216182048e-06,
+      "loss": 0.7919,
+      "step": 39921
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1097345132743363e-06,
+      "loss": 0.3989,
+      "step": 39922
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1096871049304678e-06,
+      "loss": 2.451,
+      "step": 39923
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1096396965865993e-06,
+      "loss": 0.8371,
+      "step": 39924
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1095922882427308e-06,
+      "loss": 2.6485,
+      "step": 39925
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1095448798988622e-06,
+      "loss": 2.877,
+      "step": 39926
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1094974715549937e-06,
+      "loss": 2.7019,
+      "step": 39927
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1094500632111252e-06,
+      "loss": 2.3192,
+      "step": 39928
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 1.1094026548672567e-06,
+      "loss": 1.8322,
+      "step": 39929
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1093552465233882e-06,
+      "loss": 2.6078,
+      "step": 39930
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1093078381795197e-06,
+      "loss": 1.4402,
+      "step": 39931
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.109260429835651e-06,
+      "loss": 0.902,
+      "step": 39932
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1092130214917827e-06,
+      "loss": 1.7358,
+      "step": 39933
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.109165613147914e-06,
+      "loss": 1.9102,
+      "step": 39934
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1091182048040455e-06,
+      "loss": 1.5372,
+      "step": 39935
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1090707964601772e-06,
+      "loss": 2.2059,
+      "step": 39936
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1090233881163084e-06,
+      "loss": 1.4701,
+      "step": 39937
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.10897597977244e-06,
+      "loss": 1.6428,
+      "step": 39938
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1089285714285714e-06,
+      "loss": 2.7051,
+      "step": 39939
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.108881163084703e-06,
+      "loss": 2.9462,
+      "step": 39940
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1088337547408344e-06,
+      "loss": 2.059,
+      "step": 39941
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.108786346396966e-06,
+      "loss": 2.5591,
+      "step": 39942
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1087389380530974e-06,
+      "loss": 2.5654,
+      "step": 39943
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1086915297092289e-06,
+      "loss": 2.5286,
+      "step": 39944
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1086441213653602e-06,
+      "loss": 2.1903,
+      "step": 39945
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1085967130214919e-06,
+      "loss": 2.8303,
+      "step": 39946
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1085493046776234e-06,
+      "loss": 1.914,
+      "step": 39947
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1085018963337546e-06,
+      "loss": 3.6513,
+      "step": 39948
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1084544879898861e-06,
+      "loss": 0.8728,
+      "step": 39949
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1084070796460178e-06,
+      "loss": 2.8966,
+      "step": 39950
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1083596713021491e-06,
+      "loss": 1.7243,
+      "step": 39951
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1083122629582806e-06,
+      "loss": 2.525,
+      "step": 39952
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1082648546144123e-06,
+      "loss": 1.2887,
+      "step": 39953
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1082174462705436e-06,
+      "loss": 2.4443,
+      "step": 39954
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.108170037926675e-06,
+      "loss": 2.7563,
+      "step": 39955
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1081226295828066e-06,
+      "loss": 2.4698,
+      "step": 39956
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.108075221238938e-06,
+      "loss": 0.8943,
+      "step": 39957
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1080278128950696e-06,
+      "loss": 1.4984,
+      "step": 39958
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.107980404551201e-06,
+      "loss": 3.7405,
+      "step": 39959
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1079329962073326e-06,
+      "loss": 1.9046,
+      "step": 39960
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.107885587863464e-06,
+      "loss": 2.3285,
+      "step": 39961
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1078381795195953e-06,
+      "loss": 2.1808,
+      "step": 39962
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.107790771175727e-06,
+      "loss": 2.9464,
+      "step": 39963
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1077433628318585e-06,
+      "loss": 2.1419,
+      "step": 39964
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1076959544879898e-06,
+      "loss": 3.2458,
+      "step": 39965
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1076485461441215e-06,
+      "loss": 2.3508,
+      "step": 39966
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1076011378002528e-06,
+      "loss": 2.5331,
+      "step": 39967
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1075537294563843e-06,
+      "loss": 2.0999,
+      "step": 39968
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1075063211125158e-06,
+      "loss": 2.5417,
+      "step": 39969
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1074589127686473e-06,
+      "loss": 1.2545,
+      "step": 39970
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1074115044247788e-06,
+      "loss": 2.4465,
+      "step": 39971
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1073640960809102e-06,
+      "loss": 1.8917,
+      "step": 39972
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1073166877370417e-06,
+      "loss": 1.3677,
+      "step": 39973
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1072692793931732e-06,
+      "loss": 2.3857,
+      "step": 39974
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1072218710493047e-06,
+      "loss": 2.592,
+      "step": 39975
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1071744627054362e-06,
+      "loss": 2.6864,
+      "step": 39976
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1071270543615677e-06,
+      "loss": 0.9212,
+      "step": 39977
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.107079646017699e-06,
+      "loss": 0.9945,
+      "step": 39978
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1070322376738307e-06,
+      "loss": 4.6693,
+      "step": 39979
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1069848293299622e-06,
+      "loss": 1.4129,
+      "step": 39980
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1069374209860935e-06,
+      "loss": 2.7906,
+      "step": 39981
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.106890012642225e-06,
+      "loss": 2.6881,
+      "step": 39982
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1068426042983567e-06,
+      "loss": 1.5192,
+      "step": 39983
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.106795195954488e-06,
+      "loss": 0.7582,
+      "step": 39984
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1067477876106194e-06,
+      "loss": 2.1184,
+      "step": 39985
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1067003792667511e-06,
+      "loss": 1.3664,
+      "step": 39986
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1066529709228824e-06,
+      "loss": 1.6933,
+      "step": 39987
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.106605562579014e-06,
+      "loss": 1.905,
+      "step": 39988
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1065581542351454e-06,
+      "loss": 1.8503,
+      "step": 39989
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.106510745891277e-06,
+      "loss": 2.8155,
+      "step": 39990
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1064633375474084e-06,
+      "loss": 2.6465,
+      "step": 39991
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 1.1064159292035399e-06,
+      "loss": 2.9801,
+      "step": 39992
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1063685208596714e-06,
+      "loss": 1.6619,
+      "step": 39993
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1063211125158029e-06,
+      "loss": 2.0052,
+      "step": 39994
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1062737041719341e-06,
+      "loss": 2.585,
+      "step": 39995
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1062262958280658e-06,
+      "loss": 2.2291,
+      "step": 39996
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1061788874841973e-06,
+      "loss": 2.4311,
+      "step": 39997
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1061314791403286e-06,
+      "loss": 1.335,
+      "step": 39998
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1060840707964603e-06,
+      "loss": 2.7978,
+      "step": 39999
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1060366624525916e-06,
+      "loss": 1.8271,
+      "step": 40000
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.105989254108723e-06,
+      "loss": 2.2065,
+      "step": 40001
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1059418457648546e-06,
+      "loss": 1.8605,
+      "step": 40002
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.105894437420986e-06,
+      "loss": 1.2699,
+      "step": 40003
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1058470290771176e-06,
+      "loss": 1.5358,
+      "step": 40004
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.105799620733249e-06,
+      "loss": 2.4365,
+      "step": 40005
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1057522123893806e-06,
+      "loss": 1.1783,
+      "step": 40006
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.105704804045512e-06,
+      "loss": 3.0358,
+      "step": 40007
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1056573957016435e-06,
+      "loss": 1.8581,
+      "step": 40008
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.105609987357775e-06,
+      "loss": 2.3679,
+      "step": 40009
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1055625790139065e-06,
+      "loss": 2.3711,
+      "step": 40010
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1055151706700378e-06,
+      "loss": 1.5757,
+      "step": 40011
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1054677623261695e-06,
+      "loss": 1.5617,
+      "step": 40012
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.105420353982301e-06,
+      "loss": 2.4813,
+      "step": 40013
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1053729456384323e-06,
+      "loss": 0.7442,
+      "step": 40014
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1053255372945638e-06,
+      "loss": 1.9662,
+      "step": 40015
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1052781289506955e-06,
+      "loss": 1.5415,
+      "step": 40016
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1052307206068268e-06,
+      "loss": 0.7138,
+      "step": 40017
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1051833122629583e-06,
+      "loss": 1.8238,
+      "step": 40018
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.10513590391909e-06,
+      "loss": 1.6845,
+      "step": 40019
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1050884955752212e-06,
+      "loss": 1.7726,
+      "step": 40020
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1050410872313527e-06,
+      "loss": 2.8647,
+      "step": 40021
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1049936788874842e-06,
+      "loss": 1.6088,
+      "step": 40022
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1049462705436157e-06,
+      "loss": 0.6322,
+      "step": 40023
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1048988621997472e-06,
+      "loss": 1.9884,
+      "step": 40024
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1048514538558787e-06,
+      "loss": 0.6287,
+      "step": 40025
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1048040455120102e-06,
+      "loss": 2.3082,
+      "step": 40026
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1047566371681417e-06,
+      "loss": 2.4741,
+      "step": 40027
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.104709228824273e-06,
+      "loss": 1.6993,
+      "step": 40028
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1046618204804047e-06,
+      "loss": 1.6859,
+      "step": 40029
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1046144121365362e-06,
+      "loss": 1.5661,
+      "step": 40030
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1045670037926674e-06,
+      "loss": 3.1165,
+      "step": 40031
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1045195954487991e-06,
+      "loss": 2.3217,
+      "step": 40032
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1044721871049304e-06,
+      "loss": 2.5028,
+      "step": 40033
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.104424778761062e-06,
+      "loss": 2.0321,
+      "step": 40034
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1043773704171934e-06,
+      "loss": 1.7077,
+      "step": 40035
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.104329962073325e-06,
+      "loss": 1.5174,
+      "step": 40036
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1042825537294564e-06,
+      "loss": 2.4154,
+      "step": 40037
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1042351453855879e-06,
+      "loss": 1.3885,
+      "step": 40038
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1041877370417194e-06,
+      "loss": 1.8678,
+      "step": 40039
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1041403286978509e-06,
+      "loss": 2.5838,
+      "step": 40040
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1040929203539824e-06,
+      "loss": 1.8943,
+      "step": 40041
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1040455120101139e-06,
+      "loss": 1.8047,
+      "step": 40042
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1039981036662453e-06,
+      "loss": 1.1733,
+      "step": 40043
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1039506953223766e-06,
+      "loss": 1.5563,
+      "step": 40044
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1039032869785083e-06,
+      "loss": 3.8658,
+      "step": 40045
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1038558786346398e-06,
+      "loss": 4.6815,
+      "step": 40046
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.103808470290771e-06,
+      "loss": 2.4141,
+      "step": 40047
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1037610619469026e-06,
+      "loss": 2.0764,
+      "step": 40048
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1037136536030343e-06,
+      "loss": 2.0984,
+      "step": 40049
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1036662452591656e-06,
+      "loss": 2.2699,
+      "step": 40050
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.103618836915297e-06,
+      "loss": 2.6244,
+      "step": 40051
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1035714285714288e-06,
+      "loss": 1.3831,
+      "step": 40052
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.10352402022756e-06,
+      "loss": 4.1587,
+      "step": 40053
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1034766118836915e-06,
+      "loss": 1.6841,
+      "step": 40054
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.103429203539823e-06,
+      "loss": 2.4693,
+      "step": 40055
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 1.1033817951959545e-06,
+      "loss": 1.8002,
+      "step": 40056
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.103334386852086e-06,
+      "loss": 2.521,
+      "step": 40057
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1032869785082175e-06,
+      "loss": 2.1451,
+      "step": 40058
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.103239570164349e-06,
+      "loss": 2.664,
+      "step": 40059
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1031921618204805e-06,
+      "loss": 1.8865,
+      "step": 40060
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1031447534766118e-06,
+      "loss": 2.0098,
+      "step": 40061
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1030973451327435e-06,
+      "loss": 2.0841,
+      "step": 40062
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.103049936788875e-06,
+      "loss": 0.761,
+      "step": 40063
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1030025284450063e-06,
+      "loss": 0.9847,
+      "step": 40064
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.102955120101138e-06,
+      "loss": 1.4024,
+      "step": 40065
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1029077117572692e-06,
+      "loss": 2.2381,
+      "step": 40066
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1028603034134007e-06,
+      "loss": 2.53,
+      "step": 40067
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1028128950695322e-06,
+      "loss": 0.976,
+      "step": 40068
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1027654867256637e-06,
+      "loss": 2.6552,
+      "step": 40069
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1027180783817952e-06,
+      "loss": 2.9853,
+      "step": 40070
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1026706700379267e-06,
+      "loss": 1.3298,
+      "step": 40071
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1026232616940582e-06,
+      "loss": 0.9227,
+      "step": 40072
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1025758533501897e-06,
+      "loss": 3.5858,
+      "step": 40073
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1025284450063212e-06,
+      "loss": 2.7265,
+      "step": 40074
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1024810366624527e-06,
+      "loss": 2.8079,
+      "step": 40075
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1024336283185842e-06,
+      "loss": 2.1311,
+      "step": 40076
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1023862199747154e-06,
+      "loss": 2.6075,
+      "step": 40077
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1023388116308471e-06,
+      "loss": 2.9556,
+      "step": 40078
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1022914032869786e-06,
+      "loss": 1.6843,
+      "step": 40079
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.10224399494311e-06,
+      "loss": 2.013,
+      "step": 40080
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1021965865992414e-06,
+      "loss": 1.8834,
+      "step": 40081
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1021491782553731e-06,
+      "loss": 2.2947,
+      "step": 40082
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1021017699115044e-06,
+      "loss": 2.3976,
+      "step": 40083
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1020543615676359e-06,
+      "loss": 2.5778,
+      "step": 40084
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1020069532237676e-06,
+      "loss": 3.1762,
+      "step": 40085
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1019595448798989e-06,
+      "loss": 1.5792,
+      "step": 40086
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1019121365360304e-06,
+      "loss": 1.7769,
+      "step": 40087
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1018647281921619e-06,
+      "loss": 1.9631,
+      "step": 40088
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1018173198482934e-06,
+      "loss": 1.4424,
+      "step": 40089
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1017699115044248e-06,
+      "loss": 2.4602,
+      "step": 40090
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1017225031605561e-06,
+      "loss": 1.0458,
+      "step": 40091
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1016750948166878e-06,
+      "loss": 2.2767,
+      "step": 40092
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1016276864728193e-06,
+      "loss": 2.2097,
+      "step": 40093
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1015802781289506e-06,
+      "loss": 2.4405,
+      "step": 40094
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1015328697850823e-06,
+      "loss": 3.4884,
+      "step": 40095
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1014854614412138e-06,
+      "loss": 0.3768,
+      "step": 40096
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.101438053097345e-06,
+      "loss": 2.0064,
+      "step": 40097
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1013906447534766e-06,
+      "loss": 1.1667,
+      "step": 40098
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.101343236409608e-06,
+      "loss": 2.2208,
+      "step": 40099
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1012958280657396e-06,
+      "loss": 2.7072,
+      "step": 40100
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.101248419721871e-06,
+      "loss": 1.2127,
+      "step": 40101
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1012010113780025e-06,
+      "loss": 1.8597,
+      "step": 40102
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.101153603034134e-06,
+      "loss": 2.4703,
+      "step": 40103
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1011061946902655e-06,
+      "loss": 1.7491,
+      "step": 40104
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.101058786346397e-06,
+      "loss": 2.3561,
+      "step": 40105
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1010113780025285e-06,
+      "loss": 3.4939,
+      "step": 40106
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.10096396965866e-06,
+      "loss": 2.3536,
+      "step": 40107
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1009165613147915e-06,
+      "loss": 1.6642,
+      "step": 40108
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.100869152970923e-06,
+      "loss": 2.3051,
+      "step": 40109
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1008217446270543e-06,
+      "loss": 1.7709,
+      "step": 40110
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1007743362831858e-06,
+      "loss": 1.3612,
+      "step": 40111
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1007269279393175e-06,
+      "loss": 2.7545,
+      "step": 40112
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1006795195954487e-06,
+      "loss": 2.111,
+      "step": 40113
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1006321112515802e-06,
+      "loss": 3.6122,
+      "step": 40114
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.100584702907712e-06,
+      "loss": 0.6868,
+      "step": 40115
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1005372945638432e-06,
+      "loss": 2.0195,
+      "step": 40116
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1004898862199747e-06,
+      "loss": 2.6489,
+      "step": 40117
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1004424778761062e-06,
+      "loss": 2.1981,
+      "step": 40118
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 1.1003950695322377e-06,
+      "loss": 1.4243,
+      "step": 40119
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1003476611883692e-06,
+      "loss": 1.941,
+      "step": 40120
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1003002528445007e-06,
+      "loss": 1.6105,
+      "step": 40121
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1002528445006322e-06,
+      "loss": 0.7944,
+      "step": 40122
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1002054361567637e-06,
+      "loss": 1.9522,
+      "step": 40123
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.100158027812895e-06,
+      "loss": 2.8057,
+      "step": 40124
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1001106194690266e-06,
+      "loss": 2.5922,
+      "step": 40125
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1000632111251581e-06,
+      "loss": 3.2616,
+      "step": 40126
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.1000158027812894e-06,
+      "loss": 1.62,
+      "step": 40127
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0999683944374211e-06,
+      "loss": 2.1905,
+      "step": 40128
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0999209860935526e-06,
+      "loss": 2.9791,
+      "step": 40129
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0998735777496839e-06,
+      "loss": 2.2367,
+      "step": 40130
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0998261694058154e-06,
+      "loss": 3.6684,
+      "step": 40131
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0997787610619469e-06,
+      "loss": 2.0523,
+      "step": 40132
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0997313527180784e-06,
+      "loss": 2.4867,
+      "step": 40133
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0996839443742099e-06,
+      "loss": 1.9347,
+      "step": 40134
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0996365360303414e-06,
+      "loss": 0.8054,
+      "step": 40135
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0995891276864728e-06,
+      "loss": 1.3147,
+      "step": 40136
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0995417193426043e-06,
+      "loss": 0.867,
+      "step": 40137
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0994943109987358e-06,
+      "loss": 1.9715,
+      "step": 40138
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0994469026548673e-06,
+      "loss": 0.6971,
+      "step": 40139
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0993994943109988e-06,
+      "loss": 1.7245,
+      "step": 40140
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0993520859671303e-06,
+      "loss": 1.3876,
+      "step": 40141
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0993046776232618e-06,
+      "loss": 1.7162,
+      "step": 40142
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.099257269279393e-06,
+      "loss": 1.6623,
+      "step": 40143
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0992098609355246e-06,
+      "loss": 1.9135,
+      "step": 40144
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0991624525916563e-06,
+      "loss": 3.343,
+      "step": 40145
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0991150442477876e-06,
+      "loss": 2.0268,
+      "step": 40146
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.099067635903919e-06,
+      "loss": 0.6479,
+      "step": 40147
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0990202275600508e-06,
+      "loss": 2.0562,
+      "step": 40148
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.098972819216182e-06,
+      "loss": 3.267,
+      "step": 40149
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0989254108723135e-06,
+      "loss": 1.3383,
+      "step": 40150
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.098878002528445e-06,
+      "loss": 1.0389,
+      "step": 40151
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0988305941845765e-06,
+      "loss": 1.9963,
+      "step": 40152
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.098783185840708e-06,
+      "loss": 2.1142,
+      "step": 40153
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0987357774968395e-06,
+      "loss": 0.8441,
+      "step": 40154
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.098688369152971e-06,
+      "loss": 1.8172,
+      "step": 40155
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0986409608091025e-06,
+      "loss": 2.0344,
+      "step": 40156
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0985935524652338e-06,
+      "loss": 2.3279,
+      "step": 40157
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0985461441213655e-06,
+      "loss": 1.1191,
+      "step": 40158
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.098498735777497e-06,
+      "loss": 1.7003,
+      "step": 40159
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0984513274336282e-06,
+      "loss": 3.2455,
+      "step": 40160
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.09840391908976e-06,
+      "loss": 3.2262,
+      "step": 40161
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0983565107458914e-06,
+      "loss": 2.3792,
+      "step": 40162
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0983091024020227e-06,
+      "loss": 2.308,
+      "step": 40163
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0982616940581542e-06,
+      "loss": 1.2139,
+      "step": 40164
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0982142857142857e-06,
+      "loss": 2.9005,
+      "step": 40165
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0981668773704172e-06,
+      "loss": 2.052,
+      "step": 40166
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0981194690265487e-06,
+      "loss": 1.5605,
+      "step": 40167
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0980720606826802e-06,
+      "loss": 3.838,
+      "step": 40168
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0980246523388117e-06,
+      "loss": 1.8863,
+      "step": 40169
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0979772439949432e-06,
+      "loss": 1.7583,
+      "step": 40170
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0979298356510746e-06,
+      "loss": 1.7193,
+      "step": 40171
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0978824273072061e-06,
+      "loss": 0.8643,
+      "step": 40172
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0978350189633376e-06,
+      "loss": 1.4232,
+      "step": 40173
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0977876106194691e-06,
+      "loss": 2.1647,
+      "step": 40174
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0977402022756006e-06,
+      "loss": 1.9897,
+      "step": 40175
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.097692793931732e-06,
+      "loss": 1.6778,
+      "step": 40176
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0976453855878634e-06,
+      "loss": 2.8971,
+      "step": 40177
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.097597977243995e-06,
+      "loss": 2.2049,
+      "step": 40178
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0975505689001264e-06,
+      "loss": 1.795,
+      "step": 40179
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0975031605562579e-06,
+      "loss": 1.2147,
+      "step": 40180
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0974557522123896e-06,
+      "loss": 2.1363,
+      "step": 40181
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 1.0974083438685209e-06,
+      "loss": 1.4855,
+      "step": 40182
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0973609355246523e-06,
+      "loss": 0.4996,
+      "step": 40183
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0973135271807838e-06,
+      "loss": 1.0697,
+      "step": 40184
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0972661188369153e-06,
+      "loss": 1.0496,
+      "step": 40185
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0972187104930468e-06,
+      "loss": 1.4544,
+      "step": 40186
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0971713021491783e-06,
+      "loss": 1.5012,
+      "step": 40187
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0971238938053098e-06,
+      "loss": 1.9546,
+      "step": 40188
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0970764854614413e-06,
+      "loss": 1.7019,
+      "step": 40189
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0970290771175726e-06,
+      "loss": 1.3545,
+      "step": 40190
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0969816687737043e-06,
+      "loss": 1.8962,
+      "step": 40191
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0969342604298358e-06,
+      "loss": 1.8409,
+      "step": 40192
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096886852085967e-06,
+      "loss": 1.3765,
+      "step": 40193
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0968394437420988e-06,
+      "loss": 1.4103,
+      "step": 40194
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0967920353982303e-06,
+      "loss": 2.7251,
+      "step": 40195
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0967446270543615e-06,
+      "loss": 1.5869,
+      "step": 40196
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096697218710493e-06,
+      "loss": 1.7145,
+      "step": 40197
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0966498103666245e-06,
+      "loss": 2.3916,
+      "step": 40198
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096602402022756e-06,
+      "loss": 1.9788,
+      "step": 40199
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0965549936788875e-06,
+      "loss": 2.3767,
+      "step": 40200
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096507585335019e-06,
+      "loss": 2.7949,
+      "step": 40201
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0964601769911505e-06,
+      "loss": 1.4315,
+      "step": 40202
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096412768647282e-06,
+      "loss": 3.1521,
+      "step": 40203
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0963653603034135e-06,
+      "loss": 2.4207,
+      "step": 40204
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096317951959545e-06,
+      "loss": 2.9971,
+      "step": 40205
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0962705436156765e-06,
+      "loss": 4.0911,
+      "step": 40206
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096223135271808e-06,
+      "loss": 1.2936,
+      "step": 40207
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0961757269279394e-06,
+      "loss": 3.652,
+      "step": 40208
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0961283185840707e-06,
+      "loss": 2.4485,
+      "step": 40209
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0960809102402022e-06,
+      "loss": 2.4423,
+      "step": 40210
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.096033501896334e-06,
+      "loss": 2.9995,
+      "step": 40211
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0959860935524652e-06,
+      "loss": 1.6196,
+      "step": 40212
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0959386852085967e-06,
+      "loss": 2.263,
+      "step": 40213
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0958912768647284e-06,
+      "loss": 0.8831,
+      "step": 40214
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0958438685208597e-06,
+      "loss": 3.6696,
+      "step": 40215
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0957964601769912e-06,
+      "loss": 0.4405,
+      "step": 40216
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0957490518331227e-06,
+      "loss": 1.1964,
+      "step": 40217
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0957016434892541e-06,
+      "loss": 2.1002,
+      "step": 40218
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0956542351453856e-06,
+      "loss": 2.6704,
+      "step": 40219
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.095606826801517e-06,
+      "loss": 1.2856,
+      "step": 40220
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0955594184576486e-06,
+      "loss": 1.3199,
+      "step": 40221
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0955120101137801e-06,
+      "loss": 2.4562,
+      "step": 40222
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0954646017699114e-06,
+      "loss": 1.3518,
+      "step": 40223
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.095417193426043e-06,
+      "loss": 1.5707,
+      "step": 40224
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0953697850821746e-06,
+      "loss": 1.18,
+      "step": 40225
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0953223767383059e-06,
+      "loss": 1.7736,
+      "step": 40226
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0952749683944374e-06,
+      "loss": 1.7343,
+      "step": 40227
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.095227560050569e-06,
+      "loss": 2.3742,
+      "step": 40228
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0951801517067003e-06,
+      "loss": 2.1844,
+      "step": 40229
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0951327433628318e-06,
+      "loss": 1.7114,
+      "step": 40230
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0950853350189633e-06,
+      "loss": 2.1716,
+      "step": 40231
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0950379266750948e-06,
+      "loss": 1.7887,
+      "step": 40232
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0949905183312263e-06,
+      "loss": 1.9106,
+      "step": 40233
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0949431099873578e-06,
+      "loss": 2.2729,
+      "step": 40234
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0948957016434893e-06,
+      "loss": 3.4084,
+      "step": 40235
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0948482932996208e-06,
+      "loss": 2.206,
+      "step": 40236
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0948008849557523e-06,
+      "loss": 1.195,
+      "step": 40237
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0947534766118838e-06,
+      "loss": 2.6019,
+      "step": 40238
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0947060682680153e-06,
+      "loss": 1.6693,
+      "step": 40239
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0946586599241466e-06,
+      "loss": 3.3112,
+      "step": 40240
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0946112515802783e-06,
+      "loss": 2.3845,
+      "step": 40241
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0945638432364095e-06,
+      "loss": 1.5417,
+      "step": 40242
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.094516434892541e-06,
+      "loss": 2.6948,
+      "step": 40243
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0944690265486727e-06,
+      "loss": 2.109,
+      "step": 40244
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.094421618204804e-06,
+      "loss": 1.0088,
+      "step": 40245
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 1.0943742098609355e-06,
+      "loss": 1.9917,
+      "step": 40246
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.094326801517067e-06,
+      "loss": 2.6465,
+      "step": 40247
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0942793931731985e-06,
+      "loss": 1.4726,
+      "step": 40248
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.09423198482933e-06,
+      "loss": 1.5483,
+      "step": 40249
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0941845764854615e-06,
+      "loss": 1.1906,
+      "step": 40250
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.094137168141593e-06,
+      "loss": 1.9597,
+      "step": 40251
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0940897597977245e-06,
+      "loss": 2.1083,
+      "step": 40252
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0940423514538557e-06,
+      "loss": 1.1561,
+      "step": 40253
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0939949431099874e-06,
+      "loss": 2.1062,
+      "step": 40254
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.093947534766119e-06,
+      "loss": 2.2706,
+      "step": 40255
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0939001264222502e-06,
+      "loss": 1.7579,
+      "step": 40256
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.093852718078382e-06,
+      "loss": 0.8129,
+      "step": 40257
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0938053097345134e-06,
+      "loss": 2.428,
+      "step": 40258
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0937579013906447e-06,
+      "loss": 2.7798,
+      "step": 40259
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0937104930467762e-06,
+      "loss": 1.3234,
+      "step": 40260
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0936630847029077e-06,
+      "loss": 2.3476,
+      "step": 40261
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0936156763590392e-06,
+      "loss": 1.5588,
+      "step": 40262
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0935682680151707e-06,
+      "loss": 0.9912,
+      "step": 40263
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0935208596713022e-06,
+      "loss": 1.9041,
+      "step": 40264
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0934734513274336e-06,
+      "loss": 2.6984,
+      "step": 40265
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0934260429835651e-06,
+      "loss": 1.6739,
+      "step": 40266
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0933786346396966e-06,
+      "loss": 1.71,
+      "step": 40267
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0933312262958281e-06,
+      "loss": 2.5071,
+      "step": 40268
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0932838179519596e-06,
+      "loss": 1.8121,
+      "step": 40269
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.093236409608091e-06,
+      "loss": 1.859,
+      "step": 40270
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0931890012642226e-06,
+      "loss": 1.8472,
+      "step": 40271
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0931415929203539e-06,
+      "loss": 2.2266,
+      "step": 40272
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0930941845764854e-06,
+      "loss": 3.1733,
+      "step": 40273
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.093046776232617e-06,
+      "loss": 1.5982,
+      "step": 40274
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0929993678887484e-06,
+      "loss": 1.2991,
+      "step": 40275
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0929519595448798e-06,
+      "loss": 2.858,
+      "step": 40276
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0929045512010115e-06,
+      "loss": 3.2808,
+      "step": 40277
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0928571428571428e-06,
+      "loss": 1.6765,
+      "step": 40278
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0928097345132743e-06,
+      "loss": 1.9025,
+      "step": 40279
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0927623261694058e-06,
+      "loss": 2.7826,
+      "step": 40280
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0927149178255373e-06,
+      "loss": 2.1654,
+      "step": 40281
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0926675094816688e-06,
+      "loss": 1.7994,
+      "step": 40282
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0926201011378003e-06,
+      "loss": 0.9049,
+      "step": 40283
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0925726927939318e-06,
+      "loss": 1.7043,
+      "step": 40284
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0925252844500633e-06,
+      "loss": 2.3112,
+      "step": 40285
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0924778761061946e-06,
+      "loss": 0.6685,
+      "step": 40286
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0924304677623263e-06,
+      "loss": 2.3478,
+      "step": 40287
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0923830594184578e-06,
+      "loss": 2.2929,
+      "step": 40288
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.092335651074589e-06,
+      "loss": 2.4715,
+      "step": 40289
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0922882427307207e-06,
+      "loss": 0.5423,
+      "step": 40290
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0922408343868522e-06,
+      "loss": 1.7365,
+      "step": 40291
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0921934260429835e-06,
+      "loss": 1.9766,
+      "step": 40292
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.092146017699115e-06,
+      "loss": 2.8621,
+      "step": 40293
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0920986093552465e-06,
+      "loss": 3.7944,
+      "step": 40294
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.092051201011378e-06,
+      "loss": 1.8951,
+      "step": 40295
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0920037926675095e-06,
+      "loss": 2.4029,
+      "step": 40296
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.091956384323641e-06,
+      "loss": 1.2718,
+      "step": 40297
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0919089759797725e-06,
+      "loss": 2.3357,
+      "step": 40298
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.091861567635904e-06,
+      "loss": 1.6539,
+      "step": 40299
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0918141592920354e-06,
+      "loss": 1.5062,
+      "step": 40300
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.091766750948167e-06,
+      "loss": 3.0094,
+      "step": 40301
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0917193426042984e-06,
+      "loss": 1.6046,
+      "step": 40302
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.09167193426043e-06,
+      "loss": 3.4746,
+      "step": 40303
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0916245259165614e-06,
+      "loss": 2.6024,
+      "step": 40304
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0915771175726927e-06,
+      "loss": 2.4561,
+      "step": 40305
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0915297092288242e-06,
+      "loss": 1.3744,
+      "step": 40306
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0914823008849559e-06,
+      "loss": 1.5965,
+      "step": 40307
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0914348925410872e-06,
+      "loss": 1.0438,
+      "step": 40308
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 1.0913874841972187e-06,
+      "loss": 1.9911,
+      "step": 40309
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0913400758533504e-06,
+      "loss": 1.8952,
+      "step": 40310
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0912926675094816e-06,
+      "loss": 1.9796,
+      "step": 40311
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0912452591656131e-06,
+      "loss": 3.2749,
+      "step": 40312
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0911978508217446e-06,
+      "loss": 2.7035,
+      "step": 40313
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0911504424778761e-06,
+      "loss": 1.5924,
+      "step": 40314
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0911030341340076e-06,
+      "loss": 2.1873,
+      "step": 40315
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0910556257901391e-06,
+      "loss": 1.2621,
+      "step": 40316
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0910082174462706e-06,
+      "loss": 2.1374,
+      "step": 40317
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.090960809102402e-06,
+      "loss": 2.0701,
+      "step": 40318
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0909134007585334e-06,
+      "loss": 1.0911,
+      "step": 40319
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.090865992414665e-06,
+      "loss": 0.5705,
+      "step": 40320
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0908185840707966e-06,
+      "loss": 1.9152,
+      "step": 40321
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0907711757269279e-06,
+      "loss": 2.2254,
+      "step": 40322
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0907237673830596e-06,
+      "loss": 1.9261,
+      "step": 40323
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.090676359039191e-06,
+      "loss": 3.5038,
+      "step": 40324
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0906289506953223e-06,
+      "loss": 1.547,
+      "step": 40325
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0905815423514538e-06,
+      "loss": 0.7566,
+      "step": 40326
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0905341340075853e-06,
+      "loss": 3.0066,
+      "step": 40327
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0904867256637168e-06,
+      "loss": 2.1695,
+      "step": 40328
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0904393173198483e-06,
+      "loss": 2.7297,
+      "step": 40329
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0903919089759798e-06,
+      "loss": 1.9122,
+      "step": 40330
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0903445006321113e-06,
+      "loss": 1.5084,
+      "step": 40331
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0902970922882428e-06,
+      "loss": 1.1426,
+      "step": 40332
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0902496839443743e-06,
+      "loss": 1.7101,
+      "step": 40333
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0902022756005058e-06,
+      "loss": 1.2887,
+      "step": 40334
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0901548672566372e-06,
+      "loss": 1.3952,
+      "step": 40335
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0901074589127687e-06,
+      "loss": 1.2838,
+      "step": 40336
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0900600505689002e-06,
+      "loss": 1.8265,
+      "step": 40337
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0900126422250315e-06,
+      "loss": 1.7718,
+      "step": 40338
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089965233881163e-06,
+      "loss": 1.8512,
+      "step": 40339
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0899178255372947e-06,
+      "loss": 2.3624,
+      "step": 40340
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089870417193426e-06,
+      "loss": 1.5834,
+      "step": 40341
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0898230088495575e-06,
+      "loss": 2.0281,
+      "step": 40342
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0897756005056892e-06,
+      "loss": 1.5309,
+      "step": 40343
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0897281921618205e-06,
+      "loss": 2.0238,
+      "step": 40344
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089680783817952e-06,
+      "loss": 1.2111,
+      "step": 40345
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0896333754740835e-06,
+      "loss": 1.9108,
+      "step": 40346
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089585967130215e-06,
+      "loss": 0.9058,
+      "step": 40347
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0895385587863464e-06,
+      "loss": 2.2427,
+      "step": 40348
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089491150442478e-06,
+      "loss": 1.3444,
+      "step": 40349
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0894437420986094e-06,
+      "loss": 2.2804,
+      "step": 40350
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089396333754741e-06,
+      "loss": 2.7256,
+      "step": 40351
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0893489254108722e-06,
+      "loss": 3.2767,
+      "step": 40352
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.089301517067004e-06,
+      "loss": 2.2504,
+      "step": 40353
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0892541087231354e-06,
+      "loss": 0.9899,
+      "step": 40354
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0892067003792667e-06,
+      "loss": 2.2307,
+      "step": 40355
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0891592920353984e-06,
+      "loss": 2.5162,
+      "step": 40356
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0891118836915299e-06,
+      "loss": 2.858,
+      "step": 40357
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0890644753476611e-06,
+      "loss": 1.621,
+      "step": 40358
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0890170670037926e-06,
+      "loss": 2.0688,
+      "step": 40359
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0889696586599241e-06,
+      "loss": 2.0557,
+      "step": 40360
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0889222503160556e-06,
+      "loss": 1.9349,
+      "step": 40361
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0888748419721871e-06,
+      "loss": 2.8325,
+      "step": 40362
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0888274336283186e-06,
+      "loss": 2.6219,
+      "step": 40363
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.08878002528445e-06,
+      "loss": 1.4226,
+      "step": 40364
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0887326169405816e-06,
+      "loss": 1.6467,
+      "step": 40365
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.088685208596713e-06,
+      "loss": 2.6822,
+      "step": 40366
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0886378002528446e-06,
+      "loss": 1.0596,
+      "step": 40367
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.088590391908976e-06,
+      "loss": 4.0158,
+      "step": 40368
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0885429835651073e-06,
+      "loss": 2.3833,
+      "step": 40369
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.088495575221239e-06,
+      "loss": 2.0297,
+      "step": 40370
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0884481668773703e-06,
+      "loss": 1.3487,
+      "step": 40371
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 1.0884007585335018e-06,
+      "loss": 1.2115,
+      "step": 40372
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0883533501896335e-06,
+      "loss": 2.1899,
+      "step": 40373
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0883059418457648e-06,
+      "loss": 2.1184,
+      "step": 40374
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0882585335018963e-06,
+      "loss": 1.0479,
+      "step": 40375
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0882111251580278e-06,
+      "loss": 2.5825,
+      "step": 40376
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0881637168141593e-06,
+      "loss": 1.9922,
+      "step": 40377
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0881163084702908e-06,
+      "loss": 1.4048,
+      "step": 40378
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0880689001264223e-06,
+      "loss": 1.6222,
+      "step": 40379
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0880214917825538e-06,
+      "loss": 2.9385,
+      "step": 40380
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0879740834386853e-06,
+      "loss": 1.2937,
+      "step": 40381
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0879266750948165e-06,
+      "loss": 2.5645,
+      "step": 40382
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0878792667509482e-06,
+      "loss": 1.8297,
+      "step": 40383
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0878318584070797e-06,
+      "loss": 2.4397,
+      "step": 40384
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.087784450063211e-06,
+      "loss": 2.4607,
+      "step": 40385
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0877370417193427e-06,
+      "loss": 1.6848,
+      "step": 40386
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0876896333754742e-06,
+      "loss": 2.7613,
+      "step": 40387
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0876422250316055e-06,
+      "loss": 2.2422,
+      "step": 40388
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.087594816687737e-06,
+      "loss": 2.3926,
+      "step": 40389
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0875474083438687e-06,
+      "loss": 3.2706,
+      "step": 40390
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0875e-06,
+      "loss": 2.0641,
+      "step": 40391
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0874525916561315e-06,
+      "loss": 1.9793,
+      "step": 40392
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.087405183312263e-06,
+      "loss": 3.2286,
+      "step": 40393
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0873577749683944e-06,
+      "loss": 1.1788,
+      "step": 40394
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.087310366624526e-06,
+      "loss": 1.2638,
+      "step": 40395
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0872629582806574e-06,
+      "loss": 2.5227,
+      "step": 40396
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.087215549936789e-06,
+      "loss": 2.4639,
+      "step": 40397
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0871681415929204e-06,
+      "loss": 1.8086,
+      "step": 40398
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.087120733249052e-06,
+      "loss": 0.9614,
+      "step": 40399
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0870733249051834e-06,
+      "loss": 3.8888,
+      "step": 40400
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0870259165613149e-06,
+      "loss": 2.1714,
+      "step": 40401
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0869785082174462e-06,
+      "loss": 1.4984,
+      "step": 40402
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0869310998735779e-06,
+      "loss": 3.1942,
+      "step": 40403
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0868836915297092e-06,
+      "loss": 2.6274,
+      "step": 40404
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0868362831858406e-06,
+      "loss": 1.4885,
+      "step": 40405
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0867888748419723e-06,
+      "loss": 1.9465,
+      "step": 40406
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0867414664981036e-06,
+      "loss": 4.1165,
+      "step": 40407
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0866940581542351e-06,
+      "loss": 1.4969,
+      "step": 40408
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0866466498103666e-06,
+      "loss": 1.5522,
+      "step": 40409
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.086599241466498e-06,
+      "loss": 2.1357,
+      "step": 40410
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0865518331226296e-06,
+      "loss": 1.9033,
+      "step": 40411
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.086504424778761e-06,
+      "loss": 2.3586,
+      "step": 40412
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0864570164348926e-06,
+      "loss": 0.9864,
+      "step": 40413
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.086409608091024e-06,
+      "loss": 2.9026,
+      "step": 40414
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0863621997471554e-06,
+      "loss": 1.6678,
+      "step": 40415
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.086314791403287e-06,
+      "loss": 1.6314,
+      "step": 40416
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0862673830594185e-06,
+      "loss": 2.1781,
+      "step": 40417
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0862199747155498e-06,
+      "loss": 3.2835,
+      "step": 40418
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0861725663716815e-06,
+      "loss": 1.0567,
+      "step": 40419
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.086125158027813e-06,
+      "loss": 1.9235,
+      "step": 40420
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0860777496839443e-06,
+      "loss": 2.0029,
+      "step": 40421
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0860303413400758e-06,
+      "loss": 0.3185,
+      "step": 40422
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0859829329962075e-06,
+      "loss": 2.3569,
+      "step": 40423
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0859355246523388e-06,
+      "loss": 3.017,
+      "step": 40424
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0858881163084703e-06,
+      "loss": 2.4694,
+      "step": 40425
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0858407079646018e-06,
+      "loss": 2.0964,
+      "step": 40426
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0857932996207333e-06,
+      "loss": 1.2631,
+      "step": 40427
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0857458912768648e-06,
+      "loss": 1.6742,
+      "step": 40428
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0856984829329962e-06,
+      "loss": 2.3867,
+      "step": 40429
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0856510745891277e-06,
+      "loss": 2.4413,
+      "step": 40430
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0856036662452592e-06,
+      "loss": 2.988,
+      "step": 40431
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0855562579013907e-06,
+      "loss": 1.4576,
+      "step": 40432
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0855088495575222e-06,
+      "loss": 1.4405,
+      "step": 40433
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0854614412136537e-06,
+      "loss": 0.7265,
+      "step": 40434
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.085414032869785e-06,
+      "loss": 1.2571,
+      "step": 40435
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 1.0853666245259167e-06,
+      "loss": 1.5825,
+      "step": 40436
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.085319216182048e-06,
+      "loss": 2.4357,
+      "step": 40437
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0852718078381795e-06,
+      "loss": 2.0709,
+      "step": 40438
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0852243994943112e-06,
+      "loss": 2.6523,
+      "step": 40439
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0851769911504424e-06,
+      "loss": 1.0837,
+      "step": 40440
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.085129582806574e-06,
+      "loss": 1.7602,
+      "step": 40441
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0850821744627054e-06,
+      "loss": 3.066,
+      "step": 40442
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.085034766118837e-06,
+      "loss": 0.7728,
+      "step": 40443
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0849873577749684e-06,
+      "loss": 3.6709,
+      "step": 40444
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0849399494311e-06,
+      "loss": 3.4837,
+      "step": 40445
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0848925410872314e-06,
+      "loss": 1.2523,
+      "step": 40446
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0848451327433629e-06,
+      "loss": 2.6873,
+      "step": 40447
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0847977243994942e-06,
+      "loss": 0.9184,
+      "step": 40448
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0847503160556259e-06,
+      "loss": 1.8259,
+      "step": 40449
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0847029077117574e-06,
+      "loss": 3.2735,
+      "step": 40450
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0846554993678886e-06,
+      "loss": 1.7701,
+      "step": 40451
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0846080910240204e-06,
+      "loss": 2.3691,
+      "step": 40452
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0845606826801518e-06,
+      "loss": 3.6189,
+      "step": 40453
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0845132743362831e-06,
+      "loss": 2.0387,
+      "step": 40454
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0844658659924146e-06,
+      "loss": 1.5889,
+      "step": 40455
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0844184576485463e-06,
+      "loss": 1.0004,
+      "step": 40456
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0843710493046776e-06,
+      "loss": 1.5551,
+      "step": 40457
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.084323640960809e-06,
+      "loss": 1.8621,
+      "step": 40458
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0842762326169406e-06,
+      "loss": 0.9319,
+      "step": 40459
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.084228824273072e-06,
+      "loss": 2.3061,
+      "step": 40460
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0841814159292036e-06,
+      "loss": 2.2206,
+      "step": 40461
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.084134007585335e-06,
+      "loss": 3.0412,
+      "step": 40462
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0840865992414666e-06,
+      "loss": 0.9648,
+      "step": 40463
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.084039190897598e-06,
+      "loss": 2.5682,
+      "step": 40464
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0839917825537295e-06,
+      "loss": 2.3627,
+      "step": 40465
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.083944374209861e-06,
+      "loss": 2.378,
+      "step": 40466
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0838969658659925e-06,
+      "loss": 2.5835,
+      "step": 40467
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0838495575221238e-06,
+      "loss": 2.222,
+      "step": 40468
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0838021491782555e-06,
+      "loss": 2.9591,
+      "step": 40469
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0837547408343868e-06,
+      "loss": 2.261,
+      "step": 40470
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0837073324905183e-06,
+      "loss": 2.8534,
+      "step": 40471
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.08365992414665e-06,
+      "loss": 1.803,
+      "step": 40472
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0836125158027813e-06,
+      "loss": 3.4216,
+      "step": 40473
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0835651074589128e-06,
+      "loss": 1.5433,
+      "step": 40474
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0835176991150442e-06,
+      "loss": 1.4912,
+      "step": 40475
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0834702907711757e-06,
+      "loss": 1.707,
+      "step": 40476
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0834228824273072e-06,
+      "loss": 1.9825,
+      "step": 40477
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0833754740834387e-06,
+      "loss": 2.1152,
+      "step": 40478
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0833280657395702e-06,
+      "loss": 2.3129,
+      "step": 40479
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0832806573957017e-06,
+      "loss": 1.6129,
+      "step": 40480
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.083233249051833e-06,
+      "loss": 1.3595,
+      "step": 40481
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0831858407079647e-06,
+      "loss": 1.9613,
+      "step": 40482
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0831384323640962e-06,
+      "loss": 2.1088,
+      "step": 40483
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0830910240202275e-06,
+      "loss": 1.5964,
+      "step": 40484
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0830436156763592e-06,
+      "loss": 2.1009,
+      "step": 40485
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0829962073324907e-06,
+      "loss": 2.9984,
+      "step": 40486
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.082948798988622e-06,
+      "loss": 1.1173,
+      "step": 40487
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0829013906447534e-06,
+      "loss": 2.1187,
+      "step": 40488
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0828539823008851e-06,
+      "loss": 1.751,
+      "step": 40489
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0828065739570164e-06,
+      "loss": 2.3696,
+      "step": 40490
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.082759165613148e-06,
+      "loss": 2.2065,
+      "step": 40491
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0827117572692794e-06,
+      "loss": 1.6767,
+      "step": 40492
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.082664348925411e-06,
+      "loss": 2.112,
+      "step": 40493
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0826169405815424e-06,
+      "loss": 1.8853,
+      "step": 40494
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0825695322376739e-06,
+      "loss": 2.6738,
+      "step": 40495
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0825221238938054e-06,
+      "loss": 1.4556,
+      "step": 40496
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0824747155499369e-06,
+      "loss": 1.6934,
+      "step": 40497
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0824273072060684e-06,
+      "loss": 2.2626,
+      "step": 40498
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 1.0823798988621998e-06,
+      "loss": 1.4235,
+      "step": 40499
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0823324905183313e-06,
+      "loss": 1.2981,
+      "step": 40500
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0822850821744626e-06,
+      "loss": 0.9605,
+      "step": 40501
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0822376738305943e-06,
+      "loss": 1.8044,
+      "step": 40502
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0821902654867256e-06,
+      "loss": 2.3387,
+      "step": 40503
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.082142857142857e-06,
+      "loss": 1.0939,
+      "step": 40504
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0820954487989888e-06,
+      "loss": 2.6718,
+      "step": 40505
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.08204804045512e-06,
+      "loss": 2.2246,
+      "step": 40506
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0820006321112516e-06,
+      "loss": 2.5959,
+      "step": 40507
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.081953223767383e-06,
+      "loss": 1.3844,
+      "step": 40508
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0819058154235146e-06,
+      "loss": 1.9721,
+      "step": 40509
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.081858407079646e-06,
+      "loss": 2.2524,
+      "step": 40510
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0818109987357775e-06,
+      "loss": 1.0532,
+      "step": 40511
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.081763590391909e-06,
+      "loss": 2.2544,
+      "step": 40512
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0817161820480405e-06,
+      "loss": 2.5599,
+      "step": 40513
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0816687737041718e-06,
+      "loss": 1.6508,
+      "step": 40514
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0816213653603035e-06,
+      "loss": 1.7732,
+      "step": 40515
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.081573957016435e-06,
+      "loss": 1.5113,
+      "step": 40516
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0815265486725663e-06,
+      "loss": 2.507,
+      "step": 40517
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0814791403286978e-06,
+      "loss": 1.7425,
+      "step": 40518
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0814317319848295e-06,
+      "loss": 1.781,
+      "step": 40519
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0813843236409608e-06,
+      "loss": 1.879,
+      "step": 40520
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0813369152970923e-06,
+      "loss": 1.6602,
+      "step": 40521
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.081289506953224e-06,
+      "loss": 2.0666,
+      "step": 40522
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0812420986093552e-06,
+      "loss": 1.5711,
+      "step": 40523
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0811946902654867e-06,
+      "loss": 1.376,
+      "step": 40524
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0811472819216182e-06,
+      "loss": 2.644,
+      "step": 40525
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0810998735777497e-06,
+      "loss": 3.1705,
+      "step": 40526
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0810524652338812e-06,
+      "loss": 2.3942,
+      "step": 40527
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0810050568900127e-06,
+      "loss": 1.3288,
+      "step": 40528
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0809576485461442e-06,
+      "loss": 2.0657,
+      "step": 40529
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0809102402022757e-06,
+      "loss": 0.4551,
+      "step": 40530
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.080862831858407e-06,
+      "loss": 2.417,
+      "step": 40531
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0808154235145387e-06,
+      "loss": 3.6783,
+      "step": 40532
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0807680151706702e-06,
+      "loss": 2.2462,
+      "step": 40533
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0807206068268014e-06,
+      "loss": 2.1897,
+      "step": 40534
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0806731984829331e-06,
+      "loss": 0.9479,
+      "step": 40535
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0806257901390644e-06,
+      "loss": 1.489,
+      "step": 40536
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.080578381795196e-06,
+      "loss": 2.2143,
+      "step": 40537
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0805309734513274e-06,
+      "loss": 2.8667,
+      "step": 40538
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.080483565107459e-06,
+      "loss": 2.5018,
+      "step": 40539
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0804361567635904e-06,
+      "loss": 2.3575,
+      "step": 40540
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0803887484197219e-06,
+      "loss": 2.3799,
+      "step": 40541
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0803413400758534e-06,
+      "loss": 3.5416,
+      "step": 40542
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0802939317319849e-06,
+      "loss": 1.5024,
+      "step": 40543
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0802465233881164e-06,
+      "loss": 2.433,
+      "step": 40544
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0801991150442479e-06,
+      "loss": 2.0063,
+      "step": 40545
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0801517067003793e-06,
+      "loss": 1.7758,
+      "step": 40546
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0801042983565106e-06,
+      "loss": 1.1476,
+      "step": 40547
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0800568900126423e-06,
+      "loss": 2.3922,
+      "step": 40548
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0800094816687738e-06,
+      "loss": 1.6752,
+      "step": 40549
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.079962073324905e-06,
+      "loss": 2.1609,
+      "step": 40550
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0799146649810366e-06,
+      "loss": 1.4547,
+      "step": 40551
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0798672566371683e-06,
+      "loss": 2.4893,
+      "step": 40552
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0798198482932996e-06,
+      "loss": 1.8659,
+      "step": 40553
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.079772439949431e-06,
+      "loss": 1.8789,
+      "step": 40554
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0797250316055628e-06,
+      "loss": 1.658,
+      "step": 40555
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.079677623261694e-06,
+      "loss": 1.5078,
+      "step": 40556
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0796302149178255e-06,
+      "loss": 1.8844,
+      "step": 40557
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.079582806573957e-06,
+      "loss": 1.8845,
+      "step": 40558
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0795353982300885e-06,
+      "loss": 0.2717,
+      "step": 40559
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.07948798988622e-06,
+      "loss": 2.3475,
+      "step": 40560
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.0794405815423515e-06,
+      "loss": 0.4731,
+      "step": 40561
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 1.079393173198483e-06,
+      "loss": 1.1047,
+      "step": 40562
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0793457648546145e-06,
+      "loss": 2.7232,
+      "step": 40563
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0792983565107458e-06,
+      "loss": 2.1396,
+      "step": 40564
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0792509481668775e-06,
+      "loss": 1.2496,
+      "step": 40565
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.079203539823009e-06,
+      "loss": 1.8663,
+      "step": 40566
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0791561314791403e-06,
+      "loss": 2.3214,
+      "step": 40567
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.079108723135272e-06,
+      "loss": 0.6647,
+      "step": 40568
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0790613147914032e-06,
+      "loss": 2.4327,
+      "step": 40569
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0790139064475347e-06,
+      "loss": 2.7973,
+      "step": 40570
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0789664981036662e-06,
+      "loss": 2.8316,
+      "step": 40571
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0789190897597977e-06,
+      "loss": 1.9272,
+      "step": 40572
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0788716814159292e-06,
+      "loss": 1.9211,
+      "step": 40573
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0788242730720607e-06,
+      "loss": 2.1019,
+      "step": 40574
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0787768647281922e-06,
+      "loss": 1.4131,
+      "step": 40575
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0787294563843237e-06,
+      "loss": 1.396,
+      "step": 40576
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0786820480404552e-06,
+      "loss": 1.9206,
+      "step": 40577
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0786346396965867e-06,
+      "loss": 2.3748,
+      "step": 40578
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0785872313527182e-06,
+      "loss": 1.5226,
+      "step": 40579
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0785398230088494e-06,
+      "loss": 2.6095,
+      "step": 40580
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0784924146649811e-06,
+      "loss": 3.2765,
+      "step": 40581
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0784450063211126e-06,
+      "loss": 0.8488,
+      "step": 40582
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.078397597977244e-06,
+      "loss": 2.352,
+      "step": 40583
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0783501896333754e-06,
+      "loss": 2.3062,
+      "step": 40584
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0783027812895071e-06,
+      "loss": 1.5415,
+      "step": 40585
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0782553729456384e-06,
+      "loss": 3.5684,
+      "step": 40586
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0782079646017699e-06,
+      "loss": 1.6505,
+      "step": 40587
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0781605562579016e-06,
+      "loss": 1.7289,
+      "step": 40588
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0781131479140329e-06,
+      "loss": 1.2187,
+      "step": 40589
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0780657395701644e-06,
+      "loss": 2.5057,
+      "step": 40590
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0780183312262959e-06,
+      "loss": 1.3725,
+      "step": 40591
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0779709228824274e-06,
+      "loss": 2.5034,
+      "step": 40592
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0779235145385588e-06,
+      "loss": 2.5975,
+      "step": 40593
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0778761061946903e-06,
+      "loss": 1.0436,
+      "step": 40594
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0778286978508218e-06,
+      "loss": 2.8519,
+      "step": 40595
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0777812895069533e-06,
+      "loss": 2.6099,
+      "step": 40596
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0777338811630846e-06,
+      "loss": 2.0282,
+      "step": 40597
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0776864728192163e-06,
+      "loss": 1.1761,
+      "step": 40598
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0776390644753478e-06,
+      "loss": 0.9807,
+      "step": 40599
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077591656131479e-06,
+      "loss": 2.7009,
+      "step": 40600
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0775442477876108e-06,
+      "loss": 2.9377,
+      "step": 40601
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077496839443742e-06,
+      "loss": 1.596,
+      "step": 40602
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0774494310998736e-06,
+      "loss": 1.7034,
+      "step": 40603
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077402022756005e-06,
+      "loss": 1.879,
+      "step": 40604
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0773546144121365e-06,
+      "loss": 3.1185,
+      "step": 40605
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077307206068268e-06,
+      "loss": 1.6209,
+      "step": 40606
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0772597977243995e-06,
+      "loss": 0.7195,
+      "step": 40607
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077212389380531e-06,
+      "loss": 1.3093,
+      "step": 40608
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0771649810366625e-06,
+      "loss": 1.7911,
+      "step": 40609
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077117572692794e-06,
+      "loss": 2.5623,
+      "step": 40610
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0770701643489255e-06,
+      "loss": 0.9203,
+      "step": 40611
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.077022756005057e-06,
+      "loss": 2.316,
+      "step": 40612
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0769753476611883e-06,
+      "loss": 2.5558,
+      "step": 40613
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.07692793931732e-06,
+      "loss": 2.5657,
+      "step": 40614
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0768805309734515e-06,
+      "loss": 2.0523,
+      "step": 40615
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0768331226295827e-06,
+      "loss": 3.5053,
+      "step": 40616
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0767857142857142e-06,
+      "loss": 0.9766,
+      "step": 40617
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.076738305941846e-06,
+      "loss": 1.4773,
+      "step": 40618
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0766908975979772e-06,
+      "loss": 2.3261,
+      "step": 40619
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0766434892541087e-06,
+      "loss": 1.9445,
+      "step": 40620
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0765960809102404e-06,
+      "loss": 3.1563,
+      "step": 40621
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0765486725663717e-06,
+      "loss": 0.6403,
+      "step": 40622
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0765012642225032e-06,
+      "loss": 1.9935,
+      "step": 40623
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0764538558786347e-06,
+      "loss": 2.5072,
+      "step": 40624
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0764064475347662e-06,
+      "loss": 1.378,
+      "step": 40625
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 1.0763590391908977e-06,
+      "loss": 2.1903,
+      "step": 40626
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0763116308470292e-06,
+      "loss": 4.1062,
+      "step": 40627
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0762642225031606e-06,
+      "loss": 3.1621,
+      "step": 40628
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0762168141592921e-06,
+      "loss": 1.2845,
+      "step": 40629
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0761694058154234e-06,
+      "loss": 1.8511,
+      "step": 40630
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0761219974715551e-06,
+      "loss": 2.0113,
+      "step": 40631
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0760745891276866e-06,
+      "loss": 3.4039,
+      "step": 40632
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.076027180783818e-06,
+      "loss": 1.5703,
+      "step": 40633
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0759797724399496e-06,
+      "loss": 1.4989,
+      "step": 40634
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0759323640960809e-06,
+      "loss": 1.6055,
+      "step": 40635
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0758849557522124e-06,
+      "loss": 1.1662,
+      "step": 40636
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0758375474083439e-06,
+      "loss": 2.3069,
+      "step": 40637
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0757901390644754e-06,
+      "loss": 2.6393,
+      "step": 40638
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0757427307206068e-06,
+      "loss": 2.133,
+      "step": 40639
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0756953223767383e-06,
+      "loss": 2.4953,
+      "step": 40640
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0756479140328698e-06,
+      "loss": 4.2564,
+      "step": 40641
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0756005056890013e-06,
+      "loss": 3.2724,
+      "step": 40642
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0755530973451328e-06,
+      "loss": 2.97,
+      "step": 40643
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0755056890012643e-06,
+      "loss": 2.894,
+      "step": 40644
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0754582806573958e-06,
+      "loss": 1.5295,
+      "step": 40645
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.075410872313527e-06,
+      "loss": 1.3691,
+      "step": 40646
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0753634639696586e-06,
+      "loss": 1.4261,
+      "step": 40647
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0753160556257903e-06,
+      "loss": 3.1108,
+      "step": 40648
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0752686472819216e-06,
+      "loss": 2.956,
+      "step": 40649
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.075221238938053e-06,
+      "loss": 2.1695,
+      "step": 40650
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0751738305941848e-06,
+      "loss": 2.9819,
+      "step": 40651
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.075126422250316e-06,
+      "loss": 0.9834,
+      "step": 40652
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0750790139064475e-06,
+      "loss": 1.2399,
+      "step": 40653
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.075031605562579e-06,
+      "loss": 2.6819,
+      "step": 40654
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0749841972187105e-06,
+      "loss": 1.5672,
+      "step": 40655
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.074936788874842e-06,
+      "loss": 1.8067,
+      "step": 40656
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0748893805309735e-06,
+      "loss": 1.8329,
+      "step": 40657
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.074841972187105e-06,
+      "loss": 1.7068,
+      "step": 40658
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0747945638432365e-06,
+      "loss": 2.1667,
+      "step": 40659
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0747471554993678e-06,
+      "loss": 1.4632,
+      "step": 40660
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0746997471554995e-06,
+      "loss": 1.9091,
+      "step": 40661
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.074652338811631e-06,
+      "loss": 2.4546,
+      "step": 40662
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0746049304677622e-06,
+      "loss": 2.5627,
+      "step": 40663
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.074557522123894e-06,
+      "loss": 1.1406,
+      "step": 40664
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0745101137800254e-06,
+      "loss": 1.2834,
+      "step": 40665
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0744627054361567e-06,
+      "loss": 3.4074,
+      "step": 40666
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0744152970922882e-06,
+      "loss": 3.3703,
+      "step": 40667
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0743678887484197e-06,
+      "loss": 2.1322,
+      "step": 40668
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0743204804045512e-06,
+      "loss": 1.5582,
+      "step": 40669
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0742730720606827e-06,
+      "loss": 5.1529,
+      "step": 40670
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0742256637168142e-06,
+      "loss": 1.6039,
+      "step": 40671
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0741782553729457e-06,
+      "loss": 1.3301,
+      "step": 40672
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0741308470290772e-06,
+      "loss": 1.9503,
+      "step": 40673
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0740834386852087e-06,
+      "loss": 0.8461,
+      "step": 40674
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0740360303413401e-06,
+      "loss": 2.8071,
+      "step": 40675
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0739886219974716e-06,
+      "loss": 2.292,
+      "step": 40676
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0739412136536031e-06,
+      "loss": 3.2577,
+      "step": 40677
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0738938053097346e-06,
+      "loss": 1.9597,
+      "step": 40678
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.073846396965866e-06,
+      "loss": 1.9476,
+      "step": 40679
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0737989886219974e-06,
+      "loss": 0.9552,
+      "step": 40680
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.073751580278129e-06,
+      "loss": 2.1518,
+      "step": 40681
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0737041719342604e-06,
+      "loss": 0.3909,
+      "step": 40682
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0736567635903919e-06,
+      "loss": 3.0659,
+      "step": 40683
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0736093552465236e-06,
+      "loss": 1.8227,
+      "step": 40684
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0735619469026549e-06,
+      "loss": 0.5383,
+      "step": 40685
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0735145385587863e-06,
+      "loss": 2.5366,
+      "step": 40686
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0734671302149178e-06,
+      "loss": 1.2428,
+      "step": 40687
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0734197218710493e-06,
+      "loss": 1.3578,
+      "step": 40688
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 1.0733723135271808e-06,
+      "loss": 0.6373,
+      "step": 40689
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0733249051833123e-06,
+      "loss": 2.4726,
+      "step": 40690
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0732774968394438e-06,
+      "loss": 1.1389,
+      "step": 40691
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0732300884955753e-06,
+      "loss": 1.5655,
+      "step": 40692
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0731826801517066e-06,
+      "loss": 2.4651,
+      "step": 40693
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0731352718078383e-06,
+      "loss": 1.9092,
+      "step": 40694
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0730878634639698e-06,
+      "loss": 3.4756,
+      "step": 40695
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.073040455120101e-06,
+      "loss": 1.3521,
+      "step": 40696
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0729930467762328e-06,
+      "loss": 2.0279,
+      "step": 40697
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0729456384323643e-06,
+      "loss": 2.1503,
+      "step": 40698
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0728982300884955e-06,
+      "loss": 4.1111,
+      "step": 40699
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.072850821744627e-06,
+      "loss": 3.093,
+      "step": 40700
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0728034134007585e-06,
+      "loss": 1.1598,
+      "step": 40701
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.07275600505689e-06,
+      "loss": 0.8637,
+      "step": 40702
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0727085967130215e-06,
+      "loss": 2.36,
+      "step": 40703
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.072661188369153e-06,
+      "loss": 2.7841,
+      "step": 40704
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0726137800252845e-06,
+      "loss": 2.1457,
+      "step": 40705
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.072566371681416e-06,
+      "loss": 1.6902,
+      "step": 40706
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0725189633375475e-06,
+      "loss": 3.1705,
+      "step": 40707
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.072471554993679e-06,
+      "loss": 2.8899,
+      "step": 40708
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0724241466498102e-06,
+      "loss": 2.0214,
+      "step": 40709
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.072376738305942e-06,
+      "loss": 3.0533,
+      "step": 40710
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0723293299620734e-06,
+      "loss": 1.9483,
+      "step": 40711
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0722819216182047e-06,
+      "loss": 2.0978,
+      "step": 40712
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0722345132743362e-06,
+      "loss": 1.3995,
+      "step": 40713
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.072187104930468e-06,
+      "loss": 1.961,
+      "step": 40714
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0721396965865992e-06,
+      "loss": 1.0605,
+      "step": 40715
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0720922882427307e-06,
+      "loss": 5.0675,
+      "step": 40716
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0720448798988624e-06,
+      "loss": 2.4254,
+      "step": 40717
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0719974715549937e-06,
+      "loss": 2.4768,
+      "step": 40718
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0719500632111252e-06,
+      "loss": 0.7313,
+      "step": 40719
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0719026548672567e-06,
+      "loss": 1.6298,
+      "step": 40720
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0718552465233881e-06,
+      "loss": 1.3441,
+      "step": 40721
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0718078381795196e-06,
+      "loss": 2.344,
+      "step": 40722
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0717604298356511e-06,
+      "loss": 3.58,
+      "step": 40723
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0717130214917826e-06,
+      "loss": 3.7697,
+      "step": 40724
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0716656131479141e-06,
+      "loss": 1.5155,
+      "step": 40725
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0716182048040454e-06,
+      "loss": 1.286,
+      "step": 40726
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.071570796460177e-06,
+      "loss": 2.971,
+      "step": 40727
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0715233881163086e-06,
+      "loss": 3.9887,
+      "step": 40728
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0714759797724399e-06,
+      "loss": 2.3581,
+      "step": 40729
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0714285714285716e-06,
+      "loss": 0.8718,
+      "step": 40730
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0713811630847029e-06,
+      "loss": 1.5331,
+      "step": 40731
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0713337547408343e-06,
+      "loss": 3.3666,
+      "step": 40732
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0712863463969658e-06,
+      "loss": 1.1878,
+      "step": 40733
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0712389380530973e-06,
+      "loss": 0.5513,
+      "step": 40734
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0711915297092288e-06,
+      "loss": 1.1732,
+      "step": 40735
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0711441213653603e-06,
+      "loss": 0.9617,
+      "step": 40736
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0710967130214918e-06,
+      "loss": 3.1254,
+      "step": 40737
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0710493046776233e-06,
+      "loss": 3.2503,
+      "step": 40738
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0710018963337548e-06,
+      "loss": 1.5981,
+      "step": 40739
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0709544879898863e-06,
+      "loss": 1.7052,
+      "step": 40740
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0709070796460178e-06,
+      "loss": 1.9891,
+      "step": 40741
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.070859671302149e-06,
+      "loss": 3.0786,
+      "step": 40742
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0708122629582808e-06,
+      "loss": 1.8005,
+      "step": 40743
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0707648546144123e-06,
+      "loss": 2.7358,
+      "step": 40744
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0707174462705435e-06,
+      "loss": 1.9595,
+      "step": 40745
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.070670037926675e-06,
+      "loss": 1.3761,
+      "step": 40746
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0706226295828067e-06,
+      "loss": 1.6391,
+      "step": 40747
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.070575221238938e-06,
+      "loss": 0.5498,
+      "step": 40748
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0705278128950695e-06,
+      "loss": 2.494,
+      "step": 40749
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0704804045512012e-06,
+      "loss": 1.7359,
+      "step": 40750
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.0704329962073325e-06,
+      "loss": 2.1343,
+      "step": 40751
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 1.070385587863464e-06,
+      "loss": 1.3809,
+      "step": 40752
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0703381795195955e-06,
+      "loss": 2.6879,
+      "step": 40753
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.070290771175727e-06,
+      "loss": 2.2457,
+      "step": 40754
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0702433628318585e-06,
+      "loss": 2.4074,
+      "step": 40755
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.07019595448799e-06,
+      "loss": 1.8519,
+      "step": 40756
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0701485461441214e-06,
+      "loss": 2.4989,
+      "step": 40757
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.070101137800253e-06,
+      "loss": 1.8498,
+      "step": 40758
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0700537294563842e-06,
+      "loss": 1.9627,
+      "step": 40759
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.070006321112516e-06,
+      "loss": 2.1281,
+      "step": 40760
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0699589127686474e-06,
+      "loss": 1.9039,
+      "step": 40761
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0699115044247787e-06,
+      "loss": 1.518,
+      "step": 40762
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0698640960809104e-06,
+      "loss": 3.5148,
+      "step": 40763
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0698166877370417e-06,
+      "loss": 1.3806,
+      "step": 40764
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0697692793931732e-06,
+      "loss": 1.6575,
+      "step": 40765
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0697218710493047e-06,
+      "loss": 2.3539,
+      "step": 40766
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0696744627054362e-06,
+      "loss": 2.0694,
+      "step": 40767
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0696270543615676e-06,
+      "loss": 2.627,
+      "step": 40768
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0695796460176991e-06,
+      "loss": 1.4265,
+      "step": 40769
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0695322376738306e-06,
+      "loss": 2.0502,
+      "step": 40770
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0694848293299621e-06,
+      "loss": 0.9749,
+      "step": 40771
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0694374209860936e-06,
+      "loss": 1.4676,
+      "step": 40772
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.069390012642225e-06,
+      "loss": 1.4858,
+      "step": 40773
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0693426042983566e-06,
+      "loss": 1.6123,
+      "step": 40774
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0692951959544879e-06,
+      "loss": 1.0307,
+      "step": 40775
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0692477876106196e-06,
+      "loss": 1.2794,
+      "step": 40776
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.069200379266751e-06,
+      "loss": 1.2275,
+      "step": 40777
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0691529709228824e-06,
+      "loss": 2.8482,
+      "step": 40778
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0691055625790138e-06,
+      "loss": 3.3794,
+      "step": 40779
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0690581542351456e-06,
+      "loss": 2.2946,
+      "step": 40780
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0690107458912768e-06,
+      "loss": 1.7691,
+      "step": 40781
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0689633375474083e-06,
+      "loss": 2.2268,
+      "step": 40782
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.06891592920354e-06,
+      "loss": 2.3366,
+      "step": 40783
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0688685208596713e-06,
+      "loss": 2.7942,
+      "step": 40784
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0688211125158028e-06,
+      "loss": 3.0903,
+      "step": 40785
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0687737041719343e-06,
+      "loss": 2.1254,
+      "step": 40786
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0687262958280658e-06,
+      "loss": 2.7006,
+      "step": 40787
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0686788874841973e-06,
+      "loss": 1.4185,
+      "step": 40788
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0686314791403286e-06,
+      "loss": 1.7415,
+      "step": 40789
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0685840707964603e-06,
+      "loss": 1.3717,
+      "step": 40790
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0685366624525918e-06,
+      "loss": 2.2314,
+      "step": 40791
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.068489254108723e-06,
+      "loss": 3.2768,
+      "step": 40792
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0684418457648547e-06,
+      "loss": 2.3186,
+      "step": 40793
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0683944374209862e-06,
+      "loss": 1.7979,
+      "step": 40794
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0683470290771175e-06,
+      "loss": 0.607,
+      "step": 40795
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.068299620733249e-06,
+      "loss": 0.5053,
+      "step": 40796
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0682522123893805e-06,
+      "loss": 2.6963,
+      "step": 40797
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.068204804045512e-06,
+      "loss": 2.2747,
+      "step": 40798
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0681573957016435e-06,
+      "loss": 1.9239,
+      "step": 40799
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.068109987357775e-06,
+      "loss": 2.3299,
+      "step": 40800
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0680625790139065e-06,
+      "loss": 1.1113,
+      "step": 40801
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.068015170670038e-06,
+      "loss": 3.6908,
+      "step": 40802
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0679677623261694e-06,
+      "loss": 2.6112,
+      "step": 40803
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.067920353982301e-06,
+      "loss": 1.5241,
+      "step": 40804
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0678729456384324e-06,
+      "loss": 1.7669,
+      "step": 40805
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.067825537294564e-06,
+      "loss": 3.1663,
+      "step": 40806
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0677781289506954e-06,
+      "loss": 1.1047,
+      "step": 40807
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0677307206068267e-06,
+      "loss": 2.1896,
+      "step": 40808
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0676833122629582e-06,
+      "loss": 2.3441,
+      "step": 40809
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0676359039190899e-06,
+      "loss": 1.3969,
+      "step": 40810
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0675884955752212e-06,
+      "loss": 2.1086,
+      "step": 40811
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0675410872313527e-06,
+      "loss": 1.6345,
+      "step": 40812
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0674936788874844e-06,
+      "loss": 0.3987,
+      "step": 40813
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0674462705436156e-06,
+      "loss": 1.2882,
+      "step": 40814
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0673988621997471e-06,
+      "loss": 1.1602,
+      "step": 40815
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 1.0673514538558786e-06,
+      "loss": 3.2935,
+      "step": 40816
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0673040455120101e-06,
+      "loss": 2.5539,
+      "step": 40817
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0672566371681416e-06,
+      "loss": 2.806,
+      "step": 40818
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0672092288242731e-06,
+      "loss": 1.3636,
+      "step": 40819
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0671618204804046e-06,
+      "loss": 1.4186,
+      "step": 40820
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.067114412136536e-06,
+      "loss": 1.515,
+      "step": 40821
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0670670037926674e-06,
+      "loss": 2.6723,
+      "step": 40822
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.067019595448799e-06,
+      "loss": 1.2524,
+      "step": 40823
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0669721871049306e-06,
+      "loss": 1.8861,
+      "step": 40824
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0669247787610619e-06,
+      "loss": 2.037,
+      "step": 40825
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0668773704171936e-06,
+      "loss": 2.2583,
+      "step": 40826
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.066829962073325e-06,
+      "loss": 3.0093,
+      "step": 40827
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0667825537294563e-06,
+      "loss": 2.9518,
+      "step": 40828
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0667351453855878e-06,
+      "loss": 1.8428,
+      "step": 40829
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0666877370417193e-06,
+      "loss": 3.1249,
+      "step": 40830
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0666403286978508e-06,
+      "loss": 1.1702,
+      "step": 40831
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0665929203539823e-06,
+      "loss": 2.2428,
+      "step": 40832
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0665455120101138e-06,
+      "loss": 0.7025,
+      "step": 40833
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0664981036662453e-06,
+      "loss": 1.8023,
+      "step": 40834
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0664506953223768e-06,
+      "loss": 1.6439,
+      "step": 40835
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0664032869785083e-06,
+      "loss": 2.275,
+      "step": 40836
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0663558786346398e-06,
+      "loss": 2.9046,
+      "step": 40837
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0663084702907712e-06,
+      "loss": 3.1368,
+      "step": 40838
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0662610619469027e-06,
+      "loss": 1.3349,
+      "step": 40839
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0662136536030342e-06,
+      "loss": 1.4965,
+      "step": 40840
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0661662452591655e-06,
+      "loss": 1.9989,
+      "step": 40841
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.066118836915297e-06,
+      "loss": 2.0449,
+      "step": 40842
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0660714285714287e-06,
+      "loss": 2.3731,
+      "step": 40843
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.06602402022756e-06,
+      "loss": 3.3044,
+      "step": 40844
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0659766118836915e-06,
+      "loss": 1.9053,
+      "step": 40845
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0659292035398232e-06,
+      "loss": 1.4351,
+      "step": 40846
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0658817951959545e-06,
+      "loss": 1.9556,
+      "step": 40847
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.065834386852086e-06,
+      "loss": 2.2825,
+      "step": 40848
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0657869785082175e-06,
+      "loss": 2.6558,
+      "step": 40849
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.065739570164349e-06,
+      "loss": 1.8587,
+      "step": 40850
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0656921618204804e-06,
+      "loss": 1.7323,
+      "step": 40851
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.065644753476612e-06,
+      "loss": 1.8456,
+      "step": 40852
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0655973451327434e-06,
+      "loss": 1.8098,
+      "step": 40853
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.065549936788875e-06,
+      "loss": 3.8678,
+      "step": 40854
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0655025284450062e-06,
+      "loss": 2.1441,
+      "step": 40855
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.065455120101138e-06,
+      "loss": 1.9474,
+      "step": 40856
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0654077117572694e-06,
+      "loss": 2.7776,
+      "step": 40857
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0653603034134007e-06,
+      "loss": 2.9475,
+      "step": 40858
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0653128950695324e-06,
+      "loss": 1.0934,
+      "step": 40859
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0652654867256639e-06,
+      "loss": 1.91,
+      "step": 40860
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0652180783817951e-06,
+      "loss": 2.1543,
+      "step": 40861
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0651706700379266e-06,
+      "loss": 3.2136,
+      "step": 40862
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0651232616940581e-06,
+      "loss": 2.0949,
+      "step": 40863
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0650758533501896e-06,
+      "loss": 1.0154,
+      "step": 40864
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0650284450063211e-06,
+      "loss": 0.9002,
+      "step": 40865
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0649810366624526e-06,
+      "loss": 1.738,
+      "step": 40866
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.064933628318584e-06,
+      "loss": 1.9593,
+      "step": 40867
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0648862199747156e-06,
+      "loss": 2.3913,
+      "step": 40868
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.064838811630847e-06,
+      "loss": 2.8133,
+      "step": 40869
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0647914032869786e-06,
+      "loss": 3.5013,
+      "step": 40870
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.06474399494311e-06,
+      "loss": 2.7849,
+      "step": 40871
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0646965865992416e-06,
+      "loss": 3.9018,
+      "step": 40872
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.064649178255373e-06,
+      "loss": 0.9971,
+      "step": 40873
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0646017699115043e-06,
+      "loss": 3.8602,
+      "step": 40874
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0645543615676358e-06,
+      "loss": 2.9936,
+      "step": 40875
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0645069532237675e-06,
+      "loss": 1.8638,
+      "step": 40876
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0644595448798988e-06,
+      "loss": 2.1296,
+      "step": 40877
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.0644121365360303e-06,
+      "loss": 2.0197,
+      "step": 40878
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 1.064364728192162e-06,
+      "loss": 4.4239,
+      "step": 40879
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0643173198482933e-06,
+      "loss": 2.4149,
+      "step": 40880
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0642699115044248e-06,
+      "loss": 1.0565,
+      "step": 40881
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0642225031605563e-06,
+      "loss": 2.6954,
+      "step": 40882
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0641750948166878e-06,
+      "loss": 2.8399,
+      "step": 40883
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0641276864728193e-06,
+      "loss": 2.1064,
+      "step": 40884
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0640802781289507e-06,
+      "loss": 0.9015,
+      "step": 40885
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0640328697850822e-06,
+      "loss": 1.4347,
+      "step": 40886
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0639854614412137e-06,
+      "loss": 3.1505,
+      "step": 40887
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.063938053097345e-06,
+      "loss": 1.6507,
+      "step": 40888
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0638906447534767e-06,
+      "loss": 3.0437,
+      "step": 40889
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0638432364096082e-06,
+      "loss": 2.5257,
+      "step": 40890
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0637958280657395e-06,
+      "loss": 1.1484,
+      "step": 40891
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0637484197218712e-06,
+      "loss": 2.9283,
+      "step": 40892
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0637010113780027e-06,
+      "loss": 1.4758,
+      "step": 40893
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.063653603034134e-06,
+      "loss": 3.0142,
+      "step": 40894
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0636061946902655e-06,
+      "loss": 2.0399,
+      "step": 40895
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.063558786346397e-06,
+      "loss": 2.3756,
+      "step": 40896
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0635113780025284e-06,
+      "loss": 2.1182,
+      "step": 40897
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.06346396965866e-06,
+      "loss": 2.1794,
+      "step": 40898
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0634165613147914e-06,
+      "loss": 2.5266,
+      "step": 40899
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.063369152970923e-06,
+      "loss": 3.6833,
+      "step": 40900
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0633217446270544e-06,
+      "loss": 2.6393,
+      "step": 40901
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.063274336283186e-06,
+      "loss": 2.0269,
+      "step": 40902
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0632269279393174e-06,
+      "loss": 2.3301,
+      "step": 40903
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0631795195954489e-06,
+      "loss": 1.8034,
+      "step": 40904
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0631321112515804e-06,
+      "loss": 3.4709,
+      "step": 40905
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0630847029077119e-06,
+      "loss": 1.5063,
+      "step": 40906
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0630372945638432e-06,
+      "loss": 0.6865,
+      "step": 40907
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0629898862199746e-06,
+      "loss": 1.695,
+      "step": 40908
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0629424778761063e-06,
+      "loss": 1.7143,
+      "step": 40909
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0628950695322376e-06,
+      "loss": 2.1982,
+      "step": 40910
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0628476611883691e-06,
+      "loss": 2.301,
+      "step": 40911
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0628002528445008e-06,
+      "loss": 1.676,
+      "step": 40912
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.062752844500632e-06,
+      "loss": 1.9624,
+      "step": 40913
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0627054361567636e-06,
+      "loss": 3.3468,
+      "step": 40914
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.062658027812895e-06,
+      "loss": 2.1209,
+      "step": 40915
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0626106194690266e-06,
+      "loss": 1.9411,
+      "step": 40916
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.062563211125158e-06,
+      "loss": 1.9004,
+      "step": 40917
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0625158027812894e-06,
+      "loss": 2.5999,
+      "step": 40918
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.062468394437421e-06,
+      "loss": 1.6555,
+      "step": 40919
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0624209860935525e-06,
+      "loss": 1.5357,
+      "step": 40920
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0623735777496838e-06,
+      "loss": 2.1713,
+      "step": 40921
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0623261694058155e-06,
+      "loss": 3.8966,
+      "step": 40922
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.062278761061947e-06,
+      "loss": 2.7509,
+      "step": 40923
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0622313527180783e-06,
+      "loss": 1.3317,
+      "step": 40924
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0621839443742098e-06,
+      "loss": 1.2165,
+      "step": 40925
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0621365360303415e-06,
+      "loss": 0.995,
+      "step": 40926
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0620891276864728e-06,
+      "loss": 1.3274,
+      "step": 40927
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0620417193426043e-06,
+      "loss": 2.8654,
+      "step": 40928
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0619943109987358e-06,
+      "loss": 1.6225,
+      "step": 40929
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0619469026548673e-06,
+      "loss": 1.997,
+      "step": 40930
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0618994943109988e-06,
+      "loss": 1.6916,
+      "step": 40931
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0618520859671302e-06,
+      "loss": 2.435,
+      "step": 40932
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0618046776232617e-06,
+      "loss": 2.3932,
+      "step": 40933
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0617572692793932e-06,
+      "loss": 2.3328,
+      "step": 40934
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0617098609355247e-06,
+      "loss": 1.9995,
+      "step": 40935
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0616624525916562e-06,
+      "loss": 2.1952,
+      "step": 40936
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0616150442477877e-06,
+      "loss": 2.3246,
+      "step": 40937
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.061567635903919e-06,
+      "loss": 2.1468,
+      "step": 40938
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0615202275600507e-06,
+      "loss": 2.5074,
+      "step": 40939
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.061472819216182e-06,
+      "loss": 1.2687,
+      "step": 40940
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0614254108723135e-06,
+      "loss": 2.9331,
+      "step": 40941
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 1.0613780025284452e-06,
+      "loss": 1.7126,
+      "step": 40942
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0613305941845764e-06,
+      "loss": 1.5549,
+      "step": 40943
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.061283185840708e-06,
+      "loss": 1.9722,
+      "step": 40944
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0612357774968394e-06,
+      "loss": 2.5187,
+      "step": 40945
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.061188369152971e-06,
+      "loss": 1.7323,
+      "step": 40946
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0611409608091024e-06,
+      "loss": 1.6071,
+      "step": 40947
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.061093552465234e-06,
+      "loss": 1.7163,
+      "step": 40948
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0610461441213654e-06,
+      "loss": 3.2362,
+      "step": 40949
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0609987357774969e-06,
+      "loss": 1.7314,
+      "step": 40950
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0609513274336282e-06,
+      "loss": 2.491,
+      "step": 40951
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0609039190897599e-06,
+      "loss": 2.5576,
+      "step": 40952
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0608565107458914e-06,
+      "loss": 1.8637,
+      "step": 40953
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0608091024020226e-06,
+      "loss": 1.9945,
+      "step": 40954
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0607616940581544e-06,
+      "loss": 1.4335,
+      "step": 40955
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0607142857142858e-06,
+      "loss": 1.4895,
+      "step": 40956
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0606668773704171e-06,
+      "loss": 2.6691,
+      "step": 40957
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0606194690265486e-06,
+      "loss": 3.1934,
+      "step": 40958
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0605720606826803e-06,
+      "loss": 2.122,
+      "step": 40959
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0605246523388116e-06,
+      "loss": 1.9286,
+      "step": 40960
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.060477243994943e-06,
+      "loss": 2.7692,
+      "step": 40961
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0604298356510746e-06,
+      "loss": 3.1022,
+      "step": 40962
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.060382427307206e-06,
+      "loss": 2.2509,
+      "step": 40963
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0603350189633376e-06,
+      "loss": 1.6527,
+      "step": 40964
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.060287610619469e-06,
+      "loss": 3.1624,
+      "step": 40965
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0602402022756006e-06,
+      "loss": 1.2441,
+      "step": 40966
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.060192793931732e-06,
+      "loss": 2.0068,
+      "step": 40967
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0601453855878635e-06,
+      "loss": 1.5366,
+      "step": 40968
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.060097977243995e-06,
+      "loss": 2.8757,
+      "step": 40969
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0600505689001265e-06,
+      "loss": 2.0237,
+      "step": 40970
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0600031605562578e-06,
+      "loss": 2.6086,
+      "step": 40971
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0599557522123895e-06,
+      "loss": 1.9299,
+      "step": 40972
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0599083438685208e-06,
+      "loss": 3.0253,
+      "step": 40973
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0598609355246523e-06,
+      "loss": 2.5346,
+      "step": 40974
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.059813527180784e-06,
+      "loss": 1.744,
+      "step": 40975
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0597661188369153e-06,
+      "loss": 4.3764,
+      "step": 40976
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0597187104930468e-06,
+      "loss": 2.2324,
+      "step": 40977
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0596713021491782e-06,
+      "loss": 2.1284,
+      "step": 40978
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0596238938053097e-06,
+      "loss": 1.1587,
+      "step": 40979
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0595764854614412e-06,
+      "loss": 2.452,
+      "step": 40980
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0595290771175727e-06,
+      "loss": 1.892,
+      "step": 40981
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0594816687737042e-06,
+      "loss": 2.5818,
+      "step": 40982
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0594342604298357e-06,
+      "loss": 2.6842,
+      "step": 40983
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.059386852085967e-06,
+      "loss": 1.6516,
+      "step": 40984
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0593394437420987e-06,
+      "loss": 2.5066,
+      "step": 40985
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0592920353982302e-06,
+      "loss": 1.5209,
+      "step": 40986
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0592446270543615e-06,
+      "loss": 4.3523,
+      "step": 40987
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0591972187104932e-06,
+      "loss": 2.0707,
+      "step": 40988
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0591498103666247e-06,
+      "loss": 2.3812,
+      "step": 40989
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.059102402022756e-06,
+      "loss": 1.724,
+      "step": 40990
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0590549936788874e-06,
+      "loss": 2.0403,
+      "step": 40991
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0590075853350191e-06,
+      "loss": 2.2282,
+      "step": 40992
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0589601769911504e-06,
+      "loss": 0.9288,
+      "step": 40993
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.058912768647282e-06,
+      "loss": 5.0183,
+      "step": 40994
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0588653603034134e-06,
+      "loss": 1.7522,
+      "step": 40995
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.058817951959545e-06,
+      "loss": 1.9666,
+      "step": 40996
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0587705436156764e-06,
+      "loss": 2.9463,
+      "step": 40997
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0587231352718079e-06,
+      "loss": 1.9073,
+      "step": 40998
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0586757269279394e-06,
+      "loss": 1.9879,
+      "step": 40999
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0586283185840709e-06,
+      "loss": 2.6299,
+      "step": 41000
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0585809102402024e-06,
+      "loss": 3.3814,
+      "step": 41001
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0585335018963338e-06,
+      "loss": 2.3335,
+      "step": 41002
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0584860935524653e-06,
+      "loss": 1.2672,
+      "step": 41003
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0584386852085966e-06,
+      "loss": 1.6464,
+      "step": 41004
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0583912768647283e-06,
+      "loss": 1.2146,
+      "step": 41005
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 1.0583438685208596e-06,
+      "loss": 3.4222,
+      "step": 41006
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.058296460176991e-06,
+      "loss": 2.2518,
+      "step": 41007
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0582490518331228e-06,
+      "loss": 0.8885,
+      "step": 41008
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.058201643489254e-06,
+      "loss": 2.8694,
+      "step": 41009
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0581542351453856e-06,
+      "loss": 2.053,
+      "step": 41010
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.058106826801517e-06,
+      "loss": 2.8574,
+      "step": 41011
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0580594184576486e-06,
+      "loss": 1.491,
+      "step": 41012
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.05801201011378e-06,
+      "loss": 1.0786,
+      "step": 41013
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0579646017699115e-06,
+      "loss": 2.0273,
+      "step": 41014
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.057917193426043e-06,
+      "loss": 2.9163,
+      "step": 41015
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0578697850821745e-06,
+      "loss": 1.7937,
+      "step": 41016
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0578223767383058e-06,
+      "loss": 2.3005,
+      "step": 41017
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0577749683944375e-06,
+      "loss": 1.8634,
+      "step": 41018
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.057727560050569e-06,
+      "loss": 2.742,
+      "step": 41019
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0576801517067003e-06,
+      "loss": 3.3339,
+      "step": 41020
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.057632743362832e-06,
+      "loss": 2.3995,
+      "step": 41021
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0575853350189635e-06,
+      "loss": 2.8611,
+      "step": 41022
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0575379266750948e-06,
+      "loss": 2.4717,
+      "step": 41023
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0574905183312263e-06,
+      "loss": 2.6863,
+      "step": 41024
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.057443109987358e-06,
+      "loss": 2.9926,
+      "step": 41025
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0573957016434892e-06,
+      "loss": 1.6799,
+      "step": 41026
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0573482932996207e-06,
+      "loss": 2.5442,
+      "step": 41027
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0573008849557522e-06,
+      "loss": 2.8417,
+      "step": 41028
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0572534766118837e-06,
+      "loss": 1.7305,
+      "step": 41029
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0572060682680152e-06,
+      "loss": 2.9881,
+      "step": 41030
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0571586599241467e-06,
+      "loss": 2.2296,
+      "step": 41031
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0571112515802782e-06,
+      "loss": 0.6599,
+      "step": 41032
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0570638432364097e-06,
+      "loss": 3.9961,
+      "step": 41033
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0570164348925412e-06,
+      "loss": 1.9493,
+      "step": 41034
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0569690265486727e-06,
+      "loss": 1.7092,
+      "step": 41035
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0569216182048042e-06,
+      "loss": 1.6922,
+      "step": 41036
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0568742098609354e-06,
+      "loss": 2.4634,
+      "step": 41037
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0568268015170671e-06,
+      "loss": 1.309,
+      "step": 41038
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0567793931731984e-06,
+      "loss": 1.9834,
+      "step": 41039
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.05673198482933e-06,
+      "loss": 0.9448,
+      "step": 41040
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0566845764854616e-06,
+      "loss": 0.7329,
+      "step": 41041
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.056637168141593e-06,
+      "loss": 1.7742,
+      "step": 41042
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0565897597977244e-06,
+      "loss": 0.5504,
+      "step": 41043
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0565423514538559e-06,
+      "loss": 1.9615,
+      "step": 41044
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0564949431099874e-06,
+      "loss": 2.2702,
+      "step": 41045
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0564475347661189e-06,
+      "loss": 2.9201,
+      "step": 41046
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0564001264222504e-06,
+      "loss": 1.7471,
+      "step": 41047
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0563527180783819e-06,
+      "loss": 1.3026,
+      "step": 41048
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0563053097345133e-06,
+      "loss": 3.6418,
+      "step": 41049
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0562579013906446e-06,
+      "loss": 1.8404,
+      "step": 41050
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0562104930467763e-06,
+      "loss": 2.5816,
+      "step": 41051
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0561630847029078e-06,
+      "loss": 1.7142,
+      "step": 41052
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.056115676359039e-06,
+      "loss": 1.2286,
+      "step": 41053
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0560682680151708e-06,
+      "loss": 2.5865,
+      "step": 41054
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0560208596713023e-06,
+      "loss": 2.3509,
+      "step": 41055
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0559734513274336e-06,
+      "loss": 1.5141,
+      "step": 41056
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.055926042983565e-06,
+      "loss": 0.2698,
+      "step": 41057
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0558786346396968e-06,
+      "loss": 0.6598,
+      "step": 41058
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.055831226295828e-06,
+      "loss": 1.9909,
+      "step": 41059
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0557838179519595e-06,
+      "loss": 1.2867,
+      "step": 41060
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.055736409608091e-06,
+      "loss": 0.7571,
+      "step": 41061
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0556890012642225e-06,
+      "loss": 1.2696,
+      "step": 41062
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.055641592920354e-06,
+      "loss": 2.7904,
+      "step": 41063
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0555941845764855e-06,
+      "loss": 3.1319,
+      "step": 41064
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.055546776232617e-06,
+      "loss": 2.1126,
+      "step": 41065
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0554993678887485e-06,
+      "loss": 1.6802,
+      "step": 41066
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0554519595448798e-06,
+      "loss": 1.651,
+      "step": 41067
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.0554045512010115e-06,
+      "loss": 2.2669,
+      "step": 41068
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 1.055357142857143e-06,
+      "loss": 1.8826,
+      "step": 41069
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0553097345132743e-06,
+      "loss": 2.2024,
+      "step": 41070
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.055262326169406e-06,
+      "loss": 2.6825,
+      "step": 41071
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0552149178255372e-06,
+      "loss": 1.642,
+      "step": 41072
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0551675094816687e-06,
+      "loss": 2.6915,
+      "step": 41073
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0551201011378002e-06,
+      "loss": 2.7738,
+      "step": 41074
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0550726927939317e-06,
+      "loss": 2.0345,
+      "step": 41075
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0550252844500632e-06,
+      "loss": 2.2091,
+      "step": 41076
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0549778761061947e-06,
+      "loss": 2.3466,
+      "step": 41077
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0549304677623262e-06,
+      "loss": 3.385,
+      "step": 41078
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0548830594184577e-06,
+      "loss": 2.0171,
+      "step": 41079
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0548356510745892e-06,
+      "loss": 3.03,
+      "step": 41080
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0547882427307207e-06,
+      "loss": 1.3653,
+      "step": 41081
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0547408343868522e-06,
+      "loss": 2.5565,
+      "step": 41082
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0546934260429834e-06,
+      "loss": 2.0918,
+      "step": 41083
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0546460176991151e-06,
+      "loss": 2.3831,
+      "step": 41084
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0545986093552466e-06,
+      "loss": 1.9665,
+      "step": 41085
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.054551201011378e-06,
+      "loss": 1.5986,
+      "step": 41086
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0545037926675094e-06,
+      "loss": 2.4059,
+      "step": 41087
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0544563843236411e-06,
+      "loss": 3.5978,
+      "step": 41088
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0544089759797724e-06,
+      "loss": 1.057,
+      "step": 41089
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0543615676359039e-06,
+      "loss": 1.921,
+      "step": 41090
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0543141592920356e-06,
+      "loss": 1.7306,
+      "step": 41091
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0542667509481669e-06,
+      "loss": 1.0606,
+      "step": 41092
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0542193426042984e-06,
+      "loss": 2.4248,
+      "step": 41093
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0541719342604299e-06,
+      "loss": 4.0878,
+      "step": 41094
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0541245259165614e-06,
+      "loss": 1.9042,
+      "step": 41095
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0540771175726928e-06,
+      "loss": 3.0332,
+      "step": 41096
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0540297092288243e-06,
+      "loss": 1.8992,
+      "step": 41097
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0539823008849558e-06,
+      "loss": 1.8524,
+      "step": 41098
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0539348925410873e-06,
+      "loss": 3.877,
+      "step": 41099
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0538874841972186e-06,
+      "loss": 2.0028,
+      "step": 41100
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0538400758533503e-06,
+      "loss": 2.67,
+      "step": 41101
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0537926675094818e-06,
+      "loss": 1.4654,
+      "step": 41102
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053745259165613e-06,
+      "loss": 0.9613,
+      "step": 41103
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0536978508217448e-06,
+      "loss": 2.0107,
+      "step": 41104
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053650442477876e-06,
+      "loss": 2.6367,
+      "step": 41105
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0536030341340076e-06,
+      "loss": 2.4291,
+      "step": 41106
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053555625790139e-06,
+      "loss": 2.5792,
+      "step": 41107
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0535082174462705e-06,
+      "loss": 2.9298,
+      "step": 41108
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053460809102402e-06,
+      "loss": 2.035,
+      "step": 41109
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0534134007585335e-06,
+      "loss": 2.1065,
+      "step": 41110
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053365992414665e-06,
+      "loss": 1.4329,
+      "step": 41111
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0533185840707965e-06,
+      "loss": 1.4108,
+      "step": 41112
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053271175726928e-06,
+      "loss": 1.6981,
+      "step": 41113
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0532237673830595e-06,
+      "loss": 1.7535,
+      "step": 41114
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053176359039191e-06,
+      "loss": 2.3964,
+      "step": 41115
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0531289506953223e-06,
+      "loss": 2.3113,
+      "step": 41116
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.053081542351454e-06,
+      "loss": 3.2746,
+      "step": 41117
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0530341340075855e-06,
+      "loss": 2.3607,
+      "step": 41118
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0529867256637167e-06,
+      "loss": 1.0832,
+      "step": 41119
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0529393173198482e-06,
+      "loss": 2.8483,
+      "step": 41120
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.05289190897598e-06,
+      "loss": 2.6902,
+      "step": 41121
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0528445006321112e-06,
+      "loss": 1.5015,
+      "step": 41122
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0527970922882427e-06,
+      "loss": 2.3924,
+      "step": 41123
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0527496839443744e-06,
+      "loss": 1.4394,
+      "step": 41124
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0527022756005057e-06,
+      "loss": 3.9547,
+      "step": 41125
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0526548672566372e-06,
+      "loss": 3.7724,
+      "step": 41126
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0526074589127687e-06,
+      "loss": 3.0295,
+      "step": 41127
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0525600505689002e-06,
+      "loss": 1.981,
+      "step": 41128
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0525126422250317e-06,
+      "loss": 2.7928,
+      "step": 41129
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0524652338811632e-06,
+      "loss": 1.8832,
+      "step": 41130
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0524178255372946e-06,
+      "loss": 1.3705,
+      "step": 41131
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 1.0523704171934261e-06,
+      "loss": 1.7668,
+      "step": 41132
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0523230088495574e-06,
+      "loss": 1.8825,
+      "step": 41133
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0522756005056891e-06,
+      "loss": 3.6179,
+      "step": 41134
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0522281921618206e-06,
+      "loss": 3.1814,
+      "step": 41135
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.052180783817952e-06,
+      "loss": 2.5143,
+      "step": 41136
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0521333754740836e-06,
+      "loss": 2.4753,
+      "step": 41137
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0520859671302149e-06,
+      "loss": 3.0827,
+      "step": 41138
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0520385587863464e-06,
+      "loss": 3.0554,
+      "step": 41139
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0519911504424779e-06,
+      "loss": 1.6733,
+      "step": 41140
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0519437420986094e-06,
+      "loss": 4.0399,
+      "step": 41141
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0518963337547408e-06,
+      "loss": 1.9126,
+      "step": 41142
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0518489254108723e-06,
+      "loss": 1.2971,
+      "step": 41143
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0518015170670038e-06,
+      "loss": 1.4682,
+      "step": 41144
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0517541087231353e-06,
+      "loss": 2.5644,
+      "step": 41145
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0517067003792666e-06,
+      "loss": 1.2653,
+      "step": 41146
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0516592920353983e-06,
+      "loss": 2.7015,
+      "step": 41147
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0516118836915298e-06,
+      "loss": 0.8034,
+      "step": 41148
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.051564475347661e-06,
+      "loss": 2.699,
+      "step": 41149
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0515170670037928e-06,
+      "loss": 1.3183,
+      "step": 41150
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0514696586599243e-06,
+      "loss": 1.5052,
+      "step": 41151
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0514222503160556e-06,
+      "loss": 1.3103,
+      "step": 41152
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.051374841972187e-06,
+      "loss": 1.7334,
+      "step": 41153
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0513274336283188e-06,
+      "loss": 2.1733,
+      "step": 41154
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.05128002528445e-06,
+      "loss": 1.5095,
+      "step": 41155
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0512326169405815e-06,
+      "loss": 1.0304,
+      "step": 41156
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.051185208596713e-06,
+      "loss": 1.8246,
+      "step": 41157
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0511378002528445e-06,
+      "loss": 1.2279,
+      "step": 41158
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.051090391908976e-06,
+      "loss": 2.9036,
+      "step": 41159
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0510429835651075e-06,
+      "loss": 2.1952,
+      "step": 41160
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.050995575221239e-06,
+      "loss": 0.6005,
+      "step": 41161
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0509481668773705e-06,
+      "loss": 1.9211,
+      "step": 41162
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.050900758533502e-06,
+      "loss": 1.8397,
+      "step": 41163
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0508533501896335e-06,
+      "loss": 2.7569,
+      "step": 41164
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.050805941845765e-06,
+      "loss": 2.8454,
+      "step": 41165
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0507585335018962e-06,
+      "loss": 3.6682,
+      "step": 41166
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.050711125158028e-06,
+      "loss": 1.3364,
+      "step": 41167
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0506637168141592e-06,
+      "loss": 4.7574,
+      "step": 41168
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0506163084702907e-06,
+      "loss": 2.3004,
+      "step": 41169
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0505689001264224e-06,
+      "loss": 3.1928,
+      "step": 41170
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0505214917825537e-06,
+      "loss": 1.4375,
+      "step": 41171
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0504740834386852e-06,
+      "loss": 2.295,
+      "step": 41172
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0504266750948167e-06,
+      "loss": 2.3183,
+      "step": 41173
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0503792667509482e-06,
+      "loss": 0.6247,
+      "step": 41174
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0503318584070797e-06,
+      "loss": 2.6922,
+      "step": 41175
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0502844500632112e-06,
+      "loss": 1.6894,
+      "step": 41176
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0502370417193427e-06,
+      "loss": 2.0666,
+      "step": 41177
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0501896333754741e-06,
+      "loss": 2.8489,
+      "step": 41178
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0501422250316054e-06,
+      "loss": 1.5813,
+      "step": 41179
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0500948166877371e-06,
+      "loss": 1.9701,
+      "step": 41180
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0500474083438686e-06,
+      "loss": 1.0083,
+      "step": 41181
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.05e-06,
+      "loss": 2.0944,
+      "step": 41182
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0499525916561316e-06,
+      "loss": 1.9517,
+      "step": 41183
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.049905183312263e-06,
+      "loss": 3.0704,
+      "step": 41184
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0498577749683944e-06,
+      "loss": 1.3752,
+      "step": 41185
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0498103666245259e-06,
+      "loss": 1.5439,
+      "step": 41186
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0497629582806576e-06,
+      "loss": 2.2611,
+      "step": 41187
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0497155499367889e-06,
+      "loss": 1.8549,
+      "step": 41188
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0496681415929203e-06,
+      "loss": 2.024,
+      "step": 41189
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0496207332490518e-06,
+      "loss": 0.7197,
+      "step": 41190
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0495733249051833e-06,
+      "loss": 3.5316,
+      "step": 41191
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0495259165613148e-06,
+      "loss": 2.7118,
+      "step": 41192
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0494785082174463e-06,
+      "loss": 2.8578,
+      "step": 41193
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0494310998735778e-06,
+      "loss": 0.9329,
+      "step": 41194
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0493836915297093e-06,
+      "loss": 1.6975,
+      "step": 41195
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 1.0493362831858406e-06,
+      "loss": 4.1501,
+      "step": 41196
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0492888748419723e-06,
+      "loss": 1.9626,
+      "step": 41197
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0492414664981038e-06,
+      "loss": 1.78,
+      "step": 41198
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.049194058154235e-06,
+      "loss": 2.4024,
+      "step": 41199
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0491466498103668e-06,
+      "loss": 1.0525,
+      "step": 41200
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.049099241466498e-06,
+      "loss": 2.5422,
+      "step": 41201
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0490518331226295e-06,
+      "loss": 2.307,
+      "step": 41202
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.049004424778761e-06,
+      "loss": 3.9659,
+      "step": 41203
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0489570164348925e-06,
+      "loss": 0.8952,
+      "step": 41204
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.048909608091024e-06,
+      "loss": 1.9485,
+      "step": 41205
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0488621997471555e-06,
+      "loss": 2.3431,
+      "step": 41206
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.048814791403287e-06,
+      "loss": 1.2199,
+      "step": 41207
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0487673830594185e-06,
+      "loss": 3.6363,
+      "step": 41208
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.04871997471555e-06,
+      "loss": 2.975,
+      "step": 41209
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0486725663716815e-06,
+      "loss": 2.8236,
+      "step": 41210
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.048625158027813e-06,
+      "loss": 4.4307,
+      "step": 41211
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0485777496839442e-06,
+      "loss": 0.3058,
+      "step": 41212
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.048530341340076e-06,
+      "loss": 1.8195,
+      "step": 41213
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0484829329962074e-06,
+      "loss": 3.2676,
+      "step": 41214
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0484355246523387e-06,
+      "loss": 2.1811,
+      "step": 41215
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0483881163084702e-06,
+      "loss": 3.0827,
+      "step": 41216
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.048340707964602e-06,
+      "loss": 0.6672,
+      "step": 41217
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0482932996207332e-06,
+      "loss": 2.3275,
+      "step": 41218
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0482458912768647e-06,
+      "loss": 1.6674,
+      "step": 41219
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0481984829329964e-06,
+      "loss": 2.7748,
+      "step": 41220
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0481510745891277e-06,
+      "loss": 1.663,
+      "step": 41221
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0481036662452592e-06,
+      "loss": 1.6021,
+      "step": 41222
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0480562579013907e-06,
+      "loss": 2.3965,
+      "step": 41223
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0480088495575221e-06,
+      "loss": 1.6942,
+      "step": 41224
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0479614412136536e-06,
+      "loss": 1.7216,
+      "step": 41225
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0479140328697851e-06,
+      "loss": 2.8801,
+      "step": 41226
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0478666245259166e-06,
+      "loss": 1.416,
+      "step": 41227
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0478192161820481e-06,
+      "loss": 1.1426,
+      "step": 41228
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0477718078381794e-06,
+      "loss": 1.7996,
+      "step": 41229
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.047724399494311e-06,
+      "loss": 2.6656,
+      "step": 41230
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0476769911504426e-06,
+      "loss": 1.7183,
+      "step": 41231
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0476295828065739e-06,
+      "loss": 0.899,
+      "step": 41232
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0475821744627056e-06,
+      "loss": 1.6413,
+      "step": 41233
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0475347661188369e-06,
+      "loss": 1.0059,
+      "step": 41234
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0474873577749683e-06,
+      "loss": 2.388,
+      "step": 41235
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0474399494310998e-06,
+      "loss": 0.4937,
+      "step": 41236
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0473925410872313e-06,
+      "loss": 1.1667,
+      "step": 41237
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0473451327433628e-06,
+      "loss": 3.1097,
+      "step": 41238
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0472977243994943e-06,
+      "loss": 1.3503,
+      "step": 41239
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0472503160556258e-06,
+      "loss": 2.5941,
+      "step": 41240
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0472029077117573e-06,
+      "loss": 2.9383,
+      "step": 41241
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0471554993678888e-06,
+      "loss": 2.8778,
+      "step": 41242
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0471080910240203e-06,
+      "loss": 2.5448,
+      "step": 41243
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0470606826801518e-06,
+      "loss": 2.3937,
+      "step": 41244
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.047013274336283e-06,
+      "loss": 0.8897,
+      "step": 41245
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0469658659924148e-06,
+      "loss": 1.619,
+      "step": 41246
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0469184576485463e-06,
+      "loss": 1.8087,
+      "step": 41247
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0468710493046775e-06,
+      "loss": 1.7707,
+      "step": 41248
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.046823640960809e-06,
+      "loss": 2.1975,
+      "step": 41249
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0467762326169407e-06,
+      "loss": 1.9238,
+      "step": 41250
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.046728824273072e-06,
+      "loss": 1.3599,
+      "step": 41251
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0466814159292035e-06,
+      "loss": 2.0589,
+      "step": 41252
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0466340075853352e-06,
+      "loss": 1.5882,
+      "step": 41253
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0465865992414665e-06,
+      "loss": 4.3133,
+      "step": 41254
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.046539190897598e-06,
+      "loss": 2.203,
+      "step": 41255
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0464917825537295e-06,
+      "loss": 3.0703,
+      "step": 41256
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.046444374209861e-06,
+      "loss": 1.4486,
+      "step": 41257
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.0463969658659925e-06,
+      "loss": 1.257,
+      "step": 41258
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 1.046349557522124e-06,
+      "loss": 1.3995,
+      "step": 41259
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0463021491782554e-06,
+      "loss": 1.5265,
+      "step": 41260
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.046254740834387e-06,
+      "loss": 2.9614,
+      "step": 41261
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0462073324905182e-06,
+      "loss": 1.5286,
+      "step": 41262
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.04615992414665e-06,
+      "loss": 1.3977,
+      "step": 41263
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0461125158027814e-06,
+      "loss": 1.2907,
+      "step": 41264
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0460651074589127e-06,
+      "loss": 1.2417,
+      "step": 41265
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0460176991150444e-06,
+      "loss": 1.2934,
+      "step": 41266
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0459702907711757e-06,
+      "loss": 1.9611,
+      "step": 41267
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0459228824273072e-06,
+      "loss": 2.7047,
+      "step": 41268
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0458754740834387e-06,
+      "loss": 1.5089,
+      "step": 41269
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0458280657395702e-06,
+      "loss": 2.1244,
+      "step": 41270
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0457806573957016e-06,
+      "loss": 1.4257,
+      "step": 41271
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0457332490518331e-06,
+      "loss": 2.9045,
+      "step": 41272
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0456858407079646e-06,
+      "loss": 1.0161,
+      "step": 41273
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0456384323640961e-06,
+      "loss": 2.7767,
+      "step": 41274
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0455910240202276e-06,
+      "loss": 0.8279,
+      "step": 41275
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.045543615676359e-06,
+      "loss": 1.6881,
+      "step": 41276
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0454962073324906e-06,
+      "loss": 1.4497,
+      "step": 41277
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0454487989886219e-06,
+      "loss": 1.6683,
+      "step": 41278
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0454013906447536e-06,
+      "loss": 1.6305,
+      "step": 41279
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.045353982300885e-06,
+      "loss": 2.7135,
+      "step": 41280
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0453065739570164e-06,
+      "loss": 2.5738,
+      "step": 41281
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0452591656131478e-06,
+      "loss": 3.0494,
+      "step": 41282
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0452117572692796e-06,
+      "loss": 2.4275,
+      "step": 41283
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0451643489254108e-06,
+      "loss": 2.2536,
+      "step": 41284
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0451169405815423e-06,
+      "loss": 1.7862,
+      "step": 41285
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.045069532237674e-06,
+      "loss": 1.1717,
+      "step": 41286
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0450221238938053e-06,
+      "loss": 2.2572,
+      "step": 41287
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0449747155499368e-06,
+      "loss": 2.4721,
+      "step": 41288
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0449273072060683e-06,
+      "loss": 1.4298,
+      "step": 41289
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0448798988621998e-06,
+      "loss": 1.4403,
+      "step": 41290
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0448324905183313e-06,
+      "loss": 1.9216,
+      "step": 41291
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0447850821744628e-06,
+      "loss": 3.4888,
+      "step": 41292
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0447376738305943e-06,
+      "loss": 2.554,
+      "step": 41293
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0446902654867258e-06,
+      "loss": 2.6046,
+      "step": 41294
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.044642857142857e-06,
+      "loss": 0.2638,
+      "step": 41295
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0445954487989887e-06,
+      "loss": 3.0901,
+      "step": 41296
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0445480404551202e-06,
+      "loss": 0.8733,
+      "step": 41297
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0445006321112515e-06,
+      "loss": 2.4888,
+      "step": 41298
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0444532237673832e-06,
+      "loss": 1.8808,
+      "step": 41299
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0444058154235145e-06,
+      "loss": 2.1193,
+      "step": 41300
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.044358407079646e-06,
+      "loss": 2.2883,
+      "step": 41301
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0443109987357775e-06,
+      "loss": 2.8369,
+      "step": 41302
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.044263590391909e-06,
+      "loss": 1.2932,
+      "step": 41303
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0442161820480405e-06,
+      "loss": 1.6665,
+      "step": 41304
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.044168773704172e-06,
+      "loss": 0.9358,
+      "step": 41305
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0441213653603034e-06,
+      "loss": 2.3787,
+      "step": 41306
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.044073957016435e-06,
+      "loss": 3.9026,
+      "step": 41307
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0440265486725664e-06,
+      "loss": 2.5292,
+      "step": 41308
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.043979140328698e-06,
+      "loss": 2.084,
+      "step": 41309
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0439317319848294e-06,
+      "loss": 2.8222,
+      "step": 41310
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0438843236409607e-06,
+      "loss": 1.6841,
+      "step": 41311
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0438369152970924e-06,
+      "loss": 1.5393,
+      "step": 41312
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0437895069532239e-06,
+      "loss": 3.0253,
+      "step": 41313
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0437420986093552e-06,
+      "loss": 2.5794,
+      "step": 41314
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0436946902654867e-06,
+      "loss": 0.9724,
+      "step": 41315
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0436472819216184e-06,
+      "loss": 1.353,
+      "step": 41316
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0435998735777496e-06,
+      "loss": 1.0812,
+      "step": 41317
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0435524652338811e-06,
+      "loss": 2.5781,
+      "step": 41318
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0435050568900128e-06,
+      "loss": 2.1545,
+      "step": 41319
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0434576485461441e-06,
+      "loss": 1.7428,
+      "step": 41320
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0434102402022756e-06,
+      "loss": 1.6663,
+      "step": 41321
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 1.0433628318584071e-06,
+      "loss": 1.657,
+      "step": 41322
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0433154235145386e-06,
+      "loss": 3.021,
+      "step": 41323
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.04326801517067e-06,
+      "loss": 1.3558,
+      "step": 41324
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0432206068268016e-06,
+      "loss": 2.6833,
+      "step": 41325
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.043173198482933e-06,
+      "loss": 1.3101,
+      "step": 41326
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0431257901390646e-06,
+      "loss": 1.9727,
+      "step": 41327
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0430783817951959e-06,
+      "loss": 1.4484,
+      "step": 41328
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0430309734513276e-06,
+      "loss": 1.1306,
+      "step": 41329
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.042983565107459e-06,
+      "loss": 2.4694,
+      "step": 41330
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0429361567635903e-06,
+      "loss": 3.1777,
+      "step": 41331
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.042888748419722e-06,
+      "loss": 1.9501,
+      "step": 41332
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0428413400758533e-06,
+      "loss": 3.2442,
+      "step": 41333
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0427939317319848e-06,
+      "loss": 2.5236,
+      "step": 41334
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0427465233881163e-06,
+      "loss": 0.3085,
+      "step": 41335
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0426991150442478e-06,
+      "loss": 1.6014,
+      "step": 41336
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0426517067003793e-06,
+      "loss": 3.4616,
+      "step": 41337
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0426042983565108e-06,
+      "loss": 0.9585,
+      "step": 41338
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0425568900126423e-06,
+      "loss": 0.3878,
+      "step": 41339
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0425094816687738e-06,
+      "loss": 0.9955,
+      "step": 41340
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0424620733249052e-06,
+      "loss": 3.9317,
+      "step": 41341
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0424146649810367e-06,
+      "loss": 1.7452,
+      "step": 41342
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0423672566371682e-06,
+      "loss": 3.0491,
+      "step": 41343
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0423198482932995e-06,
+      "loss": 2.9165,
+      "step": 41344
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.042272439949431e-06,
+      "loss": 2.4685,
+      "step": 41345
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0422250316055627e-06,
+      "loss": 2.6777,
+      "step": 41346
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.042177623261694e-06,
+      "loss": 2.0412,
+      "step": 41347
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0421302149178255e-06,
+      "loss": 2.6859,
+      "step": 41348
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0420828065739572e-06,
+      "loss": 2.7306,
+      "step": 41349
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0420353982300885e-06,
+      "loss": 2.8581,
+      "step": 41350
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.04198798988622e-06,
+      "loss": 1.2372,
+      "step": 41351
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0419405815423515e-06,
+      "loss": 1.4502,
+      "step": 41352
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.041893173198483e-06,
+      "loss": 0.862,
+      "step": 41353
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0418457648546144e-06,
+      "loss": 2.2153,
+      "step": 41354
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.041798356510746e-06,
+      "loss": 1.7546,
+      "step": 41355
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0417509481668774e-06,
+      "loss": 2.5933,
+      "step": 41356
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.041703539823009e-06,
+      "loss": 0.7367,
+      "step": 41357
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0416561314791402e-06,
+      "loss": 1.1972,
+      "step": 41358
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.041608723135272e-06,
+      "loss": 2.5185,
+      "step": 41359
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0415613147914034e-06,
+      "loss": 1.8635,
+      "step": 41360
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0415139064475347e-06,
+      "loss": 2.3073,
+      "step": 41361
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0414664981036664e-06,
+      "loss": 2.0816,
+      "step": 41362
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0414190897597979e-06,
+      "loss": 1.5736,
+      "step": 41363
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0413716814159291e-06,
+      "loss": 2.0592,
+      "step": 41364
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0413242730720606e-06,
+      "loss": 2.181,
+      "step": 41365
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0412768647281921e-06,
+      "loss": 3.437,
+      "step": 41366
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0412294563843236e-06,
+      "loss": 2.5068,
+      "step": 41367
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0411820480404551e-06,
+      "loss": 3.0736,
+      "step": 41368
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0411346396965866e-06,
+      "loss": 0.5805,
+      "step": 41369
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.041087231352718e-06,
+      "loss": 0.9504,
+      "step": 41370
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0410398230088496e-06,
+      "loss": 0.4731,
+      "step": 41371
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.040992414664981e-06,
+      "loss": 2.3317,
+      "step": 41372
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0409450063211126e-06,
+      "loss": 2.3372,
+      "step": 41373
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.040897597977244e-06,
+      "loss": 2.1201,
+      "step": 41374
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0408501896333756e-06,
+      "loss": 2.0353,
+      "step": 41375
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.040802781289507e-06,
+      "loss": 2.8167,
+      "step": 41376
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0407553729456383e-06,
+      "loss": 1.3145,
+      "step": 41377
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0407079646017698e-06,
+      "loss": 0.3888,
+      "step": 41378
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0406605562579015e-06,
+      "loss": 3.2571,
+      "step": 41379
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0406131479140328e-06,
+      "loss": 1.8858,
+      "step": 41380
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0405657395701643e-06,
+      "loss": 1.4159,
+      "step": 41381
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.040518331226296e-06,
+      "loss": 1.6524,
+      "step": 41382
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0404709228824273e-06,
+      "loss": 3.0065,
+      "step": 41383
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0404235145385588e-06,
+      "loss": 1.9575,
+      "step": 41384
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0403761061946903e-06,
+      "loss": 1.4935,
+      "step": 41385
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 1.0403286978508218e-06,
+      "loss": 2.9203,
+      "step": 41386
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0402812895069533e-06,
+      "loss": 3.113,
+      "step": 41387
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0402338811630847e-06,
+      "loss": 2.4369,
+      "step": 41388
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0401864728192162e-06,
+      "loss": 1.5316,
+      "step": 41389
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0401390644753477e-06,
+      "loss": 2.4265,
+      "step": 41390
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.040091656131479e-06,
+      "loss": 1.9072,
+      "step": 41391
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0400442477876107e-06,
+      "loss": 2.0616,
+      "step": 41392
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0399968394437422e-06,
+      "loss": 4.8385,
+      "step": 41393
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0399494310998735e-06,
+      "loss": 0.9359,
+      "step": 41394
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0399020227560052e-06,
+      "loss": 1.9186,
+      "step": 41395
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0398546144121367e-06,
+      "loss": 1.1538,
+      "step": 41396
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.039807206068268e-06,
+      "loss": 1.7285,
+      "step": 41397
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0397597977243995e-06,
+      "loss": 1.6891,
+      "step": 41398
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.039712389380531e-06,
+      "loss": 2.246,
+      "step": 41399
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0396649810366624e-06,
+      "loss": 1.3325,
+      "step": 41400
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.039617572692794e-06,
+      "loss": 1.4737,
+      "step": 41401
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0395701643489254e-06,
+      "loss": 1.9248,
+      "step": 41402
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.039522756005057e-06,
+      "loss": 1.1844,
+      "step": 41403
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0394753476611884e-06,
+      "loss": 2.8503,
+      "step": 41404
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.03942793931732e-06,
+      "loss": 2.7583,
+      "step": 41405
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0393805309734514e-06,
+      "loss": 2.5951,
+      "step": 41406
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0393331226295829e-06,
+      "loss": 2.6673,
+      "step": 41407
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0392857142857144e-06,
+      "loss": 0.5237,
+      "step": 41408
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0392383059418459e-06,
+      "loss": 2.1792,
+      "step": 41409
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0391908975979772e-06,
+      "loss": 2.525,
+      "step": 41410
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0391434892541086e-06,
+      "loss": 2.2549,
+      "step": 41411
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0390960809102403e-06,
+      "loss": 0.9483,
+      "step": 41412
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0390486725663716e-06,
+      "loss": 3.0113,
+      "step": 41413
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0390012642225031e-06,
+      "loss": 2.3484,
+      "step": 41414
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0389538558786348e-06,
+      "loss": 1.8719,
+      "step": 41415
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.038906447534766e-06,
+      "loss": 2.9577,
+      "step": 41416
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0388590391908976e-06,
+      "loss": 1.4959,
+      "step": 41417
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.038811630847029e-06,
+      "loss": 1.8997,
+      "step": 41418
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0387642225031606e-06,
+      "loss": 0.3366,
+      "step": 41419
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.038716814159292e-06,
+      "loss": 3.8049,
+      "step": 41420
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0386694058154236e-06,
+      "loss": 1.6279,
+      "step": 41421
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.038621997471555e-06,
+      "loss": 1.4571,
+      "step": 41422
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0385745891276865e-06,
+      "loss": 1.5808,
+      "step": 41423
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0385271807838178e-06,
+      "loss": 1.5581,
+      "step": 41424
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0384797724399495e-06,
+      "loss": 1.4754,
+      "step": 41425
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.038432364096081e-06,
+      "loss": 1.7884,
+      "step": 41426
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0383849557522123e-06,
+      "loss": 1.8193,
+      "step": 41427
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.038337547408344e-06,
+      "loss": 3.406,
+      "step": 41428
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0382901390644755e-06,
+      "loss": 2.0487,
+      "step": 41429
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0382427307206068e-06,
+      "loss": 1.6325,
+      "step": 41430
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0381953223767383e-06,
+      "loss": 1.9831,
+      "step": 41431
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0381479140328698e-06,
+      "loss": 1.2617,
+      "step": 41432
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0381005056890013e-06,
+      "loss": 3.4243,
+      "step": 41433
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0380530973451328e-06,
+      "loss": 2.2356,
+      "step": 41434
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0380056890012642e-06,
+      "loss": 1.9509,
+      "step": 41435
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0379582806573957e-06,
+      "loss": 1.9982,
+      "step": 41436
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0379108723135272e-06,
+      "loss": 2.5535,
+      "step": 41437
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0378634639696587e-06,
+      "loss": 1.6312,
+      "step": 41438
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0378160556257902e-06,
+      "loss": 0.744,
+      "step": 41439
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0377686472819217e-06,
+      "loss": 1.9313,
+      "step": 41440
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0377212389380532e-06,
+      "loss": 1.3874,
+      "step": 41441
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0376738305941847e-06,
+      "loss": 2.2838,
+      "step": 41442
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.037626422250316e-06,
+      "loss": 1.2668,
+      "step": 41443
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0375790139064475e-06,
+      "loss": 1.858,
+      "step": 41444
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0375316055625792e-06,
+      "loss": 1.8001,
+      "step": 41445
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0374841972187104e-06,
+      "loss": 2.486,
+      "step": 41446
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.037436788874842e-06,
+      "loss": 4.3433,
+      "step": 41447
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.0373893805309736e-06,
+      "loss": 1.4863,
+      "step": 41448
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 1.037341972187105e-06,
+      "loss": 1.7448,
+      "step": 41449
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0372945638432364e-06,
+      "loss": 0.7152,
+      "step": 41450
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.037247155499368e-06,
+      "loss": 2.1065,
+      "step": 41451
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0371997471554994e-06,
+      "loss": 0.7703,
+      "step": 41452
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0371523388116309e-06,
+      "loss": 1.7264,
+      "step": 41453
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0371049304677624e-06,
+      "loss": 2.0085,
+      "step": 41454
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0370575221238939e-06,
+      "loss": 2.6048,
+      "step": 41455
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0370101137800254e-06,
+      "loss": 2.8819,
+      "step": 41456
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0369627054361566e-06,
+      "loss": 1.3641,
+      "step": 41457
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0369152970922884e-06,
+      "loss": 3.5189,
+      "step": 41458
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0368678887484198e-06,
+      "loss": 1.452,
+      "step": 41459
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0368204804045511e-06,
+      "loss": 2.8712,
+      "step": 41460
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0367730720606828e-06,
+      "loss": 1.8733,
+      "step": 41461
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0367256637168143e-06,
+      "loss": 0.965,
+      "step": 41462
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0366782553729456e-06,
+      "loss": 1.6602,
+      "step": 41463
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.036630847029077e-06,
+      "loss": 2.0432,
+      "step": 41464
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0365834386852086e-06,
+      "loss": 1.7686,
+      "step": 41465
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.03653603034134e-06,
+      "loss": 1.1036,
+      "step": 41466
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0364886219974716e-06,
+      "loss": 2.8007,
+      "step": 41467
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.036441213653603e-06,
+      "loss": 3.3974,
+      "step": 41468
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0363938053097346e-06,
+      "loss": 2.0427,
+      "step": 41469
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.036346396965866e-06,
+      "loss": 2.436,
+      "step": 41470
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0362989886219975e-06,
+      "loss": 1.9219,
+      "step": 41471
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.036251580278129e-06,
+      "loss": 3.1069,
+      "step": 41472
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0362041719342605e-06,
+      "loss": 2.7786,
+      "step": 41473
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0361567635903918e-06,
+      "loss": 2.7204,
+      "step": 41474
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0361093552465235e-06,
+      "loss": 0.7509,
+      "step": 41475
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0360619469026548e-06,
+      "loss": 1.9643,
+      "step": 41476
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0360145385587863e-06,
+      "loss": 2.6642,
+      "step": 41477
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.035967130214918e-06,
+      "loss": 1.5685,
+      "step": 41478
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0359197218710493e-06,
+      "loss": 1.0601,
+      "step": 41479
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0358723135271808e-06,
+      "loss": 3.1271,
+      "step": 41480
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0358249051833122e-06,
+      "loss": 1.247,
+      "step": 41481
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0357774968394437e-06,
+      "loss": 3.982,
+      "step": 41482
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0357300884955752e-06,
+      "loss": 1.8651,
+      "step": 41483
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0356826801517067e-06,
+      "loss": 0.5468,
+      "step": 41484
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0356352718078382e-06,
+      "loss": 3.4116,
+      "step": 41485
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0355878634639697e-06,
+      "loss": 3.5411,
+      "step": 41486
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.035540455120101e-06,
+      "loss": 2.4334,
+      "step": 41487
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0354930467762327e-06,
+      "loss": 3.0171,
+      "step": 41488
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0354456384323642e-06,
+      "loss": 2.3156,
+      "step": 41489
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0353982300884955e-06,
+      "loss": 2.4588,
+      "step": 41490
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0353508217446272e-06,
+      "loss": 2.2527,
+      "step": 41491
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0353034134007587e-06,
+      "loss": 2.3743,
+      "step": 41492
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.03525600505689e-06,
+      "loss": 2.9654,
+      "step": 41493
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0352085967130214e-06,
+      "loss": 2.7762,
+      "step": 41494
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0351611883691531e-06,
+      "loss": 2.0095,
+      "step": 41495
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0351137800252844e-06,
+      "loss": 2.4529,
+      "step": 41496
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.035066371681416e-06,
+      "loss": 2.0448,
+      "step": 41497
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0350189633375474e-06,
+      "loss": 2.7363,
+      "step": 41498
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.034971554993679e-06,
+      "loss": 2.7444,
+      "step": 41499
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0349241466498104e-06,
+      "loss": 1.9396,
+      "step": 41500
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0348767383059419e-06,
+      "loss": 1.0831,
+      "step": 41501
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0348293299620734e-06,
+      "loss": 2.7962,
+      "step": 41502
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0347819216182049e-06,
+      "loss": 2.4284,
+      "step": 41503
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0347345132743364e-06,
+      "loss": 1.1101,
+      "step": 41504
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0346871049304678e-06,
+      "loss": 1.7641,
+      "step": 41505
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0346396965865993e-06,
+      "loss": 0.7091,
+      "step": 41506
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0345922882427306e-06,
+      "loss": 1.7991,
+      "step": 41507
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0345448798988623e-06,
+      "loss": 0.7366,
+      "step": 41508
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0344974715549936e-06,
+      "loss": 2.9485,
+      "step": 41509
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.034450063211125e-06,
+      "loss": 3.5499,
+      "step": 41510
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.0344026548672568e-06,
+      "loss": 2.9096,
+      "step": 41511
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 1.034355246523388e-06,
+      "loss": 0.9877,
+      "step": 41512
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0343078381795196e-06,
+      "loss": 2.2458,
+      "step": 41513
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.034260429835651e-06,
+      "loss": 3.8124,
+      "step": 41514
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0342130214917826e-06,
+      "loss": 2.8507,
+      "step": 41515
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.034165613147914e-06,
+      "loss": 3.6542,
+      "step": 41516
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0341182048040455e-06,
+      "loss": 2.6648,
+      "step": 41517
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.034070796460177e-06,
+      "loss": 2.9444,
+      "step": 41518
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0340233881163085e-06,
+      "loss": 3.2522,
+      "step": 41519
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0339759797724398e-06,
+      "loss": 2.5849,
+      "step": 41520
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0339285714285715e-06,
+      "loss": 0.9118,
+      "step": 41521
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.033881163084703e-06,
+      "loss": 1.5436,
+      "step": 41522
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0338337547408343e-06,
+      "loss": 1.3454,
+      "step": 41523
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.033786346396966e-06,
+      "loss": 2.9777,
+      "step": 41524
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0337389380530975e-06,
+      "loss": 0.4264,
+      "step": 41525
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0336915297092288e-06,
+      "loss": 1.0854,
+      "step": 41526
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0336441213653603e-06,
+      "loss": 2.2271,
+      "step": 41527
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.033596713021492e-06,
+      "loss": 1.0747,
+      "step": 41528
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0335493046776232e-06,
+      "loss": 1.3865,
+      "step": 41529
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0335018963337547e-06,
+      "loss": 4.0894,
+      "step": 41530
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0334544879898862e-06,
+      "loss": 2.3535,
+      "step": 41531
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0334070796460177e-06,
+      "loss": 1.5988,
+      "step": 41532
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0333596713021492e-06,
+      "loss": 1.373,
+      "step": 41533
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0333122629582807e-06,
+      "loss": 2.3178,
+      "step": 41534
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0332648546144122e-06,
+      "loss": 1.7353,
+      "step": 41535
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0332174462705437e-06,
+      "loss": 1.8311,
+      "step": 41536
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0331700379266752e-06,
+      "loss": 1.699,
+      "step": 41537
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0331226295828067e-06,
+      "loss": 1.5364,
+      "step": 41538
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0330752212389382e-06,
+      "loss": 1.5462,
+      "step": 41539
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0330278128950694e-06,
+      "loss": 1.9477,
+      "step": 41540
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0329804045512011e-06,
+      "loss": 2.0273,
+      "step": 41541
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0329329962073324e-06,
+      "loss": 0.9877,
+      "step": 41542
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.032885587863464e-06,
+      "loss": 1.9321,
+      "step": 41543
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0328381795195956e-06,
+      "loss": 1.7836,
+      "step": 41544
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.032790771175727e-06,
+      "loss": 2.3467,
+      "step": 41545
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0327433628318584e-06,
+      "loss": 2.1274,
+      "step": 41546
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0326959544879899e-06,
+      "loss": 2.4841,
+      "step": 41547
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0326485461441214e-06,
+      "loss": 1.6418,
+      "step": 41548
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0326011378002529e-06,
+      "loss": 0.4821,
+      "step": 41549
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0325537294563844e-06,
+      "loss": 1.1019,
+      "step": 41550
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0325063211125159e-06,
+      "loss": 1.9718,
+      "step": 41551
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0324589127686473e-06,
+      "loss": 2.5204,
+      "step": 41552
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0324115044247786e-06,
+      "loss": 1.2188,
+      "step": 41553
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0323640960809103e-06,
+      "loss": 2.4969,
+      "step": 41554
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0323166877370418e-06,
+      "loss": 1.6106,
+      "step": 41555
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.032269279393173e-06,
+      "loss": 4.0044,
+      "step": 41556
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0322218710493048e-06,
+      "loss": 1.5104,
+      "step": 41557
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0321744627054363e-06,
+      "loss": 1.5503,
+      "step": 41558
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0321270543615676e-06,
+      "loss": 1.5144,
+      "step": 41559
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.032079646017699e-06,
+      "loss": 2.1967,
+      "step": 41560
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0320322376738308e-06,
+      "loss": 4.3351,
+      "step": 41561
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.031984829329962e-06,
+      "loss": 1.2937,
+      "step": 41562
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0319374209860935e-06,
+      "loss": 1.4133,
+      "step": 41563
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.031890012642225e-06,
+      "loss": 1.786,
+      "step": 41564
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0318426042983565e-06,
+      "loss": 2.951,
+      "step": 41565
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.031795195954488e-06,
+      "loss": 2.2228,
+      "step": 41566
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0317477876106195e-06,
+      "loss": 2.0774,
+      "step": 41567
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.031700379266751e-06,
+      "loss": 3.4378,
+      "step": 41568
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0316529709228825e-06,
+      "loss": 1.7159,
+      "step": 41569
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.031605562579014e-06,
+      "loss": 2.3324,
+      "step": 41570
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0315581542351455e-06,
+      "loss": 1.0141,
+      "step": 41571
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.031510745891277e-06,
+      "loss": 2.1005,
+      "step": 41572
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0314633375474083e-06,
+      "loss": 1.5837,
+      "step": 41573
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.03141592920354e-06,
+      "loss": 2.0757,
+      "step": 41574
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0313685208596712e-06,
+      "loss": 1.9576,
+      "step": 41575
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 1.0313211125158027e-06,
+      "loss": 1.6379,
+      "step": 41576
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0312737041719344e-06,
+      "loss": 1.0185,
+      "step": 41577
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0312262958280657e-06,
+      "loss": 3.3079,
+      "step": 41578
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0311788874841972e-06,
+      "loss": 3.0604,
+      "step": 41579
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0311314791403287e-06,
+      "loss": 1.4142,
+      "step": 41580
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0310840707964602e-06,
+      "loss": 2.1176,
+      "step": 41581
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0310366624525917e-06,
+      "loss": 1.3707,
+      "step": 41582
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0309892541087232e-06,
+      "loss": 3.0047,
+      "step": 41583
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0309418457648547e-06,
+      "loss": 1.7364,
+      "step": 41584
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0308944374209862e-06,
+      "loss": 1.7092,
+      "step": 41585
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0308470290771174e-06,
+      "loss": 2.167,
+      "step": 41586
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0307996207332491e-06,
+      "loss": 3.2613,
+      "step": 41587
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0307522123893806e-06,
+      "loss": 0.5217,
+      "step": 41588
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.030704804045512e-06,
+      "loss": 1.6148,
+      "step": 41589
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0306573957016436e-06,
+      "loss": 1.6605,
+      "step": 41590
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0306099873577751e-06,
+      "loss": 2.6145,
+      "step": 41591
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0305625790139064e-06,
+      "loss": 0.7887,
+      "step": 41592
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0305151706700379e-06,
+      "loss": 1.4916,
+      "step": 41593
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0304677623261694e-06,
+      "loss": 2.4475,
+      "step": 41594
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0304203539823009e-06,
+      "loss": 2.4833,
+      "step": 41595
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0303729456384324e-06,
+      "loss": 3.7945,
+      "step": 41596
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0303255372945639e-06,
+      "loss": 2.4973,
+      "step": 41597
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0302781289506954e-06,
+      "loss": 1.4341,
+      "step": 41598
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0302307206068268e-06,
+      "loss": 3.1831,
+      "step": 41599
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0301833122629583e-06,
+      "loss": 0.642,
+      "step": 41600
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0301359039190898e-06,
+      "loss": 1.2276,
+      "step": 41601
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0300884955752213e-06,
+      "loss": 1.4047,
+      "step": 41602
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0300410872313528e-06,
+      "loss": 1.6415,
+      "step": 41603
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0299936788874843e-06,
+      "loss": 2.0106,
+      "step": 41604
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0299462705436156e-06,
+      "loss": 1.7326,
+      "step": 41605
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.029898862199747e-06,
+      "loss": 1.2925,
+      "step": 41606
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0298514538558788e-06,
+      "loss": 1.4807,
+      "step": 41607
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.02980404551201e-06,
+      "loss": 2.6454,
+      "step": 41608
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0297566371681416e-06,
+      "loss": 2.8904,
+      "step": 41609
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0297092288242733e-06,
+      "loss": 3.2354,
+      "step": 41610
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0296618204804045e-06,
+      "loss": 2.4133,
+      "step": 41611
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.029614412136536e-06,
+      "loss": 2.1641,
+      "step": 41612
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0295670037926675e-06,
+      "loss": 1.8985,
+      "step": 41613
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.029519595448799e-06,
+      "loss": 1.8427,
+      "step": 41614
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0294721871049305e-06,
+      "loss": 1.6668,
+      "step": 41615
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0294247787610618e-06,
+      "loss": 2.6446,
+      "step": 41616
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0293773704171935e-06,
+      "loss": 2.804,
+      "step": 41617
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.029329962073325e-06,
+      "loss": 2.3538,
+      "step": 41618
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0292825537294563e-06,
+      "loss": 1.8448,
+      "step": 41619
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.029235145385588e-06,
+      "loss": 1.884,
+      "step": 41620
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0291877370417195e-06,
+      "loss": 2.6973,
+      "step": 41621
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0291403286978507e-06,
+      "loss": 4.0022,
+      "step": 41622
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0290929203539822e-06,
+      "loss": 2.8471,
+      "step": 41623
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.029045512010114e-06,
+      "loss": 0.4585,
+      "step": 41624
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0289981036662452e-06,
+      "loss": 1.9417,
+      "step": 41625
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0289506953223767e-06,
+      "loss": 1.5087,
+      "step": 41626
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0289032869785082e-06,
+      "loss": 2.4313,
+      "step": 41627
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0288558786346397e-06,
+      "loss": 1.8759,
+      "step": 41628
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0288084702907712e-06,
+      "loss": 1.8989,
+      "step": 41629
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0287610619469027e-06,
+      "loss": 1.9363,
+      "step": 41630
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0287136536030342e-06,
+      "loss": 2.4467,
+      "step": 41631
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0286662452591657e-06,
+      "loss": 2.0339,
+      "step": 41632
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0286188369152972e-06,
+      "loss": 3.3486,
+      "step": 41633
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0285714285714286e-06,
+      "loss": 3.3643,
+      "step": 41634
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0285240202275601e-06,
+      "loss": 2.7274,
+      "step": 41635
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0284766118836914e-06,
+      "loss": 1.5446,
+      "step": 41636
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0284292035398231e-06,
+      "loss": 2.0044,
+      "step": 41637
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.0283817951959544e-06,
+      "loss": 3.6112,
+      "step": 41638
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 1.028334386852086e-06,
+      "loss": 2.0062,
+      "step": 41639
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0282869785082176e-06,
+      "loss": 2.9266,
+      "step": 41640
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0282395701643489e-06,
+      "loss": 1.8044,
+      "step": 41641
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0281921618204804e-06,
+      "loss": 1.5762,
+      "step": 41642
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0281447534766119e-06,
+      "loss": 3.2468,
+      "step": 41643
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0280973451327434e-06,
+      "loss": 1.4425,
+      "step": 41644
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0280499367888748e-06,
+      "loss": 1.7395,
+      "step": 41645
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0280025284450063e-06,
+      "loss": 2.1884,
+      "step": 41646
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0279551201011378e-06,
+      "loss": 2.7775,
+      "step": 41647
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0279077117572693e-06,
+      "loss": 2.3874,
+      "step": 41648
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0278603034134006e-06,
+      "loss": 2.6251,
+      "step": 41649
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0278128950695323e-06,
+      "loss": 1.7536,
+      "step": 41650
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0277654867256638e-06,
+      "loss": 2.5813,
+      "step": 41651
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.027718078381795e-06,
+      "loss": 1.5117,
+      "step": 41652
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0276706700379268e-06,
+      "loss": 1.9533,
+      "step": 41653
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0276232616940583e-06,
+      "loss": 2.5108,
+      "step": 41654
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0275758533501896e-06,
+      "loss": 2.9062,
+      "step": 41655
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.027528445006321e-06,
+      "loss": 2.8185,
+      "step": 41656
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0274810366624528e-06,
+      "loss": 1.6106,
+      "step": 41657
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.027433628318584e-06,
+      "loss": 0.939,
+      "step": 41658
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0273862199747155e-06,
+      "loss": 3.0921,
+      "step": 41659
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.027338811630847e-06,
+      "loss": 1.212,
+      "step": 41660
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0272914032869785e-06,
+      "loss": 1.4895,
+      "step": 41661
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.02724399494311e-06,
+      "loss": 2.2623,
+      "step": 41662
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0271965865992415e-06,
+      "loss": 3.2372,
+      "step": 41663
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.027149178255373e-06,
+      "loss": 2.2063,
+      "step": 41664
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0271017699115045e-06,
+      "loss": 2.8361,
+      "step": 41665
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.027054361567636e-06,
+      "loss": 2.371,
+      "step": 41666
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0270069532237675e-06,
+      "loss": 3.4629,
+      "step": 41667
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.026959544879899e-06,
+      "loss": 2.7108,
+      "step": 41668
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0269121365360302e-06,
+      "loss": 2.8095,
+      "step": 41669
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.026864728192162e-06,
+      "loss": 1.8451,
+      "step": 41670
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0268173198482932e-06,
+      "loss": 1.1578,
+      "step": 41671
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0267699115044247e-06,
+      "loss": 1.8324,
+      "step": 41672
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0267225031605564e-06,
+      "loss": 3.8,
+      "step": 41673
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0266750948166877e-06,
+      "loss": 0.9179,
+      "step": 41674
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0266276864728192e-06,
+      "loss": 2.187,
+      "step": 41675
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0265802781289507e-06,
+      "loss": 1.1854,
+      "step": 41676
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0265328697850822e-06,
+      "loss": 1.6551,
+      "step": 41677
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0264854614412137e-06,
+      "loss": 1.9526,
+      "step": 41678
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0264380530973452e-06,
+      "loss": 1.7532,
+      "step": 41679
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0263906447534767e-06,
+      "loss": 3.328,
+      "step": 41680
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0263432364096081e-06,
+      "loss": 1.3865,
+      "step": 41681
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0262958280657394e-06,
+      "loss": 1.486,
+      "step": 41682
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0262484197218711e-06,
+      "loss": 2.2188,
+      "step": 41683
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0262010113780026e-06,
+      "loss": 1.3318,
+      "step": 41684
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.026153603034134e-06,
+      "loss": 2.0691,
+      "step": 41685
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0261061946902656e-06,
+      "loss": 2.0446,
+      "step": 41686
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.026058786346397e-06,
+      "loss": 2.6379,
+      "step": 41687
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0260113780025284e-06,
+      "loss": 3.0929,
+      "step": 41688
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0259639696586599e-06,
+      "loss": 2.8369,
+      "step": 41689
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0259165613147916e-06,
+      "loss": 1.5298,
+      "step": 41690
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0258691529709229e-06,
+      "loss": 1.749,
+      "step": 41691
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0258217446270543e-06,
+      "loss": 1.795,
+      "step": 41692
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0257743362831858e-06,
+      "loss": 0.9062,
+      "step": 41693
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0257269279393173e-06,
+      "loss": 1.6726,
+      "step": 41694
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0256795195954488e-06,
+      "loss": 1.24,
+      "step": 41695
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0256321112515803e-06,
+      "loss": 1.6137,
+      "step": 41696
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0255847029077118e-06,
+      "loss": 1.3831,
+      "step": 41697
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0255372945638433e-06,
+      "loss": 1.2768,
+      "step": 41698
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0254898862199748e-06,
+      "loss": 1.6137,
+      "step": 41699
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0254424778761063e-06,
+      "loss": 2.1308,
+      "step": 41700
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.0253950695322378e-06,
+      "loss": 1.994,
+      "step": 41701
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 1.025347661188369e-06,
+      "loss": 2.2414,
+      "step": 41702
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0253002528445008e-06,
+      "loss": 1.8687,
+      "step": 41703
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.025252844500632e-06,
+      "loss": 1.7784,
+      "step": 41704
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0252054361567635e-06,
+      "loss": 1.1449,
+      "step": 41705
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0251580278128952e-06,
+      "loss": 2.5071,
+      "step": 41706
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0251106194690265e-06,
+      "loss": 1.6215,
+      "step": 41707
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.025063211125158e-06,
+      "loss": 1.9952,
+      "step": 41708
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0250158027812895e-06,
+      "loss": 2.4461,
+      "step": 41709
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.024968394437421e-06,
+      "loss": 2.3769,
+      "step": 41710
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0249209860935525e-06,
+      "loss": 2.1148,
+      "step": 41711
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.024873577749684e-06,
+      "loss": 1.5786,
+      "step": 41712
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0248261694058155e-06,
+      "loss": 1.5042,
+      "step": 41713
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.024778761061947e-06,
+      "loss": 3.7678,
+      "step": 41714
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0247313527180782e-06,
+      "loss": 3.3525,
+      "step": 41715
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.02468394437421e-06,
+      "loss": 1.34,
+      "step": 41716
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0246365360303414e-06,
+      "loss": 2.3263,
+      "step": 41717
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0245891276864727e-06,
+      "loss": 2.602,
+      "step": 41718
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0245417193426044e-06,
+      "loss": 2.0504,
+      "step": 41719
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.024494310998736e-06,
+      "loss": 3.0585,
+      "step": 41720
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0244469026548672e-06,
+      "loss": 2.2862,
+      "step": 41721
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0243994943109987e-06,
+      "loss": 2.9934,
+      "step": 41722
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0243520859671304e-06,
+      "loss": 2.9065,
+      "step": 41723
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0243046776232617e-06,
+      "loss": 2.1811,
+      "step": 41724
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0242572692793932e-06,
+      "loss": 2.5785,
+      "step": 41725
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0242098609355247e-06,
+      "loss": 2.439,
+      "step": 41726
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0241624525916561e-06,
+      "loss": 1.2756,
+      "step": 41727
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0241150442477876e-06,
+      "loss": 2.4256,
+      "step": 41728
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0240676359039191e-06,
+      "loss": 1.8074,
+      "step": 41729
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0240202275600506e-06,
+      "loss": 2.1275,
+      "step": 41730
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0239728192161821e-06,
+      "loss": 2.1197,
+      "step": 41731
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0239254108723136e-06,
+      "loss": 2.0615,
+      "step": 41732
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.023878002528445e-06,
+      "loss": 1.9864,
+      "step": 41733
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0238305941845766e-06,
+      "loss": 2.3644,
+      "step": 41734
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0237831858407079e-06,
+      "loss": 2.6196,
+      "step": 41735
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0237357774968396e-06,
+      "loss": 1.1695,
+      "step": 41736
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0236883691529709e-06,
+      "loss": 2.3487,
+      "step": 41737
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0236409608091023e-06,
+      "loss": 1.4737,
+      "step": 41738
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.023593552465234e-06,
+      "loss": 1.593,
+      "step": 41739
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0235461441213653e-06,
+      "loss": 1.7018,
+      "step": 41740
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0234987357774968e-06,
+      "loss": 1.8846,
+      "step": 41741
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0234513274336283e-06,
+      "loss": 2.6506,
+      "step": 41742
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0234039190897598e-06,
+      "loss": 0.9627,
+      "step": 41743
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0233565107458913e-06,
+      "loss": 2.4554,
+      "step": 41744
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0233091024020228e-06,
+      "loss": 3.0153,
+      "step": 41745
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0232616940581543e-06,
+      "loss": 0.2725,
+      "step": 41746
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0232142857142858e-06,
+      "loss": 3.5221,
+      "step": 41747
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.023166877370417e-06,
+      "loss": 2.5471,
+      "step": 41748
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0231194690265488e-06,
+      "loss": 4.3311,
+      "step": 41749
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0230720606826803e-06,
+      "loss": 1.6254,
+      "step": 41750
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0230246523388115e-06,
+      "loss": 1.7079,
+      "step": 41751
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022977243994943e-06,
+      "loss": 1.2382,
+      "step": 41752
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0229298356510747e-06,
+      "loss": 2.3809,
+      "step": 41753
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022882427307206e-06,
+      "loss": 1.0109,
+      "step": 41754
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0228350189633375e-06,
+      "loss": 2.7822,
+      "step": 41755
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0227876106194692e-06,
+      "loss": 1.8585,
+      "step": 41756
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0227402022756005e-06,
+      "loss": 3.2676,
+      "step": 41757
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022692793931732e-06,
+      "loss": 0.9545,
+      "step": 41758
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0226453855878635e-06,
+      "loss": 2.3526,
+      "step": 41759
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022597977243995e-06,
+      "loss": 0.9698,
+      "step": 41760
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0225505689001265e-06,
+      "loss": 1.3989,
+      "step": 41761
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022503160556258e-06,
+      "loss": 2.2539,
+      "step": 41762
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0224557522123894e-06,
+      "loss": 2.01,
+      "step": 41763
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022408343868521e-06,
+      "loss": 2.1478,
+      "step": 41764
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.0223609355246522e-06,
+      "loss": 1.8419,
+      "step": 41765
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 1.022313527180784e-06,
+      "loss": 0.4418,
+      "step": 41766
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0222661188369154e-06,
+      "loss": 1.5547,
+      "step": 41767
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0222187104930467e-06,
+      "loss": 1.4863,
+      "step": 41768
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0221713021491784e-06,
+      "loss": 3.3623,
+      "step": 41769
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0221238938053097e-06,
+      "loss": 1.7115,
+      "step": 41770
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0220764854614412e-06,
+      "loss": 0.8448,
+      "step": 41771
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0220290771175727e-06,
+      "loss": 1.8834,
+      "step": 41772
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0219816687737042e-06,
+      "loss": 1.6504,
+      "step": 41773
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0219342604298356e-06,
+      "loss": 2.7534,
+      "step": 41774
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0218868520859671e-06,
+      "loss": 0.7744,
+      "step": 41775
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0218394437420986e-06,
+      "loss": 0.6388,
+      "step": 41776
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0217920353982301e-06,
+      "loss": 2.0317,
+      "step": 41777
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0217446270543616e-06,
+      "loss": 1.5864,
+      "step": 41778
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.021697218710493e-06,
+      "loss": 2.0263,
+      "step": 41779
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0216498103666246e-06,
+      "loss": 1.6647,
+      "step": 41780
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0216024020227559e-06,
+      "loss": 2.3749,
+      "step": 41781
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0215549936788876e-06,
+      "loss": 2.8831,
+      "step": 41782
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.021507585335019e-06,
+      "loss": 1.9515,
+      "step": 41783
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0214601769911504e-06,
+      "loss": 2.0599,
+      "step": 41784
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0214127686472818e-06,
+      "loss": 2.1302,
+      "step": 41785
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0213653603034136e-06,
+      "loss": 2.2047,
+      "step": 41786
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0213179519595448e-06,
+      "loss": 2.5045,
+      "step": 41787
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0212705436156763e-06,
+      "loss": 2.2373,
+      "step": 41788
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.021223135271808e-06,
+      "loss": 1.5562,
+      "step": 41789
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0211757269279393e-06,
+      "loss": 2.8695,
+      "step": 41790
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0211283185840708e-06,
+      "loss": 2.9632,
+      "step": 41791
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0210809102402023e-06,
+      "loss": 2.0688,
+      "step": 41792
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0210335018963338e-06,
+      "loss": 2.1476,
+      "step": 41793
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0209860935524653e-06,
+      "loss": 2.2752,
+      "step": 41794
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0209386852085968e-06,
+      "loss": 1.7743,
+      "step": 41795
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0208912768647283e-06,
+      "loss": 3.8678,
+      "step": 41796
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0208438685208598e-06,
+      "loss": 1.9737,
+      "step": 41797
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.020796460176991e-06,
+      "loss": 2.7756,
+      "step": 41798
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0207490518331227e-06,
+      "loss": 1.6707,
+      "step": 41799
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0207016434892542e-06,
+      "loss": 2.7472,
+      "step": 41800
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0206542351453855e-06,
+      "loss": 1.5534,
+      "step": 41801
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0206068268015172e-06,
+      "loss": 1.578,
+      "step": 41802
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0205594184576485e-06,
+      "loss": 1.1073,
+      "step": 41803
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.02051201011378e-06,
+      "loss": 2.6836,
+      "step": 41804
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0204646017699115e-06,
+      "loss": 2.4766,
+      "step": 41805
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.020417193426043e-06,
+      "loss": 2.4771,
+      "step": 41806
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0203697850821745e-06,
+      "loss": 3.0086,
+      "step": 41807
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.020322376738306e-06,
+      "loss": 1.2085,
+      "step": 41808
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0202749683944374e-06,
+      "loss": 2.0702,
+      "step": 41809
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.020227560050569e-06,
+      "loss": 1.668,
+      "step": 41810
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0201801517067004e-06,
+      "loss": 1.9037,
+      "step": 41811
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.020132743362832e-06,
+      "loss": 2.4715,
+      "step": 41812
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0200853350189634e-06,
+      "loss": 1.7807,
+      "step": 41813
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0200379266750947e-06,
+      "loss": 1.852,
+      "step": 41814
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0199905183312264e-06,
+      "loss": 2.5056,
+      "step": 41815
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0199431099873579e-06,
+      "loss": 0.8426,
+      "step": 41816
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0198957016434892e-06,
+      "loss": 1.7055,
+      "step": 41817
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0198482932996207e-06,
+      "loss": 2.5332,
+      "step": 41818
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0198008849557524e-06,
+      "loss": 1.4462,
+      "step": 41819
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0197534766118836e-06,
+      "loss": 2.6357,
+      "step": 41820
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0197060682680151e-06,
+      "loss": 1.6174,
+      "step": 41821
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0196586599241468e-06,
+      "loss": 2.2358,
+      "step": 41822
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0196112515802781e-06,
+      "loss": 2.858,
+      "step": 41823
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0195638432364096e-06,
+      "loss": 1.1577,
+      "step": 41824
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0195164348925411e-06,
+      "loss": 2.3525,
+      "step": 41825
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0194690265486726e-06,
+      "loss": 1.5107,
+      "step": 41826
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.019421618204804e-06,
+      "loss": 1.3881,
+      "step": 41827
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.0193742098609356e-06,
+      "loss": 1.3695,
+      "step": 41828
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 1.019326801517067e-06,
+      "loss": 1.939,
+      "step": 41829
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0192793931731986e-06,
+      "loss": 1.6964,
+      "step": 41830
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0192319848293299e-06,
+      "loss": 1.1665,
+      "step": 41831
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0191845764854616e-06,
+      "loss": 1.5908,
+      "step": 41832
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.019137168141593e-06,
+      "loss": 2.0136,
+      "step": 41833
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0190897597977243e-06,
+      "loss": 1.4486,
+      "step": 41834
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.019042351453856e-06,
+      "loss": 1.6295,
+      "step": 41835
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0189949431099873e-06,
+      "loss": 4.0501,
+      "step": 41836
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0189475347661188e-06,
+      "loss": 3.5538,
+      "step": 41837
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0189001264222503e-06,
+      "loss": 1.6547,
+      "step": 41838
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0188527180783818e-06,
+      "loss": 1.8076,
+      "step": 41839
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0188053097345133e-06,
+      "loss": 0.97,
+      "step": 41840
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0187579013906448e-06,
+      "loss": 1.8553,
+      "step": 41841
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0187104930467763e-06,
+      "loss": 0.282,
+      "step": 41842
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0186630847029078e-06,
+      "loss": 1.7461,
+      "step": 41843
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0186156763590393e-06,
+      "loss": 3.5259,
+      "step": 41844
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0185682680151707e-06,
+      "loss": 1.1036,
+      "step": 41845
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0185208596713022e-06,
+      "loss": 1.1596,
+      "step": 41846
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0184734513274335e-06,
+      "loss": 1.6833,
+      "step": 41847
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0184260429835652e-06,
+      "loss": 0.7184,
+      "step": 41848
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0183786346396967e-06,
+      "loss": 2.7406,
+      "step": 41849
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.018331226295828e-06,
+      "loss": 0.6091,
+      "step": 41850
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0182838179519595e-06,
+      "loss": 1.2942,
+      "step": 41851
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0182364096080912e-06,
+      "loss": 2.5379,
+      "step": 41852
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0181890012642225e-06,
+      "loss": 1.4335,
+      "step": 41853
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.018141592920354e-06,
+      "loss": 1.0459,
+      "step": 41854
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0180941845764857e-06,
+      "loss": 2.5745,
+      "step": 41855
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.018046776232617e-06,
+      "loss": 3.454,
+      "step": 41856
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0179993678887484e-06,
+      "loss": 1.868,
+      "step": 41857
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.01795195954488e-06,
+      "loss": 1.3274,
+      "step": 41858
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0179045512010114e-06,
+      "loss": 1.4369,
+      "step": 41859
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.017857142857143e-06,
+      "loss": 1.5158,
+      "step": 41860
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0178097345132744e-06,
+      "loss": 1.2519,
+      "step": 41861
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.017762326169406e-06,
+      "loss": 3.1624,
+      "step": 41862
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0177149178255374e-06,
+      "loss": 1.2524,
+      "step": 41863
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0176675094816687e-06,
+      "loss": 3.3746,
+      "step": 41864
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0176201011378004e-06,
+      "loss": 2.1468,
+      "step": 41865
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0175726927939319e-06,
+      "loss": 1.3404,
+      "step": 41866
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0175252844500631e-06,
+      "loss": 3.0808,
+      "step": 41867
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0174778761061949e-06,
+      "loss": 1.5391,
+      "step": 41868
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0174304677623261e-06,
+      "loss": 1.3258,
+      "step": 41869
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0173830594184576e-06,
+      "loss": 0.613,
+      "step": 41870
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0173356510745891e-06,
+      "loss": 2.4678,
+      "step": 41871
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0172882427307206e-06,
+      "loss": 2.7724,
+      "step": 41872
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.017240834386852e-06,
+      "loss": 1.1874,
+      "step": 41873
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0171934260429836e-06,
+      "loss": 2.128,
+      "step": 41874
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.017146017699115e-06,
+      "loss": 1.7207,
+      "step": 41875
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0170986093552466e-06,
+      "loss": 2.5766,
+      "step": 41876
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.017051201011378e-06,
+      "loss": 1.3048,
+      "step": 41877
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0170037926675096e-06,
+      "loss": 1.7689,
+      "step": 41878
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.016956384323641e-06,
+      "loss": 2.0677,
+      "step": 41879
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0169089759797723e-06,
+      "loss": 2.7323,
+      "step": 41880
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.016861567635904e-06,
+      "loss": 1.9205,
+      "step": 41881
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0168141592920355e-06,
+      "loss": 1.8811,
+      "step": 41882
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0167667509481668e-06,
+      "loss": 2.4129,
+      "step": 41883
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0167193426042983e-06,
+      "loss": 2.0029,
+      "step": 41884
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.01667193426043e-06,
+      "loss": 1.0621,
+      "step": 41885
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0166245259165613e-06,
+      "loss": 2.5602,
+      "step": 41886
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0165771175726928e-06,
+      "loss": 2.4372,
+      "step": 41887
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0165297092288245e-06,
+      "loss": 2.6509,
+      "step": 41888
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0164823008849558e-06,
+      "loss": 2.0187,
+      "step": 41889
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0164348925410873e-06,
+      "loss": 2.097,
+      "step": 41890
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0163874841972187e-06,
+      "loss": 2.0811,
+      "step": 41891
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 1.0163400758533502e-06,
+      "loss": 0.8609,
+      "step": 41892
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0162926675094817e-06,
+      "loss": 1.9031,
+      "step": 41893
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.016245259165613e-06,
+      "loss": 2.0315,
+      "step": 41894
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0161978508217447e-06,
+      "loss": 1.8488,
+      "step": 41895
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0161504424778762e-06,
+      "loss": 1.7558,
+      "step": 41896
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0161030341340075e-06,
+      "loss": 1.9421,
+      "step": 41897
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0160556257901392e-06,
+      "loss": 1.0959,
+      "step": 41898
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0160082174462707e-06,
+      "loss": 1.7192,
+      "step": 41899
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.015960809102402e-06,
+      "loss": 3.7006,
+      "step": 41900
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0159134007585335e-06,
+      "loss": 1.0004,
+      "step": 41901
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.015865992414665e-06,
+      "loss": 0.5815,
+      "step": 41902
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0158185840707964e-06,
+      "loss": 0.6306,
+      "step": 41903
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.015771175726928e-06,
+      "loss": 2.0985,
+      "step": 41904
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0157237673830594e-06,
+      "loss": 2.0122,
+      "step": 41905
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.015676359039191e-06,
+      "loss": 1.836,
+      "step": 41906
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0156289506953224e-06,
+      "loss": 1.8554,
+      "step": 41907
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.015581542351454e-06,
+      "loss": 1.9579,
+      "step": 41908
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0155341340075854e-06,
+      "loss": 2.1265,
+      "step": 41909
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0154867256637169e-06,
+      "loss": 2.5275,
+      "step": 41910
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0154393173198484e-06,
+      "loss": 2.1278,
+      "step": 41911
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0153919089759799e-06,
+      "loss": 1.0567,
+      "step": 41912
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0153445006321112e-06,
+      "loss": 1.6859,
+      "step": 41913
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0152970922882426e-06,
+      "loss": 2.025,
+      "step": 41914
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0152496839443743e-06,
+      "loss": 2.4574,
+      "step": 41915
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0152022756005056e-06,
+      "loss": 2.44,
+      "step": 41916
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0151548672566371e-06,
+      "loss": 2.892,
+      "step": 41917
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0151074589127688e-06,
+      "loss": 2.1478,
+      "step": 41918
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0150600505689e-06,
+      "loss": 1.8124,
+      "step": 41919
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0150126422250316e-06,
+      "loss": 1.568,
+      "step": 41920
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.014965233881163e-06,
+      "loss": 2.3773,
+      "step": 41921
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0149178255372946e-06,
+      "loss": 3.0511,
+      "step": 41922
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.014870417193426e-06,
+      "loss": 1.6863,
+      "step": 41923
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0148230088495576e-06,
+      "loss": 1.6581,
+      "step": 41924
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.014775600505689e-06,
+      "loss": 2.8953,
+      "step": 41925
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0147281921618205e-06,
+      "loss": 1.0213,
+      "step": 41926
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0146807838179518e-06,
+      "loss": 1.3992,
+      "step": 41927
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0146333754740835e-06,
+      "loss": 3.4696,
+      "step": 41928
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.014585967130215e-06,
+      "loss": 2.2113,
+      "step": 41929
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0145385587863463e-06,
+      "loss": 1.3039,
+      "step": 41930
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.014491150442478e-06,
+      "loss": 1.657,
+      "step": 41931
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0144437420986095e-06,
+      "loss": 2.8986,
+      "step": 41932
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0143963337547408e-06,
+      "loss": 1.4325,
+      "step": 41933
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0143489254108723e-06,
+      "loss": 2.0491,
+      "step": 41934
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0143015170670038e-06,
+      "loss": 3.0717,
+      "step": 41935
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0142541087231353e-06,
+      "loss": 1.1877,
+      "step": 41936
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0142067003792668e-06,
+      "loss": 2.0545,
+      "step": 41937
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0141592920353982e-06,
+      "loss": 2.5442,
+      "step": 41938
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0141118836915297e-06,
+      "loss": 3.139,
+      "step": 41939
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0140644753476612e-06,
+      "loss": 3.3073,
+      "step": 41940
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0140170670037927e-06,
+      "loss": 2.5098,
+      "step": 41941
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0139696586599242e-06,
+      "loss": 0.7414,
+      "step": 41942
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0139222503160557e-06,
+      "loss": 1.6237,
+      "step": 41943
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0138748419721872e-06,
+      "loss": 3.2334,
+      "step": 41944
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0138274336283187e-06,
+      "loss": 2.0689,
+      "step": 41945
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.01378002528445e-06,
+      "loss": 2.0033,
+      "step": 41946
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0137326169405815e-06,
+      "loss": 3.1636,
+      "step": 41947
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0136852085967132e-06,
+      "loss": 1.6557,
+      "step": 41948
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0136378002528444e-06,
+      "loss": 1.783,
+      "step": 41949
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.013590391908976e-06,
+      "loss": 1.901,
+      "step": 41950
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0135429835651076e-06,
+      "loss": 2.8387,
+      "step": 41951
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.013495575221239e-06,
+      "loss": 2.8027,
+      "step": 41952
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0134481668773704e-06,
+      "loss": 1.5703,
+      "step": 41953
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.013400758533502e-06,
+      "loss": 2.9009,
+      "step": 41954
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0133533501896334e-06,
+      "loss": 2.8584,
+      "step": 41955
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 1.0133059418457649e-06,
+      "loss": 2.2587,
+      "step": 41956
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0132585335018964e-06,
+      "loss": 3.1729,
+      "step": 41957
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0132111251580279e-06,
+      "loss": 1.2281,
+      "step": 41958
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0131637168141594e-06,
+      "loss": 2.8277,
+      "step": 41959
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0131163084702906e-06,
+      "loss": 2.5461,
+      "step": 41960
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0130689001264224e-06,
+      "loss": 3.0889,
+      "step": 41961
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0130214917825538e-06,
+      "loss": 1.0942,
+      "step": 41962
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0129740834386851e-06,
+      "loss": 1.2994,
+      "step": 41963
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0129266750948168e-06,
+      "loss": 2.0454,
+      "step": 41964
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0128792667509483e-06,
+      "loss": 2.6518,
+      "step": 41965
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0128318584070796e-06,
+      "loss": 3.6517,
+      "step": 41966
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.012784450063211e-06,
+      "loss": 2.0059,
+      "step": 41967
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0127370417193426e-06,
+      "loss": 1.6006,
+      "step": 41968
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.012689633375474e-06,
+      "loss": 2.2144,
+      "step": 41969
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0126422250316056e-06,
+      "loss": 1.5014,
+      "step": 41970
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.012594816687737e-06,
+      "loss": 2.6131,
+      "step": 41971
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0125474083438686e-06,
+      "loss": 0.8079,
+      "step": 41972
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0125e-06,
+      "loss": 1.6296,
+      "step": 41973
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0124525916561315e-06,
+      "loss": 2.2605,
+      "step": 41974
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.012405183312263e-06,
+      "loss": 1.8538,
+      "step": 41975
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0123577749683945e-06,
+      "loss": 2.4609,
+      "step": 41976
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.012310366624526e-06,
+      "loss": 1.5858,
+      "step": 41977
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0122629582806575e-06,
+      "loss": 1.5203,
+      "step": 41978
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0122155499367888e-06,
+      "loss": 2.8746,
+      "step": 41979
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0121681415929203e-06,
+      "loss": 2.6658,
+      "step": 41980
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.012120733249052e-06,
+      "loss": 3.9775,
+      "step": 41981
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0120733249051833e-06,
+      "loss": 2.5973,
+      "step": 41982
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0120259165613148e-06,
+      "loss": 1.3915,
+      "step": 41983
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0119785082174465e-06,
+      "loss": 1.8466,
+      "step": 41984
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0119310998735777e-06,
+      "loss": 1.5247,
+      "step": 41985
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0118836915297092e-06,
+      "loss": 2.0864,
+      "step": 41986
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0118362831858407e-06,
+      "loss": 2.0322,
+      "step": 41987
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0117888748419722e-06,
+      "loss": 0.8344,
+      "step": 41988
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0117414664981037e-06,
+      "loss": 4.4574,
+      "step": 41989
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0116940581542352e-06,
+      "loss": 2.146,
+      "step": 41990
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0116466498103667e-06,
+      "loss": 1.7185,
+      "step": 41991
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0115992414664982e-06,
+      "loss": 2.2969,
+      "step": 41992
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0115518331226295e-06,
+      "loss": 0.7411,
+      "step": 41993
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0115044247787612e-06,
+      "loss": 1.9657,
+      "step": 41994
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0114570164348927e-06,
+      "loss": 3.0501,
+      "step": 41995
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.011409608091024e-06,
+      "loss": 2.6497,
+      "step": 41996
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0113621997471556e-06,
+      "loss": 2.8125,
+      "step": 41997
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0113147914032871e-06,
+      "loss": 2.6063,
+      "step": 41998
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0112673830594184e-06,
+      "loss": 1.4805,
+      "step": 41999
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.01121997471555e-06,
+      "loss": 1.591,
+      "step": 42000
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0111725663716814e-06,
+      "loss": 2.118,
+      "step": 42001
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.011125158027813e-06,
+      "loss": 2.7867,
+      "step": 42002
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0110777496839444e-06,
+      "loss": 3.0066,
+      "step": 42003
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0110303413400759e-06,
+      "loss": 2.189,
+      "step": 42004
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0109829329962074e-06,
+      "loss": 2.1754,
+      "step": 42005
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0109355246523389e-06,
+      "loss": 1.0527,
+      "step": 42006
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0108881163084704e-06,
+      "loss": 1.296,
+      "step": 42007
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0108407079646018e-06,
+      "loss": 2.3127,
+      "step": 42008
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0107932996207333e-06,
+      "loss": 2.0359,
+      "step": 42009
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0107458912768648e-06,
+      "loss": 2.6254,
+      "step": 42010
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0106984829329963e-06,
+      "loss": 1.2501,
+      "step": 42011
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0106510745891276e-06,
+      "loss": 2.8935,
+      "step": 42012
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.010603666245259e-06,
+      "loss": 2.783,
+      "step": 42013
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0105562579013908e-06,
+      "loss": 1.7901,
+      "step": 42014
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.010508849557522e-06,
+      "loss": 2.2774,
+      "step": 42015
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0104614412136536e-06,
+      "loss": 3.628,
+      "step": 42016
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0104140328697853e-06,
+      "loss": 2.5329,
+      "step": 42017
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.0103666245259166e-06,
+      "loss": 1.9581,
+      "step": 42018
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 1.010319216182048e-06,
+      "loss": 1.7723,
+      "step": 42019
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0102718078381795e-06,
+      "loss": 1.7325,
+      "step": 42020
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.010224399494311e-06,
+      "loss": 1.7797,
+      "step": 42021
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0101769911504425e-06,
+      "loss": 1.9057,
+      "step": 42022
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.010129582806574e-06,
+      "loss": 3.1921,
+      "step": 42023
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0100821744627055e-06,
+      "loss": 2.3381,
+      "step": 42024
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.010034766118837e-06,
+      "loss": 0.4633,
+      "step": 42025
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0099873577749683e-06,
+      "loss": 2.0472,
+      "step": 42026
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0099399494311e-06,
+      "loss": 0.9619,
+      "step": 42027
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0098925410872315e-06,
+      "loss": 1.8631,
+      "step": 42028
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0098451327433628e-06,
+      "loss": 1.6576,
+      "step": 42029
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0097977243994945e-06,
+      "loss": 2.5017,
+      "step": 42030
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0097503160556257e-06,
+      "loss": 2.1079,
+      "step": 42031
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0097029077117572e-06,
+      "loss": 1.0754,
+      "step": 42032
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0096554993678887e-06,
+      "loss": 1.1037,
+      "step": 42033
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0096080910240202e-06,
+      "loss": 2.3154,
+      "step": 42034
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0095606826801517e-06,
+      "loss": 1.6232,
+      "step": 42035
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0095132743362832e-06,
+      "loss": 0.8547,
+      "step": 42036
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0094658659924147e-06,
+      "loss": 1.2283,
+      "step": 42037
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0094184576485462e-06,
+      "loss": 2.754,
+      "step": 42038
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0093710493046777e-06,
+      "loss": 1.5284,
+      "step": 42039
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0093236409608092e-06,
+      "loss": 1.4553,
+      "step": 42040
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0092762326169407e-06,
+      "loss": 1.882,
+      "step": 42041
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.009228824273072e-06,
+      "loss": 2.5958,
+      "step": 42042
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0091814159292034e-06,
+      "loss": 0.4529,
+      "step": 42043
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0091340075853351e-06,
+      "loss": 1.7109,
+      "step": 42044
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0090865992414664e-06,
+      "loss": 1.1768,
+      "step": 42045
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.009039190897598e-06,
+      "loss": 1.4774,
+      "step": 42046
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0089917825537296e-06,
+      "loss": 1.0878,
+      "step": 42047
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.008944374209861e-06,
+      "loss": 2.052,
+      "step": 42048
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0088969658659924e-06,
+      "loss": 1.7864,
+      "step": 42049
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0088495575221239e-06,
+      "loss": 2.1058,
+      "step": 42050
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0088021491782554e-06,
+      "loss": 2.0956,
+      "step": 42051
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0087547408343869e-06,
+      "loss": 3.2074,
+      "step": 42052
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0087073324905184e-06,
+      "loss": 0.7727,
+      "step": 42053
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0086599241466499e-06,
+      "loss": 2.0784,
+      "step": 42054
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0086125158027813e-06,
+      "loss": 2.4121,
+      "step": 42055
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0085651074589126e-06,
+      "loss": 2.1779,
+      "step": 42056
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0085176991150443e-06,
+      "loss": 1.966,
+      "step": 42057
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0084702907711758e-06,
+      "loss": 2.4505,
+      "step": 42058
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.008422882427307e-06,
+      "loss": 2.6746,
+      "step": 42059
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0083754740834388e-06,
+      "loss": 0.7862,
+      "step": 42060
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0083280657395703e-06,
+      "loss": 3.554,
+      "step": 42061
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0082806573957016e-06,
+      "loss": 0.9108,
+      "step": 42062
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.008233249051833e-06,
+      "loss": 1.5565,
+      "step": 42063
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0081858407079646e-06,
+      "loss": 1.3916,
+      "step": 42064
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.008138432364096e-06,
+      "loss": 3.244,
+      "step": 42065
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0080910240202275e-06,
+      "loss": 1.0001,
+      "step": 42066
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.008043615676359e-06,
+      "loss": 2.4507,
+      "step": 42067
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0079962073324905e-06,
+      "loss": 3.6246,
+      "step": 42068
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.007948798988622e-06,
+      "loss": 3.6576,
+      "step": 42069
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0079013906447535e-06,
+      "loss": 3.2913,
+      "step": 42070
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.007853982300885e-06,
+      "loss": 1.7336,
+      "step": 42071
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0078065739570165e-06,
+      "loss": 1.8624,
+      "step": 42072
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.007759165613148e-06,
+      "loss": 2.5334,
+      "step": 42073
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0077117572692795e-06,
+      "loss": 1.4457,
+      "step": 42074
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0076643489254108e-06,
+      "loss": 2.4993,
+      "step": 42075
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0076169405815423e-06,
+      "loss": 1.462,
+      "step": 42076
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.007569532237674e-06,
+      "loss": 2.0047,
+      "step": 42077
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0075221238938052e-06,
+      "loss": 1.9855,
+      "step": 42078
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0074747155499367e-06,
+      "loss": 1.3181,
+      "step": 42079
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0074273072060684e-06,
+      "loss": 0.9322,
+      "step": 42080
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0073798988621997e-06,
+      "loss": 1.4946,
+      "step": 42081
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 1.0073324905183312e-06,
+      "loss": 2.9833,
+      "step": 42082
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0072850821744627e-06,
+      "loss": 0.8442,
+      "step": 42083
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0072376738305942e-06,
+      "loss": 2.2223,
+      "step": 42084
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0071902654867257e-06,
+      "loss": 1.2332,
+      "step": 42085
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0071428571428572e-06,
+      "loss": 2.0222,
+      "step": 42086
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0070954487989887e-06,
+      "loss": 2.1438,
+      "step": 42087
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0070480404551202e-06,
+      "loss": 2.5852,
+      "step": 42088
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0070006321112514e-06,
+      "loss": 2.7641,
+      "step": 42089
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0069532237673831e-06,
+      "loss": 2.1899,
+      "step": 42090
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0069058154235146e-06,
+      "loss": 3.0436,
+      "step": 42091
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.006858407079646e-06,
+      "loss": 1.7093,
+      "step": 42092
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0068109987357776e-06,
+      "loss": 1.8397,
+      "step": 42093
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0067635903919091e-06,
+      "loss": 1.3972,
+      "step": 42094
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0067161820480404e-06,
+      "loss": 3.0254,
+      "step": 42095
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0066687737041719e-06,
+      "loss": 1.7293,
+      "step": 42096
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0066213653603034e-06,
+      "loss": 1.6108,
+      "step": 42097
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0065739570164349e-06,
+      "loss": 2.3089,
+      "step": 42098
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0065265486725664e-06,
+      "loss": 2.4528,
+      "step": 42099
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0064791403286979e-06,
+      "loss": 1.4384,
+      "step": 42100
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0064317319848294e-06,
+      "loss": 1.6722,
+      "step": 42101
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0063843236409608e-06,
+      "loss": 3.3342,
+      "step": 42102
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0063369152970923e-06,
+      "loss": 1.7021,
+      "step": 42103
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0062895069532238e-06,
+      "loss": 1.4853,
+      "step": 42104
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0062420986093553e-06,
+      "loss": 3.4485,
+      "step": 42105
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0061946902654868e-06,
+      "loss": 2.1961,
+      "step": 42106
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0061472819216183e-06,
+      "loss": 2.9,
+      "step": 42107
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0060998735777496e-06,
+      "loss": 3.1441,
+      "step": 42108
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.006052465233881e-06,
+      "loss": 2.3045,
+      "step": 42109
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0060050568900128e-06,
+      "loss": 1.5941,
+      "step": 42110
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.005957648546144e-06,
+      "loss": 1.4271,
+      "step": 42111
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0059102402022756e-06,
+      "loss": 1.5453,
+      "step": 42112
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0058628318584073e-06,
+      "loss": 0.9734,
+      "step": 42113
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0058154235145385e-06,
+      "loss": 0.716,
+      "step": 42114
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.00576801517067e-06,
+      "loss": 2.0261,
+      "step": 42115
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0057206068268015e-06,
+      "loss": 2.2501,
+      "step": 42116
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.005673198482933e-06,
+      "loss": 3.5341,
+      "step": 42117
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0056257901390645e-06,
+      "loss": 1.7233,
+      "step": 42118
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.005578381795196e-06,
+      "loss": 2.0709,
+      "step": 42119
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0055309734513275e-06,
+      "loss": 2.5294,
+      "step": 42120
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.005483565107459e-06,
+      "loss": 2.223,
+      "step": 42121
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0054361567635903e-06,
+      "loss": 3.0802,
+      "step": 42122
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.005388748419722e-06,
+      "loss": 1.9328,
+      "step": 42123
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0053413400758535e-06,
+      "loss": 3.0251,
+      "step": 42124
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0052939317319847e-06,
+      "loss": 2.0019,
+      "step": 42125
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0052465233881164e-06,
+      "loss": 3.3399,
+      "step": 42126
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.005199115044248e-06,
+      "loss": 2.2796,
+      "step": 42127
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0051517067003792e-06,
+      "loss": 2.3719,
+      "step": 42128
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0051042983565107e-06,
+      "loss": 1.1531,
+      "step": 42129
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0050568900126422e-06,
+      "loss": 2.2553,
+      "step": 42130
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0050094816687737e-06,
+      "loss": 1.413,
+      "step": 42131
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0049620733249052e-06,
+      "loss": 3.0607,
+      "step": 42132
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0049146649810367e-06,
+      "loss": 1.7584,
+      "step": 42133
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0048672566371682e-06,
+      "loss": 1.5074,
+      "step": 42134
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0048198482932997e-06,
+      "loss": 2.1725,
+      "step": 42135
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0047724399494312e-06,
+      "loss": 0.7887,
+      "step": 42136
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0047250316055626e-06,
+      "loss": 1.0871,
+      "step": 42137
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0046776232616941e-06,
+      "loss": 2.2857,
+      "step": 42138
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0046302149178256e-06,
+      "loss": 0.9418,
+      "step": 42139
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0045828065739571e-06,
+      "loss": 1.8519,
+      "step": 42140
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0045353982300884e-06,
+      "loss": 3.4074,
+      "step": 42141
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.00448798988622e-06,
+      "loss": 3.6321,
+      "step": 42142
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0044405815423516e-06,
+      "loss": 0.8153,
+      "step": 42143
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0043931731984829e-06,
+      "loss": 1.588,
+      "step": 42144
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.0043457648546144e-06,
+      "loss": 0.8275,
+      "step": 42145
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 1.004298356510746e-06,
+      "loss": 1.4566,
+      "step": 42146
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0042509481668774e-06,
+      "loss": 1.4924,
+      "step": 42147
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0042035398230088e-06,
+      "loss": 2.2157,
+      "step": 42148
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0041561314791403e-06,
+      "loss": 2.2312,
+      "step": 42149
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0041087231352718e-06,
+      "loss": 1.5548,
+      "step": 42150
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0040613147914033e-06,
+      "loss": 0.7417,
+      "step": 42151
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0040139064475348e-06,
+      "loss": 4.021,
+      "step": 42152
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0039664981036663e-06,
+      "loss": 0.8023,
+      "step": 42153
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0039190897597978e-06,
+      "loss": 2.6152,
+      "step": 42154
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003871681415929e-06,
+      "loss": 2.0103,
+      "step": 42155
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0038242730720608e-06,
+      "loss": 1.6922,
+      "step": 42156
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0037768647281923e-06,
+      "loss": 1.8204,
+      "step": 42157
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0037294563843236e-06,
+      "loss": 1.8936,
+      "step": 42158
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0036820480404553e-06,
+      "loss": 1.6973,
+      "step": 42159
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0036346396965868e-06,
+      "loss": 1.7572,
+      "step": 42160
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003587231352718e-06,
+      "loss": 1.8774,
+      "step": 42161
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0035398230088495e-06,
+      "loss": 1.0989,
+      "step": 42162
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003492414664981e-06,
+      "loss": 1.763,
+      "step": 42163
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0034450063211125e-06,
+      "loss": 2.0082,
+      "step": 42164
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003397597977244e-06,
+      "loss": 3.0309,
+      "step": 42165
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0033501896333755e-06,
+      "loss": 1.662,
+      "step": 42166
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003302781289507e-06,
+      "loss": 2.0449,
+      "step": 42167
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0032553729456385e-06,
+      "loss": 1.7024,
+      "step": 42168
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.00320796460177e-06,
+      "loss": 1.4177,
+      "step": 42169
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0031605562579015e-06,
+      "loss": 0.7697,
+      "step": 42170
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003113147914033e-06,
+      "loss": 2.0525,
+      "step": 42171
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0030657395701642e-06,
+      "loss": 3.5034,
+      "step": 42172
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.003018331226296e-06,
+      "loss": 1.3752,
+      "step": 42173
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0029709228824272e-06,
+      "loss": 2.3496,
+      "step": 42174
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0029235145385587e-06,
+      "loss": 2.5442,
+      "step": 42175
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0028761061946904e-06,
+      "loss": 1.5918,
+      "step": 42176
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0028286978508217e-06,
+      "loss": 2.116,
+      "step": 42177
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0027812895069532e-06,
+      "loss": 1.9177,
+      "step": 42178
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0027338811630847e-06,
+      "loss": 2.791,
+      "step": 42179
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0026864728192162e-06,
+      "loss": 1.4958,
+      "step": 42180
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0026390644753477e-06,
+      "loss": 1.0759,
+      "step": 42181
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0025916561314792e-06,
+      "loss": 1.7784,
+      "step": 42182
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0025442477876107e-06,
+      "loss": 1.0981,
+      "step": 42183
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0024968394437421e-06,
+      "loss": 1.3819,
+      "step": 42184
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0024494310998734e-06,
+      "loss": 1.6414,
+      "step": 42185
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0024020227560051e-06,
+      "loss": 2.6472,
+      "step": 42186
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0023546144121366e-06,
+      "loss": 2.3057,
+      "step": 42187
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.002307206068268e-06,
+      "loss": 2.5092,
+      "step": 42188
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0022597977243996e-06,
+      "loss": 2.645,
+      "step": 42189
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.002212389380531e-06,
+      "loss": 1.8842,
+      "step": 42190
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0021649810366624e-06,
+      "loss": 1.0549,
+      "step": 42191
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0021175726927939e-06,
+      "loss": 1.8956,
+      "step": 42192
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0020701643489256e-06,
+      "loss": 2.4385,
+      "step": 42193
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0020227560050569e-06,
+      "loss": 2.7261,
+      "step": 42194
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0019753476611883e-06,
+      "loss": 0.9989,
+      "step": 42195
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0019279393173198e-06,
+      "loss": 1.1303,
+      "step": 42196
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0018805309734513e-06,
+      "loss": 1.119,
+      "step": 42197
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0018331226295828e-06,
+      "loss": 1.4877,
+      "step": 42198
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0017857142857143e-06,
+      "loss": 1.8557,
+      "step": 42199
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0017383059418458e-06,
+      "loss": 2.9655,
+      "step": 42200
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0016908975979773e-06,
+      "loss": 2.8189,
+      "step": 42201
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0016434892541088e-06,
+      "loss": 2.1825,
+      "step": 42202
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0015960809102403e-06,
+      "loss": 4.0769,
+      "step": 42203
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0015486725663718e-06,
+      "loss": 2.1015,
+      "step": 42204
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.001501264222503e-06,
+      "loss": 1.5136,
+      "step": 42205
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0014538558786348e-06,
+      "loss": 1.5154,
+      "step": 42206
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.001406447534766e-06,
+      "loss": 1.6598,
+      "step": 42207
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0013590391908975e-06,
+      "loss": 0.773,
+      "step": 42208
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 1.0013116308470292e-06,
+      "loss": 2.6154,
+      "step": 42209
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0012642225031605e-06,
+      "loss": 1.7745,
+      "step": 42210
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.001216814159292e-06,
+      "loss": 1.7896,
+      "step": 42211
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0011694058154235e-06,
+      "loss": 1.7731,
+      "step": 42212
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.001121997471555e-06,
+      "loss": 1.3758,
+      "step": 42213
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0010745891276865e-06,
+      "loss": 2.2774,
+      "step": 42214
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.001027180783818e-06,
+      "loss": 2.5437,
+      "step": 42215
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0009797724399495e-06,
+      "loss": 3.6205,
+      "step": 42216
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.000932364096081e-06,
+      "loss": 1.3824,
+      "step": 42217
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0008849557522122e-06,
+      "loss": 0.9844,
+      "step": 42218
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.000837547408344e-06,
+      "loss": 1.4415,
+      "step": 42219
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0007901390644754e-06,
+      "loss": 1.4986,
+      "step": 42220
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0007427307206067e-06,
+      "loss": 3.1594,
+      "step": 42221
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0006953223767384e-06,
+      "loss": 1.8974,
+      "step": 42222
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.00064791403287e-06,
+      "loss": 2.475,
+      "step": 42223
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0006005056890012e-06,
+      "loss": 1.9896,
+      "step": 42224
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0005530973451327e-06,
+      "loss": 1.0568,
+      "step": 42225
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0005056890012644e-06,
+      "loss": 1.8771,
+      "step": 42226
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0004582806573957e-06,
+      "loss": 3.4088,
+      "step": 42227
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0004108723135272e-06,
+      "loss": 2.6351,
+      "step": 42228
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0003634639696587e-06,
+      "loss": 0.7754,
+      "step": 42229
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0003160556257901e-06,
+      "loss": 2.6957,
+      "step": 42230
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0002686472819216e-06,
+      "loss": 1.5121,
+      "step": 42231
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0002212389380531e-06,
+      "loss": 2.8051,
+      "step": 42232
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0001738305941846e-06,
+      "loss": 2.3101,
+      "step": 42233
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0001264222503161e-06,
+      "loss": 1.0263,
+      "step": 42234
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.0000790139064476e-06,
+      "loss": 2.2172,
+      "step": 42235
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 1.000031605562579e-06,
+      "loss": 1.8129,
+      "step": 42236
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.999841972187106e-07,
+      "loss": 1.7769,
+      "step": 42237
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.999367888748419e-07,
+      "loss": 3.3328,
+      "step": 42238
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.998893805309736e-07,
+      "loss": 2.5829,
+      "step": 42239
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.998419721871049e-07,
+      "loss": 1.6627,
+      "step": 42240
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.997945638432364e-07,
+      "loss": 3.0579,
+      "step": 42241
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.99747155499368e-07,
+      "loss": 1.3711,
+      "step": 42242
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.996997471554993e-07,
+      "loss": 1.9637,
+      "step": 42243
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.996523388116308e-07,
+      "loss": 2.0875,
+      "step": 42244
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.996049304677623e-07,
+      "loss": 1.7439,
+      "step": 42245
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.995575221238938e-07,
+      "loss": 1.9999,
+      "step": 42246
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.995101137800253e-07,
+      "loss": 1.4467,
+      "step": 42247
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.994627054361568e-07,
+      "loss": 2.4133,
+      "step": 42248
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.994152970922883e-07,
+      "loss": 1.8185,
+      "step": 42249
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.993678887484198e-07,
+      "loss": 2.1662,
+      "step": 42250
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.99320480404551e-07,
+      "loss": 1.2485,
+      "step": 42251
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.992730720606828e-07,
+      "loss": 2.374,
+      "step": 42252
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.992256637168143e-07,
+      "loss": 1.6225,
+      "step": 42253
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.991782553729455e-07,
+      "loss": 2.891,
+      "step": 42254
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.991308470290772e-07,
+      "loss": 1.9843,
+      "step": 42255
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.990834386852087e-07,
+      "loss": 2.0287,
+      "step": 42256
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.9903603034134e-07,
+      "loss": 2.0387,
+      "step": 42257
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.989886219974715e-07,
+      "loss": 1.6483,
+      "step": 42258
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.989412136536032e-07,
+      "loss": 1.7431,
+      "step": 42259
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.988938053097345e-07,
+      "loss": 1.0376,
+      "step": 42260
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.98846396965866e-07,
+      "loss": 2.4848,
+      "step": 42261
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.987989886219975e-07,
+      "loss": 1.3812,
+      "step": 42262
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.98751580278129e-07,
+      "loss": 2.1172,
+      "step": 42263
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.987041719342605e-07,
+      "loss": 1.0918,
+      "step": 42264
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.98656763590392e-07,
+      "loss": 1.9357,
+      "step": 42265
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.986093552465234e-07,
+      "loss": 2.5837,
+      "step": 42266
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.98561946902655e-07,
+      "loss": 3.473,
+      "step": 42267
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.985145385587864e-07,
+      "loss": 1.9203,
+      "step": 42268
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.98467130214918e-07,
+      "loss": 1.8968,
+      "step": 42269
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.984197218710494e-07,
+      "loss": 1.7349,
+      "step": 42270
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.983723135271807e-07,
+      "loss": 0.7277,
+      "step": 42271
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.983249051833124e-07,
+      "loss": 2.2885,
+      "step": 42272
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.982774968394437e-07,
+      "loss": 2.7595,
+      "step": 42273
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.982300884955752e-07,
+      "loss": 1.5423,
+      "step": 42274
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.981826801517069e-07,
+      "loss": 1.3,
+      "step": 42275
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.981352718078382e-07,
+      "loss": 1.767,
+      "step": 42276
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.980878634639696e-07,
+      "loss": 1.663,
+      "step": 42277
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.980404551201011e-07,
+      "loss": 1.2508,
+      "step": 42278
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.979930467762326e-07,
+      "loss": 1.4868,
+      "step": 42279
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.979456384323641e-07,
+      "loss": 1.4667,
+      "step": 42280
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.978982300884956e-07,
+      "loss": 2.0109,
+      "step": 42281
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.978508217446271e-07,
+      "loss": 1.3425,
+      "step": 42282
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.978034134007586e-07,
+      "loss": 2.0224,
+      "step": 42283
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.977560050568899e-07,
+      "loss": 1.429,
+      "step": 42284
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.977085967130216e-07,
+      "loss": 1.327,
+      "step": 42285
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.97661188369153e-07,
+      "loss": 2.3154,
+      "step": 42286
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.976137800252844e-07,
+      "loss": 1.9502,
+      "step": 42287
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.97566371681416e-07,
+      "loss": 2.8006,
+      "step": 42288
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.975189633375476e-07,
+      "loss": 2.3868,
+      "step": 42289
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.974715549936788e-07,
+      "loss": 2.025,
+      "step": 42290
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.974241466498103e-07,
+      "loss": 2.6716,
+      "step": 42291
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.97376738305942e-07,
+      "loss": 2.0415,
+      "step": 42292
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.973293299620733e-07,
+      "loss": 1.5899,
+      "step": 42293
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.972819216182048e-07,
+      "loss": 1.481,
+      "step": 42294
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.972345132743363e-07,
+      "loss": 2.5049,
+      "step": 42295
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.971871049304678e-07,
+      "loss": 1.9203,
+      "step": 42296
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.971396965865993e-07,
+      "loss": 1.1929,
+      "step": 42297
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.970922882427308e-07,
+      "loss": 1.4814,
+      "step": 42298
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.970448798988623e-07,
+      "loss": 1.4254,
+      "step": 42299
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.969974715549938e-07,
+      "loss": 1.8127,
+      "step": 42300
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.969500632111252e-07,
+      "loss": 2.0443,
+      "step": 42301
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.969026548672567e-07,
+      "loss": 1.7465,
+      "step": 42302
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.968552465233882e-07,
+      "loss": 0.5318,
+      "step": 42303
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.968078381795195e-07,
+      "loss": 1.9362,
+      "step": 42304
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.967604298356512e-07,
+      "loss": 1.5049,
+      "step": 42305
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.967130214917825e-07,
+      "loss": 1.9592,
+      "step": 42306
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.96665613147914e-07,
+      "loss": 2.351,
+      "step": 42307
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.966182048040457e-07,
+      "loss": 2.4604,
+      "step": 42308
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.96570796460177e-07,
+      "loss": 1.2939,
+      "step": 42309
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.965233881163085e-07,
+      "loss": 2.1158,
+      "step": 42310
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.9647597977244e-07,
+      "loss": 1.613,
+      "step": 42311
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.964285714285714e-07,
+      "loss": 3.4063,
+      "step": 42312
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.96381163084703e-07,
+      "loss": 1.1127,
+      "step": 42313
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.963337547408344e-07,
+      "loss": 1.1573,
+      "step": 42314
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.96286346396966e-07,
+      "loss": 2.6116,
+      "step": 42315
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.962389380530974e-07,
+      "loss": 3.8024,
+      "step": 42316
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.961915297092287e-07,
+      "loss": 2.6897,
+      "step": 42317
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.961441213653604e-07,
+      "loss": 1.7891,
+      "step": 42318
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.960967130214919e-07,
+      "loss": 2.2141,
+      "step": 42319
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.960493046776232e-07,
+      "loss": 1.477,
+      "step": 42320
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.960018963337547e-07,
+      "loss": 1.9945,
+      "step": 42321
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.959544879898864e-07,
+      "loss": 2.5495,
+      "step": 42322
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.959070796460176e-07,
+      "loss": 3.2196,
+      "step": 42323
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.958596713021491e-07,
+      "loss": 2.9106,
+      "step": 42324
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.958122629582808e-07,
+      "loss": 2.6208,
+      "step": 42325
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.957648546144121e-07,
+      "loss": 0.625,
+      "step": 42326
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.957174462705436e-07,
+      "loss": 1.4691,
+      "step": 42327
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.956700379266751e-07,
+      "loss": 1.943,
+      "step": 42328
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.956226295828066e-07,
+      "loss": 0.8632,
+      "step": 42329
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.95575221238938e-07,
+      "loss": 1.87,
+      "step": 42330
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.955278128950696e-07,
+      "loss": 2.2098,
+      "step": 42331
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.95480404551201e-07,
+      "loss": 2.7296,
+      "step": 42332
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.954329962073326e-07,
+      "loss": 1.7772,
+      "step": 42333
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.953855878634639e-07,
+      "loss": 2.3168,
+      "step": 42334
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.953381795195956e-07,
+      "loss": 1.8015,
+      "step": 42335
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.95290771175727e-07,
+      "loss": 1.4707,
+      "step": 42336
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.952433628318583e-07,
+      "loss": 1.3112,
+      "step": 42337
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.9519595448799e-07,
+      "loss": 2.0015,
+      "step": 42338
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.951485461441213e-07,
+      "loss": 2.1637,
+      "step": 42339
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.951011378002528e-07,
+      "loss": 3.2487,
+      "step": 42340
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.950537294563843e-07,
+      "loss": 1.8195,
+      "step": 42341
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.950063211125158e-07,
+      "loss": 1.7168,
+      "step": 42342
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.949589127686473e-07,
+      "loss": 3.3137,
+      "step": 42343
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.949115044247788e-07,
+      "loss": 2.6518,
+      "step": 42344
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.948640960809103e-07,
+      "loss": 2.9112,
+      "step": 42345
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.948166877370418e-07,
+      "loss": 2.0328,
+      "step": 42346
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.947692793931733e-07,
+      "loss": 1.2966,
+      "step": 42347
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.947218710493047e-07,
+      "loss": 1.7605,
+      "step": 42348
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.946744627054362e-07,
+      "loss": 3.5391,
+      "step": 42349
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.946270543615675e-07,
+      "loss": 2.7211,
+      "step": 42350
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.945796460176992e-07,
+      "loss": 1.8184,
+      "step": 42351
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.945322376738307e-07,
+      "loss": 0.5545,
+      "step": 42352
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.94484829329962e-07,
+      "loss": 2.793,
+      "step": 42353
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.944374209860935e-07,
+      "loss": 1.7625,
+      "step": 42354
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.943900126422252e-07,
+      "loss": 2.3067,
+      "step": 42355
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.943426042983565e-07,
+      "loss": 3.7581,
+      "step": 42356
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.94295195954488e-07,
+      "loss": 1.1143,
+      "step": 42357
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.942477876106197e-07,
+      "loss": 1.4826,
+      "step": 42358
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.94200379266751e-07,
+      "loss": 1.4829,
+      "step": 42359
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.941529709228824e-07,
+      "loss": 1.6028,
+      "step": 42360
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.94105562579014e-07,
+      "loss": 1.3043,
+      "step": 42361
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.940581542351454e-07,
+      "loss": 1.9704,
+      "step": 42362
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.94010745891277e-07,
+      "loss": 1.5323,
+      "step": 42363
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.939633375474084e-07,
+      "loss": 2.5963,
+      "step": 42364
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.9391592920354e-07,
+      "loss": 0.9885,
+      "step": 42365
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.938685208596714e-07,
+      "loss": 1.892,
+      "step": 42366
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.938211125158027e-07,
+      "loss": 2.0293,
+      "step": 42367
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.937737041719344e-07,
+      "loss": 0.6201,
+      "step": 42368
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.937262958280659e-07,
+      "loss": 3.3246,
+      "step": 42369
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.936788874841971e-07,
+      "loss": 1.8427,
+      "step": 42370
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.936314791403289e-07,
+      "loss": 2.3406,
+      "step": 42371
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.935840707964601e-07,
+      "loss": 2.6901,
+      "step": 42372
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.935366624525916e-07,
+      "loss": 1.6105,
+      "step": 42373
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.934892541087231e-07,
+      "loss": 2.2163,
+      "step": 42374
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.934418457648546e-07,
+      "loss": 2.2244,
+      "step": 42375
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.93394437420986e-07,
+      "loss": 1.3698,
+      "step": 42376
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.933470290771176e-07,
+      "loss": 0.9558,
+      "step": 42377
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.93299620733249e-07,
+      "loss": 1.9904,
+      "step": 42378
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.932522123893806e-07,
+      "loss": 2.3109,
+      "step": 42379
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.93204804045512e-07,
+      "loss": 2.0754,
+      "step": 42380
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.931573957016436e-07,
+      "loss": 2.8855,
+      "step": 42381
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.93109987357775e-07,
+      "loss": 2.2775,
+      "step": 42382
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.930625790139063e-07,
+      "loss": 1.3612,
+      "step": 42383
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.93015170670038e-07,
+      "loss": 1.89,
+      "step": 42384
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.929677623261695e-07,
+      "loss": 4.0924,
+      "step": 42385
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.929203539823008e-07,
+      "loss": 2.0584,
+      "step": 42386
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.928729456384323e-07,
+      "loss": 1.4909,
+      "step": 42387
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.92825537294564e-07,
+      "loss": 2.7635,
+      "step": 42388
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.927781289506953e-07,
+      "loss": 1.8046,
+      "step": 42389
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.927307206068268e-07,
+      "loss": 1.7867,
+      "step": 42390
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.926833122629585e-07,
+      "loss": 1.9164,
+      "step": 42391
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.926359039190898e-07,
+      "loss": 2.0274,
+      "step": 42392
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.925884955752213e-07,
+      "loss": 0.8768,
+      "step": 42393
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.925410872313527e-07,
+      "loss": 2.1297,
+      "step": 42394
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.924936788874842e-07,
+      "loss": 2.6655,
+      "step": 42395
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.924462705436157e-07,
+      "loss": 0.9003,
+      "step": 42396
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.923988621997472e-07,
+      "loss": 2.7673,
+      "step": 42397
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.923514538558787e-07,
+      "loss": 2.8094,
+      "step": 42398
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.923040455120102e-07,
+      "loss": 3.3773,
+      "step": 42399
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.922566371681415e-07,
+      "loss": 2.4945,
+      "step": 42400
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.922092288242732e-07,
+      "loss": 2.3008,
+      "step": 42401
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.921618204804047e-07,
+      "loss": 2.1783,
+      "step": 42402
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.92114412136536e-07,
+      "loss": 2.5152,
+      "step": 42403
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.920670037926677e-07,
+      "loss": 1.537,
+      "step": 42404
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.92019595448799e-07,
+      "loss": 1.7279,
+      "step": 42405
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.919721871049304e-07,
+      "loss": 3.4211,
+      "step": 42406
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.91924778761062e-07,
+      "loss": 1.8921,
+      "step": 42407
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.918773704171934e-07,
+      "loss": 2.325,
+      "step": 42408
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.91829962073325e-07,
+      "loss": 1.1684,
+      "step": 42409
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.917825537294564e-07,
+      "loss": 1.9064,
+      "step": 42410
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.91735145385588e-07,
+      "loss": 3.0546,
+      "step": 42411
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.916877370417194e-07,
+      "loss": 3.6823,
+      "step": 42412
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.916403286978509e-07,
+      "loss": 2.4375,
+      "step": 42413
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.915929203539824e-07,
+      "loss": 0.6712,
+      "step": 42414
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.915455120101139e-07,
+      "loss": 2.9076,
+      "step": 42415
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.914981036662452e-07,
+      "loss": 2.122,
+      "step": 42416
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.914506953223769e-07,
+      "loss": 0.9226,
+      "step": 42417
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.914032869785083e-07,
+      "loss": 1.4332,
+      "step": 42418
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.913558786346396e-07,
+      "loss": 2.6338,
+      "step": 42419
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.913084702907711e-07,
+      "loss": 3.3558,
+      "step": 42420
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.912610619469028e-07,
+      "loss": 1.5829,
+      "step": 42421
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.91213653603034e-07,
+      "loss": 2.864,
+      "step": 42422
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.911662452591656e-07,
+      "loss": 2.0361,
+      "step": 42423
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.911188369152973e-07,
+      "loss": 2.2084,
+      "step": 42424
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.910714285714286e-07,
+      "loss": 2.57,
+      "step": 42425
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.9102402022756e-07,
+      "loss": 2.33,
+      "step": 42426
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.909766118836916e-07,
+      "loss": 1.3027,
+      "step": 42427
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.90929203539823e-07,
+      "loss": 2.0685,
+      "step": 42428
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.908817951959546e-07,
+      "loss": 2.5898,
+      "step": 42429
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.90834386852086e-07,
+      "loss": 2.8266,
+      "step": 42430
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.907869785082175e-07,
+      "loss": 3.9404,
+      "step": 42431
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.90739570164349e-07,
+      "loss": 1.5456,
+      "step": 42432
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.906921618204803e-07,
+      "loss": 1.3323,
+      "step": 42433
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.90644753476612e-07,
+      "loss": 1.8892,
+      "step": 42434
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.905973451327435e-07,
+      "loss": 0.9141,
+      "step": 42435
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.905499367888748e-07,
+      "loss": 1.9067,
+      "step": 42436
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.905025284450065e-07,
+      "loss": 1.0894,
+      "step": 42437
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.904551201011378e-07,
+      "loss": 0.8935,
+      "step": 42438
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.904077117572693e-07,
+      "loss": 1.6509,
+      "step": 42439
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.903603034134008e-07,
+      "loss": 1.4248,
+      "step": 42440
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.903128950695322e-07,
+      "loss": 2.9707,
+      "step": 42441
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.902654867256637e-07,
+      "loss": 1.5379,
+      "step": 42442
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.902180783817952e-07,
+      "loss": 2.1186,
+      "step": 42443
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.901706700379267e-07,
+      "loss": 3.0372,
+      "step": 42444
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.901232616940582e-07,
+      "loss": 1.2285,
+      "step": 42445
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.900758533501897e-07,
+      "loss": 1.1545,
+      "step": 42446
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.900284450063212e-07,
+      "loss": 1.6135,
+      "step": 42447
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.899810366624527e-07,
+      "loss": 3.0811,
+      "step": 42448
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.89933628318584e-07,
+      "loss": 2.9326,
+      "step": 42449
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.898862199747155e-07,
+      "loss": 1.5765,
+      "step": 42450
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.898388116308472e-07,
+      "loss": 1.67,
+      "step": 42451
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.897914032869784e-07,
+      "loss": 1.9601,
+      "step": 42452
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.8974399494311e-07,
+      "loss": 2.8987,
+      "step": 42453
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.896965865992416e-07,
+      "loss": 2.3772,
+      "step": 42454
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.89649178255373e-07,
+      "loss": 1.3934,
+      "step": 42455
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.896017699115044e-07,
+      "loss": 1.8365,
+      "step": 42456
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.89554361567636e-07,
+      "loss": 3.3278,
+      "step": 42457
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.895069532237674e-07,
+      "loss": 2.2377,
+      "step": 42458
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.894595448798989e-07,
+      "loss": 1.9719,
+      "step": 42459
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.894121365360304e-07,
+      "loss": 2.2469,
+      "step": 42460
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.893647281921619e-07,
+      "loss": 1.7879,
+      "step": 42461
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.893173198482934e-07,
+      "loss": 2.0278,
+      "step": 42462
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.892699115044246e-07,
+      "loss": 3.3956,
+      "step": 42463
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.892225031605564e-07,
+      "loss": 0.9778,
+      "step": 42464
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.891750948166878e-07,
+      "loss": 1.6905,
+      "step": 42465
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.891276864728191e-07,
+      "loss": 2.7802,
+      "step": 42466
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.890802781289508e-07,
+      "loss": 1.8718,
+      "step": 42467
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.890328697850821e-07,
+      "loss": 1.8957,
+      "step": 42468
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.889854614412136e-07,
+      "loss": 2.9495,
+      "step": 42469
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.88938053097345e-07,
+      "loss": 2.6957,
+      "step": 42470
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.888906447534766e-07,
+      "loss": 1.4745,
+      "step": 42471
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.88843236409608e-07,
+      "loss": 0.6063,
+      "step": 42472
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.887958280657396e-07,
+      "loss": 4.2637,
+      "step": 42473
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.88748419721871e-07,
+      "loss": 1.3196,
+      "step": 42474
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.887010113780026e-07,
+      "loss": 2.3242,
+      "step": 42475
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.88653603034134e-07,
+      "loss": 1.5385,
+      "step": 42476
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.886061946902655e-07,
+      "loss": 2.6647,
+      "step": 42477
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.88558786346397e-07,
+      "loss": 1.4476,
+      "step": 42478
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.885113780025283e-07,
+      "loss": 2.6817,
+      "step": 42479
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.8846396965866e-07,
+      "loss": 1.6839,
+      "step": 42480
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.884165613147915e-07,
+      "loss": 1.8235,
+      "step": 42481
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.883691529709228e-07,
+      "loss": 1.4204,
+      "step": 42482
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.883217446270543e-07,
+      "loss": 1.8074,
+      "step": 42483
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.88274336283186e-07,
+      "loss": 1.4035,
+      "step": 42484
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.882269279393173e-07,
+      "loss": 3.6643,
+      "step": 42485
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.881795195954488e-07,
+      "loss": 1.0668,
+      "step": 42486
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.881321112515805e-07,
+      "loss": 2.9534,
+      "step": 42487
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.880847029077117e-07,
+      "loss": 1.4941,
+      "step": 42488
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.880372945638432e-07,
+      "loss": 2.803,
+      "step": 42489
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.879898862199747e-07,
+      "loss": 1.552,
+      "step": 42490
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.879424778761062e-07,
+      "loss": 3.6386,
+      "step": 42491
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.878950695322377e-07,
+      "loss": 2.9804,
+      "step": 42492
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.878476611883692e-07,
+      "loss": 1.8532,
+      "step": 42493
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.878002528445007e-07,
+      "loss": 4.6879,
+      "step": 42494
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.877528445006322e-07,
+      "loss": 2.6027,
+      "step": 42495
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.877054361567635e-07,
+      "loss": 2.333,
+      "step": 42496
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.876580278128952e-07,
+      "loss": 2.5219,
+      "step": 42497
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.876106194690267e-07,
+      "loss": 2.9131,
+      "step": 42498
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.87563211125158e-07,
+      "loss": 2.732,
+      "step": 42499
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.875158027812896e-07,
+      "loss": 0.5038,
+      "step": 42500
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.87468394437421e-07,
+      "loss": 2.5813,
+      "step": 42501
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.874209860935524e-07,
+      "loss": 2.325,
+      "step": 42502
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.87373577749684e-07,
+      "loss": 2.7017,
+      "step": 42503
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.873261694058154e-07,
+      "loss": 3.4014,
+      "step": 42504
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.87278761061947e-07,
+      "loss": 2.752,
+      "step": 42505
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.872313527180784e-07,
+      "loss": 2.4967,
+      "step": 42506
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.871839443742099e-07,
+      "loss": 0.8366,
+      "step": 42507
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.871365360303414e-07,
+      "loss": 2.1026,
+      "step": 42508
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.870891276864729e-07,
+      "loss": 2.9686,
+      "step": 42509
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.870417193426044e-07,
+      "loss": 1.5239,
+      "step": 42510
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.869943109987358e-07,
+      "loss": 2.8751,
+      "step": 42511
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.869469026548671e-07,
+      "loss": 2.3581,
+      "step": 42512
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.868994943109988e-07,
+      "loss": 1.6295,
+      "step": 42513
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.868520859671303e-07,
+      "loss": 2.1381,
+      "step": 42514
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.868046776232616e-07,
+      "loss": 2.0156,
+      "step": 42515
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.86757269279393e-07,
+      "loss": 2.2924,
+      "step": 42516
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.867098609355248e-07,
+      "loss": 2.2941,
+      "step": 42517
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.86662452591656e-07,
+      "loss": 1.8953,
+      "step": 42518
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.866150442477876e-07,
+      "loss": 1.791,
+      "step": 42519
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.865676359039193e-07,
+      "loss": 2.5284,
+      "step": 42520
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.865202275600506e-07,
+      "loss": 3.0571,
+      "step": 42521
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.86472819216182e-07,
+      "loss": 0.6049,
+      "step": 42522
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.864254108723135e-07,
+      "loss": 1.6952,
+      "step": 42523
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.86378002528445e-07,
+      "loss": 1.9345,
+      "step": 42524
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.863305941845765e-07,
+      "loss": 2.7312,
+      "step": 42525
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.86283185840708e-07,
+      "loss": 1.4453,
+      "step": 42526
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.862357774968395e-07,
+      "loss": 0.4735,
+      "step": 42527
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.86188369152971e-07,
+      "loss": 2.0549,
+      "step": 42528
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.861409608091023e-07,
+      "loss": 2.5952,
+      "step": 42529
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.86093552465234e-07,
+      "loss": 3.0424,
+      "step": 42530
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.860461441213655e-07,
+      "loss": 1.5491,
+      "step": 42531
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.859987357774968e-07,
+      "loss": 1.5628,
+      "step": 42532
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.859513274336285e-07,
+      "loss": 1.8351,
+      "step": 42533
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.859039190897597e-07,
+      "loss": 1.2408,
+      "step": 42534
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.858565107458912e-07,
+      "loss": 0.7955,
+      "step": 42535
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.858091024020227e-07,
+      "loss": 1.6404,
+      "step": 42536
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.857616940581542e-07,
+      "loss": 1.2058,
+      "step": 42537
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.857142857142857e-07,
+      "loss": 1.954,
+      "step": 42538
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.856668773704172e-07,
+      "loss": 1.138,
+      "step": 42539
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.856194690265487e-07,
+      "loss": 3.3515,
+      "step": 42540
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.855720606826802e-07,
+      "loss": 2.1197,
+      "step": 42541
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.855246523388117e-07,
+      "loss": 1.1798,
+      "step": 42542
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.854772439949432e-07,
+      "loss": 1.787,
+      "step": 42543
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.854298356510747e-07,
+      "loss": 4.1167,
+      "step": 42544
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.85382427307206e-07,
+      "loss": 0.7761,
+      "step": 42545
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.853350189633377e-07,
+      "loss": 2.3463,
+      "step": 42546
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.852876106194691e-07,
+      "loss": 0.9028,
+      "step": 42547
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.852402022756004e-07,
+      "loss": 1.3924,
+      "step": 42548
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.85192793931732e-07,
+      "loss": 2.4233,
+      "step": 42549
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.851453855878636e-07,
+      "loss": 0.6632,
+      "step": 42550
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.85097977243995e-07,
+      "loss": 2.3574,
+      "step": 42551
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.850505689001264e-07,
+      "loss": 2.7456,
+      "step": 42552
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.85003160556258e-07,
+      "loss": 1.8773,
+      "step": 42553
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.849557522123894e-07,
+      "loss": 0.9706,
+      "step": 42554
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.849083438685209e-07,
+      "loss": 2.8645,
+      "step": 42555
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.848609355246524e-07,
+      "loss": 1.0867,
+      "step": 42556
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.848135271807839e-07,
+      "loss": 3.072,
+      "step": 42557
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.847661188369153e-07,
+      "loss": 1.0857,
+      "step": 42558
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.847187104930468e-07,
+      "loss": 1.5805,
+      "step": 42559
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.846713021491783e-07,
+      "loss": 0.6332,
+      "step": 42560
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.846238938053098e-07,
+      "loss": 2.8419,
+      "step": 42561
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.84576485461441e-07,
+      "loss": 1.5524,
+      "step": 42562
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.845290771175728e-07,
+      "loss": 2.4213,
+      "step": 42563
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.844816687737043e-07,
+      "loss": 1.9612,
+      "step": 42564
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.844342604298356e-07,
+      "loss": 1.3052,
+      "step": 42565
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.843868520859673e-07,
+      "loss": 1.1486,
+      "step": 42566
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.843394437420986e-07,
+      "loss": 2.6758,
+      "step": 42567
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.8429203539823e-07,
+      "loss": 2.3174,
+      "step": 42568
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.842446270543615e-07,
+      "loss": 2.1638,
+      "step": 42569
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.84197218710493e-07,
+      "loss": 1.6786,
+      "step": 42570
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.841498103666245e-07,
+      "loss": 2.5087,
+      "step": 42571
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.84102402022756e-07,
+      "loss": 2.3267,
+      "step": 42572
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.840549936788875e-07,
+      "loss": 2.9316,
+      "step": 42573
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.84007585335019e-07,
+      "loss": 2.1817,
+      "step": 42574
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.839601769911505e-07,
+      "loss": 1.3041,
+      "step": 42575
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.83912768647282e-07,
+      "loss": 0.8978,
+      "step": 42576
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.838653603034135e-07,
+      "loss": 1.9565,
+      "step": 42577
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.838179519595448e-07,
+      "loss": 2.125,
+      "step": 42578
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.837705436156765e-07,
+      "loss": 0.3265,
+      "step": 42579
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.83723135271808e-07,
+      "loss": 2.5535,
+      "step": 42580
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.836757269279392e-07,
+      "loss": 3.2154,
+      "step": 42581
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.836283185840707e-07,
+      "loss": 1.3256,
+      "step": 42582
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.835809102402024e-07,
+      "loss": 2.325,
+      "step": 42583
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.835335018963337e-07,
+      "loss": 1.5122,
+      "step": 42584
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.834860935524652e-07,
+      "loss": 3.0149,
+      "step": 42585
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.83438685208597e-07,
+      "loss": 2.3556,
+      "step": 42586
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.833912768647282e-07,
+      "loss": 2.2871,
+      "step": 42587
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.833438685208597e-07,
+      "loss": 1.4247,
+      "step": 42588
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.832964601769912e-07,
+      "loss": 3.5401,
+      "step": 42589
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.832490518331227e-07,
+      "loss": 1.0637,
+      "step": 42590
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.832016434892542e-07,
+      "loss": 1.387,
+      "step": 42591
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.831542351453854e-07,
+      "loss": 2.4922,
+      "step": 42592
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.831068268015171e-07,
+      "loss": 1.1783,
+      "step": 42593
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.830594184576486e-07,
+      "loss": 3.1224,
+      "step": 42594
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.8301201011378e-07,
+      "loss": 1.31,
+      "step": 42595
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.829646017699116e-07,
+      "loss": 2.8878,
+      "step": 42596
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.829171934260431e-07,
+      "loss": 2.0006,
+      "step": 42597
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.828697850821744e-07,
+      "loss": 3.2659,
+      "step": 42598
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.828223767383059e-07,
+      "loss": 2.3061,
+      "step": 42599
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.827749683944374e-07,
+      "loss": 2.7801,
+      "step": 42600
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.827275600505689e-07,
+      "loss": 2.0104,
+      "step": 42601
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.826801517067004e-07,
+      "loss": 1.5159,
+      "step": 42602
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.826327433628319e-07,
+      "loss": 2.1713,
+      "step": 42603
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.825853350189634e-07,
+      "loss": 1.9398,
+      "step": 42604
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.825379266750948e-07,
+      "loss": 3.0382,
+      "step": 42605
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.824905183312263e-07,
+      "loss": 2.0883,
+      "step": 42606
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.824431099873578e-07,
+      "loss": 0.8687,
+      "step": 42607
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.823957016434893e-07,
+      "loss": 1.264,
+      "step": 42608
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.823482932996208e-07,
+      "loss": 1.9432,
+      "step": 42609
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.823008849557523e-07,
+      "loss": 1.7379,
+      "step": 42610
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.822534766118836e-07,
+      "loss": 3.035,
+      "step": 42611
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.82206068268015e-07,
+      "loss": 0.358,
+      "step": 42612
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.821586599241468e-07,
+      "loss": 1.8098,
+      "step": 42613
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.82111251580278e-07,
+      "loss": 1.0367,
+      "step": 42614
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.820638432364096e-07,
+      "loss": 1.086,
+      "step": 42615
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.820164348925413e-07,
+      "loss": 1.506,
+      "step": 42616
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.819690265486725e-07,
+      "loss": 4.4163,
+      "step": 42617
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.81921618204804e-07,
+      "loss": 3.6064,
+      "step": 42618
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.818742098609355e-07,
+      "loss": 2.6044,
+      "step": 42619
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.81826801517067e-07,
+      "loss": 1.9154,
+      "step": 42620
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.817793931731985e-07,
+      "loss": 2.8445,
+      "step": 42621
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.8173198482933e-07,
+      "loss": 2.4069,
+      "step": 42622
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.816845764854615e-07,
+      "loss": 2.5704,
+      "step": 42623
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.81637168141593e-07,
+      "loss": 1.7605,
+      "step": 42624
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.815897597977243e-07,
+      "loss": 2.6042,
+      "step": 42625
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.81542351453856e-07,
+      "loss": 0.9948,
+      "step": 42626
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.814949431099875e-07,
+      "loss": 1.4986,
+      "step": 42627
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.814475347661187e-07,
+      "loss": 1.712,
+      "step": 42628
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.814001264222504e-07,
+      "loss": 3.089,
+      "step": 42629
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.81352718078382e-07,
+      "loss": 1.9612,
+      "step": 42630
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.813053097345132e-07,
+      "loss": 1.9477,
+      "step": 42631
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.812579013906447e-07,
+      "loss": 2.0338,
+      "step": 42632
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.812104930467762e-07,
+      "loss": 1.7698,
+      "step": 42633
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.811630847029077e-07,
+      "loss": 0.8137,
+      "step": 42634
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.811156763590392e-07,
+      "loss": 1.1566,
+      "step": 42635
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.810682680151707e-07,
+      "loss": 1.239,
+      "step": 42636
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.810208596713022e-07,
+      "loss": 2.3899,
+      "step": 42637
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.809734513274337e-07,
+      "loss": 2.3586,
+      "step": 42638
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.809260429835652e-07,
+      "loss": 2.9086,
+      "step": 42639
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.808786346396966e-07,
+      "loss": 1.4122,
+      "step": 42640
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.808312262958281e-07,
+      "loss": 1.2164,
+      "step": 42641
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.807838179519596e-07,
+      "loss": 3.7643,
+      "step": 42642
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.807364096080911e-07,
+      "loss": 2.2159,
+      "step": 42643
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.806890012642224e-07,
+      "loss": 2.2479,
+      "step": 42644
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.80641592920354e-07,
+      "loss": 1.7346,
+      "step": 42645
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.805941845764856e-07,
+      "loss": 2.5163,
+      "step": 42646
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.805467762326169e-07,
+      "loss": 0.9177,
+      "step": 42647
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.804993678887484e-07,
+      "loss": 1.4283,
+      "step": 42648
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.8045195954488e-07,
+      "loss": 2.9797,
+      "step": 42649
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.804045512010114e-07,
+      "loss": 2.4252,
+      "step": 42650
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.803571428571428e-07,
+      "loss": 1.2302,
+      "step": 42651
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.803097345132743e-07,
+      "loss": 2.7057,
+      "step": 42652
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.802623261694058e-07,
+      "loss": 2.035,
+      "step": 42653
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.802149178255373e-07,
+      "loss": 1.9937,
+      "step": 42654
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.801675094816688e-07,
+      "loss": 2.6412,
+      "step": 42655
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.801201011378003e-07,
+      "loss": 2.1217,
+      "step": 42656
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.800726927939318e-07,
+      "loss": 1.6168,
+      "step": 42657
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.80025284450063e-07,
+      "loss": 1.5137,
+      "step": 42658
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.799778761061948e-07,
+      "loss": 3.1415,
+      "step": 42659
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.799304677623263e-07,
+      "loss": 1.2692,
+      "step": 42660
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.798830594184576e-07,
+      "loss": 2.1597,
+      "step": 42661
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.798356510745893e-07,
+      "loss": 3.4051,
+      "step": 42662
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.797882427307208e-07,
+      "loss": 2.4205,
+      "step": 42663
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.79740834386852e-07,
+      "loss": 1.552,
+      "step": 42664
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.796934260429835e-07,
+      "loss": 1.0342,
+      "step": 42665
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.79646017699115e-07,
+      "loss": 1.0312,
+      "step": 42666
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.795986093552465e-07,
+      "loss": 3.115,
+      "step": 42667
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.79551201011378e-07,
+      "loss": 2.5658,
+      "step": 42668
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.795037926675095e-07,
+      "loss": 1.3453,
+      "step": 42669
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.79456384323641e-07,
+      "loss": 1.348,
+      "step": 42670
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.794089759797725e-07,
+      "loss": 1.9985,
+      "step": 42671
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.79361567635904e-07,
+      "loss": 2.5212,
+      "step": 42672
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.793141592920355e-07,
+      "loss": 2.5075,
+      "step": 42673
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.79266750948167e-07,
+      "loss": 2.3828,
+      "step": 42674
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.792193426042984e-07,
+      "loss": 1.1562,
+      "step": 42675
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.7917193426043e-07,
+      "loss": 2.7319,
+      "step": 42676
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.791245259165612e-07,
+      "loss": 3.604,
+      "step": 42677
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.790771175726927e-07,
+      "loss": 2.5988,
+      "step": 42678
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.790297092288244e-07,
+      "loss": 1.4654,
+      "step": 42679
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.789823008849557e-07,
+      "loss": 2.1375,
+      "step": 42680
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.789348925410872e-07,
+      "loss": 2.0007,
+      "step": 42681
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.78887484197219e-07,
+      "loss": 2.149,
+      "step": 42682
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.788400758533502e-07,
+      "loss": 2.9481,
+      "step": 42683
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.787926675094817e-07,
+      "loss": 1.8708,
+      "step": 42684
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.787452591656132e-07,
+      "loss": 1.7714,
+      "step": 42685
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.786978508217447e-07,
+      "loss": 1.6624,
+      "step": 42686
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.786504424778761e-07,
+      "loss": 1.2222,
+      "step": 42687
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.786030341340076e-07,
+      "loss": 1.8274,
+      "step": 42688
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.785556257901391e-07,
+      "loss": 2.5512,
+      "step": 42689
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.785082174462706e-07,
+      "loss": 1.1541,
+      "step": 42690
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.78460809102402e-07,
+      "loss": 0.9635,
+      "step": 42691
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.784134007585336e-07,
+      "loss": 3.8067,
+      "step": 42692
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.78365992414665e-07,
+      "loss": 2.3392,
+      "step": 42693
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.783185840707964e-07,
+      "loss": 3.5387,
+      "step": 42694
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.78271175726928e-07,
+      "loss": 1.436,
+      "step": 42695
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.782237673830596e-07,
+      "loss": 1.3676,
+      "step": 42696
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.781763590391909e-07,
+      "loss": 1.1615,
+      "step": 42697
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.781289506953223e-07,
+      "loss": 1.8097,
+      "step": 42698
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.780815423514538e-07,
+      "loss": 2.3256,
+      "step": 42699
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.780341340075853e-07,
+      "loss": 1.0498,
+      "step": 42700
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.779867256637168e-07,
+      "loss": 2.124,
+      "step": 42701
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.779393173198483e-07,
+      "loss": 1.279,
+      "step": 42702
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.778919089759798e-07,
+      "loss": 1.4519,
+      "step": 42703
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.778445006321113e-07,
+      "loss": 1.1618,
+      "step": 42704
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.777970922882428e-07,
+      "loss": 0.7831,
+      "step": 42705
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.777496839443743e-07,
+      "loss": 2.6046,
+      "step": 42706
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.777022756005058e-07,
+      "loss": 1.2381,
+      "step": 42707
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.776548672566373e-07,
+      "loss": 2.207,
+      "step": 42708
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.776074589127688e-07,
+      "loss": 2.9865,
+      "step": 42709
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.775600505689e-07,
+      "loss": 1.6031,
+      "step": 42710
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.775126422250315e-07,
+      "loss": 3.3964,
+      "step": 42711
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.774652338811632e-07,
+      "loss": 2.671,
+      "step": 42712
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.774178255372945e-07,
+      "loss": 2.5011,
+      "step": 42713
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.77370417193426e-07,
+      "loss": 3.2795,
+      "step": 42714
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.773230088495577e-07,
+      "loss": 2.5448,
+      "step": 42715
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.77275600505689e-07,
+      "loss": 2.7607,
+      "step": 42716
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.772281921618205e-07,
+      "loss": 3.6595,
+      "step": 42717
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.77180783817952e-07,
+      "loss": 2.2008,
+      "step": 42718
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.771333754740835e-07,
+      "loss": 1.5864,
+      "step": 42719
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.77085967130215e-07,
+      "loss": 1.5357,
+      "step": 42720
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.770385587863462e-07,
+      "loss": 2.1333,
+      "step": 42721
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.76991150442478e-07,
+      "loss": 1.0691,
+      "step": 42722
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.769437420986094e-07,
+      "loss": 2.7419,
+      "step": 42723
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.768963337547407e-07,
+      "loss": 1.7132,
+      "step": 42724
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.768489254108724e-07,
+      "loss": 3.0781,
+      "step": 42725
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.76801517067004e-07,
+      "loss": 2.0577,
+      "step": 42726
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.767541087231352e-07,
+      "loss": 2.5091,
+      "step": 42727
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.767067003792667e-07,
+      "loss": 2.04,
+      "step": 42728
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.766592920353984e-07,
+      "loss": 1.1492,
+      "step": 42729
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.766118836915297e-07,
+      "loss": 1.3988,
+      "step": 42730
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.765644753476612e-07,
+      "loss": 1.4704,
+      "step": 42731
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.765170670037927e-07,
+      "loss": 2.7002,
+      "step": 42732
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.764696586599241e-07,
+      "loss": 1.6487,
+      "step": 42733
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.764222503160556e-07,
+      "loss": 2.0604,
+      "step": 42734
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.763748419721871e-07,
+      "loss": 2.1032,
+      "step": 42735
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.763274336283186e-07,
+      "loss": 2.3235,
+      "step": 42736
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.762800252844501e-07,
+      "loss": 2.1528,
+      "step": 42737
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.762326169405816e-07,
+      "loss": 1.4819,
+      "step": 42738
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.76185208596713e-07,
+      "loss": 2.4178,
+      "step": 42739
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.761378002528446e-07,
+      "loss": 1.093,
+      "step": 42740
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.760903919089759e-07,
+      "loss": 1.5002,
+      "step": 42741
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.760429835651076e-07,
+      "loss": 2.9388,
+      "step": 42742
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.759955752212389e-07,
+      "loss": 2.1614,
+      "step": 42743
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.759481668773704e-07,
+      "loss": 1.955,
+      "step": 42744
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.75900758533502e-07,
+      "loss": 3.1298,
+      "step": 42745
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.758533501896333e-07,
+      "loss": 2.7984,
+      "step": 42746
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.758059418457648e-07,
+      "loss": 1.4757,
+      "step": 42747
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.757585335018963e-07,
+      "loss": 2.9151,
+      "step": 42748
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.757111251580278e-07,
+      "loss": 2.8336,
+      "step": 42749
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.756637168141593e-07,
+      "loss": 1.9457,
+      "step": 42750
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.756163084702908e-07,
+      "loss": 1.8997,
+      "step": 42751
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.755689001264223e-07,
+      "loss": 2.0079,
+      "step": 42752
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.755214917825538e-07,
+      "loss": 1.3737,
+      "step": 42753
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.75474083438685e-07,
+      "loss": 1.787,
+      "step": 42754
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.754266750948168e-07,
+      "loss": 2.1274,
+      "step": 42755
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.753792667509483e-07,
+      "loss": 2.4161,
+      "step": 42756
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.753318584070795e-07,
+      "loss": 1.4532,
+      "step": 42757
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.752844500632112e-07,
+      "loss": 3.3945,
+      "step": 42758
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.752370417193427e-07,
+      "loss": 2.0914,
+      "step": 42759
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.75189633375474e-07,
+      "loss": 2.2415,
+      "step": 42760
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.751422250316055e-07,
+      "loss": 1.4622,
+      "step": 42761
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.750948166877372e-07,
+      "loss": 3.7162,
+      "step": 42762
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.750474083438685e-07,
+      "loss": 0.5487,
+      "step": 42763
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.75e-07,
+      "loss": 1.6661,
+      "step": 42764
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.749525916561315e-07,
+      "loss": 2.4158,
+      "step": 42765
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.74905183312263e-07,
+      "loss": 2.4937,
+      "step": 42766
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.748577749683945e-07,
+      "loss": 1.3917,
+      "step": 42767
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.74810366624526e-07,
+      "loss": 2.8171,
+      "step": 42768
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.747629582806574e-07,
+      "loss": 1.5559,
+      "step": 42769
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.74715549936789e-07,
+      "loss": 1.5207,
+      "step": 42770
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.746681415929204e-07,
+      "loss": 2.3982,
+      "step": 42771
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.74620733249052e-07,
+      "loss": 2.3231,
+      "step": 42772
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.745733249051834e-07,
+      "loss": 1.0987,
+      "step": 42773
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.745259165613147e-07,
+      "loss": 2.3798,
+      "step": 42774
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.744785082174464e-07,
+      "loss": 1.4859,
+      "step": 42775
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.744310998735777e-07,
+      "loss": 2.9407,
+      "step": 42776
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.743836915297092e-07,
+      "loss": 2.1473,
+      "step": 42777
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.743362831858409e-07,
+      "loss": 2.3922,
+      "step": 42778
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.742888748419722e-07,
+      "loss": 3.0327,
+      "step": 42779
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.742414664981036e-07,
+      "loss": 0.6324,
+      "step": 42780
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.741940581542351e-07,
+      "loss": 3.059,
+      "step": 42781
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.741466498103666e-07,
+      "loss": 2.7573,
+      "step": 42782
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.740992414664981e-07,
+      "loss": 3.0989,
+      "step": 42783
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.740518331226296e-07,
+      "loss": 3.0727,
+      "step": 42784
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.740044247787611e-07,
+      "loss": 3.1142,
+      "step": 42785
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.739570164348926e-07,
+      "loss": 2.0289,
+      "step": 42786
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.739096080910239e-07,
+      "loss": 2.2525,
+      "step": 42787
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.738621997471556e-07,
+      "loss": 1.2841,
+      "step": 42788
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.73814791403287e-07,
+      "loss": 1.4422,
+      "step": 42789
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.737673830594184e-07,
+      "loss": 1.6363,
+      "step": 42790
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.7371997471555e-07,
+      "loss": 3.8212,
+      "step": 42791
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.736725663716816e-07,
+      "loss": 1.5554,
+      "step": 42792
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.736251580278128e-07,
+      "loss": 1.0557,
+      "step": 42793
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.735777496839443e-07,
+      "loss": 3.6408,
+      "step": 42794
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.73530341340076e-07,
+      "loss": 1.4953,
+      "step": 42795
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.734829329962073e-07,
+      "loss": 2.0286,
+      "step": 42796
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.734355246523388e-07,
+      "loss": 1.2171,
+      "step": 42797
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.733881163084703e-07,
+      "loss": 2.3149,
+      "step": 42798
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.733407079646018e-07,
+      "loss": 1.7471,
+      "step": 42799
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.732932996207333e-07,
+      "loss": 3.2565,
+      "step": 42800
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.732458912768648e-07,
+      "loss": 2.7637,
+      "step": 42801
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.731984829329963e-07,
+      "loss": 1.653,
+      "step": 42802
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.731510745891278e-07,
+      "loss": 2.2296,
+      "step": 42803
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.731036662452592e-07,
+      "loss": 1.195,
+      "step": 42804
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.730562579013907e-07,
+      "loss": 2.483,
+      "step": 42805
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.730088495575222e-07,
+      "loss": 1.2004,
+      "step": 42806
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.729614412136535e-07,
+      "loss": 1.7769,
+      "step": 42807
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.729140328697852e-07,
+      "loss": 2.7154,
+      "step": 42808
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.728666245259165e-07,
+      "loss": 1.6467,
+      "step": 42809
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.72819216182048e-07,
+      "loss": 0.262,
+      "step": 42810
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.727718078381797e-07,
+      "loss": 2.4525,
+      "step": 42811
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.72724399494311e-07,
+      "loss": 0.6243,
+      "step": 42812
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.726769911504425e-07,
+      "loss": 1.3514,
+      "step": 42813
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.72629582806574e-07,
+      "loss": 1.0037,
+      "step": 42814
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.725821744627054e-07,
+      "loss": 1.6252,
+      "step": 42815
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.72534766118837e-07,
+      "loss": 4.3418,
+      "step": 42816
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.724873577749684e-07,
+      "loss": 1.4278,
+      "step": 42817
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.724399494311e-07,
+      "loss": 1.9075,
+      "step": 42818
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.723925410872314e-07,
+      "loss": 1.7365,
+      "step": 42819
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.723451327433627e-07,
+      "loss": 1.8386,
+      "step": 42820
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.722977243994944e-07,
+      "loss": 2.8001,
+      "step": 42821
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.722503160556259e-07,
+      "loss": 2.1179,
+      "step": 42822
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.722029077117572e-07,
+      "loss": 2.0934,
+      "step": 42823
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.721554993678889e-07,
+      "loss": 0.9574,
+      "step": 42824
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.721080910240204e-07,
+      "loss": 2.8093,
+      "step": 42825
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.720606826801517e-07,
+      "loss": 2.3008,
+      "step": 42826
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.720132743362831e-07,
+      "loss": 1.1944,
+      "step": 42827
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.719658659924148e-07,
+      "loss": 1.8484,
+      "step": 42828
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.719184576485461e-07,
+      "loss": 0.7858,
+      "step": 42829
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.718710493046776e-07,
+      "loss": 1.0271,
+      "step": 42830
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.718236409608091e-07,
+      "loss": 2.0081,
+      "step": 42831
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.717762326169406e-07,
+      "loss": 2.6125,
+      "step": 42832
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.71728824273072e-07,
+      "loss": 3.0711,
+      "step": 42833
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.716814159292036e-07,
+      "loss": 0.7657,
+      "step": 42834
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.71634007585335e-07,
+      "loss": 1.5946,
+      "step": 42835
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.715865992414666e-07,
+      "loss": 2.4626,
+      "step": 42836
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.71539190897598e-07,
+      "loss": 1.9985,
+      "step": 42837
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.714917825537296e-07,
+      "loss": 1.3149,
+      "step": 42838
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.71444374209861e-07,
+      "loss": 2.5761,
+      "step": 42839
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.713969658659923e-07,
+      "loss": 2.2327,
+      "step": 42840
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.71349557522124e-07,
+      "loss": 1.0858,
+      "step": 42841
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.713021491782553e-07,
+      "loss": 2.1921,
+      "step": 42842
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.712547408343868e-07,
+      "loss": 2.2756,
+      "step": 42843
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.712073324905185e-07,
+      "loss": 1.1751,
+      "step": 42844
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.711599241466498e-07,
+      "loss": 2.1667,
+      "step": 42845
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.711125158027813e-07,
+      "loss": 1.4537,
+      "step": 42846
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.710651074589128e-07,
+      "loss": 1.9416,
+      "step": 42847
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.710176991150443e-07,
+      "loss": 1.7151,
+      "step": 42848
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.709702907711758e-07,
+      "loss": 3.0217,
+      "step": 42849
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.709228824273073e-07,
+      "loss": 2.7663,
+      "step": 42850
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.708754740834387e-07,
+      "loss": 1.4947,
+      "step": 42851
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.708280657395702e-07,
+      "loss": 0.7454,
+      "step": 42852
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.707806573957015e-07,
+      "loss": 0.903,
+      "step": 42853
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.707332490518332e-07,
+      "loss": 2.2842,
+      "step": 42854
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.706858407079647e-07,
+      "loss": 2.845,
+      "step": 42855
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.70638432364096e-07,
+      "loss": 1.8854,
+      "step": 42856
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.705910240202277e-07,
+      "loss": 1.745,
+      "step": 42857
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.705436156763592e-07,
+      "loss": 1.6963,
+      "step": 42858
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.704962073324905e-07,
+      "loss": 1.0503,
+      "step": 42859
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.70448798988622e-07,
+      "loss": 1.4316,
+      "step": 42860
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.704013906447537e-07,
+      "loss": 1.1849,
+      "step": 42861
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.70353982300885e-07,
+      "loss": 2.2398,
+      "step": 42862
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.703065739570164e-07,
+      "loss": 2.1968,
+      "step": 42863
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.70259165613148e-07,
+      "loss": 1.7742,
+      "step": 42864
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.702117572692794e-07,
+      "loss": 1.4441,
+      "step": 42865
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.70164348925411e-07,
+      "loss": 1.4764,
+      "step": 42866
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.701169405815424e-07,
+      "loss": 0.7536,
+      "step": 42867
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.70069532237674e-07,
+      "loss": 1.668,
+      "step": 42868
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.700221238938054e-07,
+      "loss": 2.4717,
+      "step": 42869
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.699747155499367e-07,
+      "loss": 2.0505,
+      "step": 42870
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.699273072060684e-07,
+      "loss": 2.833,
+      "step": 42871
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.698798988621999e-07,
+      "loss": 3.1131,
+      "step": 42872
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.698324905183311e-07,
+      "loss": 2.2866,
+      "step": 42873
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.697850821744629e-07,
+      "loss": 3.6435,
+      "step": 42874
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.697376738305941e-07,
+      "loss": 1.8161,
+      "step": 42875
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.696902654867256e-07,
+      "loss": 1.2374,
+      "step": 42876
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.696428571428571e-07,
+      "loss": 2.2873,
+      "step": 42877
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.695954487989886e-07,
+      "loss": 1.8997,
+      "step": 42878
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.6954804045512e-07,
+      "loss": 2.3111,
+      "step": 42879
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.695006321112516e-07,
+      "loss": 1.3772,
+      "step": 42880
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.69453223767383e-07,
+      "loss": 0.8169,
+      "step": 42881
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.694058154235146e-07,
+      "loss": 1.4096,
+      "step": 42882
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.69358407079646e-07,
+      "loss": 2.0075,
+      "step": 42883
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.693109987357776e-07,
+      "loss": 1.7721,
+      "step": 42884
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.69263590391909e-07,
+      "loss": 0.6367,
+      "step": 42885
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.692161820480403e-07,
+      "loss": 1.9011,
+      "step": 42886
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.69168773704172e-07,
+      "loss": 2.2394,
+      "step": 42887
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.691213653603035e-07,
+      "loss": 2.643,
+      "step": 42888
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.690739570164348e-07,
+      "loss": 1.4423,
+      "step": 42889
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.690265486725663e-07,
+      "loss": 2.1423,
+      "step": 42890
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.68979140328698e-07,
+      "loss": 4.2727,
+      "step": 42891
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.689317319848293e-07,
+      "loss": 2.2353,
+      "step": 42892
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.688843236409608e-07,
+      "loss": 2.2014,
+      "step": 42893
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.688369152970925e-07,
+      "loss": 1.6317,
+      "step": 42894
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.687895069532238e-07,
+      "loss": 2.8578,
+      "step": 42895
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.687420986093553e-07,
+      "loss": 1.7896,
+      "step": 42896
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.686946902654867e-07,
+      "loss": 1.3665,
+      "step": 42897
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.686472819216182e-07,
+      "loss": 1.7198,
+      "step": 42898
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.685998735777497e-07,
+      "loss": 2.7618,
+      "step": 42899
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.685524652338812e-07,
+      "loss": 3.253,
+      "step": 42900
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.685050568900127e-07,
+      "loss": 2.899,
+      "step": 42901
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.684576485461442e-07,
+      "loss": 2.2499,
+      "step": 42902
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.684102402022755e-07,
+      "loss": 2.3131,
+      "step": 42903
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.683628318584072e-07,
+      "loss": 1.8596,
+      "step": 42904
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.683154235145385e-07,
+      "loss": 2.2899,
+      "step": 42905
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.6826801517067e-07,
+      "loss": 3.0244,
+      "step": 42906
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.682206068268017e-07,
+      "loss": 2.6826,
+      "step": 42907
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.68173198482933e-07,
+      "loss": 2.0404,
+      "step": 42908
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.681257901390644e-07,
+      "loss": 1.2946,
+      "step": 42909
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.68078381795196e-07,
+      "loss": 3.0492,
+      "step": 42910
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.680309734513274e-07,
+      "loss": 2.2596,
+      "step": 42911
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.67983565107459e-07,
+      "loss": 2.8913,
+      "step": 42912
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.679361567635904e-07,
+      "loss": 3.1209,
+      "step": 42913
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.67888748419722e-07,
+      "loss": 2.3027,
+      "step": 42914
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.678413400758534e-07,
+      "loss": 2.336,
+      "step": 42915
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.677939317319847e-07,
+      "loss": 1.9358,
+      "step": 42916
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.677465233881164e-07,
+      "loss": 1.1712,
+      "step": 42917
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.676991150442479e-07,
+      "loss": 1.7954,
+      "step": 42918
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.676517067003792e-07,
+      "loss": 2.4585,
+      "step": 42919
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.676042983565109e-07,
+      "loss": 1.1735,
+      "step": 42920
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.675568900126423e-07,
+      "loss": 1.9939,
+      "step": 42921
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.675094816687736e-07,
+      "loss": 1.6437,
+      "step": 42922
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.674620733249051e-07,
+      "loss": 1.8007,
+      "step": 42923
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.674146649810368e-07,
+      "loss": 2.1169,
+      "step": 42924
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.67367256637168e-07,
+      "loss": 2.0262,
+      "step": 42925
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.673198482932996e-07,
+      "loss": 0.9981,
+      "step": 42926
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.67272439949431e-07,
+      "loss": 1.7882,
+      "step": 42927
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.672250316055626e-07,
+      "loss": 1.2328,
+      "step": 42928
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.67177623261694e-07,
+      "loss": 1.7851,
+      "step": 42929
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.671302149178256e-07,
+      "loss": 2.1292,
+      "step": 42930
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.67082806573957e-07,
+      "loss": 1.1302,
+      "step": 42931
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.670353982300886e-07,
+      "loss": 2.0635,
+      "step": 42932
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.6698798988622e-07,
+      "loss": 2.5035,
+      "step": 42933
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.669405815423515e-07,
+      "loss": 2.2182,
+      "step": 42934
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.66893173198483e-07,
+      "loss": 3.4529,
+      "step": 42935
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.668457648546143e-07,
+      "loss": 2.5647,
+      "step": 42936
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.66798356510746e-07,
+      "loss": 1.6376,
+      "step": 42937
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.667509481668773e-07,
+      "loss": 3.3505,
+      "step": 42938
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.667035398230088e-07,
+      "loss": 1.8629,
+      "step": 42939
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.666561314791405e-07,
+      "loss": 3.3843,
+      "step": 42940
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.666087231352718e-07,
+      "loss": 0.9521,
+      "step": 42941
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.665613147914033e-07,
+      "loss": 2.1016,
+      "step": 42942
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.665139064475348e-07,
+      "loss": 2.4897,
+      "step": 42943
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.664664981036662e-07,
+      "loss": 1.8453,
+      "step": 42944
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.664190897597977e-07,
+      "loss": 2.1746,
+      "step": 42945
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.663716814159292e-07,
+      "loss": 2.8677,
+      "step": 42946
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.663242730720607e-07,
+      "loss": 1.2898,
+      "step": 42947
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.662768647281922e-07,
+      "loss": 2.5181,
+      "step": 42948
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.662294563843235e-07,
+      "loss": 2.3804,
+      "step": 42949
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.661820480404552e-07,
+      "loss": 2.1188,
+      "step": 42950
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.661346396965867e-07,
+      "loss": 1.2026,
+      "step": 42951
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.66087231352718e-07,
+      "loss": 1.8169,
+      "step": 42952
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.660398230088497e-07,
+      "loss": 1.6132,
+      "step": 42953
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.659924146649812e-07,
+      "loss": 1.7536,
+      "step": 42954
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.659450063211124e-07,
+      "loss": 2.0522,
+      "step": 42955
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.65897597977244e-07,
+      "loss": 1.1871,
+      "step": 42956
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.658501896333756e-07,
+      "loss": 1.8962,
+      "step": 42957
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.65802781289507e-07,
+      "loss": 3.5419,
+      "step": 42958
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.657553729456384e-07,
+      "loss": 1.2789,
+      "step": 42959
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.6570796460177e-07,
+      "loss": 1.092,
+      "step": 42960
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.656605562579014e-07,
+      "loss": 2.9057,
+      "step": 42961
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.656131479140329e-07,
+      "loss": 1.555,
+      "step": 42962
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.655657395701644e-07,
+      "loss": 1.6213,
+      "step": 42963
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.655183312262959e-07,
+      "loss": 1.7552,
+      "step": 42964
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.654709228824274e-07,
+      "loss": 3.8313,
+      "step": 42965
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.654235145385589e-07,
+      "loss": 1.5605,
+      "step": 42966
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.653761061946904e-07,
+      "loss": 2.487,
+      "step": 42967
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.653286978508218e-07,
+      "loss": 0.7479,
+      "step": 42968
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.652812895069531e-07,
+      "loss": 1.5009,
+      "step": 42969
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.652338811630848e-07,
+      "loss": 1.5536,
+      "step": 42970
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.651864728192161e-07,
+      "loss": 1.189,
+      "step": 42971
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.651390644753476e-07,
+      "loss": 2.4566,
+      "step": 42972
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.650916561314793e-07,
+      "loss": 1.3032,
+      "step": 42973
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.650442477876106e-07,
+      "loss": 1.8465,
+      "step": 42974
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.64996839443742e-07,
+      "loss": 0.5471,
+      "step": 42975
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.649494310998736e-07,
+      "loss": 2.4712,
+      "step": 42976
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.64902022756005e-07,
+      "loss": 4.3438,
+      "step": 42977
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.648546144121366e-07,
+      "loss": 0.9503,
+      "step": 42978
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.64807206068268e-07,
+      "loss": 2.4218,
+      "step": 42979
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.647597977243995e-07,
+      "loss": 1.0883,
+      "step": 42980
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.64712389380531e-07,
+      "loss": 1.9242,
+      "step": 42981
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.646649810366623e-07,
+      "loss": 0.9096,
+      "step": 42982
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.64617572692794e-07,
+      "loss": 1.4087,
+      "step": 42983
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.645701643489255e-07,
+      "loss": 1.8603,
+      "step": 42984
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.645227560050568e-07,
+      "loss": 1.3091,
+      "step": 42985
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.644753476611885e-07,
+      "loss": 1.7105,
+      "step": 42986
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.6442793931732e-07,
+      "loss": 2.3526,
+      "step": 42987
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.643805309734513e-07,
+      "loss": 2.0089,
+      "step": 42988
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.643331226295828e-07,
+      "loss": 2.7425,
+      "step": 42989
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.642857142857145e-07,
+      "loss": 2.0318,
+      "step": 42990
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.642383059418457e-07,
+      "loss": 2.4774,
+      "step": 42991
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.641908975979772e-07,
+      "loss": 2.1725,
+      "step": 42992
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.641434892541087e-07,
+      "loss": 3.2636,
+      "step": 42993
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.640960809102402e-07,
+      "loss": 2.3435,
+      "step": 42994
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.640486725663717e-07,
+      "loss": 2.4797,
+      "step": 42995
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.640012642225032e-07,
+      "loss": 2.0201,
+      "step": 42996
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.639538558786347e-07,
+      "loss": 2.9846,
+      "step": 42997
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.639064475347662e-07,
+      "loss": 2.008,
+      "step": 42998
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.638590391908975e-07,
+      "loss": 2.2737,
+      "step": 42999
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.638116308470292e-07,
+      "loss": 1.0464,
+      "step": 43000
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.637642225031607e-07,
+      "loss": 3.2201,
+      "step": 43001
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.63716814159292e-07,
+      "loss": 1.1588,
+      "step": 43002
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.636694058154236e-07,
+      "loss": 2.7542,
+      "step": 43003
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.63621997471555e-07,
+      "loss": 3.4898,
+      "step": 43004
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.635745891276864e-07,
+      "loss": 1.9489,
+      "step": 43005
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.63527180783818e-07,
+      "loss": 2.5572,
+      "step": 43006
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.634797724399494e-07,
+      "loss": 2.651,
+      "step": 43007
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.63432364096081e-07,
+      "loss": 1.7444,
+      "step": 43008
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.633849557522124e-07,
+      "loss": 1.255,
+      "step": 43009
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.633375474083439e-07,
+      "loss": 1.0632,
+      "step": 43010
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.632901390644754e-07,
+      "loss": 1.7653,
+      "step": 43011
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.632427307206069e-07,
+      "loss": 2.344,
+      "step": 43012
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.631953223767384e-07,
+      "loss": 2.3857,
+      "step": 43013
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.631479140328699e-07,
+      "loss": 2.7781,
+      "step": 43014
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.631005056890011e-07,
+      "loss": 2.453,
+      "step": 43015
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.630530973451328e-07,
+      "loss": 2.1442,
+      "step": 43016
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.630056890012643e-07,
+      "loss": 2.9492,
+      "step": 43017
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.629582806573956e-07,
+      "loss": 2.1675,
+      "step": 43018
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.62910872313527e-07,
+      "loss": 2.6482,
+      "step": 43019
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.628634639696588e-07,
+      "loss": 2.3395,
+      "step": 43020
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.6281605562579e-07,
+      "loss": 1.5618,
+      "step": 43021
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.627686472819216e-07,
+      "loss": 2.6077,
+      "step": 43022
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.627212389380533e-07,
+      "loss": 1.9352,
+      "step": 43023
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.626738305941846e-07,
+      "loss": 2.7421,
+      "step": 43024
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.62626422250316e-07,
+      "loss": 1.8644,
+      "step": 43025
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.625790139064475e-07,
+      "loss": 2.8162,
+      "step": 43026
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.62531605562579e-07,
+      "loss": 1.4481,
+      "step": 43027
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.624841972187105e-07,
+      "loss": 3.0974,
+      "step": 43028
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.62436788874842e-07,
+      "loss": 1.8451,
+      "step": 43029
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.623893805309735e-07,
+      "loss": 1.5544,
+      "step": 43030
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.62341972187105e-07,
+      "loss": 2.1267,
+      "step": 43031
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.622945638432363e-07,
+      "loss": 0.5987,
+      "step": 43032
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.62247155499368e-07,
+      "loss": 0.7541,
+      "step": 43033
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.621997471554995e-07,
+      "loss": 3.3746,
+      "step": 43034
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.621523388116308e-07,
+      "loss": 0.9632,
+      "step": 43035
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.621049304677625e-07,
+      "loss": 2.4475,
+      "step": 43036
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.620575221238937e-07,
+      "loss": 1.37,
+      "step": 43037
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.620101137800252e-07,
+      "loss": 1.4725,
+      "step": 43038
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.619627054361567e-07,
+      "loss": 1.9184,
+      "step": 43039
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.619152970922882e-07,
+      "loss": 2.0521,
+      "step": 43040
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.618678887484197e-07,
+      "loss": 1.4489,
+      "step": 43041
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.618204804045512e-07,
+      "loss": 2.3651,
+      "step": 43042
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.617730720606827e-07,
+      "loss": 1.4383,
+      "step": 43043
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.617256637168142e-07,
+      "loss": 3.0843,
+      "step": 43044
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.616782553729457e-07,
+      "loss": 2.9744,
+      "step": 43045
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.616308470290772e-07,
+      "loss": 2.1284,
+      "step": 43046
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.615834386852087e-07,
+      "loss": 2.109,
+      "step": 43047
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.6153603034134e-07,
+      "loss": 2.2061,
+      "step": 43048
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.614886219974717e-07,
+      "loss": 0.9775,
+      "step": 43049
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.614412136536031e-07,
+      "loss": 1.6942,
+      "step": 43050
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.613938053097344e-07,
+      "loss": 2.0462,
+      "step": 43051
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.61346396965866e-07,
+      "loss": 2.2494,
+      "step": 43052
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.612989886219976e-07,
+      "loss": 2.9732,
+      "step": 43053
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.61251580278129e-07,
+      "loss": 1.4584,
+      "step": 43054
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.612041719342604e-07,
+      "loss": 1.7724,
+      "step": 43055
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.61156763590392e-07,
+      "loss": 1.1656,
+      "step": 43056
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.611093552465234e-07,
+      "loss": 1.991,
+      "step": 43057
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.610619469026549e-07,
+      "loss": 2.7426,
+      "step": 43058
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.610145385587864e-07,
+      "loss": 2.0826,
+      "step": 43059
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.609671302149179e-07,
+      "loss": 2.7669,
+      "step": 43060
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.609197218710493e-07,
+      "loss": 0.3481,
+      "step": 43061
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.608723135271808e-07,
+      "loss": 1.7449,
+      "step": 43062
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.608249051833123e-07,
+      "loss": 2.5307,
+      "step": 43063
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.607774968394438e-07,
+      "loss": 3.2812,
+      "step": 43064
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.60730088495575e-07,
+      "loss": 2.4044,
+      "step": 43065
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.606826801517068e-07,
+      "loss": 2.0139,
+      "step": 43066
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.606352718078383e-07,
+      "loss": 0.67,
+      "step": 43067
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.605878634639696e-07,
+      "loss": 2.2533,
+      "step": 43068
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.605404551201013e-07,
+      "loss": 1.8672,
+      "step": 43069
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.604930467762326e-07,
+      "loss": 1.501,
+      "step": 43070
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.60445638432364e-07,
+      "loss": 1.5097,
+      "step": 43071
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.603982300884955e-07,
+      "loss": 1.6135,
+      "step": 43072
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.60350821744627e-07,
+      "loss": 1.5419,
+      "step": 43073
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.603034134007585e-07,
+      "loss": 2.8646,
+      "step": 43074
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.6025600505689e-07,
+      "loss": 3.4442,
+      "step": 43075
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.602085967130215e-07,
+      "loss": 1.8265,
+      "step": 43076
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.60161188369153e-07,
+      "loss": 1.9405,
+      "step": 43077
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.601137800252845e-07,
+      "loss": 2.8839,
+      "step": 43078
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.60066371681416e-07,
+      "loss": 1.4889,
+      "step": 43079
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.600189633375475e-07,
+      "loss": 2.6464,
+      "step": 43080
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.599715549936788e-07,
+      "loss": 1.9471,
+      "step": 43081
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.599241466498105e-07,
+      "loss": 1.7137,
+      "step": 43082
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.59876738305942e-07,
+      "loss": 2.0651,
+      "step": 43083
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.598293299620732e-07,
+      "loss": 2.8982,
+      "step": 43084
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.597819216182047e-07,
+      "loss": 3.1134,
+      "step": 43085
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.597345132743364e-07,
+      "loss": 1.2601,
+      "step": 43086
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.596871049304677e-07,
+      "loss": 3.2037,
+      "step": 43087
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.596396965865992e-07,
+      "loss": 3.6973,
+      "step": 43088
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.59592288242731e-07,
+      "loss": 0.8529,
+      "step": 43089
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.595448798988622e-07,
+      "loss": 2.4733,
+      "step": 43090
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.594974715549937e-07,
+      "loss": 2.3499,
+      "step": 43091
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.594500632111252e-07,
+      "loss": 2.5874,
+      "step": 43092
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.594026548672567e-07,
+      "loss": 1.5505,
+      "step": 43093
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.593552465233882e-07,
+      "loss": 1.4487,
+      "step": 43094
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.593078381795197e-07,
+      "loss": 2.4089,
+      "step": 43095
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.592604298356511e-07,
+      "loss": 1.179,
+      "step": 43096
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.592130214917826e-07,
+      "loss": 2.2299,
+      "step": 43097
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.59165613147914e-07,
+      "loss": 2.3024,
+      "step": 43098
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.591182048040456e-07,
+      "loss": 2.9061,
+      "step": 43099
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.590707964601771e-07,
+      "loss": 1.415,
+      "step": 43100
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.590233881163084e-07,
+      "loss": 1.3729,
+      "step": 43101
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.5897597977244e-07,
+      "loss": 1.3246,
+      "step": 43102
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.589285714285714e-07,
+      "loss": 2.4722,
+      "step": 43103
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.588811630847029e-07,
+      "loss": 1.868,
+      "step": 43104
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.588337547408344e-07,
+      "loss": 1.3044,
+      "step": 43105
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.587863463969659e-07,
+      "loss": 2.0147,
+      "step": 43106
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.587389380530974e-07,
+      "loss": 3.5435,
+      "step": 43107
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.586915297092288e-07,
+      "loss": 2.5917,
+      "step": 43108
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.586441213653603e-07,
+      "loss": 3.2659,
+      "step": 43109
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.585967130214918e-07,
+      "loss": 2.8876,
+      "step": 43110
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.585493046776233e-07,
+      "loss": 1.1044,
+      "step": 43111
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.585018963337548e-07,
+      "loss": 3.0584,
+      "step": 43112
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.584544879898863e-07,
+      "loss": 2.7649,
+      "step": 43113
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.584070796460176e-07,
+      "loss": 1.4918,
+      "step": 43114
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.583596713021493e-07,
+      "loss": 3.2792,
+      "step": 43115
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.583122629582808e-07,
+      "loss": 1.7122,
+      "step": 43116
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.58264854614412e-07,
+      "loss": 1.9363,
+      "step": 43117
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.582174462705436e-07,
+      "loss": 3.3995,
+      "step": 43118
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.581700379266753e-07,
+      "loss": 2.0713,
+      "step": 43119
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.581226295828065e-07,
+      "loss": 2.116,
+      "step": 43120
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.58075221238938e-07,
+      "loss": 1.4188,
+      "step": 43121
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.580278128950697e-07,
+      "loss": 1.1534,
+      "step": 43122
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.57980404551201e-07,
+      "loss": 0.7791,
+      "step": 43123
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.579329962073325e-07,
+      "loss": 1.2374,
+      "step": 43124
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.57885587863464e-07,
+      "loss": 2.3456,
+      "step": 43125
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.578381795195955e-07,
+      "loss": 2.6089,
+      "step": 43126
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.57790771175727e-07,
+      "loss": 1.8725,
+      "step": 43127
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.577433628318585e-07,
+      "loss": 0.5715,
+      "step": 43128
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.5769595448799e-07,
+      "loss": 1.9238,
+      "step": 43129
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.576485461441215e-07,
+      "loss": 2.629,
+      "step": 43130
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.576011378002527e-07,
+      "loss": 0.7344,
+      "step": 43131
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.575537294563844e-07,
+      "loss": 1.5099,
+      "step": 43132
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.57506321112516e-07,
+      "loss": 2.6054,
+      "step": 43133
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.574589127686472e-07,
+      "loss": 2.7155,
+      "step": 43134
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.57411504424779e-07,
+      "loss": 0.9871,
+      "step": 43135
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.573640960809102e-07,
+      "loss": 2.9655,
+      "step": 43136
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.573166877370417e-07,
+      "loss": 1.0162,
+      "step": 43137
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.572692793931732e-07,
+      "loss": 2.5144,
+      "step": 43138
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.572218710493047e-07,
+      "loss": 1.0979,
+      "step": 43139
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.571744627054362e-07,
+      "loss": 2.2928,
+      "step": 43140
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.571270543615677e-07,
+      "loss": 1.7563,
+      "step": 43141
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.570796460176992e-07,
+      "loss": 2.5794,
+      "step": 43142
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.570322376738306e-07,
+      "loss": 2.51,
+      "step": 43143
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.569848293299621e-07,
+      "loss": 2.1841,
+      "step": 43144
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.569374209860936e-07,
+      "loss": 1.8789,
+      "step": 43145
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.568900126422251e-07,
+      "loss": 3.6967,
+      "step": 43146
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.568426042983564e-07,
+      "loss": 1.7234,
+      "step": 43147
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.56795195954488e-07,
+      "loss": 2.9688,
+      "step": 43148
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.567477876106196e-07,
+      "loss": 1.9511,
+      "step": 43149
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.567003792667509e-07,
+      "loss": 2.6137,
+      "step": 43150
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.566529709228824e-07,
+      "loss": 2.7229,
+      "step": 43151
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.56605562579014e-07,
+      "loss": 2.4087,
+      "step": 43152
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.565581542351454e-07,
+      "loss": 0.4025,
+      "step": 43153
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.565107458912768e-07,
+      "loss": 1.8951,
+      "step": 43154
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.564633375474083e-07,
+      "loss": 2.9202,
+      "step": 43155
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.564159292035398e-07,
+      "loss": 2.2061,
+      "step": 43156
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.563685208596713e-07,
+      "loss": 3.3079,
+      "step": 43157
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.563211125158028e-07,
+      "loss": 2.8121,
+      "step": 43158
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.562737041719343e-07,
+      "loss": 1.7908,
+      "step": 43159
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.562262958280658e-07,
+      "loss": 1.48,
+      "step": 43160
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.56178887484197e-07,
+      "loss": 2.2714,
+      "step": 43161
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.561314791403288e-07,
+      "loss": 2.783,
+      "step": 43162
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.560840707964603e-07,
+      "loss": 1.8916,
+      "step": 43163
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.560366624525916e-07,
+      "loss": 1.8901,
+      "step": 43164
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.559892541087233e-07,
+      "loss": 1.6263,
+      "step": 43165
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.559418457648548e-07,
+      "loss": 0.3816,
+      "step": 43166
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55894437420986e-07,
+      "loss": 2.3863,
+      "step": 43167
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.558470290771175e-07,
+      "loss": 1.8298,
+      "step": 43168
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55799620733249e-07,
+      "loss": 2.135,
+      "step": 43169
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.557522123893805e-07,
+      "loss": 2.3629,
+      "step": 43170
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55704804045512e-07,
+      "loss": 2.6649,
+      "step": 43171
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.556573957016435e-07,
+      "loss": 1.2585,
+      "step": 43172
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55609987357775e-07,
+      "loss": 1.9962,
+      "step": 43173
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.555625790139065e-07,
+      "loss": 2.3011,
+      "step": 43174
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55515170670038e-07,
+      "loss": 1.2757,
+      "step": 43175
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.554677623261695e-07,
+      "loss": 0.2711,
+      "step": 43176
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55420353982301e-07,
+      "loss": 1.6775,
+      "step": 43177
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.553729456384324e-07,
+      "loss": 3.2142,
+      "step": 43178
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55325537294564e-07,
+      "loss": 2.1738,
+      "step": 43179
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.552781289506952e-07,
+      "loss": 1.7163,
+      "step": 43180
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.552307206068267e-07,
+      "loss": 3.4255,
+      "step": 43181
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.551833122629584e-07,
+      "loss": 2.5943,
+      "step": 43182
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.551359039190897e-07,
+      "loss": 1.1673,
+      "step": 43183
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.550884955752212e-07,
+      "loss": 1.3462,
+      "step": 43184
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.55041087231353e-07,
+      "loss": 2.3933,
+      "step": 43185
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.549936788874842e-07,
+      "loss": 1.7407,
+      "step": 43186
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.549462705436157e-07,
+      "loss": 3.2449,
+      "step": 43187
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.548988621997472e-07,
+      "loss": 2.4352,
+      "step": 43188
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.548514538558787e-07,
+      "loss": 2.6311,
+      "step": 43189
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.548040455120101e-07,
+      "loss": 2.2682,
+      "step": 43190
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.547566371681416e-07,
+      "loss": 3.0721,
+      "step": 43191
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.547092288242731e-07,
+      "loss": 1.8051,
+      "step": 43192
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.546618204804046e-07,
+      "loss": 1.7452,
+      "step": 43193
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.54614412136536e-07,
+      "loss": 1.4822,
+      "step": 43194
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.545670037926676e-07,
+      "loss": 3.4257,
+      "step": 43195
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.54519595448799e-07,
+      "loss": 2.0849,
+      "step": 43196
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.544721871049304e-07,
+      "loss": 2.2705,
+      "step": 43197
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.54424778761062e-07,
+      "loss": 1.8165,
+      "step": 43198
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.543773704171936e-07,
+      "loss": 2.2797,
+      "step": 43199
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.543299620733249e-07,
+      "loss": 1.246,
+      "step": 43200
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.542825537294563e-07,
+      "loss": 2.4961,
+      "step": 43201
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.542351453855878e-07,
+      "loss": 1.5028,
+      "step": 43202
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.541877370417193e-07,
+      "loss": 1.0468,
+      "step": 43203
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.541403286978508e-07,
+      "loss": 1.3926,
+      "step": 43204
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.540929203539823e-07,
+      "loss": 2.8171,
+      "step": 43205
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.540455120101138e-07,
+      "loss": 0.8207,
+      "step": 43206
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.539981036662453e-07,
+      "loss": 0.3094,
+      "step": 43207
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.539506953223768e-07,
+      "loss": 1.8127,
+      "step": 43208
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.539032869785083e-07,
+      "loss": 3.6923,
+      "step": 43209
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.538558786346398e-07,
+      "loss": 2.838,
+      "step": 43210
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.538084702907713e-07,
+      "loss": 3.0022,
+      "step": 43211
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.537610619469028e-07,
+      "loss": 1.5064,
+      "step": 43212
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.53713653603034e-07,
+      "loss": 1.738,
+      "step": 43213
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.536662452591656e-07,
+      "loss": 1.7143,
+      "step": 43214
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.536188369152971e-07,
+      "loss": 1.479,
+      "step": 43215
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.535714285714285e-07,
+      "loss": 2.2885,
+      "step": 43216
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.535240202275601e-07,
+      "loss": 1.715,
+      "step": 43217
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.534766118836916e-07,
+      "loss": 1.9625,
+      "step": 43218
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.53429203539823e-07,
+      "loss": 1.6164,
+      "step": 43219
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.533817951959545e-07,
+      "loss": 0.7978,
+      "step": 43220
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.533343868520861e-07,
+      "loss": 2.1623,
+      "step": 43221
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.532869785082175e-07,
+      "loss": 2.1801,
+      "step": 43222
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.53239570164349e-07,
+      "loss": 1.4743,
+      "step": 43223
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.531921618204803e-07,
+      "loss": 1.5173,
+      "step": 43224
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.531447534766119e-07,
+      "loss": 2.2749,
+      "step": 43225
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.530973451327434e-07,
+      "loss": 2.6626,
+      "step": 43226
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.530499367888748e-07,
+      "loss": 1.3655,
+      "step": 43227
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.530025284450063e-07,
+      "loss": 1.5299,
+      "step": 43228
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.529551201011379e-07,
+      "loss": 2.1988,
+      "step": 43229
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.529077117572693e-07,
+      "loss": 1.7485,
+      "step": 43230
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.528603034134008e-07,
+      "loss": 1.6757,
+      "step": 43231
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.528128950695324e-07,
+      "loss": 1.0802,
+      "step": 43232
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.527654867256637e-07,
+      "loss": 1.4991,
+      "step": 43233
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.527180783817953e-07,
+      "loss": 2.0999,
+      "step": 43234
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.526706700379267e-07,
+      "loss": 1.4064,
+      "step": 43235
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.526232616940581e-07,
+      "loss": 0.6261,
+      "step": 43236
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.525758533501897e-07,
+      "loss": 2.103,
+      "step": 43237
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.525284450063211e-07,
+      "loss": 0.3649,
+      "step": 43238
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.524810366624526e-07,
+      "loss": 1.503,
+      "step": 43239
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.524336283185841e-07,
+      "loss": 1.3629,
+      "step": 43240
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.523862199747155e-07,
+      "loss": 1.8216,
+      "step": 43241
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.523388116308471e-07,
+      "loss": 2.3476,
+      "step": 43242
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.522914032869786e-07,
+      "loss": 1.3993,
+      "step": 43243
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.5224399494311e-07,
+      "loss": 1.2615,
+      "step": 43244
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.521965865992416e-07,
+      "loss": 1.5854,
+      "step": 43245
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.521491782553729e-07,
+      "loss": 2.9965,
+      "step": 43246
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.521017699115045e-07,
+      "loss": 3.3367,
+      "step": 43247
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.52054361567636e-07,
+      "loss": 1.1993,
+      "step": 43248
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.520069532237673e-07,
+      "loss": 3.7317,
+      "step": 43249
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.519595448798989e-07,
+      "loss": 2.7124,
+      "step": 43250
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.519121365360304e-07,
+      "loss": 1.8406,
+      "step": 43251
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.518647281921618e-07,
+      "loss": 1.0835,
+      "step": 43252
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.518173198482933e-07,
+      "loss": 4.2335,
+      "step": 43253
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.517699115044249e-07,
+      "loss": 4.3678,
+      "step": 43254
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.517225031605563e-07,
+      "loss": 1.6268,
+      "step": 43255
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.516750948166878e-07,
+      "loss": 3.3012,
+      "step": 43256
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.516276864728192e-07,
+      "loss": 1.3262,
+      "step": 43257
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.515802781289507e-07,
+      "loss": 2.4844,
+      "step": 43258
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.515328697850823e-07,
+      "loss": 2.6943,
+      "step": 43259
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.514854614412136e-07,
+      "loss": 1.5391,
+      "step": 43260
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.514380530973451e-07,
+      "loss": 2.0989,
+      "step": 43261
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.513906447534767e-07,
+      "loss": 2.5986,
+      "step": 43262
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.513432364096081e-07,
+      "loss": 1.8138,
+      "step": 43263
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.512958280657396e-07,
+      "loss": 0.7758,
+      "step": 43264
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.512484197218711e-07,
+      "loss": 3.349,
+      "step": 43265
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.512010113780025e-07,
+      "loss": 2.919,
+      "step": 43266
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.511536030341341e-07,
+      "loss": 2.3409,
+      "step": 43267
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.511061946902655e-07,
+      "loss": 1.5176,
+      "step": 43268
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.51058786346397e-07,
+      "loss": 2.3171,
+      "step": 43269
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.510113780025286e-07,
+      "loss": 1.8546,
+      "step": 43270
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.509639696586598e-07,
+      "loss": 2.3958,
+      "step": 43271
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.509165613147914e-07,
+      "loss": 1.3115,
+      "step": 43272
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.508691529709229e-07,
+      "loss": 3.5772,
+      "step": 43273
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.508217446270543e-07,
+      "loss": 0.9362,
+      "step": 43274
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.507743362831859e-07,
+      "loss": 2.2956,
+      "step": 43275
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.507269279393174e-07,
+      "loss": 1.7236,
+      "step": 43276
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.506795195954488e-07,
+      "loss": 1.9352,
+      "step": 43277
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.506321112515803e-07,
+      "loss": 1.5017,
+      "step": 43278
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.505847029077117e-07,
+      "loss": 4.2226,
+      "step": 43279
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.505372945638433e-07,
+      "loss": 1.0063,
+      "step": 43280
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.504898862199748e-07,
+      "loss": 1.4539,
+      "step": 43281
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.504424778761062e-07,
+      "loss": 1.1895,
+      "step": 43282
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.503950695322378e-07,
+      "loss": 1.9137,
+      "step": 43283
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.503476611883692e-07,
+      "loss": 1.5368,
+      "step": 43284
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.503002528445006e-07,
+      "loss": 1.4571,
+      "step": 43285
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.502528445006321e-07,
+      "loss": 1.2854,
+      "step": 43286
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.502054361567637e-07,
+      "loss": 1.3259,
+      "step": 43287
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.501580278128951e-07,
+      "loss": 2.865,
+      "step": 43288
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.501106194690266e-07,
+      "loss": 1.3861,
+      "step": 43289
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.50063211125158e-07,
+      "loss": 1.9328,
+      "step": 43290
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.500158027812895e-07,
+      "loss": 1.8011,
+      "step": 43291
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.499683944374211e-07,
+      "loss": 3.2458,
+      "step": 43292
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.499209860935525e-07,
+      "loss": 1.8683,
+      "step": 43293
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.49873577749684e-07,
+      "loss": 2.2547,
+      "step": 43294
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.498261694058156e-07,
+      "loss": 2.716,
+      "step": 43295
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.497787610619469e-07,
+      "loss": 2.3134,
+      "step": 43296
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.497313527180784e-07,
+      "loss": 2.3003,
+      "step": 43297
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.496839443742099e-07,
+      "loss": 3.5301,
+      "step": 43298
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.496365360303413e-07,
+      "loss": 3.1337,
+      "step": 43299
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.495891276864729e-07,
+      "loss": 0.7764,
+      "step": 43300
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.495417193426043e-07,
+      "loss": 2.1109,
+      "step": 43301
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.494943109987358e-07,
+      "loss": 1.1391,
+      "step": 43302
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.494469026548674e-07,
+      "loss": 1.6009,
+      "step": 43303
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.493994943109987e-07,
+      "loss": 1.8908,
+      "step": 43304
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.493520859671303e-07,
+      "loss": 3.4766,
+      "step": 43305
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.493046776232618e-07,
+      "loss": 2.7805,
+      "step": 43306
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.492572692793931e-07,
+      "loss": 2.12,
+      "step": 43307
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.492098609355247e-07,
+      "loss": 2.4146,
+      "step": 43308
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.491624525916562e-07,
+      "loss": 2.0672,
+      "step": 43309
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.491150442477876e-07,
+      "loss": 2.3642,
+      "step": 43310
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.490676359039191e-07,
+      "loss": 1.9079,
+      "step": 43311
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.490202275600505e-07,
+      "loss": 1.6043,
+      "step": 43312
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.489728192161821e-07,
+      "loss": 1.0515,
+      "step": 43313
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.489254108723136e-07,
+      "loss": 1.6386,
+      "step": 43314
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.48878002528445e-07,
+      "loss": 2.2165,
+      "step": 43315
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.488305941845765e-07,
+      "loss": 1.4192,
+      "step": 43316
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.487831858407081e-07,
+      "loss": 1.5353,
+      "step": 43317
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.487357774968394e-07,
+      "loss": 1.9277,
+      "step": 43318
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.486883691529709e-07,
+      "loss": 2.1765,
+      "step": 43319
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.486409608091025e-07,
+      "loss": 0.7303,
+      "step": 43320
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.485935524652339e-07,
+      "loss": 0.9517,
+      "step": 43321
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.485461441213654e-07,
+      "loss": 1.9731,
+      "step": 43322
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.484987357774968e-07,
+      "loss": 1.5124,
+      "step": 43323
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.484513274336283e-07,
+      "loss": 1.1845,
+      "step": 43324
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.484039190897599e-07,
+      "loss": 1.6801,
+      "step": 43325
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.483565107458913e-07,
+      "loss": 3.4038,
+      "step": 43326
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.483091024020228e-07,
+      "loss": 2.3794,
+      "step": 43327
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.482616940581544e-07,
+      "loss": 2.4383,
+      "step": 43328
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.482142857142857e-07,
+      "loss": 1.0897,
+      "step": 43329
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.481668773704172e-07,
+      "loss": 1.4823,
+      "step": 43330
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.481194690265487e-07,
+      "loss": 1.2235,
+      "step": 43331
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.480720606826801e-07,
+      "loss": 3.1543,
+      "step": 43332
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.480246523388117e-07,
+      "loss": 2.3279,
+      "step": 43333
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.479772439949431e-07,
+      "loss": 1.4679,
+      "step": 43334
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.479298356510746e-07,
+      "loss": 2.9873,
+      "step": 43335
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.478824273072061e-07,
+      "loss": 0.9335,
+      "step": 43336
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.478350189633375e-07,
+      "loss": 1.9468,
+      "step": 43337
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.477876106194691e-07,
+      "loss": 1.3551,
+      "step": 43338
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.477402022756006e-07,
+      "loss": 1.7878,
+      "step": 43339
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.47692793931732e-07,
+      "loss": 1.6915,
+      "step": 43340
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.476453855878636e-07,
+      "loss": 2.6315,
+      "step": 43341
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.475979772439948e-07,
+      "loss": 2.8029,
+      "step": 43342
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.475505689001264e-07,
+      "loss": 0.9145,
+      "step": 43343
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.475031605562579e-07,
+      "loss": 1.4452,
+      "step": 43344
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.474557522123893e-07,
+      "loss": 2.1185,
+      "step": 43345
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.474083438685209e-07,
+      "loss": 1.2151,
+      "step": 43346
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.473609355246524e-07,
+      "loss": 1.8981,
+      "step": 43347
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.473135271807838e-07,
+      "loss": 1.7986,
+      "step": 43348
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.472661188369153e-07,
+      "loss": 2.7261,
+      "step": 43349
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.472187104930469e-07,
+      "loss": 2.113,
+      "step": 43350
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.471713021491783e-07,
+      "loss": 1.7228,
+      "step": 43351
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.471238938053098e-07,
+      "loss": 3.128,
+      "step": 43352
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.470764854614411e-07,
+      "loss": 1.9625,
+      "step": 43353
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.470290771175727e-07,
+      "loss": 1.8222,
+      "step": 43354
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.469816687737042e-07,
+      "loss": 1.6551,
+      "step": 43355
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.469342604298356e-07,
+      "loss": 2.753,
+      "step": 43356
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.468868520859671e-07,
+      "loss": 2.3743,
+      "step": 43357
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.468394437420987e-07,
+      "loss": 2.0702,
+      "step": 43358
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.467920353982301e-07,
+      "loss": 2.8749,
+      "step": 43359
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.467446270543616e-07,
+      "loss": 2.2363,
+      "step": 43360
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.466972187104932e-07,
+      "loss": 1.3984,
+      "step": 43361
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.466498103666245e-07,
+      "loss": 0.8642,
+      "step": 43362
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.466024020227561e-07,
+      "loss": 0.6689,
+      "step": 43363
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.465549936788875e-07,
+      "loss": 2.4985,
+      "step": 43364
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.465075853350189e-07,
+      "loss": 0.9254,
+      "step": 43365
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.464601769911505e-07,
+      "loss": 2.7691,
+      "step": 43366
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.464127686472819e-07,
+      "loss": 1.8662,
+      "step": 43367
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.463653603034134e-07,
+      "loss": 2.8197,
+      "step": 43368
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.463179519595449e-07,
+      "loss": 2.0866,
+      "step": 43369
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.462705436156763e-07,
+      "loss": 1.769,
+      "step": 43370
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.462231352718079e-07,
+      "loss": 3.0284,
+      "step": 43371
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.461757269279394e-07,
+      "loss": 2.28,
+      "step": 43372
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.461283185840708e-07,
+      "loss": 1.119,
+      "step": 43373
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.460809102402024e-07,
+      "loss": 2.8611,
+      "step": 43374
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.460335018963337e-07,
+      "loss": 2.4041,
+      "step": 43375
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.459860935524653e-07,
+      "loss": 2.4936,
+      "step": 43376
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.459386852085967e-07,
+      "loss": 1.8036,
+      "step": 43377
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.458912768647281e-07,
+      "loss": 0.941,
+      "step": 43378
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.458438685208597e-07,
+      "loss": 2.9722,
+      "step": 43379
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.457964601769912e-07,
+      "loss": 2.6458,
+      "step": 43380
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.457490518331226e-07,
+      "loss": 1.0749,
+      "step": 43381
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.457016434892541e-07,
+      "loss": 1.12,
+      "step": 43382
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.456542351453857e-07,
+      "loss": 2.5546,
+      "step": 43383
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.456068268015171e-07,
+      "loss": 2.5367,
+      "step": 43384
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.455594184576486e-07,
+      "loss": 3.7319,
+      "step": 43385
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.4551201011378e-07,
+      "loss": 1.3754,
+      "step": 43386
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.454646017699115e-07,
+      "loss": 1.435,
+      "step": 43387
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.454171934260431e-07,
+      "loss": 1.0114,
+      "step": 43388
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.453697850821744e-07,
+      "loss": 1.4491,
+      "step": 43389
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.453223767383059e-07,
+      "loss": 1.8215,
+      "step": 43390
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.452749683944375e-07,
+      "loss": 0.8575,
+      "step": 43391
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.452275600505689e-07,
+      "loss": 1.6518,
+      "step": 43392
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.451801517067004e-07,
+      "loss": 2.8015,
+      "step": 43393
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.451327433628319e-07,
+      "loss": 2.114,
+      "step": 43394
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.450853350189633e-07,
+      "loss": 2.1052,
+      "step": 43395
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.450379266750949e-07,
+      "loss": 1.428,
+      "step": 43396
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.449905183312263e-07,
+      "loss": 1.1093,
+      "step": 43397
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.449431099873578e-07,
+      "loss": 1.3255,
+      "step": 43398
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.448957016434894e-07,
+      "loss": 1.334,
+      "step": 43399
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.448482932996206e-07,
+      "loss": 1.0786,
+      "step": 43400
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.448008849557522e-07,
+      "loss": 3.2178,
+      "step": 43401
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.447534766118837e-07,
+      "loss": 2.5875,
+      "step": 43402
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.447060682680151e-07,
+      "loss": 0.8463,
+      "step": 43403
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.446586599241467e-07,
+      "loss": 2.4681,
+      "step": 43404
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.446112515802782e-07,
+      "loss": 2.1328,
+      "step": 43405
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.445638432364096e-07,
+      "loss": 1.3643,
+      "step": 43406
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.445164348925411e-07,
+      "loss": 0.6206,
+      "step": 43407
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.444690265486725e-07,
+      "loss": 2.7543,
+      "step": 43408
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.444216182048041e-07,
+      "loss": 2.3042,
+      "step": 43409
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.443742098609356e-07,
+      "loss": 1.9553,
+      "step": 43410
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.44326801517067e-07,
+      "loss": 2.6037,
+      "step": 43411
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.442793931731985e-07,
+      "loss": 1.9746,
+      "step": 43412
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.4423198482933e-07,
+      "loss": 1.7373,
+      "step": 43413
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.441845764854614e-07,
+      "loss": 2.4052,
+      "step": 43414
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.441371681415929e-07,
+      "loss": 1.2142,
+      "step": 43415
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.440897597977245e-07,
+      "loss": 2.5343,
+      "step": 43416
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.440423514538559e-07,
+      "loss": 2.2261,
+      "step": 43417
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.439949431099874e-07,
+      "loss": 1.611,
+      "step": 43418
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.439475347661188e-07,
+      "loss": 1.8258,
+      "step": 43419
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.439001264222503e-07,
+      "loss": 2.2412,
+      "step": 43420
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.438527180783819e-07,
+      "loss": 2.6959,
+      "step": 43421
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.438053097345133e-07,
+      "loss": 1.8781,
+      "step": 43422
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.437579013906448e-07,
+      "loss": 1.5655,
+      "step": 43423
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.437104930467763e-07,
+      "loss": 1.4673,
+      "step": 43424
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.436630847029077e-07,
+      "loss": 2.5808,
+      "step": 43425
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.436156763590392e-07,
+      "loss": 3.2131,
+      "step": 43426
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.435682680151707e-07,
+      "loss": 2.4401,
+      "step": 43427
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.435208596713021e-07,
+      "loss": 2.4234,
+      "step": 43428
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.434734513274337e-07,
+      "loss": 1.681,
+      "step": 43429
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.434260429835651e-07,
+      "loss": 1.53,
+      "step": 43430
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.433786346396966e-07,
+      "loss": 0.9863,
+      "step": 43431
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.433312262958282e-07,
+      "loss": 1.9329,
+      "step": 43432
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.432838179519595e-07,
+      "loss": 3.0449,
+      "step": 43433
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.432364096080911e-07,
+      "loss": 2.1247,
+      "step": 43434
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.431890012642226e-07,
+      "loss": 2.9921,
+      "step": 43435
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.431415929203539e-07,
+      "loss": 1.9877,
+      "step": 43436
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.430941845764855e-07,
+      "loss": 1.996,
+      "step": 43437
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.43046776232617e-07,
+      "loss": 2.0486,
+      "step": 43438
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.429993678887484e-07,
+      "loss": 1.2478,
+      "step": 43439
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.429519595448799e-07,
+      "loss": 1.3923,
+      "step": 43440
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.429045512010113e-07,
+      "loss": 1.7372,
+      "step": 43441
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.428571428571429e-07,
+      "loss": 1.2796,
+      "step": 43442
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.428097345132744e-07,
+      "loss": 2.2993,
+      "step": 43443
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.427623261694058e-07,
+      "loss": 2.9582,
+      "step": 43444
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.427149178255374e-07,
+      "loss": 2.3854,
+      "step": 43445
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.426675094816689e-07,
+      "loss": 2.0645,
+      "step": 43446
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.426201011378002e-07,
+      "loss": 1.9242,
+      "step": 43447
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.425726927939317e-07,
+      "loss": 2.4061,
+      "step": 43448
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.425252844500633e-07,
+      "loss": 2.0218,
+      "step": 43449
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.424778761061947e-07,
+      "loss": 1.8798,
+      "step": 43450
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.424304677623262e-07,
+      "loss": 1.8068,
+      "step": 43451
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.423830594184576e-07,
+      "loss": 1.4472,
+      "step": 43452
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.423356510745891e-07,
+      "loss": 1.6509,
+      "step": 43453
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.422882427307207e-07,
+      "loss": 2.9206,
+      "step": 43454
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.422408343868521e-07,
+      "loss": 1.0387,
+      "step": 43455
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.421934260429836e-07,
+      "loss": 2.2581,
+      "step": 43456
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.421460176991152e-07,
+      "loss": 1.0978,
+      "step": 43457
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.420986093552464e-07,
+      "loss": 2.4992,
+      "step": 43458
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.42051201011378e-07,
+      "loss": 1.6168,
+      "step": 43459
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.420037926675095e-07,
+      "loss": 0.7891,
+      "step": 43460
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.419563843236409e-07,
+      "loss": 1.9376,
+      "step": 43461
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.419089759797725e-07,
+      "loss": 0.9264,
+      "step": 43462
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.418615676359039e-07,
+      "loss": 2.0686,
+      "step": 43463
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.418141592920354e-07,
+      "loss": 1.6387,
+      "step": 43464
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.417667509481669e-07,
+      "loss": 3.5515,
+      "step": 43465
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.417193426042983e-07,
+      "loss": 1.5698,
+      "step": 43466
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.416719342604299e-07,
+      "loss": 1.876,
+      "step": 43467
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.416245259165614e-07,
+      "loss": 1.0481,
+      "step": 43468
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.415771175726928e-07,
+      "loss": 2.3862,
+      "step": 43469
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.415297092288244e-07,
+      "loss": 1.5959,
+      "step": 43470
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.414823008849558e-07,
+      "loss": 1.4479,
+      "step": 43471
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.414348925410872e-07,
+      "loss": 0.6156,
+      "step": 43472
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.413874841972187e-07,
+      "loss": 2.1169,
+      "step": 43473
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.413400758533501e-07,
+      "loss": 1.2217,
+      "step": 43474
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.412926675094817e-07,
+      "loss": 1.0451,
+      "step": 43475
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.412452591656132e-07,
+      "loss": 1.9382,
+      "step": 43476
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.411978508217446e-07,
+      "loss": 3.5644,
+      "step": 43477
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.411504424778761e-07,
+      "loss": 1.6217,
+      "step": 43478
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.411030341340077e-07,
+      "loss": 1.4517,
+      "step": 43479
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.410556257901391e-07,
+      "loss": 2.5811,
+      "step": 43480
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.410082174462706e-07,
+      "loss": 1.2044,
+      "step": 43481
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.409608091024022e-07,
+      "loss": 3.0193,
+      "step": 43482
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.409134007585335e-07,
+      "loss": 1.996,
+      "step": 43483
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.40865992414665e-07,
+      "loss": 1.6833,
+      "step": 43484
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.408185840707964e-07,
+      "loss": 1.4806,
+      "step": 43485
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.407711757269279e-07,
+      "loss": 1.5792,
+      "step": 43486
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.407237673830595e-07,
+      "loss": 1.5155,
+      "step": 43487
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.406763590391909e-07,
+      "loss": 2.8875,
+      "step": 43488
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.406289506953224e-07,
+      "loss": 2.9019,
+      "step": 43489
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.40581542351454e-07,
+      "loss": 2.9994,
+      "step": 43490
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.405341340075853e-07,
+      "loss": 1.368,
+      "step": 43491
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.404867256637169e-07,
+      "loss": 2.7688,
+      "step": 43492
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.404393173198484e-07,
+      "loss": 1.8061,
+      "step": 43493
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.403919089759797e-07,
+      "loss": 2.4289,
+      "step": 43494
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.403445006321113e-07,
+      "loss": 1.5114,
+      "step": 43495
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.402970922882427e-07,
+      "loss": 1.3453,
+      "step": 43496
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.402496839443742e-07,
+      "loss": 4.1078,
+      "step": 43497
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.402022756005057e-07,
+      "loss": 1.4904,
+      "step": 43498
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.401548672566371e-07,
+      "loss": 2.747,
+      "step": 43499
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.401074589127687e-07,
+      "loss": 2.082,
+      "step": 43500
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.400600505689002e-07,
+      "loss": 3.7347,
+      "step": 43501
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.400126422250316e-07,
+      "loss": 1.3449,
+      "step": 43502
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.399652338811632e-07,
+      "loss": 3.0772,
+      "step": 43503
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.399178255372947e-07,
+      "loss": 1.8997,
+      "step": 43504
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.39870417193426e-07,
+      "loss": 1.4161,
+      "step": 43505
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.398230088495575e-07,
+      "loss": 2.3327,
+      "step": 43506
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.397756005056889e-07,
+      "loss": 1.3439,
+      "step": 43507
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.397281921618205e-07,
+      "loss": 1.3912,
+      "step": 43508
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.39680783817952e-07,
+      "loss": 1.4503,
+      "step": 43509
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.396333754740834e-07,
+      "loss": 2.3556,
+      "step": 43510
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.395859671302149e-07,
+      "loss": 1.9262,
+      "step": 43511
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.395385587863465e-07,
+      "loss": 1.8667,
+      "step": 43512
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.394911504424779e-07,
+      "loss": 0.6954,
+      "step": 43513
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.394437420986094e-07,
+      "loss": 2.3164,
+      "step": 43514
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.39396333754741e-07,
+      "loss": 1.6748,
+      "step": 43515
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.393489254108724e-07,
+      "loss": 1.605,
+      "step": 43516
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.393015170670039e-07,
+      "loss": 1.3724,
+      "step": 43517
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.392541087231352e-07,
+      "loss": 3.1133,
+      "step": 43518
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.392067003792667e-07,
+      "loss": 1.3218,
+      "step": 43519
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.391592920353983e-07,
+      "loss": 1.1318,
+      "step": 43520
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.391118836915297e-07,
+      "loss": 3.1825,
+      "step": 43521
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.390644753476612e-07,
+      "loss": 2.4652,
+      "step": 43522
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.390170670037928e-07,
+      "loss": 2.3556,
+      "step": 43523
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.389696586599241e-07,
+      "loss": 1.7702,
+      "step": 43524
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.389222503160557e-07,
+      "loss": 2.0948,
+      "step": 43525
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.388748419721872e-07,
+      "loss": 2.4865,
+      "step": 43526
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.388274336283186e-07,
+      "loss": 2.4111,
+      "step": 43527
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.387800252844502e-07,
+      "loss": 0.5458,
+      "step": 43528
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.387326169405814e-07,
+      "loss": 3.5148,
+      "step": 43529
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.38685208596713e-07,
+      "loss": 1.1905,
+      "step": 43530
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.386378002528445e-07,
+      "loss": 2.6791,
+      "step": 43531
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.385903919089759e-07,
+      "loss": 1.3861,
+      "step": 43532
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.385429835651075e-07,
+      "loss": 1.2978,
+      "step": 43533
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.38495575221239e-07,
+      "loss": 1.6994,
+      "step": 43534
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.384481668773704e-07,
+      "loss": 1.8462,
+      "step": 43535
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.384007585335019e-07,
+      "loss": 0.9531,
+      "step": 43536
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.383533501896335e-07,
+      "loss": 1.0896,
+      "step": 43537
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.383059418457649e-07,
+      "loss": 1.307,
+      "step": 43538
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.382585335018964e-07,
+      "loss": 2.0483,
+      "step": 43539
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.382111251580277e-07,
+      "loss": 2.2499,
+      "step": 43540
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.381637168141593e-07,
+      "loss": 0.7256,
+      "step": 43541
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.381163084702908e-07,
+      "loss": 1.3255,
+      "step": 43542
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.380689001264222e-07,
+      "loss": 1.7212,
+      "step": 43543
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.380214917825537e-07,
+      "loss": 3.7885,
+      "step": 43544
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.379740834386853e-07,
+      "loss": 3.0689,
+      "step": 43545
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.379266750948167e-07,
+      "loss": 4.0731,
+      "step": 43546
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.378792667509482e-07,
+      "loss": 2.0799,
+      "step": 43547
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.378318584070798e-07,
+      "loss": 3.0297,
+      "step": 43548
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.377844500632111e-07,
+      "loss": 2.9356,
+      "step": 43549
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.377370417193427e-07,
+      "loss": 1.8208,
+      "step": 43550
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.376896333754741e-07,
+      "loss": 2.7219,
+      "step": 43551
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.376422250316055e-07,
+      "loss": 2.7218,
+      "step": 43552
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.375948166877371e-07,
+      "loss": 1.5722,
+      "step": 43553
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.375474083438685e-07,
+      "loss": 2.4571,
+      "step": 43554
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.375e-07,
+      "loss": 2.8716,
+      "step": 43555
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.374525916561315e-07,
+      "loss": 1.1441,
+      "step": 43556
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.374051833122629e-07,
+      "loss": 2.5647,
+      "step": 43557
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.373577749683945e-07,
+      "loss": 1.9559,
+      "step": 43558
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.37310366624526e-07,
+      "loss": 1.1278,
+      "step": 43559
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.372629582806574e-07,
+      "loss": 0.6762,
+      "step": 43560
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.37215549936789e-07,
+      "loss": 1.1603,
+      "step": 43561
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.371681415929203e-07,
+      "loss": 1.5326,
+      "step": 43562
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.371207332490519e-07,
+      "loss": 2.5811,
+      "step": 43563
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.370733249051833e-07,
+      "loss": 1.4704,
+      "step": 43564
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.370259165613147e-07,
+      "loss": 2.5153,
+      "step": 43565
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.369785082174463e-07,
+      "loss": 2.2183,
+      "step": 43566
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.369310998735778e-07,
+      "loss": 2.3468,
+      "step": 43567
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.368836915297092e-07,
+      "loss": 3.512,
+      "step": 43568
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.368362831858407e-07,
+      "loss": 0.9622,
+      "step": 43569
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.367888748419723e-07,
+      "loss": 2.4161,
+      "step": 43570
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.367414664981037e-07,
+      "loss": 3.3664,
+      "step": 43571
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.366940581542352e-07,
+      "loss": 2.1061,
+      "step": 43572
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.366466498103666e-07,
+      "loss": 1.5937,
+      "step": 43573
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.365992414664982e-07,
+      "loss": 1.0422,
+      "step": 43574
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.365518331226297e-07,
+      "loss": 1.9764,
+      "step": 43575
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.36504424778761e-07,
+      "loss": 2.717,
+      "step": 43576
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.364570164348925e-07,
+      "loss": 1.5483,
+      "step": 43577
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.364096080910241e-07,
+      "loss": 1.8696,
+      "step": 43578
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.363621997471555e-07,
+      "loss": 2.2641,
+      "step": 43579
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.36314791403287e-07,
+      "loss": 2.0833,
+      "step": 43580
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.362673830594186e-07,
+      "loss": 1.547,
+      "step": 43581
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.362199747155499e-07,
+      "loss": 2.488,
+      "step": 43582
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.361725663716815e-07,
+      "loss": 2.0869,
+      "step": 43583
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.361251580278129e-07,
+      "loss": 2.3011,
+      "step": 43584
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.360777496839444e-07,
+      "loss": 1.8182,
+      "step": 43585
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.36030341340076e-07,
+      "loss": 1.2572,
+      "step": 43586
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.359829329962072e-07,
+      "loss": 1.2538,
+      "step": 43587
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.359355246523388e-07,
+      "loss": 1.0875,
+      "step": 43588
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.358881163084703e-07,
+      "loss": 3.9892,
+      "step": 43589
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.358407079646017e-07,
+      "loss": 1.9844,
+      "step": 43590
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.357932996207333e-07,
+      "loss": 1.4352,
+      "step": 43591
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.357458912768648e-07,
+      "loss": 3.0316,
+      "step": 43592
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.356984829329962e-07,
+      "loss": 2.1153,
+      "step": 43593
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.356510745891277e-07,
+      "loss": 1.443,
+      "step": 43594
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.356036662452591e-07,
+      "loss": 2.0797,
+      "step": 43595
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.355562579013907e-07,
+      "loss": 1.7456,
+      "step": 43596
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.355088495575222e-07,
+      "loss": 2.6623,
+      "step": 43597
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.354614412136536e-07,
+      "loss": 3.3343,
+      "step": 43598
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.354140328697852e-07,
+      "loss": 3.0236,
+      "step": 43599
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.353666245259166e-07,
+      "loss": 0.6605,
+      "step": 43600
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.35319216182048e-07,
+      "loss": 1.8635,
+      "step": 43601
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.352718078381795e-07,
+      "loss": 2.0848,
+      "step": 43602
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.352243994943111e-07,
+      "loss": 2.1352,
+      "step": 43603
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.351769911504425e-07,
+      "loss": 1.9081,
+      "step": 43604
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.35129582806574e-07,
+      "loss": 2.345,
+      "step": 43605
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.350821744627054e-07,
+      "loss": 1.4658,
+      "step": 43606
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.350347661188369e-07,
+      "loss": 0.7067,
+      "step": 43607
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.349873577749685e-07,
+      "loss": 1.0388,
+      "step": 43608
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.349399494310999e-07,
+      "loss": 2.1401,
+      "step": 43609
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.348925410872314e-07,
+      "loss": 2.1085,
+      "step": 43610
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.34845132743363e-07,
+      "loss": 2.7974,
+      "step": 43611
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.347977243994943e-07,
+      "loss": 0.9162,
+      "step": 43612
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.347503160556258e-07,
+      "loss": 0.9859,
+      "step": 43613
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.347029077117573e-07,
+      "loss": 2.1207,
+      "step": 43614
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.346554993678887e-07,
+      "loss": 1.3906,
+      "step": 43615
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.346080910240203e-07,
+      "loss": 1.1579,
+      "step": 43616
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.345606826801517e-07,
+      "loss": 3.1957,
+      "step": 43617
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.345132743362832e-07,
+      "loss": 3.1243,
+      "step": 43618
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.344658659924148e-07,
+      "loss": 0.9394,
+      "step": 43619
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.344184576485461e-07,
+      "loss": 1.0273,
+      "step": 43620
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.343710493046777e-07,
+      "loss": 2.2387,
+      "step": 43621
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.343236409608092e-07,
+      "loss": 1.5042,
+      "step": 43622
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.342762326169405e-07,
+      "loss": 2.8677,
+      "step": 43623
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.342288242730721e-07,
+      "loss": 2.0855,
+      "step": 43624
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.341814159292036e-07,
+      "loss": 1.8004,
+      "step": 43625
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.34134007585335e-07,
+      "loss": 2.7292,
+      "step": 43626
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.340865992414665e-07,
+      "loss": 2.9639,
+      "step": 43627
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.340391908975979e-07,
+      "loss": 1.0291,
+      "step": 43628
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.339917825537295e-07,
+      "loss": 1.4709,
+      "step": 43629
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.33944374209861e-07,
+      "loss": 2.0598,
+      "step": 43630
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.338969658659924e-07,
+      "loss": 1.6896,
+      "step": 43631
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.33849557522124e-07,
+      "loss": 0.9038,
+      "step": 43632
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.338021491782555e-07,
+      "loss": 1.6488,
+      "step": 43633
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.337547408343868e-07,
+      "loss": 0.7546,
+      "step": 43634
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.337073324905183e-07,
+      "loss": 2.4994,
+      "step": 43635
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.336599241466499e-07,
+      "loss": 1.2666,
+      "step": 43636
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.336125158027813e-07,
+      "loss": 1.6587,
+      "step": 43637
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.335651074589128e-07,
+      "loss": 2.0512,
+      "step": 43638
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.335176991150442e-07,
+      "loss": 1.8726,
+      "step": 43639
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.334702907711757e-07,
+      "loss": 3.0701,
+      "step": 43640
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.334228824273073e-07,
+      "loss": 2.3065,
+      "step": 43641
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.333754740834387e-07,
+      "loss": 3.1282,
+      "step": 43642
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.333280657395702e-07,
+      "loss": 3.0852,
+      "step": 43643
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.332806573957018e-07,
+      "loss": 3.9863,
+      "step": 43644
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.332332490518332e-07,
+      "loss": 2.7852,
+      "step": 43645
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.331858407079646e-07,
+      "loss": 1.9306,
+      "step": 43646
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.331384323640961e-07,
+      "loss": 2.1436,
+      "step": 43647
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.330910240202275e-07,
+      "loss": 1.935,
+      "step": 43648
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.330436156763591e-07,
+      "loss": 1.177,
+      "step": 43649
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.329962073324905e-07,
+      "loss": 1.5834,
+      "step": 43650
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.32948798988622e-07,
+      "loss": 1.9128,
+      "step": 43651
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.329013906447536e-07,
+      "loss": 0.7904,
+      "step": 43652
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.328539823008849e-07,
+      "loss": 1.0543,
+      "step": 43653
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.328065739570165e-07,
+      "loss": 3.3175,
+      "step": 43654
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.32759165613148e-07,
+      "loss": 0.9155,
+      "step": 43655
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.327117572692794e-07,
+      "loss": 1.2501,
+      "step": 43656
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.32664348925411e-07,
+      "loss": 1.546,
+      "step": 43657
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.326169405815424e-07,
+      "loss": 1.3436,
+      "step": 43658
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.325695322376738e-07,
+      "loss": 2.348,
+      "step": 43659
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.325221238938053e-07,
+      "loss": 0.3239,
+      "step": 43660
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.324747155499367e-07,
+      "loss": 2.6447,
+      "step": 43661
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.324273072060683e-07,
+      "loss": 1.5856,
+      "step": 43662
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.323798988621998e-07,
+      "loss": 1.9115,
+      "step": 43663
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.323324905183312e-07,
+      "loss": 2.5955,
+      "step": 43664
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.322850821744627e-07,
+      "loss": 2.0808,
+      "step": 43665
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.322376738305943e-07,
+      "loss": 1.352,
+      "step": 43666
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.321902654867257e-07,
+      "loss": 1.0959,
+      "step": 43667
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.321428571428572e-07,
+      "loss": 2.456,
+      "step": 43668
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.320954487989888e-07,
+      "loss": 1.6058,
+      "step": 43669
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.320480404551201e-07,
+      "loss": 1.6321,
+      "step": 43670
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.320006321112516e-07,
+      "loss": 1.2367,
+      "step": 43671
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.31953223767383e-07,
+      "loss": 2.0922,
+      "step": 43672
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.319058154235145e-07,
+      "loss": 1.3337,
+      "step": 43673
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.318584070796461e-07,
+      "loss": 0.6504,
+      "step": 43674
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.318109987357775e-07,
+      "loss": 1.9572,
+      "step": 43675
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.31763590391909e-07,
+      "loss": 1.8401,
+      "step": 43676
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.317161820480406e-07,
+      "loss": 2.1312,
+      "step": 43677
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.316687737041719e-07,
+      "loss": 2.3469,
+      "step": 43678
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.316213653603035e-07,
+      "loss": 2.1871,
+      "step": 43679
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.31573957016435e-07,
+      "loss": 2.4075,
+      "step": 43680
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.315265486725663e-07,
+      "loss": 2.0459,
+      "step": 43681
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.314791403286979e-07,
+      "loss": 2.3592,
+      "step": 43682
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.314317319848293e-07,
+      "loss": 1.8826,
+      "step": 43683
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.313843236409608e-07,
+      "loss": 1.2219,
+      "step": 43684
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.313369152970923e-07,
+      "loss": 1.9892,
+      "step": 43685
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.312895069532237e-07,
+      "loss": 1.0756,
+      "step": 43686
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.312420986093553e-07,
+      "loss": 2.5761,
+      "step": 43687
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.311946902654868e-07,
+      "loss": 1.4282,
+      "step": 43688
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.311472819216182e-07,
+      "loss": 1.22,
+      "step": 43689
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.310998735777498e-07,
+      "loss": 2.7671,
+      "step": 43690
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.310524652338813e-07,
+      "loss": 2.2696,
+      "step": 43691
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.310050568900127e-07,
+      "loss": 3.751,
+      "step": 43692
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.309576485461441e-07,
+      "loss": 2.7512,
+      "step": 43693
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.309102402022755e-07,
+      "loss": 2.2192,
+      "step": 43694
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.308628318584071e-07,
+      "loss": 1.2416,
+      "step": 43695
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.308154235145386e-07,
+      "loss": 2.0177,
+      "step": 43696
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.3076801517067e-07,
+      "loss": 2.8429,
+      "step": 43697
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.307206068268015e-07,
+      "loss": 1.897,
+      "step": 43698
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.306731984829331e-07,
+      "loss": 1.6506,
+      "step": 43699
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.306257901390645e-07,
+      "loss": 0.6841,
+      "step": 43700
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.30578381795196e-07,
+      "loss": 1.1138,
+      "step": 43701
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.305309734513276e-07,
+      "loss": 2.3725,
+      "step": 43702
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.30483565107459e-07,
+      "loss": 2.9332,
+      "step": 43703
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.304361567635905e-07,
+      "loss": 0.9065,
+      "step": 43704
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.303887484197218e-07,
+      "loss": 2.4601,
+      "step": 43705
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.303413400758533e-07,
+      "loss": 1.5947,
+      "step": 43706
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.302939317319849e-07,
+      "loss": 2.1238,
+      "step": 43707
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.302465233881163e-07,
+      "loss": 3.2756,
+      "step": 43708
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.301991150442478e-07,
+      "loss": 0.5482,
+      "step": 43709
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.301517067003794e-07,
+      "loss": 2.1954,
+      "step": 43710
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.301042983565107e-07,
+      "loss": 2.0591,
+      "step": 43711
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.300568900126423e-07,
+      "loss": 1.8395,
+      "step": 43712
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.300094816687738e-07,
+      "loss": 2.1554,
+      "step": 43713
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.299620733249052e-07,
+      "loss": 1.4093,
+      "step": 43714
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.299146649810368e-07,
+      "loss": 1.0138,
+      "step": 43715
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.298672566371681e-07,
+      "loss": 1.9343,
+      "step": 43716
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.298198482932996e-07,
+      "loss": 2.813,
+      "step": 43717
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.297724399494311e-07,
+      "loss": 1.2687,
+      "step": 43718
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.297250316055625e-07,
+      "loss": 2.0263,
+      "step": 43719
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.296776232616941e-07,
+      "loss": 1.8258,
+      "step": 43720
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.296302149178256e-07,
+      "loss": 0.8666,
+      "step": 43721
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.29582806573957e-07,
+      "loss": 2.6086,
+      "step": 43722
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.295353982300886e-07,
+      "loss": 2.2628,
+      "step": 43723
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.294879898862201e-07,
+      "loss": 3.2199,
+      "step": 43724
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.294405815423515e-07,
+      "loss": 0.765,
+      "step": 43725
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.29393173198483e-07,
+      "loss": 2.7833,
+      "step": 43726
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.293457648546143e-07,
+      "loss": 1.558,
+      "step": 43727
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.292983565107459e-07,
+      "loss": 1.3496,
+      "step": 43728
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.292509481668774e-07,
+      "loss": 0.7259,
+      "step": 43729
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.292035398230088e-07,
+      "loss": 1.661,
+      "step": 43730
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.291561314791403e-07,
+      "loss": 2.8119,
+      "step": 43731
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.291087231352719e-07,
+      "loss": 2.3231,
+      "step": 43732
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.290613147914033e-07,
+      "loss": 2.115,
+      "step": 43733
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.290139064475348e-07,
+      "loss": 1.7276,
+      "step": 43734
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.289664981036664e-07,
+      "loss": 1.9156,
+      "step": 43735
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.289190897597977e-07,
+      "loss": 0.9256,
+      "step": 43736
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.288716814159293e-07,
+      "loss": 1.4427,
+      "step": 43737
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.288242730720607e-07,
+      "loss": 2.0878,
+      "step": 43738
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.287768647281921e-07,
+      "loss": 2.478,
+      "step": 43739
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.287294563843237e-07,
+      "loss": 4.0651,
+      "step": 43740
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.286820480404551e-07,
+      "loss": 1.7117,
+      "step": 43741
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.286346396965866e-07,
+      "loss": 2.9852,
+      "step": 43742
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.285872313527181e-07,
+      "loss": 1.6204,
+      "step": 43743
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.285398230088495e-07,
+      "loss": 3.3138,
+      "step": 43744
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.284924146649811e-07,
+      "loss": 1.616,
+      "step": 43745
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.284450063211126e-07,
+      "loss": 0.7791,
+      "step": 43746
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.28397597977244e-07,
+      "loss": 1.0202,
+      "step": 43747
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.283501896333756e-07,
+      "loss": 1.6361,
+      "step": 43748
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.283027812895069e-07,
+      "loss": 1.8269,
+      "step": 43749
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.282553729456385e-07,
+      "loss": 2.2626,
+      "step": 43750
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.2820796460177e-07,
+      "loss": 2.5659,
+      "step": 43751
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.281605562579013e-07,
+      "loss": 1.4757,
+      "step": 43752
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.281131479140329e-07,
+      "loss": 1.4103,
+      "step": 43753
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.280657395701644e-07,
+      "loss": 2.9161,
+      "step": 43754
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.280183312262958e-07,
+      "loss": 3.1095,
+      "step": 43755
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.279709228824273e-07,
+      "loss": 1.748,
+      "step": 43756
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.279235145385589e-07,
+      "loss": 2.4902,
+      "step": 43757
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.278761061946903e-07,
+      "loss": 3.7721,
+      "step": 43758
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.278286978508218e-07,
+      "loss": 3.6016,
+      "step": 43759
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.277812895069532e-07,
+      "loss": 1.3005,
+      "step": 43760
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.277338811630848e-07,
+      "loss": 1.9458,
+      "step": 43761
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.276864728192163e-07,
+      "loss": 0.8202,
+      "step": 43762
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.276390644753476e-07,
+      "loss": 1.7007,
+      "step": 43763
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.275916561314791e-07,
+      "loss": 1.8918,
+      "step": 43764
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.275442477876107e-07,
+      "loss": 0.9355,
+      "step": 43765
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.274968394437421e-07,
+      "loss": 4.4012,
+      "step": 43766
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.274494310998736e-07,
+      "loss": 1.9343,
+      "step": 43767
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.274020227560052e-07,
+      "loss": 1.9956,
+      "step": 43768
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.273546144121365e-07,
+      "loss": 0.8185,
+      "step": 43769
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.273072060682681e-07,
+      "loss": 1.7694,
+      "step": 43770
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.272597977243995e-07,
+      "loss": 1.0511,
+      "step": 43771
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.27212389380531e-07,
+      "loss": 1.6886,
+      "step": 43772
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.271649810366626e-07,
+      "loss": 1.7288,
+      "step": 43773
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.27117572692794e-07,
+      "loss": 1.0942,
+      "step": 43774
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.270701643489254e-07,
+      "loss": 2.2084,
+      "step": 43775
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.270227560050569e-07,
+      "loss": 2.1687,
+      "step": 43776
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.269753476611883e-07,
+      "loss": 1.8422,
+      "step": 43777
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.269279393173199e-07,
+      "loss": 2.3014,
+      "step": 43778
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.268805309734513e-07,
+      "loss": 2.4199,
+      "step": 43779
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.268331226295828e-07,
+      "loss": 2.4317,
+      "step": 43780
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.267857142857144e-07,
+      "loss": 1.3349,
+      "step": 43781
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.267383059418457e-07,
+      "loss": 1.5959,
+      "step": 43782
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.266908975979773e-07,
+      "loss": 1.6685,
+      "step": 43783
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.266434892541088e-07,
+      "loss": 3.1815,
+      "step": 43784
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.265960809102402e-07,
+      "loss": 2.6545,
+      "step": 43785
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.265486725663718e-07,
+      "loss": 1.9807,
+      "step": 43786
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.265012642225032e-07,
+      "loss": 1.3419,
+      "step": 43787
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.264538558786346e-07,
+      "loss": 1.27,
+      "step": 43788
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.264064475347661e-07,
+      "loss": 3.0092,
+      "step": 43789
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.263590391908975e-07,
+      "loss": 1.8574,
+      "step": 43790
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.263116308470291e-07,
+      "loss": 1.2801,
+      "step": 43791
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.262642225031606e-07,
+      "loss": 2.6634,
+      "step": 43792
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.26216814159292e-07,
+      "loss": 2.0571,
+      "step": 43793
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.261694058154236e-07,
+      "loss": 0.924,
+      "step": 43794
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.261219974715551e-07,
+      "loss": 1.872,
+      "step": 43795
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.260745891276865e-07,
+      "loss": 2.5007,
+      "step": 43796
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.26027180783818e-07,
+      "loss": 2.1997,
+      "step": 43797
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.259797724399496e-07,
+      "loss": 1.261,
+      "step": 43798
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.259323640960809e-07,
+      "loss": 1.3433,
+      "step": 43799
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.258849557522124e-07,
+      "loss": 1.2882,
+      "step": 43800
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.258375474083438e-07,
+      "loss": 2.2922,
+      "step": 43801
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.257901390644753e-07,
+      "loss": 2.3823,
+      "step": 43802
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.257427307206069e-07,
+      "loss": 2.6206,
+      "step": 43803
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.256953223767383e-07,
+      "loss": 1.7156,
+      "step": 43804
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.256479140328698e-07,
+      "loss": 2.1895,
+      "step": 43805
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.256005056890014e-07,
+      "loss": 3.5242,
+      "step": 43806
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.255530973451327e-07,
+      "loss": 1.4385,
+      "step": 43807
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.255056890012643e-07,
+      "loss": 1.6923,
+      "step": 43808
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.254582806573958e-07,
+      "loss": 2.0964,
+      "step": 43809
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.254108723135271e-07,
+      "loss": 1.7725,
+      "step": 43810
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.253634639696587e-07,
+      "loss": 2.1935,
+      "step": 43811
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.253160556257901e-07,
+      "loss": 2.839,
+      "step": 43812
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.252686472819216e-07,
+      "loss": 3.346,
+      "step": 43813
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.252212389380531e-07,
+      "loss": 3.442,
+      "step": 43814
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.251738305941845e-07,
+      "loss": 2.0235,
+      "step": 43815
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.251264222503161e-07,
+      "loss": 0.9782,
+      "step": 43816
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.250790139064476e-07,
+      "loss": 2.3906,
+      "step": 43817
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.25031605562579e-07,
+      "loss": 0.9184,
+      "step": 43818
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.249841972187106e-07,
+      "loss": 2.5799,
+      "step": 43819
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.249367888748421e-07,
+      "loss": 2.2189,
+      "step": 43820
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.248893805309734e-07,
+      "loss": 3.5504,
+      "step": 43821
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.248419721871049e-07,
+      "loss": 1.038,
+      "step": 43822
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.247945638432363e-07,
+      "loss": 1.1958,
+      "step": 43823
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.247471554993679e-07,
+      "loss": 1.8115,
+      "step": 43824
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.246997471554994e-07,
+      "loss": 1.2256,
+      "step": 43825
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.246523388116308e-07,
+      "loss": 1.84,
+      "step": 43826
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.246049304677623e-07,
+      "loss": 0.6869,
+      "step": 43827
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.245575221238939e-07,
+      "loss": 2.6686,
+      "step": 43828
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.245101137800253e-07,
+      "loss": 1.3214,
+      "step": 43829
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.244627054361568e-07,
+      "loss": 3.0309,
+      "step": 43830
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.244152970922884e-07,
+      "loss": 1.5048,
+      "step": 43831
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.243678887484198e-07,
+      "loss": 2.3757,
+      "step": 43832
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.243204804045512e-07,
+      "loss": 2.3813,
+      "step": 43833
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.242730720606826e-07,
+      "loss": 2.057,
+      "step": 43834
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.242256637168141e-07,
+      "loss": 2.3788,
+      "step": 43835
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.241782553729457e-07,
+      "loss": 2.6653,
+      "step": 43836
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.241308470290771e-07,
+      "loss": 0.9115,
+      "step": 43837
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.240834386852086e-07,
+      "loss": 2.7089,
+      "step": 43838
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.240360303413402e-07,
+      "loss": 0.5883,
+      "step": 43839
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.239886219974715e-07,
+      "loss": 1.4213,
+      "step": 43840
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.239412136536031e-07,
+      "loss": 2.3237,
+      "step": 43841
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.238938053097346e-07,
+      "loss": 2.6694,
+      "step": 43842
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.23846396965866e-07,
+      "loss": 2.9842,
+      "step": 43843
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.237989886219976e-07,
+      "loss": 2.9622,
+      "step": 43844
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.237515802781289e-07,
+      "loss": 2.0229,
+      "step": 43845
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.237041719342604e-07,
+      "loss": 2.5787,
+      "step": 43846
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.236567635903919e-07,
+      "loss": 3.2731,
+      "step": 43847
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.236093552465233e-07,
+      "loss": 1.3957,
+      "step": 43848
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.235619469026549e-07,
+      "loss": 1.9933,
+      "step": 43849
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.235145385587864e-07,
+      "loss": 1.6514,
+      "step": 43850
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.234671302149178e-07,
+      "loss": 3.9306,
+      "step": 43851
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.234197218710494e-07,
+      "loss": 1.8321,
+      "step": 43852
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.233723135271809e-07,
+      "loss": 2.4872,
+      "step": 43853
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.233249051833123e-07,
+      "loss": 0.9105,
+      "step": 43854
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.232774968394438e-07,
+      "loss": 2.1899,
+      "step": 43855
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.232300884955751e-07,
+      "loss": 3.0471,
+      "step": 43856
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.231826801517067e-07,
+      "loss": 2.9391,
+      "step": 43857
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.231352718078382e-07,
+      "loss": 3.3215,
+      "step": 43858
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.230878634639696e-07,
+      "loss": 2.9892,
+      "step": 43859
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.230404551201011e-07,
+      "loss": 2.0711,
+      "step": 43860
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.229930467762327e-07,
+      "loss": 2.6895,
+      "step": 43861
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.229456384323641e-07,
+      "loss": 1.9486,
+      "step": 43862
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.228982300884956e-07,
+      "loss": 1.4612,
+      "step": 43863
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.228508217446272e-07,
+      "loss": 1.4149,
+      "step": 43864
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.228034134007585e-07,
+      "loss": 3.3642,
+      "step": 43865
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.227560050568901e-07,
+      "loss": 3.0748,
+      "step": 43866
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.227085967130215e-07,
+      "loss": 1.9155,
+      "step": 43867
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.226611883691529e-07,
+      "loss": 2.4593,
+      "step": 43868
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.226137800252845e-07,
+      "loss": 4.0609,
+      "step": 43869
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.225663716814159e-07,
+      "loss": 1.9843,
+      "step": 43870
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.225189633375474e-07,
+      "loss": 2.0361,
+      "step": 43871
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.224715549936789e-07,
+      "loss": 2.6188,
+      "step": 43872
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.224241466498103e-07,
+      "loss": 1.5941,
+      "step": 43873
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.223767383059419e-07,
+      "loss": 1.093,
+      "step": 43874
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.223293299620734e-07,
+      "loss": 3.3223,
+      "step": 43875
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.222819216182048e-07,
+      "loss": 1.55,
+      "step": 43876
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.222345132743364e-07,
+      "loss": 1.8796,
+      "step": 43877
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.221871049304677e-07,
+      "loss": 2.777,
+      "step": 43878
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.221396965865993e-07,
+      "loss": 2.5383,
+      "step": 43879
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.220922882427307e-07,
+      "loss": 2.707,
+      "step": 43880
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.220448798988621e-07,
+      "loss": 3.216,
+      "step": 43881
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.219974715549937e-07,
+      "loss": 1.7534,
+      "step": 43882
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.219500632111252e-07,
+      "loss": 2.0794,
+      "step": 43883
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.219026548672566e-07,
+      "loss": 0.7017,
+      "step": 43884
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.218552465233881e-07,
+      "loss": 1.5324,
+      "step": 43885
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.218078381795197e-07,
+      "loss": 0.6279,
+      "step": 43886
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.217604298356511e-07,
+      "loss": 1.3868,
+      "step": 43887
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.217130214917826e-07,
+      "loss": 2.4619,
+      "step": 43888
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.21665613147914e-07,
+      "loss": 1.5126,
+      "step": 43889
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.216182048040456e-07,
+      "loss": 2.7833,
+      "step": 43890
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.215707964601771e-07,
+      "loss": 1.4415,
+      "step": 43891
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.215233881163084e-07,
+      "loss": 1.1517,
+      "step": 43892
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.214759797724399e-07,
+      "loss": 1.3508,
+      "step": 43893
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.214285714285715e-07,
+      "loss": 2.0149,
+      "step": 43894
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.213811630847029e-07,
+      "loss": 2.1341,
+      "step": 43895
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.213337547408344e-07,
+      "loss": 2.1512,
+      "step": 43896
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.21286346396966e-07,
+      "loss": 2.2371,
+      "step": 43897
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.212389380530973e-07,
+      "loss": 1.3865,
+      "step": 43898
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.211915297092289e-07,
+      "loss": 2.9703,
+      "step": 43899
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.211441213653603e-07,
+      "loss": 1.5985,
+      "step": 43900
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.210967130214918e-07,
+      "loss": 1.947,
+      "step": 43901
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.210493046776234e-07,
+      "loss": 1.5249,
+      "step": 43902
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.210018963337547e-07,
+      "loss": 3.0435,
+      "step": 43903
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.209544879898862e-07,
+      "loss": 2.8538,
+      "step": 43904
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.209070796460177e-07,
+      "loss": 3.3528,
+      "step": 43905
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.208596713021491e-07,
+      "loss": 0.9506,
+      "step": 43906
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.208122629582807e-07,
+      "loss": 2.0488,
+      "step": 43907
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.207648546144122e-07,
+      "loss": 2.2324,
+      "step": 43908
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.207174462705436e-07,
+      "loss": 2.9288,
+      "step": 43909
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.206700379266752e-07,
+      "loss": 2.0837,
+      "step": 43910
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.206226295828065e-07,
+      "loss": 1.4102,
+      "step": 43911
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.205752212389381e-07,
+      "loss": 1.766,
+      "step": 43912
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.205278128950696e-07,
+      "loss": 1.6476,
+      "step": 43913
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.20480404551201e-07,
+      "loss": 1.725,
+      "step": 43914
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.204329962073325e-07,
+      "loss": 3.1967,
+      "step": 43915
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.20385587863464e-07,
+      "loss": 2.5108,
+      "step": 43916
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.203381795195954e-07,
+      "loss": 0.5545,
+      "step": 43917
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.202907711757269e-07,
+      "loss": 2.2354,
+      "step": 43918
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.202433628318585e-07,
+      "loss": 2.7853,
+      "step": 43919
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.201959544879899e-07,
+      "loss": 1.0322,
+      "step": 43920
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.201485461441214e-07,
+      "loss": 2.109,
+      "step": 43921
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.201011378002528e-07,
+      "loss": 1.5746,
+      "step": 43922
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.200537294563844e-07,
+      "loss": 2.5583,
+      "step": 43923
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.200063211125159e-07,
+      "loss": 1.8789,
+      "step": 43924
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.199589127686473e-07,
+      "loss": 1.9675,
+      "step": 43925
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.199115044247788e-07,
+      "loss": 1.8215,
+      "step": 43926
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.198640960809103e-07,
+      "loss": 1.9187,
+      "step": 43927
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.198166877370417e-07,
+      "loss": 2.1982,
+      "step": 43928
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.197692793931732e-07,
+      "loss": 2.0391,
+      "step": 43929
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.197218710493048e-07,
+      "loss": 1.5892,
+      "step": 43930
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.196744627054361e-07,
+      "loss": 1.8267,
+      "step": 43931
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.196270543615677e-07,
+      "loss": 2.6715,
+      "step": 43932
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.195796460176991e-07,
+      "loss": 2.7857,
+      "step": 43933
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.195322376738306e-07,
+      "loss": 2.6191,
+      "step": 43934
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.194848293299622e-07,
+      "loss": 1.3329,
+      "step": 43935
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.194374209860935e-07,
+      "loss": 3.0731,
+      "step": 43936
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.193900126422251e-07,
+      "loss": 1.7248,
+      "step": 43937
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.193426042983566e-07,
+      "loss": 1.8302,
+      "step": 43938
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.192951959544879e-07,
+      "loss": 2.3466,
+      "step": 43939
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.192477876106195e-07,
+      "loss": 2.4019,
+      "step": 43940
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.19200379266751e-07,
+      "loss": 1.9881,
+      "step": 43941
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.191529709228824e-07,
+      "loss": 3.3579,
+      "step": 43942
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.191055625790139e-07,
+      "loss": 2.533,
+      "step": 43943
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.190581542351453e-07,
+      "loss": 2.5796,
+      "step": 43944
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.190107458912769e-07,
+      "loss": 0.8604,
+      "step": 43945
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.189633375474084e-07,
+      "loss": 2.3424,
+      "step": 43946
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.189159292035398e-07,
+      "loss": 2.6279,
+      "step": 43947
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.188685208596714e-07,
+      "loss": 3.1194,
+      "step": 43948
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.188211125158029e-07,
+      "loss": 2.1473,
+      "step": 43949
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.187737041719342e-07,
+      "loss": 1.993,
+      "step": 43950
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.187262958280657e-07,
+      "loss": 2.7871,
+      "step": 43951
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.186788874841973e-07,
+      "loss": 2.5999,
+      "step": 43952
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.186314791403287e-07,
+      "loss": 2.2219,
+      "step": 43953
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.185840707964602e-07,
+      "loss": 2.9048,
+      "step": 43954
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.185366624525916e-07,
+      "loss": 2.1478,
+      "step": 43955
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.184892541087231e-07,
+      "loss": 1.8907,
+      "step": 43956
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.184418457648547e-07,
+      "loss": 1.8285,
+      "step": 43957
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.183944374209861e-07,
+      "loss": 2.0619,
+      "step": 43958
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.183470290771176e-07,
+      "loss": 2.1686,
+      "step": 43959
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.182996207332492e-07,
+      "loss": 1.3636,
+      "step": 43960
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.182522123893806e-07,
+      "loss": 1.7046,
+      "step": 43961
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.18204804045512e-07,
+      "loss": 2.5156,
+      "step": 43962
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.181573957016435e-07,
+      "loss": 2.291,
+      "step": 43963
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.181099873577749e-07,
+      "loss": 1.7054,
+      "step": 43964
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.180625790139065e-07,
+      "loss": 0.2999,
+      "step": 43965
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.180151706700379e-07,
+      "loss": 2.3698,
+      "step": 43966
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.179677623261694e-07,
+      "loss": 1.9795,
+      "step": 43967
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.17920353982301e-07,
+      "loss": 1.8109,
+      "step": 43968
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.178729456384323e-07,
+      "loss": 1.0665,
+      "step": 43969
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.178255372945639e-07,
+      "loss": 1.8064,
+      "step": 43970
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.177781289506954e-07,
+      "loss": 1.8637,
+      "step": 43971
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.177307206068268e-07,
+      "loss": 1.3895,
+      "step": 43972
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.176833122629584e-07,
+      "loss": 2.063,
+      "step": 43973
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.176359039190898e-07,
+      "loss": 1.3627,
+      "step": 43974
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.175884955752212e-07,
+      "loss": 3.286,
+      "step": 43975
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.175410872313527e-07,
+      "loss": 1.5607,
+      "step": 43976
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.174936788874841e-07,
+      "loss": 1.9916,
+      "step": 43977
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.174462705436157e-07,
+      "loss": 1.5681,
+      "step": 43978
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.173988621997472e-07,
+      "loss": 1.9952,
+      "step": 43979
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.173514538558786e-07,
+      "loss": 1.907,
+      "step": 43980
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.173040455120102e-07,
+      "loss": 1.7541,
+      "step": 43981
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.172566371681417e-07,
+      "loss": 1.5862,
+      "step": 43982
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.172092288242731e-07,
+      "loss": 1.4387,
+      "step": 43983
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.171618204804046e-07,
+      "loss": 1.3641,
+      "step": 43984
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.171144121365362e-07,
+      "loss": 1.8331,
+      "step": 43985
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.170670037926675e-07,
+      "loss": 1.7093,
+      "step": 43986
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.17019595448799e-07,
+      "loss": 2.841,
+      "step": 43987
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.169721871049304e-07,
+      "loss": 2.8466,
+      "step": 43988
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.169247787610619e-07,
+      "loss": 2.0703,
+      "step": 43989
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.168773704171935e-07,
+      "loss": 1.3589,
+      "step": 43990
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.168299620733249e-07,
+      "loss": 4.0575,
+      "step": 43991
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.167825537294564e-07,
+      "loss": 1.912,
+      "step": 43992
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.16735145385588e-07,
+      "loss": 2.8275,
+      "step": 43993
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.166877370417194e-07,
+      "loss": 1.8579,
+      "step": 43994
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.166403286978509e-07,
+      "loss": 2.2872,
+      "step": 43995
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.165929203539824e-07,
+      "loss": 2.3062,
+      "step": 43996
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.165455120101137e-07,
+      "loss": 1.43,
+      "step": 43997
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.164981036662453e-07,
+      "loss": 1.4734,
+      "step": 43998
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.164506953223767e-07,
+      "loss": 1.5827,
+      "step": 43999
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.164032869785082e-07,
+      "loss": 2.2844,
+      "step": 44000
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.163558786346398e-07,
+      "loss": 3.6602,
+      "step": 44001
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.163084702907711e-07,
+      "loss": 0.1952,
+      "step": 44002
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.162610619469027e-07,
+      "loss": 2.5705,
+      "step": 44003
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.162136536030342e-07,
+      "loss": 1.9353,
+      "step": 44004
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.161662452591656e-07,
+      "loss": 1.7984,
+      "step": 44005
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.161188369152972e-07,
+      "loss": 2.035,
+      "step": 44006
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.160714285714287e-07,
+      "loss": 3.6235,
+      "step": 44007
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.1602402022756e-07,
+      "loss": 1.8462,
+      "step": 44008
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.159766118836915e-07,
+      "loss": 3.1986,
+      "step": 44009
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.159292035398229e-07,
+      "loss": 1.0254,
+      "step": 44010
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.158817951959545e-07,
+      "loss": 1.3957,
+      "step": 44011
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.15834386852086e-07,
+      "loss": 3.1863,
+      "step": 44012
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.157869785082174e-07,
+      "loss": 0.8823,
+      "step": 44013
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.157395701643489e-07,
+      "loss": 2.3066,
+      "step": 44014
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.156921618204805e-07,
+      "loss": 2.9047,
+      "step": 44015
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.156447534766119e-07,
+      "loss": 2.2489,
+      "step": 44016
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.155973451327434e-07,
+      "loss": 1.7875,
+      "step": 44017
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.15549936788875e-07,
+      "loss": 1.9671,
+      "step": 44018
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.155025284450064e-07,
+      "loss": 2.7788,
+      "step": 44019
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.154551201011379e-07,
+      "loss": 1.4045,
+      "step": 44020
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.154077117572692e-07,
+      "loss": 2.5928,
+      "step": 44021
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.153603034134007e-07,
+      "loss": 1.2313,
+      "step": 44022
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.153128950695323e-07,
+      "loss": 2.2322,
+      "step": 44023
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.152654867256637e-07,
+      "loss": 3.8043,
+      "step": 44024
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.152180783817952e-07,
+      "loss": 0.8225,
+      "step": 44025
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.151706700379268e-07,
+      "loss": 1.3896,
+      "step": 44026
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.151232616940581e-07,
+      "loss": 2.6351,
+      "step": 44027
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.150758533501897e-07,
+      "loss": 2.0654,
+      "step": 44028
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.150284450063212e-07,
+      "loss": 2.868,
+      "step": 44029
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.149810366624526e-07,
+      "loss": 0.6098,
+      "step": 44030
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.149336283185842e-07,
+      "loss": 1.6696,
+      "step": 44031
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.148862199747155e-07,
+      "loss": 1.7522,
+      "step": 44032
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.14838811630847e-07,
+      "loss": 3.4386,
+      "step": 44033
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.147914032869785e-07,
+      "loss": 1.8663,
+      "step": 44034
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.147439949431099e-07,
+      "loss": 1.9406,
+      "step": 44035
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.146965865992415e-07,
+      "loss": 1.0732,
+      "step": 44036
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.14649178255373e-07,
+      "loss": 4.2234,
+      "step": 44037
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.146017699115044e-07,
+      "loss": 1.0101,
+      "step": 44038
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.14554361567636e-07,
+      "loss": 2.0274,
+      "step": 44039
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.145069532237675e-07,
+      "loss": 2.5023,
+      "step": 44040
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.144595448798989e-07,
+      "loss": 1.3797,
+      "step": 44041
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.144121365360304e-07,
+      "loss": 2.0643,
+      "step": 44042
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.143647281921617e-07,
+      "loss": 1.9252,
+      "step": 44043
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.143173198482933e-07,
+      "loss": 2.5455,
+      "step": 44044
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.142699115044248e-07,
+      "loss": 2.2138,
+      "step": 44045
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.142225031605562e-07,
+      "loss": 1.7632,
+      "step": 44046
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.141750948166877e-07,
+      "loss": 0.8549,
+      "step": 44047
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.141276864728193e-07,
+      "loss": 1.8457,
+      "step": 44048
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.140802781289507e-07,
+      "loss": 1.8866,
+      "step": 44049
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.140328697850822e-07,
+      "loss": 1.9059,
+      "step": 44050
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.139854614412138e-07,
+      "loss": 2.0065,
+      "step": 44051
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.139380530973452e-07,
+      "loss": 0.9169,
+      "step": 44052
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.138906447534767e-07,
+      "loss": 2.0691,
+      "step": 44053
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.138432364096081e-07,
+      "loss": 1.0449,
+      "step": 44054
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.137958280657395e-07,
+      "loss": 0.8318,
+      "step": 44055
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.137484197218711e-07,
+      "loss": 0.5712,
+      "step": 44056
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.137010113780025e-07,
+      "loss": 2.9495,
+      "step": 44057
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.13653603034134e-07,
+      "loss": 1.8758,
+      "step": 44058
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.136061946902656e-07,
+      "loss": 1.3941,
+      "step": 44059
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.135587863463969e-07,
+      "loss": 2.5873,
+      "step": 44060
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.135113780025285e-07,
+      "loss": 2.2262,
+      "step": 44061
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.1346396965866e-07,
+      "loss": 1.8696,
+      "step": 44062
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.134165613147914e-07,
+      "loss": 2.2448,
+      "step": 44063
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.13369152970923e-07,
+      "loss": 1.2878,
+      "step": 44064
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.133217446270544e-07,
+      "loss": 2.0153,
+      "step": 44065
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.132743362831859e-07,
+      "loss": 2.8096,
+      "step": 44066
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.132269279393173e-07,
+      "loss": 2.0047,
+      "step": 44067
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.131795195954487e-07,
+      "loss": 1.9019,
+      "step": 44068
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.131321112515803e-07,
+      "loss": 2.5055,
+      "step": 44069
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.130847029077118e-07,
+      "loss": 1.4043,
+      "step": 44070
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.130372945638432e-07,
+      "loss": 1.2861,
+      "step": 44071
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.129898862199748e-07,
+      "loss": 3.3941,
+      "step": 44072
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.129424778761063e-07,
+      "loss": 1.7574,
+      "step": 44073
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.128950695322377e-07,
+      "loss": 1.1471,
+      "step": 44074
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.128476611883692e-07,
+      "loss": 3.8234,
+      "step": 44075
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.128002528445006e-07,
+      "loss": 1.7516,
+      "step": 44076
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.127528445006322e-07,
+      "loss": 1.8466,
+      "step": 44077
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.127054361567637e-07,
+      "loss": 1.5107,
+      "step": 44078
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.12658027812895e-07,
+      "loss": 2.525,
+      "step": 44079
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.126106194690265e-07,
+      "loss": 4.0546,
+      "step": 44080
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.125632111251581e-07,
+      "loss": 1.0728,
+      "step": 44081
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.125158027812895e-07,
+      "loss": 1.6112,
+      "step": 44082
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.12468394437421e-07,
+      "loss": 1.7815,
+      "step": 44083
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.124209860935526e-07,
+      "loss": 1.4232,
+      "step": 44084
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.123735777496839e-07,
+      "loss": 2.3104,
+      "step": 44085
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.123261694058155e-07,
+      "loss": 3.7298,
+      "step": 44086
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.122787610619469e-07,
+      "loss": 3.1191,
+      "step": 44087
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.122313527180784e-07,
+      "loss": 1.0246,
+      "step": 44088
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.1218394437421e-07,
+      "loss": 1.9683,
+      "step": 44089
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.121365360303414e-07,
+      "loss": 1.433,
+      "step": 44090
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.120891276864728e-07,
+      "loss": 1.7732,
+      "step": 44091
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.120417193426043e-07,
+      "loss": 2.1883,
+      "step": 44092
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.119943109987357e-07,
+      "loss": 1.4254,
+      "step": 44093
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.119469026548673e-07,
+      "loss": 0.8545,
+      "step": 44094
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.118994943109988e-07,
+      "loss": 1.3793,
+      "step": 44095
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.118520859671302e-07,
+      "loss": 2.4421,
+      "step": 44096
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.118046776232618e-07,
+      "loss": 3.6798,
+      "step": 44097
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.117572692793931e-07,
+      "loss": 1.6822,
+      "step": 44098
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.117098609355247e-07,
+      "loss": 3.388,
+      "step": 44099
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.116624525916562e-07,
+      "loss": 2.4887,
+      "step": 44100
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.116150442477876e-07,
+      "loss": 1.8797,
+      "step": 44101
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.115676359039192e-07,
+      "loss": 1.4046,
+      "step": 44102
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.115202275600506e-07,
+      "loss": 1.3479,
+      "step": 44103
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.11472819216182e-07,
+      "loss": 1.8435,
+      "step": 44104
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.114254108723135e-07,
+      "loss": 1.58,
+      "step": 44105
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.113780025284451e-07,
+      "loss": 1.978,
+      "step": 44106
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.113305941845765e-07,
+      "loss": 1.8813,
+      "step": 44107
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.11283185840708e-07,
+      "loss": 1.8484,
+      "step": 44108
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.112357774968394e-07,
+      "loss": 1.301,
+      "step": 44109
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.11188369152971e-07,
+      "loss": 2.1687,
+      "step": 44110
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.111409608091025e-07,
+      "loss": 2.6509,
+      "step": 44111
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.110935524652339e-07,
+      "loss": 2.1938,
+      "step": 44112
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.110461441213654e-07,
+      "loss": 3.1378,
+      "step": 44113
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.10998735777497e-07,
+      "loss": 2.0811,
+      "step": 44114
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.109513274336283e-07,
+      "loss": 2.3208,
+      "step": 44115
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.109039190897598e-07,
+      "loss": 1.3389,
+      "step": 44116
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.108565107458914e-07,
+      "loss": 1.539,
+      "step": 44117
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.108091024020227e-07,
+      "loss": 1.9984,
+      "step": 44118
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.107616940581543e-07,
+      "loss": 1.6641,
+      "step": 44119
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.107142857142857e-07,
+      "loss": 1.9993,
+      "step": 44120
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.106668773704172e-07,
+      "loss": 2.5004,
+      "step": 44121
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.106194690265488e-07,
+      "loss": 2.126,
+      "step": 44122
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.105720606826802e-07,
+      "loss": 3.0592,
+      "step": 44123
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.105246523388117e-07,
+      "loss": 2.7824,
+      "step": 44124
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.104772439949432e-07,
+      "loss": 1.1307,
+      "step": 44125
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.104298356510745e-07,
+      "loss": 1.8762,
+      "step": 44126
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.103824273072061e-07,
+      "loss": 1.9345,
+      "step": 44127
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.103350189633376e-07,
+      "loss": 1.7416,
+      "step": 44128
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.10287610619469e-07,
+      "loss": 1.5849,
+      "step": 44129
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.102402022756006e-07,
+      "loss": 2.2768,
+      "step": 44130
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.101927939317319e-07,
+      "loss": 1.8395,
+      "step": 44131
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.101453855878635e-07,
+      "loss": 0.7435,
+      "step": 44132
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.10097977243995e-07,
+      "loss": 3.7984,
+      "step": 44133
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.100505689001264e-07,
+      "loss": 2.2125,
+      "step": 44134
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.10003160556258e-07,
+      "loss": 3.4133,
+      "step": 44135
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.099557522123895e-07,
+      "loss": 2.387,
+      "step": 44136
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.099083438685208e-07,
+      "loss": 0.8748,
+      "step": 44137
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.098609355246523e-07,
+      "loss": 2.3893,
+      "step": 44138
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.098135271807839e-07,
+      "loss": 2.0554,
+      "step": 44139
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.097661188369153e-07,
+      "loss": 2.027,
+      "step": 44140
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.097187104930468e-07,
+      "loss": 0.9356,
+      "step": 44141
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.096713021491782e-07,
+      "loss": 2.0313,
+      "step": 44142
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.096238938053098e-07,
+      "loss": 3.5202,
+      "step": 44143
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.095764854614413e-07,
+      "loss": 1.5704,
+      "step": 44144
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.095290771175727e-07,
+      "loss": 1.5587,
+      "step": 44145
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.094816687737042e-07,
+      "loss": 2.4818,
+      "step": 44146
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.094342604298358e-07,
+      "loss": 2.16,
+      "step": 44147
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.093868520859672e-07,
+      "loss": 1.427,
+      "step": 44148
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.093394437420986e-07,
+      "loss": 2.6198,
+      "step": 44149
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.092920353982302e-07,
+      "loss": 2.1811,
+      "step": 44150
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.092446270543615e-07,
+      "loss": 1.4229,
+      "step": 44151
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.091972187104931e-07,
+      "loss": 0.9479,
+      "step": 44152
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.091498103666245e-07,
+      "loss": 2.1267,
+      "step": 44153
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.09102402022756e-07,
+      "loss": 2.0297,
+      "step": 44154
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.090549936788876e-07,
+      "loss": 1.8638,
+      "step": 44155
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.090075853350189e-07,
+      "loss": 0.4704,
+      "step": 44156
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.089601769911505e-07,
+      "loss": 2.5886,
+      "step": 44157
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.08912768647282e-07,
+      "loss": 2.1794,
+      "step": 44158
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.088653603034134e-07,
+      "loss": 2.6198,
+      "step": 44159
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.08817951959545e-07,
+      "loss": 2.2868,
+      "step": 44160
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.087705436156764e-07,
+      "loss": 1.8801,
+      "step": 44161
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.087231352718078e-07,
+      "loss": 1.1751,
+      "step": 44162
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.086757269279393e-07,
+      "loss": 1.6062,
+      "step": 44163
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.086283185840707e-07,
+      "loss": 1.2782,
+      "step": 44164
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.085809102402023e-07,
+      "loss": 1.3126,
+      "step": 44165
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.085335018963338e-07,
+      "loss": 1.3346,
+      "step": 44166
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.084860935524652e-07,
+      "loss": 1.8023,
+      "step": 44167
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.084386852085968e-07,
+      "loss": 2.7088,
+      "step": 44168
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.083912768647283e-07,
+      "loss": 1.6913,
+      "step": 44169
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.083438685208597e-07,
+      "loss": 2.9052,
+      "step": 44170
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.082964601769912e-07,
+      "loss": 2.3608,
+      "step": 44171
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.082490518331228e-07,
+      "loss": 1.1279,
+      "step": 44172
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.082016434892541e-07,
+      "loss": 1.7435,
+      "step": 44173
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.081542351453856e-07,
+      "loss": 1.9427,
+      "step": 44174
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.08106826801517e-07,
+      "loss": 1.2725,
+      "step": 44175
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.080594184576485e-07,
+      "loss": 1.3646,
+      "step": 44176
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.080120101137801e-07,
+      "loss": 3.0092,
+      "step": 44177
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.079646017699115e-07,
+      "loss": 2.0184,
+      "step": 44178
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.07917193426043e-07,
+      "loss": 2.0444,
+      "step": 44179
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.078697850821746e-07,
+      "loss": 0.6425,
+      "step": 44180
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.07822376738306e-07,
+      "loss": 1.478,
+      "step": 44181
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.077749683944375e-07,
+      "loss": 0.641,
+      "step": 44182
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.07727560050569e-07,
+      "loss": 1.2774,
+      "step": 44183
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.076801517067003e-07,
+      "loss": 1.9016,
+      "step": 44184
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.076327433628319e-07,
+      "loss": 1.8635,
+      "step": 44185
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.075853350189633e-07,
+      "loss": 1.1926,
+      "step": 44186
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.075379266750948e-07,
+      "loss": 1.3879,
+      "step": 44187
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.074905183312264e-07,
+      "loss": 2.7961,
+      "step": 44188
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.074431099873577e-07,
+      "loss": 1.2459,
+      "step": 44189
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.073957016434893e-07,
+      "loss": 1.8151,
+      "step": 44190
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.073482932996208e-07,
+      "loss": 2.0918,
+      "step": 44191
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.073008849557522e-07,
+      "loss": 2.995,
+      "step": 44192
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.072534766118838e-07,
+      "loss": 1.6084,
+      "step": 44193
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.072060682680153e-07,
+      "loss": 1.6114,
+      "step": 44194
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.071586599241467e-07,
+      "loss": 1.7768,
+      "step": 44195
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.071112515802781e-07,
+      "loss": 3.1693,
+      "step": 44196
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.070638432364095e-07,
+      "loss": 0.5547,
+      "step": 44197
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.070164348925411e-07,
+      "loss": 3.0367,
+      "step": 44198
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.069690265486726e-07,
+      "loss": 3.8006,
+      "step": 44199
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.06921618204804e-07,
+      "loss": 3.4569,
+      "step": 44200
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.068742098609356e-07,
+      "loss": 2.7609,
+      "step": 44201
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.068268015170671e-07,
+      "loss": 2.4879,
+      "step": 44202
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.067793931731985e-07,
+      "loss": 1.4897,
+      "step": 44203
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.0673198482933e-07,
+      "loss": 2.3933,
+      "step": 44204
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.066845764854616e-07,
+      "loss": 0.8245,
+      "step": 44205
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.06637168141593e-07,
+      "loss": 1.6027,
+      "step": 44206
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.065897597977245e-07,
+      "loss": 1.8351,
+      "step": 44207
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.065423514538558e-07,
+      "loss": 1.8481,
+      "step": 44208
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.064949431099873e-07,
+      "loss": 1.2781,
+      "step": 44209
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.064475347661189e-07,
+      "loss": 1.8102,
+      "step": 44210
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.064001264222503e-07,
+      "loss": 3.0839,
+      "step": 44211
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.063527180783818e-07,
+      "loss": 1.2319,
+      "step": 44212
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.063053097345134e-07,
+      "loss": 2.2819,
+      "step": 44213
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.062579013906447e-07,
+      "loss": 1.2672,
+      "step": 44214
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.062104930467763e-07,
+      "loss": 0.9431,
+      "step": 44215
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.061630847029077e-07,
+      "loss": 2.0667,
+      "step": 44216
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.061156763590392e-07,
+      "loss": 2.0354,
+      "step": 44217
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.060682680151708e-07,
+      "loss": 1.5915,
+      "step": 44218
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.060208596713021e-07,
+      "loss": 3.4662,
+      "step": 44219
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.059734513274336e-07,
+      "loss": 3.0453,
+      "step": 44220
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.059260429835651e-07,
+      "loss": 3.4219,
+      "step": 44221
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.058786346396965e-07,
+      "loss": 2.4093,
+      "step": 44222
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.058312262958281e-07,
+      "loss": 2.1815,
+      "step": 44223
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.057838179519596e-07,
+      "loss": 1.1724,
+      "step": 44224
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.05736409608091e-07,
+      "loss": 1.6534,
+      "step": 44225
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.056890012642226e-07,
+      "loss": 1.5559,
+      "step": 44226
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.056415929203539e-07,
+      "loss": 1.9171,
+      "step": 44227
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.055941845764855e-07,
+      "loss": 2.2641,
+      "step": 44228
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.05546776232617e-07,
+      "loss": 2.1689,
+      "step": 44229
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.054993678887483e-07,
+      "loss": 3.336,
+      "step": 44230
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.0545195954488e-07,
+      "loss": 2.4512,
+      "step": 44231
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.054045512010114e-07,
+      "loss": 1.2278,
+      "step": 44232
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.053571428571428e-07,
+      "loss": 1.5006,
+      "step": 44233
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.053097345132743e-07,
+      "loss": 1.439,
+      "step": 44234
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.052623261694059e-07,
+      "loss": 2.5388,
+      "step": 44235
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 9.052149178255373e-07,
+      "loss": 1.9905,
+      "step": 44236
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.051675094816688e-07,
+      "loss": 1.1936,
+      "step": 44237
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.051201011378002e-07,
+      "loss": 0.874,
+      "step": 44238
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.050726927939318e-07,
+      "loss": 2.6165,
+      "step": 44239
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.050252844500633e-07,
+      "loss": 2.2085,
+      "step": 44240
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.049778761061947e-07,
+      "loss": 1.726,
+      "step": 44241
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.049304677623261e-07,
+      "loss": 0.9205,
+      "step": 44242
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.048830594184577e-07,
+      "loss": 2.8565,
+      "step": 44243
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.048356510745891e-07,
+      "loss": 1.7433,
+      "step": 44244
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.047882427307206e-07,
+      "loss": 2.0192,
+      "step": 44245
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.047408343868522e-07,
+      "loss": 3.1141,
+      "step": 44246
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.046934260429835e-07,
+      "loss": 2.3129,
+      "step": 44247
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.046460176991151e-07,
+      "loss": 1.6433,
+      "step": 44248
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.045986093552465e-07,
+      "loss": 2.3046,
+      "step": 44249
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.04551201011378e-07,
+      "loss": 0.6131,
+      "step": 44250
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.045037926675096e-07,
+      "loss": 1.3513,
+      "step": 44251
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.04456384323641e-07,
+      "loss": 1.957,
+      "step": 44252
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.044089759797725e-07,
+      "loss": 3.2539,
+      "step": 44253
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.04361567635904e-07,
+      "loss": 2.0736,
+      "step": 44254
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.043141592920353e-07,
+      "loss": 0.3317,
+      "step": 44255
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.042667509481669e-07,
+      "loss": 3.2582,
+      "step": 44256
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.042193426042984e-07,
+      "loss": 3.8252,
+      "step": 44257
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.041719342604298e-07,
+      "loss": 1.9259,
+      "step": 44258
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.041245259165614e-07,
+      "loss": 1.4581,
+      "step": 44259
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.040771175726927e-07,
+      "loss": 0.6527,
+      "step": 44260
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.040297092288243e-07,
+      "loss": 2.1529,
+      "step": 44261
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.039823008849558e-07,
+      "loss": 1.1584,
+      "step": 44262
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.039348925410872e-07,
+      "loss": 1.9396,
+      "step": 44263
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.038874841972188e-07,
+      "loss": 1.1176,
+      "step": 44264
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.038400758533503e-07,
+      "loss": 1.8482,
+      "step": 44265
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.037926675094816e-07,
+      "loss": 2.3712,
+      "step": 44266
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.037452591656131e-07,
+      "loss": 1.5076,
+      "step": 44267
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.036978508217447e-07,
+      "loss": 3.1704,
+      "step": 44268
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.036504424778761e-07,
+      "loss": 1.5634,
+      "step": 44269
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.036030341340076e-07,
+      "loss": 2.8328,
+      "step": 44270
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.03555625790139e-07,
+      "loss": 2.9031,
+      "step": 44271
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.035082174462706e-07,
+      "loss": 0.9921,
+      "step": 44272
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.034608091024021e-07,
+      "loss": 1.1275,
+      "step": 44273
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.034134007585335e-07,
+      "loss": 2.4817,
+      "step": 44274
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.03365992414665e-07,
+      "loss": 3.4073,
+      "step": 44275
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.033185840707966e-07,
+      "loss": 2.3189,
+      "step": 44276
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.03271175726928e-07,
+      "loss": 1.2158,
+      "step": 44277
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.032237673830594e-07,
+      "loss": 2.0608,
+      "step": 44278
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.03176359039191e-07,
+      "loss": 1.451,
+      "step": 44279
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.031289506953223e-07,
+      "loss": 1.0811,
+      "step": 44280
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.030815423514539e-07,
+      "loss": 1.2186,
+      "step": 44281
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.030341340075853e-07,
+      "loss": 2.3836,
+      "step": 44282
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.029867256637168e-07,
+      "loss": 2.6359,
+      "step": 44283
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.029393173198484e-07,
+      "loss": 3.3926,
+      "step": 44284
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.028919089759797e-07,
+      "loss": 1.5637,
+      "step": 44285
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.028445006321113e-07,
+      "loss": 2.6403,
+      "step": 44286
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.027970922882428e-07,
+      "loss": 0.8727,
+      "step": 44287
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.027496839443742e-07,
+      "loss": 1.796,
+      "step": 44288
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.027022756005058e-07,
+      "loss": 2.8686,
+      "step": 44289
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.026548672566372e-07,
+      "loss": 1.4959,
+      "step": 44290
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.026074589127686e-07,
+      "loss": 2.1069,
+      "step": 44291
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.025600505689001e-07,
+      "loss": 2.9522,
+      "step": 44292
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.025126422250315e-07,
+      "loss": 2.4239,
+      "step": 44293
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.024652338811631e-07,
+      "loss": 1.8582,
+      "step": 44294
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.024178255372946e-07,
+      "loss": 0.7356,
+      "step": 44295
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.02370417193426e-07,
+      "loss": 2.0714,
+      "step": 44296
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.023230088495576e-07,
+      "loss": 3.1727,
+      "step": 44297
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.022756005056891e-07,
+      "loss": 1.0428,
+      "step": 44298
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.022281921618205e-07,
+      "loss": 1.8594,
+      "step": 44299
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.02180783817952e-07,
+      "loss": 1.7859,
+      "step": 44300
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.021333754740836e-07,
+      "loss": 3.6893,
+      "step": 44301
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.020859671302149e-07,
+      "loss": 1.3211,
+      "step": 44302
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.020385587863464e-07,
+      "loss": 3.3489,
+      "step": 44303
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.019911504424778e-07,
+      "loss": 1.6285,
+      "step": 44304
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.019437420986093e-07,
+      "loss": 2.3524,
+      "step": 44305
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.018963337547409e-07,
+      "loss": 1.3578,
+      "step": 44306
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.018489254108723e-07,
+      "loss": 3.6782,
+      "step": 44307
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.018015170670038e-07,
+      "loss": 2.5414,
+      "step": 44308
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.017541087231354e-07,
+      "loss": 0.6268,
+      "step": 44309
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.017067003792668e-07,
+      "loss": 1.6646,
+      "step": 44310
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.016592920353983e-07,
+      "loss": 2.5868,
+      "step": 44311
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.016118836915298e-07,
+      "loss": 2.6807,
+      "step": 44312
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.015644753476611e-07,
+      "loss": 2.7838,
+      "step": 44313
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.015170670037927e-07,
+      "loss": 2.5482,
+      "step": 44314
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.014696586599241e-07,
+      "loss": 1.5156,
+      "step": 44315
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.014222503160556e-07,
+      "loss": 3.2029,
+      "step": 44316
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.013748419721872e-07,
+      "loss": 1.981,
+      "step": 44317
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.013274336283185e-07,
+      "loss": 1.4417,
+      "step": 44318
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.012800252844501e-07,
+      "loss": 1.7081,
+      "step": 44319
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.012326169405816e-07,
+      "loss": 1.1959,
+      "step": 44320
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.01185208596713e-07,
+      "loss": 1.731,
+      "step": 44321
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.011378002528446e-07,
+      "loss": 1.7986,
+      "step": 44322
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.010903919089761e-07,
+      "loss": 1.1619,
+      "step": 44323
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.010429835651074e-07,
+      "loss": 2.0978,
+      "step": 44324
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.009955752212389e-07,
+      "loss": 2.1769,
+      "step": 44325
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.009481668773703e-07,
+      "loss": 2.1221,
+      "step": 44326
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.009007585335019e-07,
+      "loss": 1.6047,
+      "step": 44327
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.008533501896334e-07,
+      "loss": 1.2347,
+      "step": 44328
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.008059418457648e-07,
+      "loss": 2.144,
+      "step": 44329
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.007585335018964e-07,
+      "loss": 4.2061,
+      "step": 44330
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.007111251580279e-07,
+      "loss": 2.5514,
+      "step": 44331
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.5449378330373001,
+      "eval_loss": 1.9364993572235107,
+      "eval_runtime": 156.8041,
+      "eval_samples_per_second": 17.952,
+      "eval_steps_per_second": 4.49,
+      "step": 44331
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.006637168141593e-07,
+      "loss": 2.968,
+      "step": 44332
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.006163084702908e-07,
+      "loss": 1.9731,
+      "step": 44333
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.005689001264224e-07,
+      "loss": 1.3453,
+      "step": 44334
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.005214917825538e-07,
+      "loss": 1.1258,
+      "step": 44335
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.004740834386852e-07,
+      "loss": 1.8926,
+      "step": 44336
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.004266750948166e-07,
+      "loss": 3.5927,
+      "step": 44337
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.003792667509481e-07,
+      "loss": 3.1913,
+      "step": 44338
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.003318584070797e-07,
+      "loss": 2.1792,
+      "step": 44339
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.002844500632111e-07,
+      "loss": 1.5981,
+      "step": 44340
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.002370417193426e-07,
+      "loss": 1.7261,
+      "step": 44341
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.001896333754742e-07,
+      "loss": 2.0292,
+      "step": 44342
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.001422250316056e-07,
+      "loss": 1.6201,
+      "step": 44343
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.000948166877371e-07,
+      "loss": 1.9499,
+      "step": 44344
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.000474083438686e-07,
+      "loss": 0.5301,
+      "step": 44345
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9e-07,
+      "loss": 3.11,
+      "step": 44346
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.999525916561316e-07,
+      "loss": 0.6565,
+      "step": 44347
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.999051833122629e-07,
+      "loss": 4.6844,
+      "step": 44348
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.998577749683944e-07,
+      "loss": 2.4557,
+      "step": 44349
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.99810366624526e-07,
+      "loss": 0.882,
+      "step": 44350
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.997629582806573e-07,
+      "loss": 1.823,
+      "step": 44351
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.997155499367889e-07,
+      "loss": 1.5769,
+      "step": 44352
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.996681415929204e-07,
+      "loss": 1.6175,
+      "step": 44353
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.996207332490518e-07,
+      "loss": 1.7827,
+      "step": 44354
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.995733249051834e-07,
+      "loss": 2.8857,
+      "step": 44355
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.995259165613149e-07,
+      "loss": 2.0711,
+      "step": 44356
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.994785082174463e-07,
+      "loss": 1.4352,
+      "step": 44357
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.994310998735778e-07,
+      "loss": 0.5691,
+      "step": 44358
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.993836915297091e-07,
+      "loss": 3.1619,
+      "step": 44359
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.993362831858407e-07,
+      "loss": 1.5517,
+      "step": 44360
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.992888748419722e-07,
+      "loss": 1.9371,
+      "step": 44361
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 8.992414664981036e-07,
+      "loss": 1.8676,
+      "step": 44362
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.991940581542351e-07,
+      "loss": 1.5085,
+      "step": 44363
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.991466498103667e-07,
+      "loss": 1.3473,
+      "step": 44364
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.990992414664981e-07,
+      "loss": 2.0765,
+      "step": 44365
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.990518331226296e-07,
+      "loss": 2.5794,
+      "step": 44366
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.990044247787612e-07,
+      "loss": 1.7143,
+      "step": 44367
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.989570164348926e-07,
+      "loss": 1.45,
+      "step": 44368
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.989096080910241e-07,
+      "loss": 1.1742,
+      "step": 44369
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.988621997471555e-07,
+      "loss": 1.9494,
+      "step": 44370
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.988147914032869e-07,
+      "loss": 2.5578,
+      "step": 44371
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.987673830594185e-07,
+      "loss": 2.5937,
+      "step": 44372
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.987199747155499e-07,
+      "loss": 1.6863,
+      "step": 44373
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.986725663716814e-07,
+      "loss": 0.627,
+      "step": 44374
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.98625158027813e-07,
+      "loss": 1.3177,
+      "step": 44375
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.985777496839443e-07,
+      "loss": 3.0512,
+      "step": 44376
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.985303413400759e-07,
+      "loss": 2.6776,
+      "step": 44377
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.984829329962074e-07,
+      "loss": 1.8164,
+      "step": 44378
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.984355246523388e-07,
+      "loss": 1.5502,
+      "step": 44379
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.983881163084704e-07,
+      "loss": 1.6756,
+      "step": 44380
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.983407079646018e-07,
+      "loss": 2.3124,
+      "step": 44381
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.982932996207333e-07,
+      "loss": 1.1596,
+      "step": 44382
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.982458912768647e-07,
+      "loss": 0.9412,
+      "step": 44383
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.981984829329961e-07,
+      "loss": 2.8056,
+      "step": 44384
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.981510745891277e-07,
+      "loss": 1.8263,
+      "step": 44385
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.981036662452592e-07,
+      "loss": 2.7532,
+      "step": 44386
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.980562579013906e-07,
+      "loss": 1.6545,
+      "step": 44387
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.980088495575222e-07,
+      "loss": 1.7499,
+      "step": 44388
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.979614412136537e-07,
+      "loss": 1.2114,
+      "step": 44389
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.979140328697851e-07,
+      "loss": 1.2921,
+      "step": 44390
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.978666245259166e-07,
+      "loss": 3.0399,
+      "step": 44391
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.97819216182048e-07,
+      "loss": 1.5329,
+      "step": 44392
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.977718078381796e-07,
+      "loss": 2.055,
+      "step": 44393
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.977243994943111e-07,
+      "loss": 1.7148,
+      "step": 44394
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.976769911504424e-07,
+      "loss": 1.6229,
+      "step": 44395
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.976295828065739e-07,
+      "loss": 1.4303,
+      "step": 44396
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.975821744627055e-07,
+      "loss": 3.1975,
+      "step": 44397
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.975347661188369e-07,
+      "loss": 1.2451,
+      "step": 44398
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.974873577749684e-07,
+      "loss": 3.8375,
+      "step": 44399
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.974399494311e-07,
+      "loss": 1.3845,
+      "step": 44400
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.973925410872314e-07,
+      "loss": 2.0677,
+      "step": 44401
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.973451327433629e-07,
+      "loss": 2.0663,
+      "step": 44402
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.972977243994943e-07,
+      "loss": 1.6445,
+      "step": 44403
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.972503160556258e-07,
+      "loss": 1.721,
+      "step": 44404
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.972029077117574e-07,
+      "loss": 2.0277,
+      "step": 44405
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.971554993678887e-07,
+      "loss": 3.6742,
+      "step": 44406
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.971080910240202e-07,
+      "loss": 2.0786,
+      "step": 44407
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.970606826801518e-07,
+      "loss": 1.8421,
+      "step": 44408
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.970132743362831e-07,
+      "loss": 0.8382,
+      "step": 44409
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.969658659924147e-07,
+      "loss": 1.1381,
+      "step": 44410
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.969184576485462e-07,
+      "loss": 0.8358,
+      "step": 44411
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.968710493046776e-07,
+      "loss": 2.3613,
+      "step": 44412
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.968236409608092e-07,
+      "loss": 1.3791,
+      "step": 44413
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.967762326169406e-07,
+      "loss": 2.3356,
+      "step": 44414
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.967288242730721e-07,
+      "loss": 0.8615,
+      "step": 44415
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.966814159292036e-07,
+      "loss": 1.0244,
+      "step": 44416
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.96634007585335e-07,
+      "loss": 1.7486,
+      "step": 44417
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.965865992414665e-07,
+      "loss": 2.1801,
+      "step": 44418
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.96539190897598e-07,
+      "loss": 1.8183,
+      "step": 44419
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.964917825537294e-07,
+      "loss": 3.1526,
+      "step": 44420
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.96444374209861e-07,
+      "loss": 2.0626,
+      "step": 44421
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.963969658659925e-07,
+      "loss": 1.5759,
+      "step": 44422
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.963495575221239e-07,
+      "loss": 1.8843,
+      "step": 44423
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.963021491782554e-07,
+      "loss": 2.1927,
+      "step": 44424
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 8.962547408343868e-07,
+      "loss": 3.0932,
+      "step": 44425
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.962073324905184e-07,
+      "loss": 2.276,
+      "step": 44426
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.961599241466499e-07,
+      "loss": 2.0417,
+      "step": 44427
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.961125158027813e-07,
+      "loss": 3.3365,
+      "step": 44428
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.960651074589128e-07,
+      "loss": 1.9511,
+      "step": 44429
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.960176991150443e-07,
+      "loss": 1.8512,
+      "step": 44430
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.959702907711757e-07,
+      "loss": 2.3824,
+      "step": 44431
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.959228824273072e-07,
+      "loss": 2.5795,
+      "step": 44432
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.958754740834388e-07,
+      "loss": 2.5043,
+      "step": 44433
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.958280657395701e-07,
+      "loss": 1.0334,
+      "step": 44434
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.957806573957017e-07,
+      "loss": 2.4649,
+      "step": 44435
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.957332490518331e-07,
+      "loss": 1.4431,
+      "step": 44436
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.956858407079646e-07,
+      "loss": 1.4335,
+      "step": 44437
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.956384323640962e-07,
+      "loss": 2.3012,
+      "step": 44438
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.955910240202276e-07,
+      "loss": 2.3234,
+      "step": 44439
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.955436156763591e-07,
+      "loss": 3.2388,
+      "step": 44440
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.954962073324906e-07,
+      "loss": 2.2675,
+      "step": 44441
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.954487989886219e-07,
+      "loss": 1.8866,
+      "step": 44442
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.954013906447535e-07,
+      "loss": 2.0925,
+      "step": 44443
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.95353982300885e-07,
+      "loss": 2.2022,
+      "step": 44444
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.953065739570164e-07,
+      "loss": 1.9767,
+      "step": 44445
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.95259165613148e-07,
+      "loss": 2.0481,
+      "step": 44446
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.952117572692793e-07,
+      "loss": 3.2857,
+      "step": 44447
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.951643489254109e-07,
+      "loss": 3.7556,
+      "step": 44448
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.951169405815424e-07,
+      "loss": 2.3854,
+      "step": 44449
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.950695322376738e-07,
+      "loss": 0.6,
+      "step": 44450
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.950221238938054e-07,
+      "loss": 1.812,
+      "step": 44451
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.949747155499369e-07,
+      "loss": 1.8775,
+      "step": 44452
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.949273072060682e-07,
+      "loss": 1.3871,
+      "step": 44453
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.948798988621997e-07,
+      "loss": 2.5031,
+      "step": 44454
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.948324905183313e-07,
+      "loss": 3.0767,
+      "step": 44455
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.947850821744627e-07,
+      "loss": 1.9122,
+      "step": 44456
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.947376738305942e-07,
+      "loss": 2.1925,
+      "step": 44457
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.946902654867256e-07,
+      "loss": 0.8422,
+      "step": 44458
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.946428571428572e-07,
+      "loss": 2.0098,
+      "step": 44459
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.945954487989887e-07,
+      "loss": 1.526,
+      "step": 44460
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.945480404551201e-07,
+      "loss": 3.5163,
+      "step": 44461
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.945006321112516e-07,
+      "loss": 2.7079,
+      "step": 44462
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.944532237673832e-07,
+      "loss": 1.5136,
+      "step": 44463
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.944058154235146e-07,
+      "loss": 1.4178,
+      "step": 44464
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.94358407079646e-07,
+      "loss": 1.8915,
+      "step": 44465
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.943109987357776e-07,
+      "loss": 1.511,
+      "step": 44466
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.942635903919089e-07,
+      "loss": 1.7039,
+      "step": 44467
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.942161820480405e-07,
+      "loss": 0.9369,
+      "step": 44468
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.941687737041719e-07,
+      "loss": 2.566,
+      "step": 44469
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.941213653603034e-07,
+      "loss": 2.3374,
+      "step": 44470
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.94073957016435e-07,
+      "loss": 1.8389,
+      "step": 44471
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.940265486725664e-07,
+      "loss": 1.8805,
+      "step": 44472
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.939791403286979e-07,
+      "loss": 2.7285,
+      "step": 44473
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.939317319848294e-07,
+      "loss": 1.6782,
+      "step": 44474
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.938843236409608e-07,
+      "loss": 3.1994,
+      "step": 44475
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.938369152970924e-07,
+      "loss": 2.0405,
+      "step": 44476
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.937895069532238e-07,
+      "loss": 1.4244,
+      "step": 44477
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.937420986093552e-07,
+      "loss": 1.916,
+      "step": 44478
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.936946902654868e-07,
+      "loss": 3.3807,
+      "step": 44479
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.936472819216181e-07,
+      "loss": 1.2782,
+      "step": 44480
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.935998735777497e-07,
+      "loss": 3.0103,
+      "step": 44481
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.935524652338812e-07,
+      "loss": 1.3295,
+      "step": 44482
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.935050568900126e-07,
+      "loss": 0.8252,
+      "step": 44483
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.934576485461442e-07,
+      "loss": 1.4098,
+      "step": 44484
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.934102402022757e-07,
+      "loss": 1.0258,
+      "step": 44485
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.933628318584071e-07,
+      "loss": 2.9468,
+      "step": 44486
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.933154235145386e-07,
+      "loss": 1.4568,
+      "step": 44487
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.932680151706702e-07,
+      "loss": 1.786,
+      "step": 44488
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 8.932206068268015e-07,
+      "loss": 2.1085,
+      "step": 44489
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.93173198482933e-07,
+      "loss": 1.5902,
+      "step": 44490
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.931257901390644e-07,
+      "loss": 1.8327,
+      "step": 44491
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.930783817951959e-07,
+      "loss": 1.135,
+      "step": 44492
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.930309734513275e-07,
+      "loss": 2.6389,
+      "step": 44493
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.929835651074589e-07,
+      "loss": 1.303,
+      "step": 44494
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.929361567635904e-07,
+      "loss": 1.0474,
+      "step": 44495
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.92888748419722e-07,
+      "loss": 1.8134,
+      "step": 44496
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.928413400758534e-07,
+      "loss": 1.3602,
+      "step": 44497
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.927939317319849e-07,
+      "loss": 2.6394,
+      "step": 44498
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.927465233881164e-07,
+      "loss": 3.0868,
+      "step": 44499
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.926991150442477e-07,
+      "loss": 2.0305,
+      "step": 44500
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.926517067003793e-07,
+      "loss": 2.4266,
+      "step": 44501
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.926042983565107e-07,
+      "loss": 3.2397,
+      "step": 44502
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.925568900126422e-07,
+      "loss": 3.2682,
+      "step": 44503
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.925094816687738e-07,
+      "loss": 1.5593,
+      "step": 44504
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.924620733249051e-07,
+      "loss": 1.589,
+      "step": 44505
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.924146649810367e-07,
+      "loss": 2.7707,
+      "step": 44506
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.923672566371682e-07,
+      "loss": 1.5457,
+      "step": 44507
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.923198482932996e-07,
+      "loss": 2.7809,
+      "step": 44508
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.922724399494312e-07,
+      "loss": 3.1911,
+      "step": 44509
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.922250316055627e-07,
+      "loss": 2.4549,
+      "step": 44510
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.92177623261694e-07,
+      "loss": 1.7054,
+      "step": 44511
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.921302149178255e-07,
+      "loss": 1.762,
+      "step": 44512
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.920828065739569e-07,
+      "loss": 0.8432,
+      "step": 44513
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.920353982300885e-07,
+      "loss": 1.6825,
+      "step": 44514
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.9198798988622e-07,
+      "loss": 2.0348,
+      "step": 44515
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.919405815423514e-07,
+      "loss": 2.1551,
+      "step": 44516
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.91893173198483e-07,
+      "loss": 2.4188,
+      "step": 44517
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.918457648546145e-07,
+      "loss": 0.8343,
+      "step": 44518
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.917983565107459e-07,
+      "loss": 2.9255,
+      "step": 44519
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.917509481668774e-07,
+      "loss": 2.4892,
+      "step": 44520
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.91703539823009e-07,
+      "loss": 2.3405,
+      "step": 44521
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.916561314791404e-07,
+      "loss": 2.1708,
+      "step": 44522
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.916087231352719e-07,
+      "loss": 2.2181,
+      "step": 44523
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.915613147914032e-07,
+      "loss": 2.7721,
+      "step": 44524
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.915139064475347e-07,
+      "loss": 2.3132,
+      "step": 44525
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.914664981036663e-07,
+      "loss": 1.0267,
+      "step": 44526
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.914190897597977e-07,
+      "loss": 1.1747,
+      "step": 44527
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.913716814159292e-07,
+      "loss": 1.5936,
+      "step": 44528
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.913242730720608e-07,
+      "loss": 2.1505,
+      "step": 44529
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.912768647281922e-07,
+      "loss": 1.5465,
+      "step": 44530
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.912294563843237e-07,
+      "loss": 1.9591,
+      "step": 44531
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.911820480404552e-07,
+      "loss": 2.6481,
+      "step": 44532
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.911346396965866e-07,
+      "loss": 1.8449,
+      "step": 44533
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.910872313527182e-07,
+      "loss": 2.2176,
+      "step": 44534
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.910398230088495e-07,
+      "loss": 0.7135,
+      "step": 44535
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.90992414664981e-07,
+      "loss": 1.3963,
+      "step": 44536
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.909450063211126e-07,
+      "loss": 1.479,
+      "step": 44537
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.908975979772439e-07,
+      "loss": 3.1703,
+      "step": 44538
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.908501896333755e-07,
+      "loss": 3.1118,
+      "step": 44539
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.90802781289507e-07,
+      "loss": 3.3677,
+      "step": 44540
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.907553729456384e-07,
+      "loss": 1.7395,
+      "step": 44541
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.9070796460177e-07,
+      "loss": 2.8287,
+      "step": 44542
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.906605562579015e-07,
+      "loss": 2.1125,
+      "step": 44543
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.906131479140329e-07,
+      "loss": 3.094,
+      "step": 44544
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.905657395701644e-07,
+      "loss": 1.6455,
+      "step": 44545
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.905183312262957e-07,
+      "loss": 2.5094,
+      "step": 44546
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.904709228824273e-07,
+      "loss": 1.2238,
+      "step": 44547
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.904235145385588e-07,
+      "loss": 2.09,
+      "step": 44548
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.903761061946902e-07,
+      "loss": 2.0542,
+      "step": 44549
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.903286978508218e-07,
+      "loss": 3.2091,
+      "step": 44550
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.902812895069533e-07,
+      "loss": 2.1999,
+      "step": 44551
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 8.902338811630847e-07,
+      "loss": 2.1527,
+      "step": 44552
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.901864728192162e-07,
+      "loss": 2.6359,
+      "step": 44553
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.901390644753478e-07,
+      "loss": 2.9647,
+      "step": 44554
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.900916561314792e-07,
+      "loss": 2.7818,
+      "step": 44555
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.900442477876107e-07,
+      "loss": 1.7245,
+      "step": 44556
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.899968394437421e-07,
+      "loss": 3.2198,
+      "step": 44557
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.899494310998735e-07,
+      "loss": 2.3066,
+      "step": 44558
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.899020227560051e-07,
+      "loss": 1.0495,
+      "step": 44559
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.898546144121365e-07,
+      "loss": 2.7452,
+      "step": 44560
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.89807206068268e-07,
+      "loss": 1.7493,
+      "step": 44561
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.897597977243996e-07,
+      "loss": 2.8375,
+      "step": 44562
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.897123893805309e-07,
+      "loss": 1.8099,
+      "step": 44563
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.896649810366625e-07,
+      "loss": 1.5823,
+      "step": 44564
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.89617572692794e-07,
+      "loss": 2.1966,
+      "step": 44565
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.895701643489254e-07,
+      "loss": 3.2094,
+      "step": 44566
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.89522756005057e-07,
+      "loss": 2.9477,
+      "step": 44567
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.894753476611884e-07,
+      "loss": 1.6469,
+      "step": 44568
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.894279393173199e-07,
+      "loss": 2.8955,
+      "step": 44569
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.893805309734513e-07,
+      "loss": 1.244,
+      "step": 44570
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.893331226295827e-07,
+      "loss": 1.3257,
+      "step": 44571
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.892857142857143e-07,
+      "loss": 1.371,
+      "step": 44572
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.892383059418458e-07,
+      "loss": 1.2524,
+      "step": 44573
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.891908975979772e-07,
+      "loss": 2.623,
+      "step": 44574
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.891434892541088e-07,
+      "loss": 0.8217,
+      "step": 44575
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.890960809102403e-07,
+      "loss": 1.066,
+      "step": 44576
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.890486725663717e-07,
+      "loss": 2.2089,
+      "step": 44577
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.890012642225032e-07,
+      "loss": 4.1431,
+      "step": 44578
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.889538558786346e-07,
+      "loss": 1.0202,
+      "step": 44579
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.889064475347662e-07,
+      "loss": 2.9432,
+      "step": 44580
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.888590391908977e-07,
+      "loss": 0.2189,
+      "step": 44581
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.88811630847029e-07,
+      "loss": 1.5934,
+      "step": 44582
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.887642225031605e-07,
+      "loss": 1.5815,
+      "step": 44583
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.887168141592921e-07,
+      "loss": 2.7168,
+      "step": 44584
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.886694058154235e-07,
+      "loss": 2.3097,
+      "step": 44585
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.88621997471555e-07,
+      "loss": 1.7213,
+      "step": 44586
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.885745891276866e-07,
+      "loss": 1.4024,
+      "step": 44587
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.88527180783818e-07,
+      "loss": 2.3519,
+      "step": 44588
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.884797724399495e-07,
+      "loss": 5.3104,
+      "step": 44589
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.884323640960809e-07,
+      "loss": 2.259,
+      "step": 44590
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.883849557522124e-07,
+      "loss": 3.3601,
+      "step": 44591
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.88337547408344e-07,
+      "loss": 1.166,
+      "step": 44592
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.882901390644754e-07,
+      "loss": 1.4444,
+      "step": 44593
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.882427307206068e-07,
+      "loss": 2.9563,
+      "step": 44594
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.881953223767384e-07,
+      "loss": 2.4175,
+      "step": 44595
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.881479140328697e-07,
+      "loss": 1.1294,
+      "step": 44596
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.881005056890013e-07,
+      "loss": 2.5836,
+      "step": 44597
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.880530973451328e-07,
+      "loss": 3.0191,
+      "step": 44598
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.880056890012642e-07,
+      "loss": 1.6721,
+      "step": 44599
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.879582806573958e-07,
+      "loss": 2.5455,
+      "step": 44600
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.879108723135272e-07,
+      "loss": 1.3151,
+      "step": 44601
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.878634639696587e-07,
+      "loss": 1.3748,
+      "step": 44602
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.878160556257902e-07,
+      "loss": 2.2069,
+      "step": 44603
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.877686472819216e-07,
+      "loss": 2.0155,
+      "step": 44604
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.877212389380532e-07,
+      "loss": 1.9983,
+      "step": 44605
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.876738305941846e-07,
+      "loss": 1.6176,
+      "step": 44606
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.87626422250316e-07,
+      "loss": 2.4257,
+      "step": 44607
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.875790139064476e-07,
+      "loss": 2.4226,
+      "step": 44608
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.875316055625791e-07,
+      "loss": 1.8845,
+      "step": 44609
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.874841972187105e-07,
+      "loss": 2.7685,
+      "step": 44610
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.87436788874842e-07,
+      "loss": 2.3603,
+      "step": 44611
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.873893805309734e-07,
+      "loss": 2.3593,
+      "step": 44612
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.87341972187105e-07,
+      "loss": 2.5837,
+      "step": 44613
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.872945638432365e-07,
+      "loss": 0.5857,
+      "step": 44614
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 8.872471554993679e-07,
+      "loss": 1.642,
+      "step": 44615
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.871997471554994e-07,
+      "loss": 1.1802,
+      "step": 44616
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.87152338811631e-07,
+      "loss": 1.8065,
+      "step": 44617
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.871049304677623e-07,
+      "loss": 0.8324,
+      "step": 44618
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.870575221238938e-07,
+      "loss": 1.6313,
+      "step": 44619
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.870101137800254e-07,
+      "loss": 0.4422,
+      "step": 44620
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.869627054361568e-07,
+      "loss": 2.7263,
+      "step": 44621
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.869152970922883e-07,
+      "loss": 1.4625,
+      "step": 44622
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.868678887484197e-07,
+      "loss": 1.4874,
+      "step": 44623
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.868204804045512e-07,
+      "loss": 1.9264,
+      "step": 44624
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.867730720606828e-07,
+      "loss": 1.447,
+      "step": 44625
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.867256637168142e-07,
+      "loss": 1.1351,
+      "step": 44626
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.866782553729457e-07,
+      "loss": 1.7473,
+      "step": 44627
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.866308470290773e-07,
+      "loss": 1.573,
+      "step": 44628
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.865834386852085e-07,
+      "loss": 3.7559,
+      "step": 44629
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.865360303413401e-07,
+      "loss": 0.9832,
+      "step": 44630
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.864886219974716e-07,
+      "loss": 3.1823,
+      "step": 44631
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.86441213653603e-07,
+      "loss": 2.8057,
+      "step": 44632
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.863938053097346e-07,
+      "loss": 1.8093,
+      "step": 44633
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.863463969658659e-07,
+      "loss": 1.5237,
+      "step": 44634
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.862989886219975e-07,
+      "loss": 2.4014,
+      "step": 44635
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.86251580278129e-07,
+      "loss": 3.2087,
+      "step": 44636
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.862041719342604e-07,
+      "loss": 2.5568,
+      "step": 44637
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.86156763590392e-07,
+      "loss": 2.3685,
+      "step": 44638
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.861093552465235e-07,
+      "loss": 2.2589,
+      "step": 44639
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.860619469026548e-07,
+      "loss": 1.5788,
+      "step": 44640
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.860145385587863e-07,
+      "loss": 1.8384,
+      "step": 44641
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.859671302149179e-07,
+      "loss": 1.0355,
+      "step": 44642
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.859197218710493e-07,
+      "loss": 2.043,
+      "step": 44643
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.858723135271808e-07,
+      "loss": 1.4748,
+      "step": 44644
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.858249051833122e-07,
+      "loss": 1.7295,
+      "step": 44645
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.857774968394438e-07,
+      "loss": 3.0672,
+      "step": 44646
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.857300884955753e-07,
+      "loss": 0.4638,
+      "step": 44647
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.856826801517067e-07,
+      "loss": 1.2029,
+      "step": 44648
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.856352718078382e-07,
+      "loss": 0.9548,
+      "step": 44649
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.855878634639698e-07,
+      "loss": 1.1632,
+      "step": 44650
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.855404551201012e-07,
+      "loss": 2.1283,
+      "step": 44651
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.854930467762326e-07,
+      "loss": 1.4506,
+      "step": 44652
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.854456384323642e-07,
+      "loss": 2.0922,
+      "step": 44653
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.853982300884955e-07,
+      "loss": 1.6229,
+      "step": 44654
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.853508217446271e-07,
+      "loss": 1.7111,
+      "step": 44655
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.853034134007585e-07,
+      "loss": 2.8586,
+      "step": 44656
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.8525600505689e-07,
+      "loss": 1.5919,
+      "step": 44657
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.852085967130216e-07,
+      "loss": 1.6849,
+      "step": 44658
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.85161188369153e-07,
+      "loss": 1.7956,
+      "step": 44659
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.851137800252845e-07,
+      "loss": 3.7775,
+      "step": 44660
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.85066371681416e-07,
+      "loss": 1.7218,
+      "step": 44661
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.850189633375474e-07,
+      "loss": 1.364,
+      "step": 44662
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.84971554993679e-07,
+      "loss": 2.1496,
+      "step": 44663
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.849241466498103e-07,
+      "loss": 1.9684,
+      "step": 44664
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.848767383059418e-07,
+      "loss": 3.3336,
+      "step": 44665
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.848293299620734e-07,
+      "loss": 2.2521,
+      "step": 44666
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.847819216182047e-07,
+      "loss": 3.0953,
+      "step": 44667
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.847345132743363e-07,
+      "loss": 1.6625,
+      "step": 44668
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.846871049304678e-07,
+      "loss": 2.7441,
+      "step": 44669
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.846396965865992e-07,
+      "loss": 1.8234,
+      "step": 44670
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.845922882427308e-07,
+      "loss": 2.7828,
+      "step": 44671
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.845448798988623e-07,
+      "loss": 1.5972,
+      "step": 44672
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.844974715549937e-07,
+      "loss": 1.4236,
+      "step": 44673
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.844500632111252e-07,
+      "loss": 1.2956,
+      "step": 44674
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.844026548672565e-07,
+      "loss": 1.153,
+      "step": 44675
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.843552465233881e-07,
+      "loss": 2.7812,
+      "step": 44676
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.843078381795196e-07,
+      "loss": 2.8011,
+      "step": 44677
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.84260429835651e-07,
+      "loss": 0.8802,
+      "step": 44678
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 8.842130214917826e-07,
+      "loss": 1.8189,
+      "step": 44679
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.841656131479141e-07,
+      "loss": 1.6172,
+      "step": 44680
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.841182048040455e-07,
+      "loss": 2.4146,
+      "step": 44681
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.84070796460177e-07,
+      "loss": 2.3149,
+      "step": 44682
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.840233881163086e-07,
+      "loss": 2.6013,
+      "step": 44683
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.8397597977244e-07,
+      "loss": 0.2984,
+      "step": 44684
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.839285714285715e-07,
+      "loss": 2.299,
+      "step": 44685
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.838811630847029e-07,
+      "loss": 3.3301,
+      "step": 44686
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.838337547408343e-07,
+      "loss": 1.76,
+      "step": 44687
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.837863463969659e-07,
+      "loss": 1.567,
+      "step": 44688
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.837389380530973e-07,
+      "loss": 0.7799,
+      "step": 44689
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.836915297092288e-07,
+      "loss": 2.7064,
+      "step": 44690
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.836441213653604e-07,
+      "loss": 1.3245,
+      "step": 44691
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.835967130214918e-07,
+      "loss": 2.5005,
+      "step": 44692
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.835493046776233e-07,
+      "loss": 1.5913,
+      "step": 44693
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.835018963337548e-07,
+      "loss": 2.2676,
+      "step": 44694
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.834544879898862e-07,
+      "loss": 1.4369,
+      "step": 44695
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.834070796460178e-07,
+      "loss": 3.3189,
+      "step": 44696
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.833596713021492e-07,
+      "loss": 2.2088,
+      "step": 44697
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.833122629582807e-07,
+      "loss": 1.3333,
+      "step": 44698
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.832648546144123e-07,
+      "loss": 1.6955,
+      "step": 44699
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.832174462705435e-07,
+      "loss": 2.631,
+      "step": 44700
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.831700379266751e-07,
+      "loss": 3.1108,
+      "step": 44701
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.831226295828066e-07,
+      "loss": 2.2321,
+      "step": 44702
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.83075221238938e-07,
+      "loss": 1.7001,
+      "step": 44703
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.830278128950696e-07,
+      "loss": 2.1042,
+      "step": 44704
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.829804045512011e-07,
+      "loss": 1.7802,
+      "step": 44705
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.829329962073325e-07,
+      "loss": 1.9539,
+      "step": 44706
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.82885587863464e-07,
+      "loss": 2.4615,
+      "step": 44707
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.828381795195954e-07,
+      "loss": 0.6135,
+      "step": 44708
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.82790771175727e-07,
+      "loss": 1.8331,
+      "step": 44709
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.827433628318585e-07,
+      "loss": 1.2698,
+      "step": 44710
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.826959544879898e-07,
+      "loss": 2.2424,
+      "step": 44711
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.826485461441213e-07,
+      "loss": 1.4534,
+      "step": 44712
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.826011378002529e-07,
+      "loss": 3.304,
+      "step": 44713
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.825537294563843e-07,
+      "loss": 1.4016,
+      "step": 44714
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.825063211125158e-07,
+      "loss": 1.0408,
+      "step": 44715
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.824589127686474e-07,
+      "loss": 1.0387,
+      "step": 44716
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.824115044247788e-07,
+      "loss": 3.0161,
+      "step": 44717
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.823640960809103e-07,
+      "loss": 2.1484,
+      "step": 44718
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.823166877370417e-07,
+      "loss": 0.844,
+      "step": 44719
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.822692793931732e-07,
+      "loss": 1.1186,
+      "step": 44720
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.822218710493048e-07,
+      "loss": 0.7016,
+      "step": 44721
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.821744627054361e-07,
+      "loss": 0.6059,
+      "step": 44722
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.821270543615676e-07,
+      "loss": 1.098,
+      "step": 44723
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.820796460176992e-07,
+      "loss": 1.3541,
+      "step": 44724
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.820322376738305e-07,
+      "loss": 2.89,
+      "step": 44725
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.819848293299621e-07,
+      "loss": 1.6598,
+      "step": 44726
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.819374209860936e-07,
+      "loss": 2.8211,
+      "step": 44727
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.81890012642225e-07,
+      "loss": 1.9485,
+      "step": 44728
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.818426042983566e-07,
+      "loss": 2.2193,
+      "step": 44729
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.81795195954488e-07,
+      "loss": 2.9162,
+      "step": 44730
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.817477876106195e-07,
+      "loss": 2.1911,
+      "step": 44731
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.81700379266751e-07,
+      "loss": 1.0909,
+      "step": 44732
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.816529709228823e-07,
+      "loss": 2.3648,
+      "step": 44733
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.81605562579014e-07,
+      "loss": 0.2114,
+      "step": 44734
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.815581542351454e-07,
+      "loss": 2.0712,
+      "step": 44735
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.815107458912768e-07,
+      "loss": 2.2208,
+      "step": 44736
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.814633375474084e-07,
+      "loss": 3.3008,
+      "step": 44737
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.814159292035399e-07,
+      "loss": 2.9861,
+      "step": 44738
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.813685208596713e-07,
+      "loss": 1.3193,
+      "step": 44739
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.813211125158028e-07,
+      "loss": 0.8443,
+      "step": 44740
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.812737041719342e-07,
+      "loss": 0.4003,
+      "step": 44741
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 8.812262958280658e-07,
+      "loss": 1.112,
+      "step": 44742
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.811788874841973e-07,
+      "loss": 3.0129,
+      "step": 44743
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.811314791403287e-07,
+      "loss": 2.9101,
+      "step": 44744
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.810840707964601e-07,
+      "loss": 1.8186,
+      "step": 44745
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.810366624525917e-07,
+      "loss": 1.8084,
+      "step": 44746
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.809892541087231e-07,
+      "loss": 2.0971,
+      "step": 44747
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.809418457648546e-07,
+      "loss": 1.1257,
+      "step": 44748
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.808944374209862e-07,
+      "loss": 1.1883,
+      "step": 44749
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.808470290771176e-07,
+      "loss": 1.7863,
+      "step": 44750
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.807996207332491e-07,
+      "loss": 0.7845,
+      "step": 44751
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.807522123893805e-07,
+      "loss": 3.7935,
+      "step": 44752
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.80704804045512e-07,
+      "loss": 1.5119,
+      "step": 44753
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.806573957016436e-07,
+      "loss": 1.0814,
+      "step": 44754
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.80609987357775e-07,
+      "loss": 2.3682,
+      "step": 44755
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.805625790139065e-07,
+      "loss": 1.4867,
+      "step": 44756
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.805151706700381e-07,
+      "loss": 0.7884,
+      "step": 44757
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.804677623261693e-07,
+      "loss": 1.4074,
+      "step": 44758
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.804203539823009e-07,
+      "loss": 2.4345,
+      "step": 44759
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.803729456384324e-07,
+      "loss": 1.2092,
+      "step": 44760
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.803255372945638e-07,
+      "loss": 2.2724,
+      "step": 44761
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.802781289506954e-07,
+      "loss": 1.4654,
+      "step": 44762
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.802307206068268e-07,
+      "loss": 1.4819,
+      "step": 44763
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.801833122629583e-07,
+      "loss": 3.7766,
+      "step": 44764
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.801359039190898e-07,
+      "loss": 2.3226,
+      "step": 44765
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.800884955752212e-07,
+      "loss": 2.719,
+      "step": 44766
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.800410872313528e-07,
+      "loss": 2.1767,
+      "step": 44767
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.799936788874843e-07,
+      "loss": 1.9514,
+      "step": 44768
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.799462705436156e-07,
+      "loss": 1.2037,
+      "step": 44769
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.798988621997472e-07,
+      "loss": 3.3528,
+      "step": 44770
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.798514538558787e-07,
+      "loss": 0.9912,
+      "step": 44771
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.798040455120101e-07,
+      "loss": 3.042,
+      "step": 44772
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.797566371681416e-07,
+      "loss": 1.0261,
+      "step": 44773
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.79709228824273e-07,
+      "loss": 2.9999,
+      "step": 44774
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.796618204804046e-07,
+      "loss": 1.6752,
+      "step": 44775
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.796144121365361e-07,
+      "loss": 2.0787,
+      "step": 44776
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.795670037926675e-07,
+      "loss": 2.0226,
+      "step": 44777
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.79519595448799e-07,
+      "loss": 1.7252,
+      "step": 44778
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.794721871049306e-07,
+      "loss": 2.0926,
+      "step": 44779
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.79424778761062e-07,
+      "loss": 2.3833,
+      "step": 44780
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.793773704171934e-07,
+      "loss": 1.0182,
+      "step": 44781
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.79329962073325e-07,
+      "loss": 2.4224,
+      "step": 44782
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.792825537294563e-07,
+      "loss": 2.1326,
+      "step": 44783
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.792351453855879e-07,
+      "loss": 2.7513,
+      "step": 44784
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.791877370417193e-07,
+      "loss": 2.075,
+      "step": 44785
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.791403286978508e-07,
+      "loss": 2.374,
+      "step": 44786
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.790929203539824e-07,
+      "loss": 1.1365,
+      "step": 44787
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.790455120101138e-07,
+      "loss": 1.1481,
+      "step": 44788
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.789981036662453e-07,
+      "loss": 1.923,
+      "step": 44789
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.789506953223768e-07,
+      "loss": 1.3432,
+      "step": 44790
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.789032869785082e-07,
+      "loss": 3.4323,
+      "step": 44791
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.788558786346398e-07,
+      "loss": 1.3091,
+      "step": 44792
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.788084702907712e-07,
+      "loss": 3.2265,
+      "step": 44793
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.787610619469026e-07,
+      "loss": 2.3407,
+      "step": 44794
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.787136536030342e-07,
+      "loss": 1.626,
+      "step": 44795
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.786662452591655e-07,
+      "loss": 2.3146,
+      "step": 44796
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.786188369152971e-07,
+      "loss": 2.0325,
+      "step": 44797
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.785714285714286e-07,
+      "loss": 1.2205,
+      "step": 44798
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.7852402022756e-07,
+      "loss": 0.6898,
+      "step": 44799
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.784766118836916e-07,
+      "loss": 3.5878,
+      "step": 44800
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.784292035398231e-07,
+      "loss": 0.2839,
+      "step": 44801
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.783817951959545e-07,
+      "loss": 1.7887,
+      "step": 44802
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.78334386852086e-07,
+      "loss": 1.1462,
+      "step": 44803
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.782869785082176e-07,
+      "loss": 1.5449,
+      "step": 44804
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 8.782395701643489e-07,
+      "loss": 3.2863,
+      "step": 44805
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.781921618204804e-07,
+      "loss": 1.312,
+      "step": 44806
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.781447534766118e-07,
+      "loss": 2.0522,
+      "step": 44807
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.780973451327434e-07,
+      "loss": 2.0664,
+      "step": 44808
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.780499367888749e-07,
+      "loss": 2.4171,
+      "step": 44809
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.780025284450063e-07,
+      "loss": 2.4193,
+      "step": 44810
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.779551201011378e-07,
+      "loss": 1.5142,
+      "step": 44811
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.779077117572694e-07,
+      "loss": 1.7683,
+      "step": 44812
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.778603034134008e-07,
+      "loss": 2.3042,
+      "step": 44813
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.778128950695323e-07,
+      "loss": 3.3409,
+      "step": 44814
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.777654867256639e-07,
+      "loss": 0.8845,
+      "step": 44815
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.777180783817951e-07,
+      "loss": 3.0189,
+      "step": 44816
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.776706700379267e-07,
+      "loss": 1.9809,
+      "step": 44817
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.776232616940581e-07,
+      "loss": 1.4785,
+      "step": 44818
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.775758533501896e-07,
+      "loss": 0.9134,
+      "step": 44819
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.775284450063212e-07,
+      "loss": 1.8472,
+      "step": 44820
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.774810366624526e-07,
+      "loss": 0.7326,
+      "step": 44821
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.774336283185841e-07,
+      "loss": 2.5201,
+      "step": 44822
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.773862199747156e-07,
+      "loss": 2.9359,
+      "step": 44823
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.77338811630847e-07,
+      "loss": 2.3268,
+      "step": 44824
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.772914032869786e-07,
+      "loss": 1.6185,
+      "step": 44825
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.772439949431101e-07,
+      "loss": 1.4626,
+      "step": 44826
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.771965865992414e-07,
+      "loss": 1.3323,
+      "step": 44827
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.77149178255373e-07,
+      "loss": 1.612,
+      "step": 44828
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.771017699115043e-07,
+      "loss": 2.2561,
+      "step": 44829
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.770543615676359e-07,
+      "loss": 1.4407,
+      "step": 44830
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.770069532237674e-07,
+      "loss": 1.9304,
+      "step": 44831
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.769595448798988e-07,
+      "loss": 0.7431,
+      "step": 44832
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.769121365360304e-07,
+      "loss": 1.47,
+      "step": 44833
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.768647281921619e-07,
+      "loss": 1.5602,
+      "step": 44834
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.768173198482933e-07,
+      "loss": 2.3201,
+      "step": 44835
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.767699115044248e-07,
+      "loss": 1.305,
+      "step": 44836
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.767225031605564e-07,
+      "loss": 3.2327,
+      "step": 44837
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.766750948166878e-07,
+      "loss": 0.3742,
+      "step": 44838
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.766276864728192e-07,
+      "loss": 2.4998,
+      "step": 44839
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.765802781289506e-07,
+      "loss": 2.7247,
+      "step": 44840
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.765328697850821e-07,
+      "loss": 1.7377,
+      "step": 44841
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.764854614412137e-07,
+      "loss": 2.4951,
+      "step": 44842
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.764380530973451e-07,
+      "loss": 0.5985,
+      "step": 44843
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.763906447534766e-07,
+      "loss": 2.2863,
+      "step": 44844
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.763432364096082e-07,
+      "loss": 0.7414,
+      "step": 44845
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.762958280657396e-07,
+      "loss": 1.4604,
+      "step": 44846
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.762484197218711e-07,
+      "loss": 1.7344,
+      "step": 44847
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.762010113780026e-07,
+      "loss": 1.6102,
+      "step": 44848
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.76153603034134e-07,
+      "loss": 2.0117,
+      "step": 44849
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.761061946902656e-07,
+      "loss": 2.4331,
+      "step": 44850
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.760587863463969e-07,
+      "loss": 2.9514,
+      "step": 44851
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.760113780025284e-07,
+      "loss": 2.3997,
+      "step": 44852
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.7596396965866e-07,
+      "loss": 1.7521,
+      "step": 44853
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.759165613147913e-07,
+      "loss": 3.2074,
+      "step": 44854
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.758691529709229e-07,
+      "loss": 2.6879,
+      "step": 44855
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.758217446270544e-07,
+      "loss": 3.501,
+      "step": 44856
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.757743362831858e-07,
+      "loss": 2.3963,
+      "step": 44857
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.757269279393174e-07,
+      "loss": 2.3309,
+      "step": 44858
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.756795195954489e-07,
+      "loss": 3.6397,
+      "step": 44859
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.756321112515803e-07,
+      "loss": 2.8588,
+      "step": 44860
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.755847029077118e-07,
+      "loss": 0.7671,
+      "step": 44861
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.755372945638431e-07,
+      "loss": 1.635,
+      "step": 44862
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.754898862199747e-07,
+      "loss": 2.9473,
+      "step": 44863
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.754424778761062e-07,
+      "loss": 2.067,
+      "step": 44864
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.753950695322376e-07,
+      "loss": 2.2878,
+      "step": 44865
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.753476611883692e-07,
+      "loss": 2.8436,
+      "step": 44866
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.753002528445007e-07,
+      "loss": 1.6971,
+      "step": 44867
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.752528445006321e-07,
+      "loss": 1.8401,
+      "step": 44868
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 8.752054361567636e-07,
+      "loss": 1.4836,
+      "step": 44869
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.751580278128952e-07,
+      "loss": 1.6061,
+      "step": 44870
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.751106194690266e-07,
+      "loss": 1.5077,
+      "step": 44871
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.750632111251581e-07,
+      "loss": 0.9027,
+      "step": 44872
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.750158027812895e-07,
+      "loss": 1.4446,
+      "step": 44873
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.749683944374209e-07,
+      "loss": 2.5675,
+      "step": 44874
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.749209860935525e-07,
+      "loss": 1.437,
+      "step": 44875
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.748735777496839e-07,
+      "loss": 1.2253,
+      "step": 44876
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.748261694058154e-07,
+      "loss": 0.4862,
+      "step": 44877
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.74778761061947e-07,
+      "loss": 0.9465,
+      "step": 44878
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.747313527180784e-07,
+      "loss": 2.5745,
+      "step": 44879
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.746839443742099e-07,
+      "loss": 2.0773,
+      "step": 44880
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.746365360303414e-07,
+      "loss": 2.0765,
+      "step": 44881
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.745891276864728e-07,
+      "loss": 2.0481,
+      "step": 44882
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.745417193426044e-07,
+      "loss": 1.5596,
+      "step": 44883
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.744943109987358e-07,
+      "loss": 1.1517,
+      "step": 44884
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.744469026548673e-07,
+      "loss": 1.9673,
+      "step": 44885
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.743994943109989e-07,
+      "loss": 1.1644,
+      "step": 44886
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.743520859671301e-07,
+      "loss": 1.7088,
+      "step": 44887
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.743046776232617e-07,
+      "loss": 0.1816,
+      "step": 44888
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.742572692793932e-07,
+      "loss": 2.1743,
+      "step": 44889
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.742098609355246e-07,
+      "loss": 1.2609,
+      "step": 44890
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.741624525916562e-07,
+      "loss": 3.189,
+      "step": 44891
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.741150442477877e-07,
+      "loss": 2.561,
+      "step": 44892
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.740676359039191e-07,
+      "loss": 1.925,
+      "step": 44893
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.740202275600506e-07,
+      "loss": 2.0511,
+      "step": 44894
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.73972819216182e-07,
+      "loss": 3.6146,
+      "step": 44895
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.739254108723136e-07,
+      "loss": 2.7097,
+      "step": 44896
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.738780025284451e-07,
+      "loss": 0.9391,
+      "step": 44897
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.738305941845764e-07,
+      "loss": 1.8481,
+      "step": 44898
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.73783185840708e-07,
+      "loss": 2.585,
+      "step": 44899
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.737357774968395e-07,
+      "loss": 3.9282,
+      "step": 44900
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.736883691529709e-07,
+      "loss": 2.6247,
+      "step": 44901
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.736409608091024e-07,
+      "loss": 1.981,
+      "step": 44902
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.73593552465234e-07,
+      "loss": 2.6547,
+      "step": 44903
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.735461441213654e-07,
+      "loss": 2.0126,
+      "step": 44904
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.734987357774969e-07,
+      "loss": 3.0311,
+      "step": 44905
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.734513274336283e-07,
+      "loss": 1.3003,
+      "step": 44906
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.734039190897598e-07,
+      "loss": 2.3162,
+      "step": 44907
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.733565107458914e-07,
+      "loss": 1.1529,
+      "step": 44908
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.733091024020227e-07,
+      "loss": 1.4086,
+      "step": 44909
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.732616940581542e-07,
+      "loss": 1.7754,
+      "step": 44910
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.732142857142858e-07,
+      "loss": 1.7328,
+      "step": 44911
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.731668773704171e-07,
+      "loss": 0.8565,
+      "step": 44912
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.731194690265487e-07,
+      "loss": 2.2793,
+      "step": 44913
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.730720606826802e-07,
+      "loss": 1.8901,
+      "step": 44914
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.730246523388116e-07,
+      "loss": 1.269,
+      "step": 44915
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.729772439949432e-07,
+      "loss": 1.9244,
+      "step": 44916
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.729298356510746e-07,
+      "loss": 1.8943,
+      "step": 44917
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.728824273072061e-07,
+      "loss": 2.9762,
+      "step": 44918
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.728350189633376e-07,
+      "loss": 2.3015,
+      "step": 44919
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.72787610619469e-07,
+      "loss": 1.4063,
+      "step": 44920
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.727402022756005e-07,
+      "loss": 1.7405,
+      "step": 44921
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.72692793931732e-07,
+      "loss": 0.8387,
+      "step": 44922
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.726453855878634e-07,
+      "loss": 2.3963,
+      "step": 44923
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.72597977243995e-07,
+      "loss": 2.4822,
+      "step": 44924
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.725505689001265e-07,
+      "loss": 3.0266,
+      "step": 44925
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.725031605562579e-07,
+      "loss": 1.9258,
+      "step": 44926
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.724557522123894e-07,
+      "loss": 1.7306,
+      "step": 44927
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.724083438685208e-07,
+      "loss": 2.7986,
+      "step": 44928
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.723609355246524e-07,
+      "loss": 1.004,
+      "step": 44929
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.723135271807839e-07,
+      "loss": 2.5364,
+      "step": 44930
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.722661188369153e-07,
+      "loss": 2.8739,
+      "step": 44931
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 8.722187104930468e-07,
+      "loss": 2.0386,
+      "step": 44932
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.721713021491783e-07,
+      "loss": 2.063,
+      "step": 44933
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.721238938053097e-07,
+      "loss": 2.0336,
+      "step": 44934
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.720764854614412e-07,
+      "loss": 2.1556,
+      "step": 44935
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.720290771175728e-07,
+      "loss": 1.819,
+      "step": 44936
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.719816687737042e-07,
+      "loss": 1.7422,
+      "step": 44937
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.719342604298357e-07,
+      "loss": 1.4175,
+      "step": 44938
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.718868520859671e-07,
+      "loss": 1.9273,
+      "step": 44939
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.718394437420986e-07,
+      "loss": 2.3418,
+      "step": 44940
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.717920353982302e-07,
+      "loss": 0.5938,
+      "step": 44941
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.717446270543616e-07,
+      "loss": 0.9489,
+      "step": 44942
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.716972187104931e-07,
+      "loss": 2.603,
+      "step": 44943
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.716498103666247e-07,
+      "loss": 2.8298,
+      "step": 44944
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.716024020227559e-07,
+      "loss": 0.3907,
+      "step": 44945
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.715549936788875e-07,
+      "loss": 1.2351,
+      "step": 44946
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.71507585335019e-07,
+      "loss": 2.3789,
+      "step": 44947
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.714601769911504e-07,
+      "loss": 2.9288,
+      "step": 44948
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.71412768647282e-07,
+      "loss": 1.4835,
+      "step": 44949
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.713653603034134e-07,
+      "loss": 2.1346,
+      "step": 44950
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.713179519595449e-07,
+      "loss": 3.0449,
+      "step": 44951
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.712705436156764e-07,
+      "loss": 2.263,
+      "step": 44952
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.712231352718078e-07,
+      "loss": 3.0302,
+      "step": 44953
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.711757269279394e-07,
+      "loss": 1.3232,
+      "step": 44954
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.711283185840709e-07,
+      "loss": 1.1007,
+      "step": 44955
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.710809102402022e-07,
+      "loss": 2.0967,
+      "step": 44956
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.710335018963338e-07,
+      "loss": 1.7711,
+      "step": 44957
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.709860935524653e-07,
+      "loss": 2.5195,
+      "step": 44958
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.709386852085967e-07,
+      "loss": 3.0661,
+      "step": 44959
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.708912768647282e-07,
+      "loss": 2.4095,
+      "step": 44960
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.708438685208596e-07,
+      "loss": 1.1759,
+      "step": 44961
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.707964601769912e-07,
+      "loss": 0.5477,
+      "step": 44962
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.707490518331227e-07,
+      "loss": 3.2643,
+      "step": 44963
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.707016434892541e-07,
+      "loss": 1.7593,
+      "step": 44964
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.706542351453856e-07,
+      "loss": 1.9076,
+      "step": 44965
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.706068268015172e-07,
+      "loss": 0.8476,
+      "step": 44966
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.705594184576486e-07,
+      "loss": 2.0624,
+      "step": 44967
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.7051201011378e-07,
+      "loss": 1.9541,
+      "step": 44968
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.704646017699116e-07,
+      "loss": 1.2924,
+      "step": 44969
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.70417193426043e-07,
+      "loss": 0.9962,
+      "step": 44970
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.703697850821745e-07,
+      "loss": 1.6672,
+      "step": 44971
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.703223767383059e-07,
+      "loss": 2.9554,
+      "step": 44972
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.702749683944374e-07,
+      "loss": 1.3527,
+      "step": 44973
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.70227560050569e-07,
+      "loss": 0.8522,
+      "step": 44974
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.701801517067004e-07,
+      "loss": 1.9137,
+      "step": 44975
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.701327433628319e-07,
+      "loss": 1.2241,
+      "step": 44976
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.700853350189635e-07,
+      "loss": 2.0818,
+      "step": 44977
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.700379266750948e-07,
+      "loss": 1.104,
+      "step": 44978
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.699905183312264e-07,
+      "loss": 2.287,
+      "step": 44979
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.699431099873578e-07,
+      "loss": 2.28,
+      "step": 44980
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.698957016434892e-07,
+      "loss": 1.4661,
+      "step": 44981
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.698482932996208e-07,
+      "loss": 1.542,
+      "step": 44982
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.698008849557521e-07,
+      "loss": 2.4695,
+      "step": 44983
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.697534766118837e-07,
+      "loss": 1.9353,
+      "step": 44984
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.697060682680152e-07,
+      "loss": 2.7507,
+      "step": 44985
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.696586599241466e-07,
+      "loss": 2.1777,
+      "step": 44986
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.696112515802782e-07,
+      "loss": 2.171,
+      "step": 44987
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.695638432364097e-07,
+      "loss": 2.4393,
+      "step": 44988
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.695164348925411e-07,
+      "loss": 1.6052,
+      "step": 44989
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.694690265486726e-07,
+      "loss": 0.9342,
+      "step": 44990
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.694216182048042e-07,
+      "loss": 1.4453,
+      "step": 44991
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.693742098609355e-07,
+      "loss": 0.4539,
+      "step": 44992
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.69326801517067e-07,
+      "loss": 1.7071,
+      "step": 44993
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.692793931731984e-07,
+      "loss": 1.8245,
+      "step": 44994
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 8.6923198482933e-07,
+      "loss": 2.9214,
+      "step": 44995
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.691845764854615e-07,
+      "loss": 2.2622,
+      "step": 44996
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.691371681415929e-07,
+      "loss": 1.4327,
+      "step": 44997
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.690897597977244e-07,
+      "loss": 3.7594,
+      "step": 44998
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.69042351453856e-07,
+      "loss": 1.532,
+      "step": 44999
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.689949431099874e-07,
+      "loss": 2.839,
+      "step": 45000
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.689475347661189e-07,
+      "loss": 1.6532,
+      "step": 45001
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.689001264222505e-07,
+      "loss": 3.1676,
+      "step": 45002
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.688527180783817e-07,
+      "loss": 1.255,
+      "step": 45003
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.688053097345133e-07,
+      "loss": 2.5122,
+      "step": 45004
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.687579013906447e-07,
+      "loss": 2.1088,
+      "step": 45005
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.687104930467762e-07,
+      "loss": 1.8638,
+      "step": 45006
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.686630847029078e-07,
+      "loss": 2.2523,
+      "step": 45007
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.686156763590392e-07,
+      "loss": 1.8197,
+      "step": 45008
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.685682680151707e-07,
+      "loss": 1.3686,
+      "step": 45009
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.685208596713022e-07,
+      "loss": 1.8656,
+      "step": 45010
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.684734513274336e-07,
+      "loss": 2.6471,
+      "step": 45011
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.684260429835652e-07,
+      "loss": 3.7607,
+      "step": 45012
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.683786346396967e-07,
+      "loss": 1.8849,
+      "step": 45013
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.68331226295828e-07,
+      "loss": 1.676,
+      "step": 45014
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.682838179519596e-07,
+      "loss": 1.4087,
+      "step": 45015
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.682364096080909e-07,
+      "loss": 2.0206,
+      "step": 45016
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.681890012642225e-07,
+      "loss": 2.4087,
+      "step": 45017
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.68141592920354e-07,
+      "loss": 2.1506,
+      "step": 45018
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.680941845764854e-07,
+      "loss": 2.6724,
+      "step": 45019
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.68046776232617e-07,
+      "loss": 1.1608,
+      "step": 45020
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.679993678887485e-07,
+      "loss": 2.607,
+      "step": 45021
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.679519595448799e-07,
+      "loss": 2.7241,
+      "step": 45022
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.679045512010114e-07,
+      "loss": 2.1145,
+      "step": 45023
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.67857142857143e-07,
+      "loss": 1.5069,
+      "step": 45024
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.678097345132744e-07,
+      "loss": 1.0121,
+      "step": 45025
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.677623261694059e-07,
+      "loss": 1.3169,
+      "step": 45026
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.677149178255372e-07,
+      "loss": 2.0334,
+      "step": 45027
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.676675094816688e-07,
+      "loss": 1.0382,
+      "step": 45028
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.676201011378003e-07,
+      "loss": 1.9491,
+      "step": 45029
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.675726927939317e-07,
+      "loss": 2.582,
+      "step": 45030
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.675252844500632e-07,
+      "loss": 1.9877,
+      "step": 45031
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.674778761061948e-07,
+      "loss": 1.8085,
+      "step": 45032
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.674304677623262e-07,
+      "loss": 1.5807,
+      "step": 45033
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.673830594184577e-07,
+      "loss": 2.6606,
+      "step": 45034
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.673356510745893e-07,
+      "loss": 2.8279,
+      "step": 45035
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.672882427307206e-07,
+      "loss": 1.1959,
+      "step": 45036
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.672408343868522e-07,
+      "loss": 1.625,
+      "step": 45037
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.671934260429835e-07,
+      "loss": 1.6436,
+      "step": 45038
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.67146017699115e-07,
+      "loss": 2.7365,
+      "step": 45039
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.670986093552466e-07,
+      "loss": 1.7494,
+      "step": 45040
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.67051201011378e-07,
+      "loss": 2.7207,
+      "step": 45041
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.670037926675095e-07,
+      "loss": 2.0658,
+      "step": 45042
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.66956384323641e-07,
+      "loss": 2.4637,
+      "step": 45043
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.669089759797724e-07,
+      "loss": 1.986,
+      "step": 45044
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.66861567635904e-07,
+      "loss": 2.4257,
+      "step": 45045
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.668141592920355e-07,
+      "loss": 1.9785,
+      "step": 45046
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.667667509481669e-07,
+      "loss": 2.6748,
+      "step": 45047
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.667193426042985e-07,
+      "loss": 2.1248,
+      "step": 45048
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.666719342604297e-07,
+      "loss": 1.2242,
+      "step": 45049
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.666245259165613e-07,
+      "loss": 2.8735,
+      "step": 45050
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.665771175726928e-07,
+      "loss": 2.2576,
+      "step": 45051
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.665297092288242e-07,
+      "loss": 2.3496,
+      "step": 45052
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.664823008849558e-07,
+      "loss": 2.1582,
+      "step": 45053
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.664348925410873e-07,
+      "loss": 1.0164,
+      "step": 45054
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.663874841972187e-07,
+      "loss": 2.7009,
+      "step": 45055
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.663400758533502e-07,
+      "loss": 1.5819,
+      "step": 45056
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.662926675094818e-07,
+      "loss": 3.3842,
+      "step": 45057
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.662452591656132e-07,
+      "loss": 3.9544,
+      "step": 45058
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 8.661978508217447e-07,
+      "loss": 3.0965,
+      "step": 45059
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.661504424778761e-07,
+      "loss": 1.7527,
+      "step": 45060
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.661030341340075e-07,
+      "loss": 3.0966,
+      "step": 45061
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.660556257901391e-07,
+      "loss": 0.6874,
+      "step": 45062
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.660082174462705e-07,
+      "loss": 0.6716,
+      "step": 45063
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.65960809102402e-07,
+      "loss": 2.1016,
+      "step": 45064
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.659134007585336e-07,
+      "loss": 0.4372,
+      "step": 45065
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.65865992414665e-07,
+      "loss": 3.4717,
+      "step": 45066
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.658185840707965e-07,
+      "loss": 2.4934,
+      "step": 45067
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.65771175726928e-07,
+      "loss": 1.9127,
+      "step": 45068
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.657237673830594e-07,
+      "loss": 2.3784,
+      "step": 45069
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.65676359039191e-07,
+      "loss": 2.265,
+      "step": 45070
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.656289506953224e-07,
+      "loss": 3.5611,
+      "step": 45071
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.655815423514539e-07,
+      "loss": 0.9812,
+      "step": 45072
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.655341340075855e-07,
+      "loss": 1.9208,
+      "step": 45073
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.654867256637167e-07,
+      "loss": 1.7957,
+      "step": 45074
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.654393173198483e-07,
+      "loss": 2.8321,
+      "step": 45075
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.653919089759798e-07,
+      "loss": 2.79,
+      "step": 45076
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.653445006321112e-07,
+      "loss": 1.976,
+      "step": 45077
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.652970922882428e-07,
+      "loss": 1.8187,
+      "step": 45078
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.652496839443743e-07,
+      "loss": 2.922,
+      "step": 45079
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.652022756005057e-07,
+      "loss": 2.3872,
+      "step": 45080
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.651548672566372e-07,
+      "loss": 1.0855,
+      "step": 45081
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.651074589127686e-07,
+      "loss": 0.97,
+      "step": 45082
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.650600505689002e-07,
+      "loss": 1.0862,
+      "step": 45083
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.650126422250317e-07,
+      "loss": 2.4015,
+      "step": 45084
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.64965233881163e-07,
+      "loss": 3.231,
+      "step": 45085
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.649178255372946e-07,
+      "loss": 2.5411,
+      "step": 45086
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.648704171934261e-07,
+      "loss": 1.8061,
+      "step": 45087
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.648230088495575e-07,
+      "loss": 2.356,
+      "step": 45088
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.64775600505689e-07,
+      "loss": 3.3893,
+      "step": 45089
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.647281921618206e-07,
+      "loss": 2.2297,
+      "step": 45090
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.64680783817952e-07,
+      "loss": 0.2442,
+      "step": 45091
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.646333754740835e-07,
+      "loss": 2.7718,
+      "step": 45092
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.645859671302149e-07,
+      "loss": 1.996,
+      "step": 45093
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.645385587863464e-07,
+      "loss": 1.8027,
+      "step": 45094
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.64491150442478e-07,
+      "loss": 1.1127,
+      "step": 45095
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.644437420986094e-07,
+      "loss": 2.3856,
+      "step": 45096
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.643963337547408e-07,
+      "loss": 1.2263,
+      "step": 45097
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.643489254108724e-07,
+      "loss": 1.1465,
+      "step": 45098
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.643015170670038e-07,
+      "loss": 1.8892,
+      "step": 45099
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.642541087231353e-07,
+      "loss": 1.8279,
+      "step": 45100
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.642067003792667e-07,
+      "loss": 2.4917,
+      "step": 45101
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.641592920353982e-07,
+      "loss": 3.1688,
+      "step": 45102
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.641118836915298e-07,
+      "loss": 2.9835,
+      "step": 45103
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.640644753476612e-07,
+      "loss": 0.6675,
+      "step": 45104
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.640170670037927e-07,
+      "loss": 1.9106,
+      "step": 45105
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.639696586599243e-07,
+      "loss": 1.2088,
+      "step": 45106
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.639222503160556e-07,
+      "loss": 0.9067,
+      "step": 45107
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.638748419721872e-07,
+      "loss": 2.3186,
+      "step": 45108
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.638274336283186e-07,
+      "loss": 2.6976,
+      "step": 45109
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.6378002528445e-07,
+      "loss": 2.064,
+      "step": 45110
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.637326169405816e-07,
+      "loss": 2.0311,
+      "step": 45111
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.636852085967129e-07,
+      "loss": 1.8443,
+      "step": 45112
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.636378002528445e-07,
+      "loss": 1.353,
+      "step": 45113
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.63590391908976e-07,
+      "loss": 1.6567,
+      "step": 45114
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.635429835651074e-07,
+      "loss": 0.8315,
+      "step": 45115
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.63495575221239e-07,
+      "loss": 1.2592,
+      "step": 45116
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.634481668773705e-07,
+      "loss": 1.2966,
+      "step": 45117
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.634007585335019e-07,
+      "loss": 3.4094,
+      "step": 45118
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.633533501896334e-07,
+      "loss": 1.5058,
+      "step": 45119
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.63305941845765e-07,
+      "loss": 2.2546,
+      "step": 45120
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.632585335018963e-07,
+      "loss": 2.5493,
+      "step": 45121
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 8.632111251580278e-07,
+      "loss": 1.0438,
+      "step": 45122
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.631637168141592e-07,
+      "loss": 1.7924,
+      "step": 45123
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.631163084702908e-07,
+      "loss": 2.6737,
+      "step": 45124
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.630689001264223e-07,
+      "loss": 3.5455,
+      "step": 45125
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.630214917825537e-07,
+      "loss": 1.6205,
+      "step": 45126
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.629740834386852e-07,
+      "loss": 2.1094,
+      "step": 45127
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.629266750948168e-07,
+      "loss": 1.3397,
+      "step": 45128
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.628792667509482e-07,
+      "loss": 1.5948,
+      "step": 45129
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.628318584070797e-07,
+      "loss": 1.5199,
+      "step": 45130
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.627844500632113e-07,
+      "loss": 1.7111,
+      "step": 45131
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.627370417193425e-07,
+      "loss": 2.0972,
+      "step": 45132
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.626896333754741e-07,
+      "loss": 1.3713,
+      "step": 45133
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.626422250316055e-07,
+      "loss": 2.2313,
+      "step": 45134
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.62594816687737e-07,
+      "loss": 1.4092,
+      "step": 45135
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.625474083438686e-07,
+      "loss": 1.3858,
+      "step": 45136
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.625e-07,
+      "loss": 2.7077,
+      "step": 45137
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.624525916561315e-07,
+      "loss": 1.5558,
+      "step": 45138
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.62405183312263e-07,
+      "loss": 1.9665,
+      "step": 45139
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.623577749683944e-07,
+      "loss": 2.6751,
+      "step": 45140
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.62310366624526e-07,
+      "loss": 1.3017,
+      "step": 45141
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.622629582806575e-07,
+      "loss": 2.338,
+      "step": 45142
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.622155499367888e-07,
+      "loss": 2.8056,
+      "step": 45143
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.621681415929204e-07,
+      "loss": 1.2947,
+      "step": 45144
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.621207332490517e-07,
+      "loss": 2.7237,
+      "step": 45145
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.620733249051833e-07,
+      "loss": 2.0446,
+      "step": 45146
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.620259165613148e-07,
+      "loss": 2.9528,
+      "step": 45147
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.619785082174462e-07,
+      "loss": 1.0158,
+      "step": 45148
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.619310998735778e-07,
+      "loss": 1.3792,
+      "step": 45149
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.618836915297093e-07,
+      "loss": 2.5739,
+      "step": 45150
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.618362831858407e-07,
+      "loss": 1.4196,
+      "step": 45151
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.617888748419722e-07,
+      "loss": 1.9591,
+      "step": 45152
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.617414664981038e-07,
+      "loss": 2.1071,
+      "step": 45153
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.616940581542352e-07,
+      "loss": 2.3039,
+      "step": 45154
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.616466498103666e-07,
+      "loss": 2.014,
+      "step": 45155
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.61599241466498e-07,
+      "loss": 2.0031,
+      "step": 45156
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.615518331226296e-07,
+      "loss": 1.9838,
+      "step": 45157
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.615044247787611e-07,
+      "loss": 0.9814,
+      "step": 45158
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.614570164348925e-07,
+      "loss": 2.9606,
+      "step": 45159
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.61409608091024e-07,
+      "loss": 1.3558,
+      "step": 45160
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.613621997471556e-07,
+      "loss": 2.1554,
+      "step": 45161
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.61314791403287e-07,
+      "loss": 1.8276,
+      "step": 45162
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.612673830594185e-07,
+      "loss": 2.0897,
+      "step": 45163
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.612199747155501e-07,
+      "loss": 3.4027,
+      "step": 45164
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.611725663716814e-07,
+      "loss": 3.1546,
+      "step": 45165
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.61125158027813e-07,
+      "loss": 1.2059,
+      "step": 45166
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.610777496839443e-07,
+      "loss": 2.0723,
+      "step": 45167
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.610303413400758e-07,
+      "loss": 2.0977,
+      "step": 45168
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.609829329962074e-07,
+      "loss": 0.8659,
+      "step": 45169
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.609355246523388e-07,
+      "loss": 1.5828,
+      "step": 45170
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.608881163084703e-07,
+      "loss": 1.0364,
+      "step": 45171
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.608407079646018e-07,
+      "loss": 2.1198,
+      "step": 45172
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.607932996207332e-07,
+      "loss": 3.4649,
+      "step": 45173
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.607458912768648e-07,
+      "loss": 1.7051,
+      "step": 45174
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.606984829329963e-07,
+      "loss": 1.8737,
+      "step": 45175
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.606510745891277e-07,
+      "loss": 1.3619,
+      "step": 45176
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.606036662452593e-07,
+      "loss": 1.0861,
+      "step": 45177
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.605562579013905e-07,
+      "loss": 2.1855,
+      "step": 45178
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.605088495575221e-07,
+      "loss": 2.6462,
+      "step": 45179
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.604614412136536e-07,
+      "loss": 2.88,
+      "step": 45180
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.60414032869785e-07,
+      "loss": 1.4569,
+      "step": 45181
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.603666245259166e-07,
+      "loss": 0.8654,
+      "step": 45182
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.603192161820481e-07,
+      "loss": 1.9999,
+      "step": 45183
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.602718078381795e-07,
+      "loss": 1.3027,
+      "step": 45184
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 8.60224399494311e-07,
+      "loss": 2.3651,
+      "step": 45185
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.601769911504426e-07,
+      "loss": 2.3821,
+      "step": 45186
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.60129582806574e-07,
+      "loss": 2.812,
+      "step": 45187
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.600821744627055e-07,
+      "loss": 2.5107,
+      "step": 45188
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.600347661188369e-07,
+      "loss": 3.5259,
+      "step": 45189
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.599873577749683e-07,
+      "loss": 2.8722,
+      "step": 45190
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.599399494310999e-07,
+      "loss": 3.8766,
+      "step": 45191
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.598925410872313e-07,
+      "loss": 2.6723,
+      "step": 45192
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.598451327433628e-07,
+      "loss": 1.2451,
+      "step": 45193
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.597977243994944e-07,
+      "loss": 1.2732,
+      "step": 45194
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.597503160556258e-07,
+      "loss": 2.4872,
+      "step": 45195
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.597029077117573e-07,
+      "loss": 2.4218,
+      "step": 45196
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.596554993678888e-07,
+      "loss": 1.3542,
+      "step": 45197
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.596080910240202e-07,
+      "loss": 1.9169,
+      "step": 45198
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.595606826801518e-07,
+      "loss": 1.4448,
+      "step": 45199
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.595132743362832e-07,
+      "loss": 3.5989,
+      "step": 45200
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.594658659924147e-07,
+      "loss": 1.6858,
+      "step": 45201
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.594184576485463e-07,
+      "loss": 3.7765,
+      "step": 45202
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.593710493046775e-07,
+      "loss": 1.0762,
+      "step": 45203
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.593236409608091e-07,
+      "loss": 1.5704,
+      "step": 45204
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.592762326169406e-07,
+      "loss": 1.2745,
+      "step": 45205
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.59228824273072e-07,
+      "loss": 2.8937,
+      "step": 45206
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.591814159292036e-07,
+      "loss": 2.361,
+      "step": 45207
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.591340075853351e-07,
+      "loss": 1.1088,
+      "step": 45208
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.590865992414665e-07,
+      "loss": 1.8067,
+      "step": 45209
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.59039190897598e-07,
+      "loss": 1.2584,
+      "step": 45210
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.589917825537294e-07,
+      "loss": 3.5061,
+      "step": 45211
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.58944374209861e-07,
+      "loss": 1.2052,
+      "step": 45212
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.588969658659925e-07,
+      "loss": 2.1822,
+      "step": 45213
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.588495575221238e-07,
+      "loss": 2.4413,
+      "step": 45214
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.588021491782554e-07,
+      "loss": 1.3261,
+      "step": 45215
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.587547408343869e-07,
+      "loss": 1.518,
+      "step": 45216
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.587073324905183e-07,
+      "loss": 1.6066,
+      "step": 45217
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.586599241466498e-07,
+      "loss": 2.0088,
+      "step": 45218
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.586125158027814e-07,
+      "loss": 2.914,
+      "step": 45219
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.585651074589128e-07,
+      "loss": 2.8888,
+      "step": 45220
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.585176991150443e-07,
+      "loss": 3.119,
+      "step": 45221
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.584702907711757e-07,
+      "loss": 2.4926,
+      "step": 45222
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.584228824273072e-07,
+      "loss": 1.7888,
+      "step": 45223
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.583754740834388e-07,
+      "loss": 1.1301,
+      "step": 45224
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.583280657395701e-07,
+      "loss": 1.5024,
+      "step": 45225
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.582806573957016e-07,
+      "loss": 0.9604,
+      "step": 45226
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.582332490518332e-07,
+      "loss": 2.2316,
+      "step": 45227
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.581858407079646e-07,
+      "loss": 1.7589,
+      "step": 45228
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.581384323640961e-07,
+      "loss": 2.704,
+      "step": 45229
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.580910240202276e-07,
+      "loss": 1.0173,
+      "step": 45230
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.58043615676359e-07,
+      "loss": 3.1494,
+      "step": 45231
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.579962073324906e-07,
+      "loss": 1.9601,
+      "step": 45232
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.57948798988622e-07,
+      "loss": 1.6921,
+      "step": 45233
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.579013906447535e-07,
+      "loss": 1.9346,
+      "step": 45234
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.578539823008851e-07,
+      "loss": 2.2383,
+      "step": 45235
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.578065739570163e-07,
+      "loss": 2.0135,
+      "step": 45236
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.57759165613148e-07,
+      "loss": 3.6564,
+      "step": 45237
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.577117572692794e-07,
+      "loss": 1.8939,
+      "step": 45238
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.576643489254108e-07,
+      "loss": 1.7525,
+      "step": 45239
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.576169405815424e-07,
+      "loss": 1.0771,
+      "step": 45240
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.575695322376739e-07,
+      "loss": 0.8066,
+      "step": 45241
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.575221238938053e-07,
+      "loss": 2.2807,
+      "step": 45242
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.574747155499368e-07,
+      "loss": 2.3123,
+      "step": 45243
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.574273072060682e-07,
+      "loss": 2.2689,
+      "step": 45244
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.573798988621998e-07,
+      "loss": 2.3205,
+      "step": 45245
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.573324905183313e-07,
+      "loss": 2.2289,
+      "step": 45246
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.572850821744627e-07,
+      "loss": 2.5975,
+      "step": 45247
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.572376738305943e-07,
+      "loss": 1.4114,
+      "step": 45248
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 8.571902654867257e-07,
+      "loss": 1.9034,
+      "step": 45249
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.571428571428571e-07,
+      "loss": 2.3101,
+      "step": 45250
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.570954487989886e-07,
+      "loss": 1.1991,
+      "step": 45251
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.570480404551202e-07,
+      "loss": 3.2914,
+      "step": 45252
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.570006321112516e-07,
+      "loss": 1.9981,
+      "step": 45253
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.569532237673831e-07,
+      "loss": 3.4576,
+      "step": 45254
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.569058154235145e-07,
+      "loss": 1.7208,
+      "step": 45255
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.56858407079646e-07,
+      "loss": 0.4906,
+      "step": 45256
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.568109987357776e-07,
+      "loss": 2.044,
+      "step": 45257
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.56763590391909e-07,
+      "loss": 1.4245,
+      "step": 45258
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.567161820480405e-07,
+      "loss": 1.6814,
+      "step": 45259
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.566687737041721e-07,
+      "loss": 3.7065,
+      "step": 45260
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.566213653603033e-07,
+      "loss": 3.1571,
+      "step": 45261
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.565739570164349e-07,
+      "loss": 3.1416,
+      "step": 45262
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.565265486725664e-07,
+      "loss": 2.4668,
+      "step": 45263
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.564791403286978e-07,
+      "loss": 1.1887,
+      "step": 45264
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.564317319848294e-07,
+      "loss": 1.7697,
+      "step": 45265
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.563843236409608e-07,
+      "loss": 1.87,
+      "step": 45266
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.563369152970923e-07,
+      "loss": 1.9507,
+      "step": 45267
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.562895069532238e-07,
+      "loss": 3.0744,
+      "step": 45268
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.562420986093552e-07,
+      "loss": 0.7531,
+      "step": 45269
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.561946902654868e-07,
+      "loss": 1.3419,
+      "step": 45270
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.561472819216183e-07,
+      "loss": 1.6314,
+      "step": 45271
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.560998735777496e-07,
+      "loss": 2.4927,
+      "step": 45272
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.560524652338812e-07,
+      "loss": 1.3244,
+      "step": 45273
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.560050568900127e-07,
+      "loss": 3.2926,
+      "step": 45274
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.559576485461441e-07,
+      "loss": 2.6358,
+      "step": 45275
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.559102402022756e-07,
+      "loss": 0.5578,
+      "step": 45276
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.55862831858407e-07,
+      "loss": 0.5665,
+      "step": 45277
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.558154235145386e-07,
+      "loss": 1.5865,
+      "step": 45278
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.557680151706701e-07,
+      "loss": 1.3433,
+      "step": 45279
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.557206068268015e-07,
+      "loss": 2.0523,
+      "step": 45280
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.55673198482933e-07,
+      "loss": 1.8299,
+      "step": 45281
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.556257901390646e-07,
+      "loss": 2.3051,
+      "step": 45282
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.55578381795196e-07,
+      "loss": 0.9035,
+      "step": 45283
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.555309734513274e-07,
+      "loss": 2.65,
+      "step": 45284
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.55483565107459e-07,
+      "loss": 2.5121,
+      "step": 45285
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.554361567635904e-07,
+      "loss": 1.4846,
+      "step": 45286
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.553887484197219e-07,
+      "loss": 1.8744,
+      "step": 45287
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.553413400758533e-07,
+      "loss": 1.9804,
+      "step": 45288
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.552939317319848e-07,
+      "loss": 1.9974,
+      "step": 45289
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.552465233881164e-07,
+      "loss": 1.7787,
+      "step": 45290
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.551991150442478e-07,
+      "loss": 1.411,
+      "step": 45291
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.551517067003793e-07,
+      "loss": 2.2,
+      "step": 45292
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.551042983565109e-07,
+      "loss": 0.8524,
+      "step": 45293
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.550568900126422e-07,
+      "loss": 3.2373,
+      "step": 45294
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.550094816687738e-07,
+      "loss": 2.3925,
+      "step": 45295
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.549620733249052e-07,
+      "loss": 1.2366,
+      "step": 45296
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.549146649810366e-07,
+      "loss": 3.0117,
+      "step": 45297
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.548672566371682e-07,
+      "loss": 1.8845,
+      "step": 45298
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.548198482932996e-07,
+      "loss": 2.8675,
+      "step": 45299
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.547724399494311e-07,
+      "loss": 1.9652,
+      "step": 45300
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.547250316055626e-07,
+      "loss": 1.8388,
+      "step": 45301
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.54677623261694e-07,
+      "loss": 1.5796,
+      "step": 45302
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.546302149178256e-07,
+      "loss": 0.547,
+      "step": 45303
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.545828065739571e-07,
+      "loss": 1.5012,
+      "step": 45304
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.545353982300885e-07,
+      "loss": 2.0686,
+      "step": 45305
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.544879898862201e-07,
+      "loss": 1.8722,
+      "step": 45306
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.544405815423516e-07,
+      "loss": 2.7447,
+      "step": 45307
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.543931731984829e-07,
+      "loss": 1.3472,
+      "step": 45308
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.543457648546144e-07,
+      "loss": 2.6796,
+      "step": 45309
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.542983565107458e-07,
+      "loss": 2.1033,
+      "step": 45310
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.542509481668774e-07,
+      "loss": 2.001,
+      "step": 45311
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 8.542035398230089e-07,
+      "loss": 1.0828,
+      "step": 45312
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.541561314791403e-07,
+      "loss": 1.7143,
+      "step": 45313
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.541087231352718e-07,
+      "loss": 1.9303,
+      "step": 45314
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.540613147914034e-07,
+      "loss": 1.6085,
+      "step": 45315
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.540139064475348e-07,
+      "loss": 2.5124,
+      "step": 45316
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.539664981036663e-07,
+      "loss": 4.061,
+      "step": 45317
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.539190897597979e-07,
+      "loss": 1.7389,
+      "step": 45318
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.538716814159292e-07,
+      "loss": 1.4983,
+      "step": 45319
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.538242730720607e-07,
+      "loss": 2.4431,
+      "step": 45320
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.537768647281921e-07,
+      "loss": 2.9277,
+      "step": 45321
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.537294563843236e-07,
+      "loss": 1.8851,
+      "step": 45322
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.536820480404552e-07,
+      "loss": 1.3011,
+      "step": 45323
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.536346396965866e-07,
+      "loss": 1.2256,
+      "step": 45324
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.535872313527181e-07,
+      "loss": 1.3939,
+      "step": 45325
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.535398230088497e-07,
+      "loss": 1.0951,
+      "step": 45326
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.53492414664981e-07,
+      "loss": 1.8523,
+      "step": 45327
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.534450063211126e-07,
+      "loss": 1.6634,
+      "step": 45328
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.533975979772441e-07,
+      "loss": 1.6262,
+      "step": 45329
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.533501896333754e-07,
+      "loss": 1.6412,
+      "step": 45330
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.53302781289507e-07,
+      "loss": 2.7601,
+      "step": 45331
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.532553729456383e-07,
+      "loss": 1.0857,
+      "step": 45332
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.532079646017699e-07,
+      "loss": 2.2192,
+      "step": 45333
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.531605562579014e-07,
+      "loss": 0.9008,
+      "step": 45334
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.531131479140328e-07,
+      "loss": 1.2438,
+      "step": 45335
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.530657395701644e-07,
+      "loss": 2.045,
+      "step": 45336
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.530183312262959e-07,
+      "loss": 3.088,
+      "step": 45337
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.529709228824273e-07,
+      "loss": 1.8558,
+      "step": 45338
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.529235145385588e-07,
+      "loss": 2.4472,
+      "step": 45339
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.528761061946904e-07,
+      "loss": 1.2001,
+      "step": 45340
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.528286978508218e-07,
+      "loss": 1.0064,
+      "step": 45341
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.527812895069532e-07,
+      "loss": 1.6737,
+      "step": 45342
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.527338811630846e-07,
+      "loss": 1.2687,
+      "step": 45343
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.526864728192162e-07,
+      "loss": 3.5145,
+      "step": 45344
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.526390644753477e-07,
+      "loss": 1.502,
+      "step": 45345
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.525916561314791e-07,
+      "loss": 1.4537,
+      "step": 45346
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.525442477876106e-07,
+      "loss": 0.9013,
+      "step": 45347
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.524968394437422e-07,
+      "loss": 2.219,
+      "step": 45348
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.524494310998736e-07,
+      "loss": 2.0074,
+      "step": 45349
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.524020227560051e-07,
+      "loss": 1.0212,
+      "step": 45350
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.523546144121367e-07,
+      "loss": 1.5216,
+      "step": 45351
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.52307206068268e-07,
+      "loss": 2.5598,
+      "step": 45352
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.522597977243996e-07,
+      "loss": 1.0543,
+      "step": 45353
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.522123893805309e-07,
+      "loss": 2.6388,
+      "step": 45354
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.521649810366624e-07,
+      "loss": 1.3137,
+      "step": 45355
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.52117572692794e-07,
+      "loss": 3.5735,
+      "step": 45356
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.520701643489254e-07,
+      "loss": 2.6707,
+      "step": 45357
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.520227560050569e-07,
+      "loss": 3.6913,
+      "step": 45358
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.519753476611884e-07,
+      "loss": 3.2961,
+      "step": 45359
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.519279393173198e-07,
+      "loss": 3.8039,
+      "step": 45360
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.518805309734514e-07,
+      "loss": 2.3648,
+      "step": 45361
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.518331226295829e-07,
+      "loss": 1.4185,
+      "step": 45362
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.517857142857143e-07,
+      "loss": 2.0852,
+      "step": 45363
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.517383059418459e-07,
+      "loss": 1.7505,
+      "step": 45364
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.516908975979771e-07,
+      "loss": 1.7028,
+      "step": 45365
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.516434892541087e-07,
+      "loss": 2.2755,
+      "step": 45366
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.515960809102402e-07,
+      "loss": 1.4876,
+      "step": 45367
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.515486725663716e-07,
+      "loss": 3.4318,
+      "step": 45368
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.515012642225032e-07,
+      "loss": 4.6881,
+      "step": 45369
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.514538558786347e-07,
+      "loss": 1.7735,
+      "step": 45370
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.514064475347661e-07,
+      "loss": 1.241,
+      "step": 45371
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.513590391908976e-07,
+      "loss": 2.7651,
+      "step": 45372
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.513116308470292e-07,
+      "loss": 1.362,
+      "step": 45373
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.512642225031606e-07,
+      "loss": 1.1894,
+      "step": 45374
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 8.512168141592921e-07,
+      "loss": 1.8405,
+      "step": 45375
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.511694058154235e-07,
+      "loss": 2.1332,
+      "step": 45376
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.51121997471555e-07,
+      "loss": 2.347,
+      "step": 45377
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.510745891276865e-07,
+      "loss": 2.8671,
+      "step": 45378
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.510271807838179e-07,
+      "loss": 2.3073,
+      "step": 45379
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.509797724399494e-07,
+      "loss": 1.6736,
+      "step": 45380
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.50932364096081e-07,
+      "loss": 2.5206,
+      "step": 45381
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.508849557522124e-07,
+      "loss": 3.949,
+      "step": 45382
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.508375474083439e-07,
+      "loss": 0.9564,
+      "step": 45383
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.507901390644755e-07,
+      "loss": 1.7718,
+      "step": 45384
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.507427307206068e-07,
+      "loss": 1.9128,
+      "step": 45385
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.506953223767384e-07,
+      "loss": 2.3289,
+      "step": 45386
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.506479140328698e-07,
+      "loss": 1.2008,
+      "step": 45387
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.506005056890013e-07,
+      "loss": 1.3002,
+      "step": 45388
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.505530973451329e-07,
+      "loss": 2.3029,
+      "step": 45389
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.505056890012641e-07,
+      "loss": 2.2964,
+      "step": 45390
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.504582806573957e-07,
+      "loss": 1.1662,
+      "step": 45391
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.504108723135272e-07,
+      "loss": 0.5207,
+      "step": 45392
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.503634639696586e-07,
+      "loss": 1.2217,
+      "step": 45393
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.503160556257902e-07,
+      "loss": 1.946,
+      "step": 45394
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.502686472819217e-07,
+      "loss": 1.6189,
+      "step": 45395
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.502212389380531e-07,
+      "loss": 0.4424,
+      "step": 45396
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.501738305941846e-07,
+      "loss": 0.9592,
+      "step": 45397
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.50126422250316e-07,
+      "loss": 1.9748,
+      "step": 45398
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.500790139064476e-07,
+      "loss": 1.5356,
+      "step": 45399
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.500316055625791e-07,
+      "loss": 1.8897,
+      "step": 45400
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.499841972187104e-07,
+      "loss": 2.4184,
+      "step": 45401
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.49936788874842e-07,
+      "loss": 2.6852,
+      "step": 45402
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.498893805309735e-07,
+      "loss": 1.7105,
+      "step": 45403
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.498419721871049e-07,
+      "loss": 1.0798,
+      "step": 45404
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.497945638432364e-07,
+      "loss": 3.4974,
+      "step": 45405
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.49747155499368e-07,
+      "loss": 1.9309,
+      "step": 45406
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.496997471554994e-07,
+      "loss": 3.2029,
+      "step": 45407
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.496523388116309e-07,
+      "loss": 0.8297,
+      "step": 45408
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.496049304677623e-07,
+      "loss": 2.6298,
+      "step": 45409
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.495575221238938e-07,
+      "loss": 2.2307,
+      "step": 45410
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.495101137800254e-07,
+      "loss": 2.848,
+      "step": 45411
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.494627054361567e-07,
+      "loss": 1.477,
+      "step": 45412
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.494152970922882e-07,
+      "loss": 3.3688,
+      "step": 45413
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.493678887484198e-07,
+      "loss": 1.5069,
+      "step": 45414
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.493204804045512e-07,
+      "loss": 3.916,
+      "step": 45415
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.492730720606827e-07,
+      "loss": 0.7755,
+      "step": 45416
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.492256637168142e-07,
+      "loss": 1.6203,
+      "step": 45417
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.491782553729456e-07,
+      "loss": 1.2191,
+      "step": 45418
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.491308470290772e-07,
+      "loss": 1.9588,
+      "step": 45419
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.490834386852086e-07,
+      "loss": 1.7515,
+      "step": 45420
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.490360303413401e-07,
+      "loss": 1.0423,
+      "step": 45421
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.489886219974717e-07,
+      "loss": 2.6546,
+      "step": 45422
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.48941213653603e-07,
+      "loss": 2.5299,
+      "step": 45423
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.488938053097345e-07,
+      "loss": 1.6815,
+      "step": 45424
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.48846396965866e-07,
+      "loss": 0.7651,
+      "step": 45425
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.487989886219974e-07,
+      "loss": 2.5945,
+      "step": 45426
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.48751580278129e-07,
+      "loss": 1.7882,
+      "step": 45427
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.487041719342605e-07,
+      "loss": 1.0731,
+      "step": 45428
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.486567635903919e-07,
+      "loss": 2.0644,
+      "step": 45429
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.486093552465234e-07,
+      "loss": 0.8172,
+      "step": 45430
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.485619469026548e-07,
+      "loss": 3.0308,
+      "step": 45431
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.485145385587864e-07,
+      "loss": 1.2988,
+      "step": 45432
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.484671302149179e-07,
+      "loss": 2.0737,
+      "step": 45433
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.484197218710493e-07,
+      "loss": 1.6519,
+      "step": 45434
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.483723135271809e-07,
+      "loss": 2.91,
+      "step": 45435
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.483249051833123e-07,
+      "loss": 2.0062,
+      "step": 45436
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.482774968394437e-07,
+      "loss": 2.1731,
+      "step": 45437
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.482300884955752e-07,
+      "loss": 1.5946,
+      "step": 45438
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 8.481826801517068e-07,
+      "loss": 2.5404,
+      "step": 45439
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.481352718078382e-07,
+      "loss": 2.3765,
+      "step": 45440
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.480878634639697e-07,
+      "loss": 2.3858,
+      "step": 45441
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.480404551201011e-07,
+      "loss": 1.0734,
+      "step": 45442
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.479930467762326e-07,
+      "loss": 2.6503,
+      "step": 45443
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.479456384323642e-07,
+      "loss": 1.0794,
+      "step": 45444
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.478982300884956e-07,
+      "loss": 1.6568,
+      "step": 45445
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.478508217446271e-07,
+      "loss": 3.4773,
+      "step": 45446
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.478034134007587e-07,
+      "loss": 3.5547,
+      "step": 45447
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.4775600505689e-07,
+      "loss": 1.7419,
+      "step": 45448
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.477085967130215e-07,
+      "loss": 2.0011,
+      "step": 45449
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.47661188369153e-07,
+      "loss": 1.6763,
+      "step": 45450
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.476137800252844e-07,
+      "loss": 2.5807,
+      "step": 45451
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.47566371681416e-07,
+      "loss": 2.2665,
+      "step": 45452
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.475189633375474e-07,
+      "loss": 0.1773,
+      "step": 45453
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.474715549936789e-07,
+      "loss": 1.9288,
+      "step": 45454
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.474241466498105e-07,
+      "loss": 3.0931,
+      "step": 45455
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.473767383059418e-07,
+      "loss": 1.5578,
+      "step": 45456
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.473293299620734e-07,
+      "loss": 2.9691,
+      "step": 45457
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.472819216182049e-07,
+      "loss": 2.7416,
+      "step": 45458
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.472345132743362e-07,
+      "loss": 1.8425,
+      "step": 45459
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.471871049304678e-07,
+      "loss": 0.6898,
+      "step": 45460
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.471396965865993e-07,
+      "loss": 1.1172,
+      "step": 45461
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.470922882427307e-07,
+      "loss": 2.0569,
+      "step": 45462
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.470448798988622e-07,
+      "loss": 1.5311,
+      "step": 45463
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.469974715549936e-07,
+      "loss": 3.1353,
+      "step": 45464
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.469500632111252e-07,
+      "loss": 2.8291,
+      "step": 45465
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.469026548672567e-07,
+      "loss": 1.3123,
+      "step": 45466
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.468552465233881e-07,
+      "loss": 1.5762,
+      "step": 45467
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.468078381795196e-07,
+      "loss": 2.2278,
+      "step": 45468
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.467604298356512e-07,
+      "loss": 3.6379,
+      "step": 45469
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.467130214917826e-07,
+      "loss": 0.8188,
+      "step": 45470
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.46665613147914e-07,
+      "loss": 1.8166,
+      "step": 45471
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.466182048040456e-07,
+      "loss": 2.7332,
+      "step": 45472
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.46570796460177e-07,
+      "loss": 1.7947,
+      "step": 45473
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.465233881163085e-07,
+      "loss": 2.8473,
+      "step": 45474
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.464759797724399e-07,
+      "loss": 2.4263,
+      "step": 45475
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.464285714285714e-07,
+      "loss": 2.027,
+      "step": 45476
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.46381163084703e-07,
+      "loss": 1.4833,
+      "step": 45477
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.463337547408344e-07,
+      "loss": 2.246,
+      "step": 45478
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.462863463969659e-07,
+      "loss": 3.2244,
+      "step": 45479
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.462389380530975e-07,
+      "loss": 2.6698,
+      "step": 45480
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.461915297092288e-07,
+      "loss": 0.9253,
+      "step": 45481
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.461441213653604e-07,
+      "loss": 1.5102,
+      "step": 45482
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.460967130214918e-07,
+      "loss": 2.0243,
+      "step": 45483
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.460493046776232e-07,
+      "loss": 0.2426,
+      "step": 45484
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.460018963337548e-07,
+      "loss": 2.8675,
+      "step": 45485
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.459544879898862e-07,
+      "loss": 2.7578,
+      "step": 45486
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.459070796460177e-07,
+      "loss": 2.6657,
+      "step": 45487
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.458596713021492e-07,
+      "loss": 1.787,
+      "step": 45488
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.458122629582806e-07,
+      "loss": 2.1341,
+      "step": 45489
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.457648546144122e-07,
+      "loss": 2.5506,
+      "step": 45490
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.457174462705437e-07,
+      "loss": 1.0842,
+      "step": 45491
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.456700379266751e-07,
+      "loss": 3.1226,
+      "step": 45492
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.456226295828067e-07,
+      "loss": 2.5461,
+      "step": 45493
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.455752212389382e-07,
+      "loss": 0.7444,
+      "step": 45494
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.455278128950695e-07,
+      "loss": 2.6625,
+      "step": 45495
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.45480404551201e-07,
+      "loss": 2.3714,
+      "step": 45496
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.454329962073324e-07,
+      "loss": 1.5144,
+      "step": 45497
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.45385587863464e-07,
+      "loss": 2.031,
+      "step": 45498
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.453381795195955e-07,
+      "loss": 3.3745,
+      "step": 45499
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.452907711757269e-07,
+      "loss": 1.6956,
+      "step": 45500
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.452433628318584e-07,
+      "loss": 1.3055,
+      "step": 45501
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 8.4519595448799e-07,
+      "loss": 1.5722,
+      "step": 45502
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.451485461441214e-07,
+      "loss": 1.4296,
+      "step": 45503
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.451011378002529e-07,
+      "loss": 1.7582,
+      "step": 45504
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.450537294563845e-07,
+      "loss": 1.9967,
+      "step": 45505
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.450063211125158e-07,
+      "loss": 1.6709,
+      "step": 45506
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.449589127686473e-07,
+      "loss": 1.6461,
+      "step": 45507
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.449115044247787e-07,
+      "loss": 1.4299,
+      "step": 45508
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.448640960809102e-07,
+      "loss": 2.2962,
+      "step": 45509
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.448166877370418e-07,
+      "loss": 1.9347,
+      "step": 45510
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.447692793931732e-07,
+      "loss": 1.1979,
+      "step": 45511
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.447218710493047e-07,
+      "loss": 2.8023,
+      "step": 45512
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.446744627054363e-07,
+      "loss": 2.9879,
+      "step": 45513
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.446270543615676e-07,
+      "loss": 3.0371,
+      "step": 45514
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.445796460176992e-07,
+      "loss": 3.054,
+      "step": 45515
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.445322376738307e-07,
+      "loss": 1.8129,
+      "step": 45516
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.44484829329962e-07,
+      "loss": 1.7916,
+      "step": 45517
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.444374209860936e-07,
+      "loss": 2.517,
+      "step": 45518
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.44390012642225e-07,
+      "loss": 1.6851,
+      "step": 45519
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.443426042983565e-07,
+      "loss": 2.7799,
+      "step": 45520
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.44295195954488e-07,
+      "loss": 1.8723,
+      "step": 45521
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.442477876106194e-07,
+      "loss": 2.2248,
+      "step": 45522
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.44200379266751e-07,
+      "loss": 1.8204,
+      "step": 45523
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.441529709228825e-07,
+      "loss": 1.6172,
+      "step": 45524
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.441055625790139e-07,
+      "loss": 0.9947,
+      "step": 45525
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.440581542351455e-07,
+      "loss": 3.3267,
+      "step": 45526
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.44010745891277e-07,
+      "loss": 2.0252,
+      "step": 45527
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.439633375474084e-07,
+      "loss": 2.6407,
+      "step": 45528
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.439159292035399e-07,
+      "loss": 0.575,
+      "step": 45529
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.438685208596712e-07,
+      "loss": 3.0096,
+      "step": 45530
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.438211125158028e-07,
+      "loss": 2.8837,
+      "step": 45531
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.437737041719343e-07,
+      "loss": 2.0742,
+      "step": 45532
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.437262958280657e-07,
+      "loss": 1.5835,
+      "step": 45533
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.436788874841972e-07,
+      "loss": 2.2094,
+      "step": 45534
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.436314791403288e-07,
+      "loss": 0.8664,
+      "step": 45535
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.435840707964602e-07,
+      "loss": 2.3596,
+      "step": 45536
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.435366624525917e-07,
+      "loss": 1.8779,
+      "step": 45537
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.434892541087231e-07,
+      "loss": 1.1909,
+      "step": 45538
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.434418457648546e-07,
+      "loss": 2.8559,
+      "step": 45539
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.433944374209862e-07,
+      "loss": 2.2664,
+      "step": 45540
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.433470290771175e-07,
+      "loss": 1.7765,
+      "step": 45541
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.43299620733249e-07,
+      "loss": 0.6597,
+      "step": 45542
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.432522123893806e-07,
+      "loss": 2.2139,
+      "step": 45543
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.43204804045512e-07,
+      "loss": 0.8774,
+      "step": 45544
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.431573957016435e-07,
+      "loss": 4.4781,
+      "step": 45545
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.43109987357775e-07,
+      "loss": 1.1554,
+      "step": 45546
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.430625790139064e-07,
+      "loss": 1.1913,
+      "step": 45547
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.43015170670038e-07,
+      "loss": 1.9127,
+      "step": 45548
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.429677623261694e-07,
+      "loss": 1.5757,
+      "step": 45549
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.429203539823009e-07,
+      "loss": 3.6249,
+      "step": 45550
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.428729456384325e-07,
+      "loss": 1.5748,
+      "step": 45551
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.428255372945637e-07,
+      "loss": 1.907,
+      "step": 45552
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.427781289506953e-07,
+      "loss": 1.9278,
+      "step": 45553
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.427307206068268e-07,
+      "loss": 1.9306,
+      "step": 45554
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.426833122629582e-07,
+      "loss": 2.8182,
+      "step": 45555
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.426359039190898e-07,
+      "loss": 1.8103,
+      "step": 45556
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.425884955752213e-07,
+      "loss": 1.6938,
+      "step": 45557
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.425410872313527e-07,
+      "loss": 2.9083,
+      "step": 45558
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.424936788874842e-07,
+      "loss": 2.934,
+      "step": 45559
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.424462705436156e-07,
+      "loss": 2.2235,
+      "step": 45560
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.423988621997472e-07,
+      "loss": 1.6082,
+      "step": 45561
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.423514538558787e-07,
+      "loss": 1.7133,
+      "step": 45562
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.423040455120101e-07,
+      "loss": 0.7706,
+      "step": 45563
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.422566371681417e-07,
+      "loss": 0.8904,
+      "step": 45564
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 8.422092288242731e-07,
+      "loss": 2.3754,
+      "step": 45565
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.421618204804045e-07,
+      "loss": 2.1942,
+      "step": 45566
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.42114412136536e-07,
+      "loss": 2.4567,
+      "step": 45567
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.420670037926676e-07,
+      "loss": 1.0047,
+      "step": 45568
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.42019595448799e-07,
+      "loss": 0.6457,
+      "step": 45569
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.419721871049305e-07,
+      "loss": 1.454,
+      "step": 45570
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.419247787610619e-07,
+      "loss": 1.9231,
+      "step": 45571
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.418773704171934e-07,
+      "loss": 1.4548,
+      "step": 45572
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.41829962073325e-07,
+      "loss": 1.0835,
+      "step": 45573
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.417825537294564e-07,
+      "loss": 1.8205,
+      "step": 45574
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.417351453855879e-07,
+      "loss": 2.0249,
+      "step": 45575
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.416877370417195e-07,
+      "loss": 1.2763,
+      "step": 45576
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.416403286978508e-07,
+      "loss": 1.6578,
+      "step": 45577
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.415929203539823e-07,
+      "loss": 1.115,
+      "step": 45578
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.415455120101138e-07,
+      "loss": 0.7411,
+      "step": 45579
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.414981036662452e-07,
+      "loss": 2.0045,
+      "step": 45580
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.414506953223768e-07,
+      "loss": 2.1845,
+      "step": 45581
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.414032869785082e-07,
+      "loss": 2.9203,
+      "step": 45582
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.413558786346397e-07,
+      "loss": 0.997,
+      "step": 45583
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.413084702907713e-07,
+      "loss": 2.297,
+      "step": 45584
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.412610619469026e-07,
+      "loss": 2.004,
+      "step": 45585
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.412136536030342e-07,
+      "loss": 2.2355,
+      "step": 45586
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.411662452591657e-07,
+      "loss": 2.001,
+      "step": 45587
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.41118836915297e-07,
+      "loss": 0.8833,
+      "step": 45588
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.410714285714286e-07,
+      "loss": 2.2886,
+      "step": 45589
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.410240202275601e-07,
+      "loss": 2.9646,
+      "step": 45590
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.409766118836915e-07,
+      "loss": 2.536,
+      "step": 45591
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.40929203539823e-07,
+      "loss": 3.7345,
+      "step": 45592
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.408817951959544e-07,
+      "loss": 1.952,
+      "step": 45593
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.40834386852086e-07,
+      "loss": 0.9614,
+      "step": 45594
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.407869785082175e-07,
+      "loss": 1.3096,
+      "step": 45595
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.407395701643489e-07,
+      "loss": 2.6039,
+      "step": 45596
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.406921618204805e-07,
+      "loss": 2.2067,
+      "step": 45597
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.40644753476612e-07,
+      "loss": 1.5431,
+      "step": 45598
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.405973451327434e-07,
+      "loss": 2.6874,
+      "step": 45599
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.405499367888748e-07,
+      "loss": 2.0336,
+      "step": 45600
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.405025284450064e-07,
+      "loss": 1.5073,
+      "step": 45601
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.404551201011378e-07,
+      "loss": 2.0357,
+      "step": 45602
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.404077117572693e-07,
+      "loss": 1.936,
+      "step": 45603
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.403603034134007e-07,
+      "loss": 0.6245,
+      "step": 45604
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.403128950695322e-07,
+      "loss": 1.2652,
+      "step": 45605
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.402654867256638e-07,
+      "loss": 3.995,
+      "step": 45606
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.402180783817952e-07,
+      "loss": 2.13,
+      "step": 45607
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.401706700379267e-07,
+      "loss": 1.1073,
+      "step": 45608
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.401232616940583e-07,
+      "loss": 1.8751,
+      "step": 45609
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.400758533501896e-07,
+      "loss": 2.26,
+      "step": 45610
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.400284450063212e-07,
+      "loss": 3.5471,
+      "step": 45611
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.399810366624526e-07,
+      "loss": 3.2215,
+      "step": 45612
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.39933628318584e-07,
+      "loss": 1.482,
+      "step": 45613
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.398862199747156e-07,
+      "loss": 3.0205,
+      "step": 45614
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.39838811630847e-07,
+      "loss": 1.7581,
+      "step": 45615
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.397914032869785e-07,
+      "loss": 1.2447,
+      "step": 45616
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.3974399494311e-07,
+      "loss": 2.7967,
+      "step": 45617
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.396965865992414e-07,
+      "loss": 1.1733,
+      "step": 45618
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.39649178255373e-07,
+      "loss": 1.1786,
+      "step": 45619
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.396017699115045e-07,
+      "loss": 2.4036,
+      "step": 45620
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.395543615676359e-07,
+      "loss": 2.0591,
+      "step": 45621
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.395069532237675e-07,
+      "loss": 1.769,
+      "step": 45622
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.39459544879899e-07,
+      "loss": 1.3764,
+      "step": 45623
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.394121365360303e-07,
+      "loss": 1.9031,
+      "step": 45624
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.393647281921618e-07,
+      "loss": 0.7557,
+      "step": 45625
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.393173198482932e-07,
+      "loss": 0.8796,
+      "step": 45626
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.392699115044248e-07,
+      "loss": 0.8218,
+      "step": 45627
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.392225031605563e-07,
+      "loss": 2.9505,
+      "step": 45628
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 8.391750948166877e-07,
+      "loss": 1.5155,
+      "step": 45629
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.391276864728192e-07,
+      "loss": 1.4995,
+      "step": 45630
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.390802781289508e-07,
+      "loss": 3.0304,
+      "step": 45631
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.390328697850822e-07,
+      "loss": 1.8662,
+      "step": 45632
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.389854614412137e-07,
+      "loss": 2.7788,
+      "step": 45633
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.389380530973453e-07,
+      "loss": 1.9413,
+      "step": 45634
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.388906447534766e-07,
+      "loss": 1.555,
+      "step": 45635
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.388432364096081e-07,
+      "loss": 2.3985,
+      "step": 45636
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.387958280657395e-07,
+      "loss": 2.95,
+      "step": 45637
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.38748419721871e-07,
+      "loss": 2.1994,
+      "step": 45638
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.387010113780026e-07,
+      "loss": 2.6681,
+      "step": 45639
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.38653603034134e-07,
+      "loss": 1.9248,
+      "step": 45640
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.386061946902655e-07,
+      "loss": 1.6053,
+      "step": 45641
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.385587863463971e-07,
+      "loss": 1.8618,
+      "step": 45642
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.385113780025284e-07,
+      "loss": 2.8001,
+      "step": 45643
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.3846396965866e-07,
+      "loss": 1.0274,
+      "step": 45644
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.384165613147915e-07,
+      "loss": 2.1642,
+      "step": 45645
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.383691529709228e-07,
+      "loss": 2.0167,
+      "step": 45646
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.383217446270544e-07,
+      "loss": 2.95,
+      "step": 45647
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.382743362831858e-07,
+      "loss": 2.4852,
+      "step": 45648
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.382269279393173e-07,
+      "loss": 1.5347,
+      "step": 45649
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.381795195954488e-07,
+      "loss": 2.8806,
+      "step": 45650
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.381321112515802e-07,
+      "loss": 1.4788,
+      "step": 45651
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.380847029077118e-07,
+      "loss": 2.7333,
+      "step": 45652
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.380372945638433e-07,
+      "loss": 2.2006,
+      "step": 45653
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.379898862199747e-07,
+      "loss": 2.3144,
+      "step": 45654
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.379424778761063e-07,
+      "loss": 3.3954,
+      "step": 45655
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.378950695322378e-07,
+      "loss": 1.1858,
+      "step": 45656
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.378476611883692e-07,
+      "loss": 2.9464,
+      "step": 45657
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.378002528445006e-07,
+      "loss": 2.4964,
+      "step": 45658
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.37752844500632e-07,
+      "loss": 1.4822,
+      "step": 45659
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.377054361567636e-07,
+      "loss": 2.6359,
+      "step": 45660
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.376580278128951e-07,
+      "loss": 2.0466,
+      "step": 45661
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.376106194690265e-07,
+      "loss": 1.6049,
+      "step": 45662
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.37563211125158e-07,
+      "loss": 2.3436,
+      "step": 45663
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.375158027812896e-07,
+      "loss": 3.0883,
+      "step": 45664
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.37468394437421e-07,
+      "loss": 2.6443,
+      "step": 45665
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.374209860935525e-07,
+      "loss": 1.8375,
+      "step": 45666
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.373735777496841e-07,
+      "loss": 1.9908,
+      "step": 45667
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.373261694058155e-07,
+      "loss": 2.4479,
+      "step": 45668
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.37278761061947e-07,
+      "loss": 3.2177,
+      "step": 45669
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.372313527180783e-07,
+      "loss": 2.6764,
+      "step": 45670
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.371839443742098e-07,
+      "loss": 1.896,
+      "step": 45671
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.371365360303414e-07,
+      "loss": 2.2772,
+      "step": 45672
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.370891276864728e-07,
+      "loss": 2.2877,
+      "step": 45673
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.370417193426043e-07,
+      "loss": 2.5007,
+      "step": 45674
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.369943109987359e-07,
+      "loss": 0.8678,
+      "step": 45675
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.369469026548672e-07,
+      "loss": 2.4546,
+      "step": 45676
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.368994943109988e-07,
+      "loss": 1.997,
+      "step": 45677
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.368520859671303e-07,
+      "loss": 1.5499,
+      "step": 45678
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.368046776232617e-07,
+      "loss": 2.3264,
+      "step": 45679
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.367572692793933e-07,
+      "loss": 2.567,
+      "step": 45680
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.367098609355245e-07,
+      "loss": 2.7168,
+      "step": 45681
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.366624525916561e-07,
+      "loss": 2.3386,
+      "step": 45682
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.366150442477876e-07,
+      "loss": 2.4384,
+      "step": 45683
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.36567635903919e-07,
+      "loss": 2.7028,
+      "step": 45684
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.365202275600506e-07,
+      "loss": 1.6927,
+      "step": 45685
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.364728192161821e-07,
+      "loss": 2.739,
+      "step": 45686
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.364254108723135e-07,
+      "loss": 1.5487,
+      "step": 45687
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.36378002528445e-07,
+      "loss": 2.2322,
+      "step": 45688
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.363305941845766e-07,
+      "loss": 1.6608,
+      "step": 45689
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.36283185840708e-07,
+      "loss": 2.6404,
+      "step": 45690
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.362357774968395e-07,
+      "loss": 1.7116,
+      "step": 45691
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 8.361883691529709e-07,
+      "loss": 1.9094,
+      "step": 45692
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.361409608091025e-07,
+      "loss": 2.0746,
+      "step": 45693
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.360935524652339e-07,
+      "loss": 1.6582,
+      "step": 45694
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.360461441213653e-07,
+      "loss": 1.065,
+      "step": 45695
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.359987357774968e-07,
+      "loss": 1.0507,
+      "step": 45696
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.359513274336284e-07,
+      "loss": 0.9919,
+      "step": 45697
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.359039190897598e-07,
+      "loss": 1.0798,
+      "step": 45698
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.358565107458913e-07,
+      "loss": 2.1686,
+      "step": 45699
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.358091024020229e-07,
+      "loss": 2.1571,
+      "step": 45700
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.357616940581542e-07,
+      "loss": 1.4448,
+      "step": 45701
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.357142857142858e-07,
+      "loss": 1.5752,
+      "step": 45702
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.356668773704172e-07,
+      "loss": 1.8429,
+      "step": 45703
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.356194690265487e-07,
+      "loss": 1.283,
+      "step": 45704
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.355720606826803e-07,
+      "loss": 1.8851,
+      "step": 45705
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.355246523388116e-07,
+      "loss": 1.3931,
+      "step": 45706
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.354772439949431e-07,
+      "loss": 0.8634,
+      "step": 45707
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.354298356510746e-07,
+      "loss": 0.736,
+      "step": 45708
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.35382427307206e-07,
+      "loss": 1.8524,
+      "step": 45709
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.353350189633376e-07,
+      "loss": 1.1452,
+      "step": 45710
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.352876106194691e-07,
+      "loss": 1.585,
+      "step": 45711
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.352402022756005e-07,
+      "loss": 1.7051,
+      "step": 45712
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.351927939317321e-07,
+      "loss": 1.8537,
+      "step": 45713
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.351453855878634e-07,
+      "loss": 1.6718,
+      "step": 45714
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.35097977243995e-07,
+      "loss": 1.0867,
+      "step": 45715
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.350505689001265e-07,
+      "loss": 2.0608,
+      "step": 45716
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.350031605562578e-07,
+      "loss": 1.1898,
+      "step": 45717
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.349557522123894e-07,
+      "loss": 3.1536,
+      "step": 45718
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.349083438685209e-07,
+      "loss": 1.8364,
+      "step": 45719
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.348609355246523e-07,
+      "loss": 3.1585,
+      "step": 45720
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.348135271807838e-07,
+      "loss": 0.7373,
+      "step": 45721
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.347661188369154e-07,
+      "loss": 1.2126,
+      "step": 45722
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.347187104930468e-07,
+      "loss": 2.9035,
+      "step": 45723
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.346713021491783e-07,
+      "loss": 1.474,
+      "step": 45724
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.346238938053097e-07,
+      "loss": 2.0401,
+      "step": 45725
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.345764854614413e-07,
+      "loss": 1.3894,
+      "step": 45726
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.345290771175728e-07,
+      "loss": 1.6516,
+      "step": 45727
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.344816687737041e-07,
+      "loss": 1.8123,
+      "step": 45728
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.344342604298356e-07,
+      "loss": 3.242,
+      "step": 45729
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.343868520859672e-07,
+      "loss": 2.0683,
+      "step": 45730
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.343394437420986e-07,
+      "loss": 2.1902,
+      "step": 45731
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.342920353982301e-07,
+      "loss": 3.0508,
+      "step": 45732
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.342446270543617e-07,
+      "loss": 1.5893,
+      "step": 45733
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.34197218710493e-07,
+      "loss": 1.4652,
+      "step": 45734
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.341498103666246e-07,
+      "loss": 3.6408,
+      "step": 45735
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.34102402022756e-07,
+      "loss": 1.6482,
+      "step": 45736
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.340549936788875e-07,
+      "loss": 3.3181,
+      "step": 45737
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.340075853350191e-07,
+      "loss": 1.5023,
+      "step": 45738
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.339601769911504e-07,
+      "loss": 2.252,
+      "step": 45739
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.33912768647282e-07,
+      "loss": 2.1922,
+      "step": 45740
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.338653603034134e-07,
+      "loss": 2.3284,
+      "step": 45741
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.338179519595448e-07,
+      "loss": 2.5189,
+      "step": 45742
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.337705436156764e-07,
+      "loss": 2.1911,
+      "step": 45743
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.337231352718079e-07,
+      "loss": 1.6947,
+      "step": 45744
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.336757269279393e-07,
+      "loss": 1.2154,
+      "step": 45745
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.336283185840708e-07,
+      "loss": 1.3259,
+      "step": 45746
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.335809102402022e-07,
+      "loss": 1.8612,
+      "step": 45747
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.335335018963338e-07,
+      "loss": 2.8692,
+      "step": 45748
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.334860935524653e-07,
+      "loss": 3.3668,
+      "step": 45749
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.334386852085967e-07,
+      "loss": 1.946,
+      "step": 45750
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.333912768647283e-07,
+      "loss": 1.3134,
+      "step": 45751
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.333438685208597e-07,
+      "loss": 1.9662,
+      "step": 45752
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.332964601769911e-07,
+      "loss": 3.8539,
+      "step": 45753
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.332490518331226e-07,
+      "loss": 2.0587,
+      "step": 45754
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 8.332016434892542e-07,
+      "loss": 0.9809,
+      "step": 45755
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.331542351453856e-07,
+      "loss": 0.9776,
+      "step": 45756
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.331068268015171e-07,
+      "loss": 3.325,
+      "step": 45757
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.330594184576485e-07,
+      "loss": 2.3266,
+      "step": 45758
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.3301201011378e-07,
+      "loss": 2.512,
+      "step": 45759
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.329646017699116e-07,
+      "loss": 1.227,
+      "step": 45760
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.32917193426043e-07,
+      "loss": 2.1237,
+      "step": 45761
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.328697850821745e-07,
+      "loss": 2.2451,
+      "step": 45762
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.328223767383061e-07,
+      "loss": 1.334,
+      "step": 45763
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.327749683944374e-07,
+      "loss": 1.5241,
+      "step": 45764
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.327275600505689e-07,
+      "loss": 1.1404,
+      "step": 45765
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.326801517067004e-07,
+      "loss": 1.8354,
+      "step": 45766
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.326327433628318e-07,
+      "loss": 1.63,
+      "step": 45767
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.325853350189634e-07,
+      "loss": 1.005,
+      "step": 45768
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.325379266750948e-07,
+      "loss": 3.7996,
+      "step": 45769
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.324905183312263e-07,
+      "loss": 1.752,
+      "step": 45770
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.324431099873579e-07,
+      "loss": 2.7928,
+      "step": 45771
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.323957016434892e-07,
+      "loss": 0.8933,
+      "step": 45772
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.323482932996208e-07,
+      "loss": 0.7421,
+      "step": 45773
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.323008849557523e-07,
+      "loss": 2.7634,
+      "step": 45774
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.322534766118836e-07,
+      "loss": 2.2394,
+      "step": 45775
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.322060682680152e-07,
+      "loss": 2.0659,
+      "step": 45776
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.321586599241467e-07,
+      "loss": 1.9832,
+      "step": 45777
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.321112515802781e-07,
+      "loss": 0.8884,
+      "step": 45778
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.320638432364096e-07,
+      "loss": 2.6982,
+      "step": 45779
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.32016434892541e-07,
+      "loss": 3.0185,
+      "step": 45780
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.319690265486726e-07,
+      "loss": 1.4099,
+      "step": 45781
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.319216182048041e-07,
+      "loss": 1.7091,
+      "step": 45782
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.318742098609355e-07,
+      "loss": 0.7701,
+      "step": 45783
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.318268015170671e-07,
+      "loss": 2.1599,
+      "step": 45784
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.317793931731986e-07,
+      "loss": 3.0604,
+      "step": 45785
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.3173198482933e-07,
+      "loss": 2.2709,
+      "step": 45786
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.316845764854614e-07,
+      "loss": 2.5124,
+      "step": 45787
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.31637168141593e-07,
+      "loss": 1.7753,
+      "step": 45788
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.315897597977244e-07,
+      "loss": 2.9034,
+      "step": 45789
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.315423514538559e-07,
+      "loss": 2.5433,
+      "step": 45790
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.314949431099873e-07,
+      "loss": 2.5997,
+      "step": 45791
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.314475347661188e-07,
+      "loss": 1.7522,
+      "step": 45792
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.314001264222504e-07,
+      "loss": 1.591,
+      "step": 45793
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.313527180783818e-07,
+      "loss": 1.9546,
+      "step": 45794
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.313053097345133e-07,
+      "loss": 1.8334,
+      "step": 45795
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.312579013906449e-07,
+      "loss": 0.9819,
+      "step": 45796
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.312104930467763e-07,
+      "loss": 3.19,
+      "step": 45797
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.311630847029078e-07,
+      "loss": 1.2449,
+      "step": 45798
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.311156763590392e-07,
+      "loss": 1.8614,
+      "step": 45799
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.310682680151706e-07,
+      "loss": 3.3387,
+      "step": 45800
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.310208596713022e-07,
+      "loss": 0.9478,
+      "step": 45801
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.309734513274336e-07,
+      "loss": 2.4221,
+      "step": 45802
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.309260429835651e-07,
+      "loss": 0.5059,
+      "step": 45803
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.308786346396967e-07,
+      "loss": 1.1759,
+      "step": 45804
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.30831226295828e-07,
+      "loss": 1.024,
+      "step": 45805
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.307838179519596e-07,
+      "loss": 2.9798,
+      "step": 45806
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.307364096080911e-07,
+      "loss": 1.8623,
+      "step": 45807
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.306890012642225e-07,
+      "loss": 0.5968,
+      "step": 45808
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.306415929203541e-07,
+      "loss": 0.4945,
+      "step": 45809
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.305941845764856e-07,
+      "loss": 1.5101,
+      "step": 45810
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.305467762326169e-07,
+      "loss": 2.6043,
+      "step": 45811
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.304993678887484e-07,
+      "loss": 1.9247,
+      "step": 45812
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.304519595448798e-07,
+      "loss": 1.5956,
+      "step": 45813
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.304045512010114e-07,
+      "loss": 0.7433,
+      "step": 45814
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.303571428571429e-07,
+      "loss": 2.6621,
+      "step": 45815
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.303097345132743e-07,
+      "loss": 1.5828,
+      "step": 45816
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.302623261694058e-07,
+      "loss": 0.8729,
+      "step": 45817
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.302149178255374e-07,
+      "loss": 3.6221,
+      "step": 45818
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 8.301675094816688e-07,
+      "loss": 1.9029,
+      "step": 45819
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.301201011378003e-07,
+      "loss": 1.5264,
+      "step": 45820
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.300726927939319e-07,
+      "loss": 2.2473,
+      "step": 45821
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.300252844500632e-07,
+      "loss": 2.1056,
+      "step": 45822
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.299778761061947e-07,
+      "loss": 3.2135,
+      "step": 45823
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.299304677623261e-07,
+      "loss": 2.1277,
+      "step": 45824
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.298830594184576e-07,
+      "loss": 1.5937,
+      "step": 45825
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.298356510745892e-07,
+      "loss": 2.7223,
+      "step": 45826
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.297882427307206e-07,
+      "loss": 2.5376,
+      "step": 45827
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.297408343868521e-07,
+      "loss": 1.5564,
+      "step": 45828
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.296934260429837e-07,
+      "loss": 2.2412,
+      "step": 45829
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.29646017699115e-07,
+      "loss": 1.1416,
+      "step": 45830
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.295986093552466e-07,
+      "loss": 2.3284,
+      "step": 45831
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.295512010113781e-07,
+      "loss": 1.2015,
+      "step": 45832
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.295037926675094e-07,
+      "loss": 2.2383,
+      "step": 45833
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.29456384323641e-07,
+      "loss": 1.2824,
+      "step": 45834
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.294089759797724e-07,
+      "loss": 2.8731,
+      "step": 45835
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.293615676359039e-07,
+      "loss": 2.5449,
+      "step": 45836
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.293141592920354e-07,
+      "loss": 2.2156,
+      "step": 45837
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.292667509481668e-07,
+      "loss": 3.02,
+      "step": 45838
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.292193426042984e-07,
+      "loss": 2.6412,
+      "step": 45839
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.291719342604299e-07,
+      "loss": 1.9323,
+      "step": 45840
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.291245259165613e-07,
+      "loss": 1.6389,
+      "step": 45841
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.290771175726929e-07,
+      "loss": 1.9179,
+      "step": 45842
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.290297092288244e-07,
+      "loss": 2.0641,
+      "step": 45843
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.289823008849558e-07,
+      "loss": 2.0837,
+      "step": 45844
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.289348925410873e-07,
+      "loss": 2.5667,
+      "step": 45845
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.288874841972186e-07,
+      "loss": 3.6943,
+      "step": 45846
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.288400758533502e-07,
+      "loss": 1.8605,
+      "step": 45847
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.287926675094817e-07,
+      "loss": 2.0625,
+      "step": 45848
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.287452591656131e-07,
+      "loss": 2.2183,
+      "step": 45849
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.286978508217446e-07,
+      "loss": 0.9356,
+      "step": 45850
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.286504424778762e-07,
+      "loss": 3.236,
+      "step": 45851
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.286030341340076e-07,
+      "loss": 3.5152,
+      "step": 45852
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.285556257901391e-07,
+      "loss": 2.5137,
+      "step": 45853
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.285082174462707e-07,
+      "loss": 1.2541,
+      "step": 45854
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.284608091024021e-07,
+      "loss": 1.5508,
+      "step": 45855
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.284134007585336e-07,
+      "loss": 2.2284,
+      "step": 45856
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.283659924146649e-07,
+      "loss": 2.0252,
+      "step": 45857
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.283185840707964e-07,
+      "loss": 0.8468,
+      "step": 45858
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.28271175726928e-07,
+      "loss": 1.7014,
+      "step": 45859
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.282237673830594e-07,
+      "loss": 0.5741,
+      "step": 45860
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.281763590391909e-07,
+      "loss": 3.2195,
+      "step": 45861
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.281289506953225e-07,
+      "loss": 2.628,
+      "step": 45862
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.280815423514538e-07,
+      "loss": 3.0715,
+      "step": 45863
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.280341340075854e-07,
+      "loss": 1.3532,
+      "step": 45864
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.279867256637169e-07,
+      "loss": 2.1557,
+      "step": 45865
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.279393173198483e-07,
+      "loss": 1.6607,
+      "step": 45866
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.278919089759799e-07,
+      "loss": 1.7472,
+      "step": 45867
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.278445006321113e-07,
+      "loss": 1.1413,
+      "step": 45868
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.277970922882427e-07,
+      "loss": 0.6795,
+      "step": 45869
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.277496839443742e-07,
+      "loss": 2.0714,
+      "step": 45870
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.277022756005056e-07,
+      "loss": 2.9595,
+      "step": 45871
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.276548672566372e-07,
+      "loss": 1.2836,
+      "step": 45872
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.276074589127687e-07,
+      "loss": 1.243,
+      "step": 45873
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.275600505689001e-07,
+      "loss": 2.7725,
+      "step": 45874
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.275126422250317e-07,
+      "loss": 1.7223,
+      "step": 45875
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.274652338811632e-07,
+      "loss": 1.5167,
+      "step": 45876
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.274178255372946e-07,
+      "loss": 3.1125,
+      "step": 45877
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.273704171934261e-07,
+      "loss": 1.25,
+      "step": 45878
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.273230088495575e-07,
+      "loss": 1.7761,
+      "step": 45879
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.27275600505689e-07,
+      "loss": 1.194,
+      "step": 45880
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.272281921618205e-07,
+      "loss": 2.2184,
+      "step": 45881
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 8.271807838179519e-07,
+      "loss": 2.0099,
+      "step": 45882
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.271333754740834e-07,
+      "loss": 1.8206,
+      "step": 45883
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.27085967130215e-07,
+      "loss": 1.8878,
+      "step": 45884
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.270385587863464e-07,
+      "loss": 3.1117,
+      "step": 45885
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.269911504424779e-07,
+      "loss": 3.562,
+      "step": 45886
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.269437420986095e-07,
+      "loss": 2.7232,
+      "step": 45887
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.268963337547408e-07,
+      "loss": 1.6337,
+      "step": 45888
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.268489254108724e-07,
+      "loss": 1.9197,
+      "step": 45889
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.268015170670038e-07,
+      "loss": 1.5375,
+      "step": 45890
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.267541087231353e-07,
+      "loss": 1.6391,
+      "step": 45891
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.267067003792669e-07,
+      "loss": 1.7381,
+      "step": 45892
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.266592920353982e-07,
+      "loss": 2.0183,
+      "step": 45893
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.266118836915297e-07,
+      "loss": 2.1886,
+      "step": 45894
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.265644753476612e-07,
+      "loss": 2.8695,
+      "step": 45895
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.265170670037926e-07,
+      "loss": 2.0216,
+      "step": 45896
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.264696586599242e-07,
+      "loss": 2.1189,
+      "step": 45897
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.264222503160557e-07,
+      "loss": 1.7653,
+      "step": 45898
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.263748419721871e-07,
+      "loss": 2.0486,
+      "step": 45899
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.263274336283187e-07,
+      "loss": 1.4431,
+      "step": 45900
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.2628002528445e-07,
+      "loss": 1.9308,
+      "step": 45901
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.262326169405816e-07,
+      "loss": 1.9814,
+      "step": 45902
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.261852085967131e-07,
+      "loss": 1.7066,
+      "step": 45903
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.261378002528444e-07,
+      "loss": 1.5721,
+      "step": 45904
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.26090391908976e-07,
+      "loss": 1.822,
+      "step": 45905
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.260429835651075e-07,
+      "loss": 0.9226,
+      "step": 45906
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.259955752212389e-07,
+      "loss": 1.7448,
+      "step": 45907
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.259481668773704e-07,
+      "loss": 3.3816,
+      "step": 45908
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.25900758533502e-07,
+      "loss": 1.9716,
+      "step": 45909
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.258533501896334e-07,
+      "loss": 2.3107,
+      "step": 45910
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.258059418457649e-07,
+      "loss": 2.0624,
+      "step": 45911
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.257585335018963e-07,
+      "loss": 3.1944,
+      "step": 45912
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.257111251580279e-07,
+      "loss": 2.3529,
+      "step": 45913
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.256637168141594e-07,
+      "loss": 1.8063,
+      "step": 45914
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.256163084702907e-07,
+      "loss": 2.0782,
+      "step": 45915
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.255689001264222e-07,
+      "loss": 1.9412,
+      "step": 45916
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.255214917825538e-07,
+      "loss": 2.08,
+      "step": 45917
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.254740834386852e-07,
+      "loss": 1.877,
+      "step": 45918
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.254266750948167e-07,
+      "loss": 3.1567,
+      "step": 45919
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.253792667509483e-07,
+      "loss": 1.5915,
+      "step": 45920
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.253318584070796e-07,
+      "loss": 2.6001,
+      "step": 45921
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.252844500632112e-07,
+      "loss": 0.8155,
+      "step": 45922
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.252370417193426e-07,
+      "loss": 1.4329,
+      "step": 45923
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.251896333754741e-07,
+      "loss": 1.6581,
+      "step": 45924
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.251422250316057e-07,
+      "loss": 0.9621,
+      "step": 45925
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.250948166877371e-07,
+      "loss": 2.7515,
+      "step": 45926
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.250474083438685e-07,
+      "loss": 3.0726,
+      "step": 45927
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.25e-07,
+      "loss": 1.4123,
+      "step": 45928
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.249525916561314e-07,
+      "loss": 1.0852,
+      "step": 45929
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.24905183312263e-07,
+      "loss": 1.9452,
+      "step": 45930
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.248577749683945e-07,
+      "loss": 0.9661,
+      "step": 45931
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.248103666245259e-07,
+      "loss": 3.1368,
+      "step": 45932
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.247629582806575e-07,
+      "loss": 4.8083,
+      "step": 45933
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.247155499367888e-07,
+      "loss": 2.2282,
+      "step": 45934
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.246681415929204e-07,
+      "loss": 2.5551,
+      "step": 45935
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.246207332490519e-07,
+      "loss": 0.3795,
+      "step": 45936
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.245733249051833e-07,
+      "loss": 1.7013,
+      "step": 45937
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.245259165613149e-07,
+      "loss": 1.4456,
+      "step": 45938
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.244785082174464e-07,
+      "loss": 1.3558,
+      "step": 45939
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.244310998735777e-07,
+      "loss": 3.5285,
+      "step": 45940
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.243836915297092e-07,
+      "loss": 1.4539,
+      "step": 45941
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.243362831858408e-07,
+      "loss": 0.887,
+      "step": 45942
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.242888748419722e-07,
+      "loss": 2.1513,
+      "step": 45943
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.242414664981037e-07,
+      "loss": 2.6793,
+      "step": 45944
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 8.241940581542351e-07,
+      "loss": 3.0209,
+      "step": 45945
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.241466498103667e-07,
+      "loss": 1.0893,
+      "step": 45946
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.240992414664982e-07,
+      "loss": 1.3345,
+      "step": 45947
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.240518331226296e-07,
+      "loss": 2.1935,
+      "step": 45948
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.240044247787611e-07,
+      "loss": 3.5821,
+      "step": 45949
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.239570164348927e-07,
+      "loss": 2.3422,
+      "step": 45950
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.23909608091024e-07,
+      "loss": 4.1762,
+      "step": 45951
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.238621997471555e-07,
+      "loss": 2.152,
+      "step": 45952
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.238147914032871e-07,
+      "loss": 2.5475,
+      "step": 45953
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.237673830594184e-07,
+      "loss": 1.6585,
+      "step": 45954
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.2371997471555e-07,
+      "loss": 3.0045,
+      "step": 45955
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.236725663716814e-07,
+      "loss": 1.2096,
+      "step": 45956
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.236251580278129e-07,
+      "loss": 1.8142,
+      "step": 45957
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.235777496839445e-07,
+      "loss": 2.0606,
+      "step": 45958
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.235303413400758e-07,
+      "loss": 1.6866,
+      "step": 45959
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.234829329962074e-07,
+      "loss": 0.4598,
+      "step": 45960
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.234355246523389e-07,
+      "loss": 1.4368,
+      "step": 45961
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.233881163084702e-07,
+      "loss": 1.9594,
+      "step": 45962
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.233407079646018e-07,
+      "loss": 2.1329,
+      "step": 45963
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.232932996207333e-07,
+      "loss": 1.6367,
+      "step": 45964
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.232458912768647e-07,
+      "loss": 2.5952,
+      "step": 45965
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.231984829329962e-07,
+      "loss": 2.1686,
+      "step": 45966
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.231510745891276e-07,
+      "loss": 0.8895,
+      "step": 45967
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.231036662452592e-07,
+      "loss": 2.207,
+      "step": 45968
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.230562579013907e-07,
+      "loss": 1.5507,
+      "step": 45969
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.230088495575221e-07,
+      "loss": 3.0891,
+      "step": 45970
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.229614412136537e-07,
+      "loss": 2.2689,
+      "step": 45971
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.229140328697852e-07,
+      "loss": 0.966,
+      "step": 45972
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.228666245259166e-07,
+      "loss": 1.6158,
+      "step": 45973
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.22819216182048e-07,
+      "loss": 1.6702,
+      "step": 45974
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.227718078381794e-07,
+      "loss": 1.1732,
+      "step": 45975
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.22724399494311e-07,
+      "loss": 2.2589,
+      "step": 45976
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.226769911504425e-07,
+      "loss": 1.3798,
+      "step": 45977
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.226295828065739e-07,
+      "loss": 2.3092,
+      "step": 45978
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.225821744627054e-07,
+      "loss": 2.2839,
+      "step": 45979
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.22534766118837e-07,
+      "loss": 0.9407,
+      "step": 45980
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.224873577749684e-07,
+      "loss": 1.7297,
+      "step": 45981
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.224399494310999e-07,
+      "loss": 1.472,
+      "step": 45982
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.223925410872315e-07,
+      "loss": 1.8305,
+      "step": 45983
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.223451327433629e-07,
+      "loss": 1.1893,
+      "step": 45984
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.222977243994944e-07,
+      "loss": 3.964,
+      "step": 45985
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.222503160556257e-07,
+      "loss": 1.9126,
+      "step": 45986
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.222029077117572e-07,
+      "loss": 2.2636,
+      "step": 45987
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.221554993678888e-07,
+      "loss": 1.6873,
+      "step": 45988
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.221080910240202e-07,
+      "loss": 1.7055,
+      "step": 45989
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.220606826801517e-07,
+      "loss": 1.8606,
+      "step": 45990
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.220132743362833e-07,
+      "loss": 1.9374,
+      "step": 45991
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.219658659924146e-07,
+      "loss": 2.4471,
+      "step": 45992
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.219184576485462e-07,
+      "loss": 4.2804,
+      "step": 45993
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.218710493046777e-07,
+      "loss": 1.5238,
+      "step": 45994
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.218236409608091e-07,
+      "loss": 0.4771,
+      "step": 45995
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.217762326169407e-07,
+      "loss": 2.3164,
+      "step": 45996
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.21728824273072e-07,
+      "loss": 3.058,
+      "step": 45997
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.216814159292035e-07,
+      "loss": 1.2073,
+      "step": 45998
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.21634007585335e-07,
+      "loss": 2.8838,
+      "step": 45999
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.215865992414664e-07,
+      "loss": 3.1661,
+      "step": 46000
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.21539190897598e-07,
+      "loss": 0.6324,
+      "step": 46001
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.214917825537295e-07,
+      "loss": 1.2168,
+      "step": 46002
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.214443742098609e-07,
+      "loss": 2.3878,
+      "step": 46003
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.213969658659925e-07,
+      "loss": 1.0302,
+      "step": 46004
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.21349557522124e-07,
+      "loss": 2.3893,
+      "step": 46005
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.213021491782554e-07,
+      "loss": 2.73,
+      "step": 46006
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.212547408343869e-07,
+      "loss": 3.6367,
+      "step": 46007
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.212073324905183e-07,
+      "loss": 3.1678,
+      "step": 46008
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 8.211599241466498e-07,
+      "loss": 2.1944,
+      "step": 46009
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.211125158027813e-07,
+      "loss": 1.9461,
+      "step": 46010
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.210651074589127e-07,
+      "loss": 1.6725,
+      "step": 46011
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.210176991150442e-07,
+      "loss": 2.3654,
+      "step": 46012
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.209702907711758e-07,
+      "loss": 2.7198,
+      "step": 46013
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.209228824273072e-07,
+      "loss": 3.678,
+      "step": 46014
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.208754740834387e-07,
+      "loss": 1.5335,
+      "step": 46015
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.208280657395703e-07,
+      "loss": 2.5882,
+      "step": 46016
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.207806573957016e-07,
+      "loss": 1.3614,
+      "step": 46017
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.207332490518332e-07,
+      "loss": 1.4253,
+      "step": 46018
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.206858407079646e-07,
+      "loss": 0.7504,
+      "step": 46019
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.20638432364096e-07,
+      "loss": 1.613,
+      "step": 46020
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.205910240202276e-07,
+      "loss": 2.8481,
+      "step": 46021
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.20543615676359e-07,
+      "loss": 2.5991,
+      "step": 46022
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.204962073324905e-07,
+      "loss": 1.1579,
+      "step": 46023
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.20448798988622e-07,
+      "loss": 2.4997,
+      "step": 46024
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.204013906447534e-07,
+      "loss": 2.2406,
+      "step": 46025
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.20353982300885e-07,
+      "loss": 2.9535,
+      "step": 46026
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.203065739570165e-07,
+      "loss": 1.9674,
+      "step": 46027
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.202591656131479e-07,
+      "loss": 2.2396,
+      "step": 46028
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.202117572692795e-07,
+      "loss": 2.6787,
+      "step": 46029
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.201643489254108e-07,
+      "loss": 1.8879,
+      "step": 46030
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.201169405815424e-07,
+      "loss": 2.0868,
+      "step": 46031
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.200695322376739e-07,
+      "loss": 0.9289,
+      "step": 46032
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.200221238938052e-07,
+      "loss": 2.9512,
+      "step": 46033
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.199747155499368e-07,
+      "loss": 2.1323,
+      "step": 46034
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.199273072060683e-07,
+      "loss": 1.2238,
+      "step": 46035
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.198798988621997e-07,
+      "loss": 1.1848,
+      "step": 46036
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.198324905183312e-07,
+      "loss": 1.1532,
+      "step": 46037
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.197850821744628e-07,
+      "loss": 1.1741,
+      "step": 46038
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.197376738305942e-07,
+      "loss": 1.8374,
+      "step": 46039
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.196902654867257e-07,
+      "loss": 2.2841,
+      "step": 46040
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.196428571428571e-07,
+      "loss": 2.4645,
+      "step": 46041
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.195954487989887e-07,
+      "loss": 2.558,
+      "step": 46042
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.195480404551202e-07,
+      "loss": 2.1546,
+      "step": 46043
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.195006321112515e-07,
+      "loss": 2.1619,
+      "step": 46044
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.19453223767383e-07,
+      "loss": 0.9203,
+      "step": 46045
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.194058154235146e-07,
+      "loss": 1.5926,
+      "step": 46046
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.19358407079646e-07,
+      "loss": 2.9844,
+      "step": 46047
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.193109987357775e-07,
+      "loss": 2.8481,
+      "step": 46048
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.192635903919091e-07,
+      "loss": 1.7604,
+      "step": 46049
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.192161820480404e-07,
+      "loss": 1.5689,
+      "step": 46050
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.19168773704172e-07,
+      "loss": 1.366,
+      "step": 46051
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.191213653603034e-07,
+      "loss": 3.8275,
+      "step": 46052
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.190739570164349e-07,
+      "loss": 1.5092,
+      "step": 46053
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.190265486725665e-07,
+      "loss": 1.9405,
+      "step": 46054
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.189791403286979e-07,
+      "loss": 1.142,
+      "step": 46055
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.189317319848293e-07,
+      "loss": 3.0614,
+      "step": 46056
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.188843236409608e-07,
+      "loss": 2.9742,
+      "step": 46057
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.188369152970922e-07,
+      "loss": 2.562,
+      "step": 46058
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.187895069532238e-07,
+      "loss": 1.6691,
+      "step": 46059
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.187420986093553e-07,
+      "loss": 2.2146,
+      "step": 46060
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.186946902654867e-07,
+      "loss": 1.2716,
+      "step": 46061
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.186472819216183e-07,
+      "loss": 0.669,
+      "step": 46062
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.185998735777496e-07,
+      "loss": 0.8984,
+      "step": 46063
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.185524652338812e-07,
+      "loss": 3.2828,
+      "step": 46064
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.185050568900127e-07,
+      "loss": 0.9776,
+      "step": 46065
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.184576485461441e-07,
+      "loss": 2.4685,
+      "step": 46066
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.184102402022757e-07,
+      "loss": 1.4566,
+      "step": 46067
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.183628318584071e-07,
+      "loss": 2.9034,
+      "step": 46068
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.183154235145385e-07,
+      "loss": 1.1751,
+      "step": 46069
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.1826801517067e-07,
+      "loss": 2.7142,
+      "step": 46070
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.182206068268016e-07,
+      "loss": 2.1501,
+      "step": 46071
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 8.18173198482933e-07,
+      "loss": 1.5433,
+      "step": 46072
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.181257901390645e-07,
+      "loss": 0.8429,
+      "step": 46073
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.180783817951959e-07,
+      "loss": 2.2946,
+      "step": 46074
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.180309734513275e-07,
+      "loss": 1.5157,
+      "step": 46075
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.17983565107459e-07,
+      "loss": 2.0504,
+      "step": 46076
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.179361567635904e-07,
+      "loss": 1.6609,
+      "step": 46077
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.178887484197219e-07,
+      "loss": 1.7466,
+      "step": 46078
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.178413400758535e-07,
+      "loss": 2.025,
+      "step": 46079
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.177939317319848e-07,
+      "loss": 2.2701,
+      "step": 46080
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.177465233881163e-07,
+      "loss": 1.6381,
+      "step": 46081
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.176991150442479e-07,
+      "loss": 1.1827,
+      "step": 46082
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.176517067003792e-07,
+      "loss": 3.7708,
+      "step": 46083
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.176042983565108e-07,
+      "loss": 4.1123,
+      "step": 46084
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.175568900126422e-07,
+      "loss": 1.706,
+      "step": 46085
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.175094816687737e-07,
+      "loss": 3.6499,
+      "step": 46086
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.174620733249053e-07,
+      "loss": 2.3131,
+      "step": 46087
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.174146649810366e-07,
+      "loss": 3.1229,
+      "step": 46088
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.173672566371682e-07,
+      "loss": 2.9139,
+      "step": 46089
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.173198482932997e-07,
+      "loss": 3.3772,
+      "step": 46090
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.17272439949431e-07,
+      "loss": 1.0626,
+      "step": 46091
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.172250316055626e-07,
+      "loss": 1.9383,
+      "step": 46092
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.171776232616941e-07,
+      "loss": 2.077,
+      "step": 46093
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.171302149178255e-07,
+      "loss": 1.244,
+      "step": 46094
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.17082806573957e-07,
+      "loss": 1.7323,
+      "step": 46095
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.170353982300884e-07,
+      "loss": 0.8658,
+      "step": 46096
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.1698798988622e-07,
+      "loss": 1.6688,
+      "step": 46097
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.169405815423515e-07,
+      "loss": 0.4708,
+      "step": 46098
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.168931731984829e-07,
+      "loss": 2.301,
+      "step": 46099
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.168457648546145e-07,
+      "loss": 1.9814,
+      "step": 46100
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.16798356510746e-07,
+      "loss": 2.3136,
+      "step": 46101
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.167509481668774e-07,
+      "loss": 2.5254,
+      "step": 46102
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.167035398230088e-07,
+      "loss": 2.0786,
+      "step": 46103
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.166561314791404e-07,
+      "loss": 1.2536,
+      "step": 46104
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.166087231352718e-07,
+      "loss": 1.7334,
+      "step": 46105
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.165613147914033e-07,
+      "loss": 1.6497,
+      "step": 46106
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.165139064475347e-07,
+      "loss": 2.6555,
+      "step": 46107
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.164664981036662e-07,
+      "loss": 2.677,
+      "step": 46108
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.164190897597978e-07,
+      "loss": 3.3863,
+      "step": 46109
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.163716814159292e-07,
+      "loss": 1.8346,
+      "step": 46110
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.163242730720607e-07,
+      "loss": 1.3743,
+      "step": 46111
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.162768647281923e-07,
+      "loss": 2.1401,
+      "step": 46112
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.162294563843237e-07,
+      "loss": 2.9527,
+      "step": 46113
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.161820480404552e-07,
+      "loss": 1.2055,
+      "step": 46114
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.161346396965866e-07,
+      "loss": 1.2191,
+      "step": 46115
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.16087231352718e-07,
+      "loss": 1.1568,
+      "step": 46116
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.160398230088496e-07,
+      "loss": 2.0095,
+      "step": 46117
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.15992414664981e-07,
+      "loss": 1.6734,
+      "step": 46118
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.159450063211125e-07,
+      "loss": 1.448,
+      "step": 46119
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.158975979772441e-07,
+      "loss": 2.3519,
+      "step": 46120
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.158501896333754e-07,
+      "loss": 1.4783,
+      "step": 46121
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.15802781289507e-07,
+      "loss": 1.8268,
+      "step": 46122
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.157553729456385e-07,
+      "loss": 2.5434,
+      "step": 46123
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.157079646017699e-07,
+      "loss": 1.6164,
+      "step": 46124
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.156605562579015e-07,
+      "loss": 2.752,
+      "step": 46125
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.15613147914033e-07,
+      "loss": 1.4401,
+      "step": 46126
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.155657395701643e-07,
+      "loss": 3.2607,
+      "step": 46127
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.155183312262958e-07,
+      "loss": 1.0958,
+      "step": 46128
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.154709228824272e-07,
+      "loss": 1.1903,
+      "step": 46129
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.154235145385588e-07,
+      "loss": 2.8581,
+      "step": 46130
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.153761061946903e-07,
+      "loss": 3.0918,
+      "step": 46131
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.153286978508217e-07,
+      "loss": 0.9481,
+      "step": 46132
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.152812895069533e-07,
+      "loss": 2.8626,
+      "step": 46133
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.152338811630848e-07,
+      "loss": 2.1312,
+      "step": 46134
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 8.151864728192162e-07,
+      "loss": 1.0364,
+      "step": 46135
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.151390644753477e-07,
+      "loss": 1.5235,
+      "step": 46136
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.150916561314793e-07,
+      "loss": 1.8313,
+      "step": 46137
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.150442477876106e-07,
+      "loss": 2.1006,
+      "step": 46138
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.149968394437421e-07,
+      "loss": 1.9569,
+      "step": 46139
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.149494310998735e-07,
+      "loss": 0.5056,
+      "step": 46140
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.14902022756005e-07,
+      "loss": 1.7782,
+      "step": 46141
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.148546144121366e-07,
+      "loss": 4.0141,
+      "step": 46142
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.14807206068268e-07,
+      "loss": 1.6006,
+      "step": 46143
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.147597977243995e-07,
+      "loss": 1.5446,
+      "step": 46144
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.147123893805311e-07,
+      "loss": 1.6794,
+      "step": 46145
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.146649810366625e-07,
+      "loss": 1.4731,
+      "step": 46146
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.14617572692794e-07,
+      "loss": 2.3935,
+      "step": 46147
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.145701643489255e-07,
+      "loss": 2.2048,
+      "step": 46148
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.145227560050568e-07,
+      "loss": 1.9553,
+      "step": 46149
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.144753476611884e-07,
+      "loss": 1.4318,
+      "step": 46150
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.144279393173198e-07,
+      "loss": 1.4738,
+      "step": 46151
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.143805309734513e-07,
+      "loss": 1.9121,
+      "step": 46152
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.143331226295829e-07,
+      "loss": 2.2873,
+      "step": 46153
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.142857142857142e-07,
+      "loss": 2.0468,
+      "step": 46154
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.142383059418458e-07,
+      "loss": 2.8468,
+      "step": 46155
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.141908975979773e-07,
+      "loss": 1.7802,
+      "step": 46156
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.141434892541087e-07,
+      "loss": 1.855,
+      "step": 46157
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.140960809102403e-07,
+      "loss": 2.7101,
+      "step": 46158
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.140486725663718e-07,
+      "loss": 1.467,
+      "step": 46159
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.140012642225032e-07,
+      "loss": 2.1378,
+      "step": 46160
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.139538558786346e-07,
+      "loss": 2.505,
+      "step": 46161
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.13906447534766e-07,
+      "loss": 1.7135,
+      "step": 46162
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.138590391908976e-07,
+      "loss": 2.8163,
+      "step": 46163
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.138116308470291e-07,
+      "loss": 2.0557,
+      "step": 46164
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.137642225031605e-07,
+      "loss": 2.1457,
+      "step": 46165
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.13716814159292e-07,
+      "loss": 2.251,
+      "step": 46166
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.136694058154236e-07,
+      "loss": 1.619,
+      "step": 46167
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.13621997471555e-07,
+      "loss": 0.9921,
+      "step": 46168
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.135745891276865e-07,
+      "loss": 1.799,
+      "step": 46169
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.135271807838181e-07,
+      "loss": 1.7092,
+      "step": 46170
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.134797724399495e-07,
+      "loss": 1.9796,
+      "step": 46171
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.13432364096081e-07,
+      "loss": 1.6708,
+      "step": 46172
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.133849557522123e-07,
+      "loss": 3.4496,
+      "step": 46173
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.133375474083438e-07,
+      "loss": 1.4762,
+      "step": 46174
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.132901390644754e-07,
+      "loss": 1.657,
+      "step": 46175
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.132427307206068e-07,
+      "loss": 1.3603,
+      "step": 46176
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.131953223767383e-07,
+      "loss": 1.4473,
+      "step": 46177
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.131479140328699e-07,
+      "loss": 2.7296,
+      "step": 46178
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.131005056890012e-07,
+      "loss": 4.0405,
+      "step": 46179
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.130530973451328e-07,
+      "loss": 1.9605,
+      "step": 46180
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.130056890012643e-07,
+      "loss": 1.4537,
+      "step": 46181
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.129582806573957e-07,
+      "loss": 2.3761,
+      "step": 46182
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.129108723135273e-07,
+      "loss": 1.4046,
+      "step": 46183
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.128634639696587e-07,
+      "loss": 0.8362,
+      "step": 46184
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.128160556257901e-07,
+      "loss": 1.6793,
+      "step": 46185
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.127686472819216e-07,
+      "loss": 1.2922,
+      "step": 46186
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.12721238938053e-07,
+      "loss": 0.7157,
+      "step": 46187
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.126738305941846e-07,
+      "loss": 1.8865,
+      "step": 46188
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.126264222503161e-07,
+      "loss": 2.0694,
+      "step": 46189
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.125790139064475e-07,
+      "loss": 1.6215,
+      "step": 46190
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.125316055625791e-07,
+      "loss": 0.94,
+      "step": 46191
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.124841972187106e-07,
+      "loss": 2.6626,
+      "step": 46192
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.12436788874842e-07,
+      "loss": 2.2576,
+      "step": 46193
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.123893805309735e-07,
+      "loss": 2.787,
+      "step": 46194
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.123419721871049e-07,
+      "loss": 1.2594,
+      "step": 46195
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.122945638432365e-07,
+      "loss": 2.8774,
+      "step": 46196
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.122471554993679e-07,
+      "loss": 1.5547,
+      "step": 46197
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.121997471554993e-07,
+      "loss": 3.5269,
+      "step": 46198
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 8.121523388116308e-07,
+      "loss": 2.0568,
+      "step": 46199
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.121049304677624e-07,
+      "loss": 2.0419,
+      "step": 46200
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.120575221238938e-07,
+      "loss": 1.2891,
+      "step": 46201
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.120101137800253e-07,
+      "loss": 3.2771,
+      "step": 46202
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.119627054361569e-07,
+      "loss": 2.6032,
+      "step": 46203
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.119152970922883e-07,
+      "loss": 2.4719,
+      "step": 46204
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.118678887484198e-07,
+      "loss": 2.4121,
+      "step": 46205
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.118204804045512e-07,
+      "loss": 0.4405,
+      "step": 46206
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.117730720606827e-07,
+      "loss": 2.7209,
+      "step": 46207
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.117256637168143e-07,
+      "loss": 1.9427,
+      "step": 46208
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.116782553729456e-07,
+      "loss": 0.3262,
+      "step": 46209
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.116308470290771e-07,
+      "loss": 3.3108,
+      "step": 46210
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.115834386852087e-07,
+      "loss": 2.0631,
+      "step": 46211
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.1153603034134e-07,
+      "loss": 1.062,
+      "step": 46212
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.114886219974716e-07,
+      "loss": 0.9419,
+      "step": 46213
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.114412136536031e-07,
+      "loss": 2.2457,
+      "step": 46214
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.113938053097345e-07,
+      "loss": 2.8083,
+      "step": 46215
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.113463969658661e-07,
+      "loss": 1.5789,
+      "step": 46216
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.112989886219975e-07,
+      "loss": 1.48,
+      "step": 46217
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.11251580278129e-07,
+      "loss": 2.6162,
+      "step": 46218
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.112041719342605e-07,
+      "loss": 0.8978,
+      "step": 46219
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.111567635903918e-07,
+      "loss": 2.3029,
+      "step": 46220
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.111093552465234e-07,
+      "loss": 0.2523,
+      "step": 46221
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.110619469026549e-07,
+      "loss": 1.9726,
+      "step": 46222
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.110145385587863e-07,
+      "loss": 2.457,
+      "step": 46223
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.109671302149179e-07,
+      "loss": 2.1282,
+      "step": 46224
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.109197218710494e-07,
+      "loss": 1.7586,
+      "step": 46225
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.108723135271808e-07,
+      "loss": 0.5745,
+      "step": 46226
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.108249051833123e-07,
+      "loss": 1.9008,
+      "step": 46227
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.107774968394437e-07,
+      "loss": 1.4568,
+      "step": 46228
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.107300884955753e-07,
+      "loss": 2.1189,
+      "step": 46229
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.106826801517068e-07,
+      "loss": 2.1248,
+      "step": 46230
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.106352718078381e-07,
+      "loss": 1.8543,
+      "step": 46231
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.105878634639696e-07,
+      "loss": 0.9429,
+      "step": 46232
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.105404551201012e-07,
+      "loss": 1.351,
+      "step": 46233
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.104930467762326e-07,
+      "loss": 1.4384,
+      "step": 46234
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.104456384323641e-07,
+      "loss": 3.2977,
+      "step": 46235
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.103982300884957e-07,
+      "loss": 2.8781,
+      "step": 46236
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.10350821744627e-07,
+      "loss": 1.8797,
+      "step": 46237
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.103034134007586e-07,
+      "loss": 1.0628,
+      "step": 46238
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.1025600505689e-07,
+      "loss": 2.5182,
+      "step": 46239
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.102085967130215e-07,
+      "loss": 2.3381,
+      "step": 46240
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.101611883691531e-07,
+      "loss": 1.5458,
+      "step": 46241
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.101137800252845e-07,
+      "loss": 1.2833,
+      "step": 46242
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.10066371681416e-07,
+      "loss": 0.5694,
+      "step": 46243
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.100189633375474e-07,
+      "loss": 0.5892,
+      "step": 46244
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.099715549936788e-07,
+      "loss": 1.0611,
+      "step": 46245
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.099241466498104e-07,
+      "loss": 2.3458,
+      "step": 46246
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.098767383059419e-07,
+      "loss": 1.9272,
+      "step": 46247
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.098293299620733e-07,
+      "loss": 2.1116,
+      "step": 46248
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.097819216182049e-07,
+      "loss": 2.2191,
+      "step": 46249
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.097345132743362e-07,
+      "loss": 3.6407,
+      "step": 46250
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.096871049304678e-07,
+      "loss": 2.3752,
+      "step": 46251
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.096396965865993e-07,
+      "loss": 2.047,
+      "step": 46252
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.095922882427307e-07,
+      "loss": 2.3964,
+      "step": 46253
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.095448798988623e-07,
+      "loss": 0.8078,
+      "step": 46254
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.094974715549937e-07,
+      "loss": 2.0573,
+      "step": 46255
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.094500632111251e-07,
+      "loss": 1.6654,
+      "step": 46256
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.094026548672566e-07,
+      "loss": 2.4404,
+      "step": 46257
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.093552465233882e-07,
+      "loss": 2.9969,
+      "step": 46258
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.093078381795196e-07,
+      "loss": 0.8286,
+      "step": 46259
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.092604298356511e-07,
+      "loss": 2.1709,
+      "step": 46260
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.092130214917825e-07,
+      "loss": 1.8838,
+      "step": 46261
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 8.091656131479141e-07,
+      "loss": 2.8258,
+      "step": 46262
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.091182048040456e-07,
+      "loss": 3.5833,
+      "step": 46263
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.09070796460177e-07,
+      "loss": 1.0625,
+      "step": 46264
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.090233881163085e-07,
+      "loss": 2.7269,
+      "step": 46265
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.089759797724401e-07,
+      "loss": 0.9192,
+      "step": 46266
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.089285714285714e-07,
+      "loss": 1.5845,
+      "step": 46267
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.088811630847029e-07,
+      "loss": 1.9583,
+      "step": 46268
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.088337547408345e-07,
+      "loss": 1.9348,
+      "step": 46269
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.087863463969658e-07,
+      "loss": 0.8793,
+      "step": 46270
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.087389380530974e-07,
+      "loss": 1.6795,
+      "step": 46271
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.086915297092288e-07,
+      "loss": 1.8588,
+      "step": 46272
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.086441213653603e-07,
+      "loss": 2.0655,
+      "step": 46273
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.085967130214919e-07,
+      "loss": 1.9567,
+      "step": 46274
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.085493046776233e-07,
+      "loss": 3.1976,
+      "step": 46275
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.085018963337548e-07,
+      "loss": 3.1599,
+      "step": 46276
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.084544879898863e-07,
+      "loss": 2.9882,
+      "step": 46277
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.084070796460176e-07,
+      "loss": 1.5269,
+      "step": 46278
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.083596713021492e-07,
+      "loss": 2.4404,
+      "step": 46279
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.083122629582807e-07,
+      "loss": 1.8574,
+      "step": 46280
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.082648546144121e-07,
+      "loss": 1.1124,
+      "step": 46281
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.082174462705437e-07,
+      "loss": 1.2015,
+      "step": 46282
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.08170037926675e-07,
+      "loss": 3.182,
+      "step": 46283
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.081226295828066e-07,
+      "loss": 1.1117,
+      "step": 46284
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.080752212389381e-07,
+      "loss": 1.6307,
+      "step": 46285
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.080278128950695e-07,
+      "loss": 0.8934,
+      "step": 46286
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.079804045512011e-07,
+      "loss": 2.2561,
+      "step": 46287
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.079329962073326e-07,
+      "loss": 1.7808,
+      "step": 46288
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.07885587863464e-07,
+      "loss": 2.0176,
+      "step": 46289
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.078381795195954e-07,
+      "loss": 1.9598,
+      "step": 46290
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.07790771175727e-07,
+      "loss": 1.1103,
+      "step": 46291
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.077433628318584e-07,
+      "loss": 1.6679,
+      "step": 46292
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.076959544879899e-07,
+      "loss": 2.1672,
+      "step": 46293
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.076485461441213e-07,
+      "loss": 0.9372,
+      "step": 46294
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.076011378002528e-07,
+      "loss": 3.2361,
+      "step": 46295
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.075537294563844e-07,
+      "loss": 3.4269,
+      "step": 46296
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.075063211125158e-07,
+      "loss": 0.5504,
+      "step": 46297
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.074589127686473e-07,
+      "loss": 2.0178,
+      "step": 46298
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.074115044247789e-07,
+      "loss": 2.7165,
+      "step": 46299
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.073640960809103e-07,
+      "loss": 1.3848,
+      "step": 46300
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.073166877370418e-07,
+      "loss": 1.6803,
+      "step": 46301
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.072692793931732e-07,
+      "loss": 2.434,
+      "step": 46302
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.072218710493046e-07,
+      "loss": 0.4862,
+      "step": 46303
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.071744627054362e-07,
+      "loss": 2.7295,
+      "step": 46304
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.071270543615676e-07,
+      "loss": 2.0835,
+      "step": 46305
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.070796460176991e-07,
+      "loss": 1.5957,
+      "step": 46306
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.070322376738307e-07,
+      "loss": 2.2862,
+      "step": 46307
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.06984829329962e-07,
+      "loss": 4.5387,
+      "step": 46308
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.069374209860936e-07,
+      "loss": 2.4768,
+      "step": 46309
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.068900126422251e-07,
+      "loss": 1.1169,
+      "step": 46310
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.068426042983565e-07,
+      "loss": 1.7719,
+      "step": 46311
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.067951959544881e-07,
+      "loss": 2.0049,
+      "step": 46312
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.067477876106196e-07,
+      "loss": 3.0206,
+      "step": 46313
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.067003792667509e-07,
+      "loss": 3.7893,
+      "step": 46314
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.066529709228824e-07,
+      "loss": 2.3903,
+      "step": 46315
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.066055625790138e-07,
+      "loss": 2.8137,
+      "step": 46316
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.065581542351454e-07,
+      "loss": 2.6495,
+      "step": 46317
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.065107458912769e-07,
+      "loss": 1.8598,
+      "step": 46318
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.064633375474083e-07,
+      "loss": 0.8971,
+      "step": 46319
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.064159292035399e-07,
+      "loss": 2.9206,
+      "step": 46320
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.063685208596714e-07,
+      "loss": 2.1085,
+      "step": 46321
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.063211125158028e-07,
+      "loss": 1.7572,
+      "step": 46322
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.062737041719343e-07,
+      "loss": 2.4025,
+      "step": 46323
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.062262958280659e-07,
+      "loss": 2.7447,
+      "step": 46324
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 8.061788874841972e-07,
+      "loss": 2.0535,
+      "step": 46325
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.061314791403287e-07,
+      "loss": 1.6317,
+      "step": 46326
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.060840707964601e-07,
+      "loss": 3.7878,
+      "step": 46327
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.060366624525916e-07,
+      "loss": 1.9478,
+      "step": 46328
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.059892541087232e-07,
+      "loss": 2.568,
+      "step": 46329
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.059418457648546e-07,
+      "loss": 1.804,
+      "step": 46330
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.058944374209861e-07,
+      "loss": 2.5947,
+      "step": 46331
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.058470290771177e-07,
+      "loss": 1.0428,
+      "step": 46332
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.057996207332491e-07,
+      "loss": 2.3768,
+      "step": 46333
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.057522123893806e-07,
+      "loss": 2.1297,
+      "step": 46334
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.057048040455121e-07,
+      "loss": 2.1433,
+      "step": 46335
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.056573957016435e-07,
+      "loss": 2.4657,
+      "step": 46336
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.05609987357775e-07,
+      "loss": 2.21,
+      "step": 46337
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.055625790139064e-07,
+      "loss": 2.9,
+      "step": 46338
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.055151706700379e-07,
+      "loss": 1.9806,
+      "step": 46339
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.054677623261695e-07,
+      "loss": 2.405,
+      "step": 46340
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.054203539823008e-07,
+      "loss": 1.8081,
+      "step": 46341
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.053729456384324e-07,
+      "loss": 1.1695,
+      "step": 46342
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.053255372945639e-07,
+      "loss": 1.8706,
+      "step": 46343
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.052781289506953e-07,
+      "loss": 1.7913,
+      "step": 46344
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.052307206068269e-07,
+      "loss": 2.8183,
+      "step": 46345
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.051833122629584e-07,
+      "loss": 1.3938,
+      "step": 46346
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.051359039190898e-07,
+      "loss": 2.1723,
+      "step": 46347
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.050884955752213e-07,
+      "loss": 0.5155,
+      "step": 46348
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.050410872313526e-07,
+      "loss": 3.3228,
+      "step": 46349
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.049936788874842e-07,
+      "loss": 1.1413,
+      "step": 46350
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.049462705436157e-07,
+      "loss": 1.9527,
+      "step": 46351
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.048988621997471e-07,
+      "loss": 1.7383,
+      "step": 46352
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.048514538558787e-07,
+      "loss": 1.2567,
+      "step": 46353
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.048040455120102e-07,
+      "loss": 3.758,
+      "step": 46354
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.047566371681416e-07,
+      "loss": 1.5985,
+      "step": 46355
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.047092288242731e-07,
+      "loss": 2.9229,
+      "step": 46356
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.046618204804047e-07,
+      "loss": 1.0741,
+      "step": 46357
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.046144121365361e-07,
+      "loss": 1.9078,
+      "step": 46358
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.045670037926676e-07,
+      "loss": 1.4402,
+      "step": 46359
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.045195954487989e-07,
+      "loss": 3.0976,
+      "step": 46360
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.044721871049304e-07,
+      "loss": 1.8348,
+      "step": 46361
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.04424778761062e-07,
+      "loss": 2.0849,
+      "step": 46362
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.043773704171934e-07,
+      "loss": 2.6366,
+      "step": 46363
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.043299620733249e-07,
+      "loss": 2.642,
+      "step": 46364
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.042825537294565e-07,
+      "loss": 1.9968,
+      "step": 46365
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.042351453855878e-07,
+      "loss": 1.7132,
+      "step": 46366
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.041877370417194e-07,
+      "loss": 1.5542,
+      "step": 46367
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.041403286978509e-07,
+      "loss": 1.8184,
+      "step": 46368
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.040929203539823e-07,
+      "loss": 1.6467,
+      "step": 46369
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.040455120101139e-07,
+      "loss": 0.6849,
+      "step": 46370
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.039981036662453e-07,
+      "loss": 1.4828,
+      "step": 46371
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.039506953223767e-07,
+      "loss": 2.2505,
+      "step": 46372
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.039032869785082e-07,
+      "loss": 2.4075,
+      "step": 46373
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.038558786346396e-07,
+      "loss": 1.7675,
+      "step": 46374
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.038084702907712e-07,
+      "loss": 0.9313,
+      "step": 46375
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.037610619469027e-07,
+      "loss": 2.3284,
+      "step": 46376
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.037136536030341e-07,
+      "loss": 1.0217,
+      "step": 46377
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.036662452591657e-07,
+      "loss": 0.7658,
+      "step": 46378
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.036188369152972e-07,
+      "loss": 3.1356,
+      "step": 46379
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.035714285714286e-07,
+      "loss": 2.3853,
+      "step": 46380
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.035240202275601e-07,
+      "loss": 1.6748,
+      "step": 46381
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.034766118836915e-07,
+      "loss": 2.7133,
+      "step": 46382
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.034292035398231e-07,
+      "loss": 1.7164,
+      "step": 46383
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.033817951959545e-07,
+      "loss": 1.0878,
+      "step": 46384
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.033343868520859e-07,
+      "loss": 0.5774,
+      "step": 46385
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.032869785082174e-07,
+      "loss": 2.2186,
+      "step": 46386
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.03239570164349e-07,
+      "loss": 1.6484,
+      "step": 46387
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.031921618204804e-07,
+      "loss": 2.4299,
+      "step": 46388
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 8.031447534766119e-07,
+      "loss": 1.7994,
+      "step": 46389
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.030973451327435e-07,
+      "loss": 0.5492,
+      "step": 46390
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.030499367888749e-07,
+      "loss": 1.5585,
+      "step": 46391
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.030025284450064e-07,
+      "loss": 2.4473,
+      "step": 46392
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.029551201011378e-07,
+      "loss": 1.5359,
+      "step": 46393
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.029077117572693e-07,
+      "loss": 1.2136,
+      "step": 46394
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.028603034134009e-07,
+      "loss": 1.9199,
+      "step": 46395
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.028128950695322e-07,
+      "loss": 1.7091,
+      "step": 46396
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.027654867256637e-07,
+      "loss": 1.8264,
+      "step": 46397
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.027180783817953e-07,
+      "loss": 2.2324,
+      "step": 46398
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.026706700379266e-07,
+      "loss": 1.985,
+      "step": 46399
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.026232616940582e-07,
+      "loss": 2.1586,
+      "step": 46400
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.025758533501897e-07,
+      "loss": 2.3185,
+      "step": 46401
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.025284450063211e-07,
+      "loss": 2.389,
+      "step": 46402
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.024810366624527e-07,
+      "loss": 1.8351,
+      "step": 46403
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.024336283185841e-07,
+      "loss": 1.5209,
+      "step": 46404
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.023862199747156e-07,
+      "loss": 2.2262,
+      "step": 46405
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.023388116308471e-07,
+      "loss": 1.5112,
+      "step": 46406
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.022914032869784e-07,
+      "loss": 1.99,
+      "step": 46407
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.0224399494311e-07,
+      "loss": 2.0571,
+      "step": 46408
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.021965865992415e-07,
+      "loss": 3.2655,
+      "step": 46409
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.021491782553729e-07,
+      "loss": 1.8495,
+      "step": 46410
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.021017699115045e-07,
+      "loss": 1.6959,
+      "step": 46411
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.020543615676358e-07,
+      "loss": 1.6076,
+      "step": 46412
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.020069532237674e-07,
+      "loss": 1.1126,
+      "step": 46413
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.019595448798989e-07,
+      "loss": 1.4162,
+      "step": 46414
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.019121365360303e-07,
+      "loss": 0.327,
+      "step": 46415
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.018647281921619e-07,
+      "loss": 2.3006,
+      "step": 46416
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.018173198482934e-07,
+      "loss": 1.8234,
+      "step": 46417
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.017699115044247e-07,
+      "loss": 1.8111,
+      "step": 46418
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.017225031605562e-07,
+      "loss": 1.3575,
+      "step": 46419
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.016750948166878e-07,
+      "loss": 2.0003,
+      "step": 46420
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.016276864728192e-07,
+      "loss": 1.7669,
+      "step": 46421
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.015802781289507e-07,
+      "loss": 1.2694,
+      "step": 46422
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.015328697850821e-07,
+      "loss": 1.7454,
+      "step": 46423
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.014854614412137e-07,
+      "loss": 2.6079,
+      "step": 46424
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.014380530973452e-07,
+      "loss": 1.1453,
+      "step": 46425
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.013906447534766e-07,
+      "loss": 2.4261,
+      "step": 46426
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.013432364096081e-07,
+      "loss": 1.8125,
+      "step": 46427
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.012958280657397e-07,
+      "loss": 3.811,
+      "step": 46428
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.012484197218711e-07,
+      "loss": 2.512,
+      "step": 46429
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.012010113780026e-07,
+      "loss": 2.8782,
+      "step": 46430
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.011536030341341e-07,
+      "loss": 2.9242,
+      "step": 46431
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.011061946902654e-07,
+      "loss": 2.1056,
+      "step": 46432
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.01058786346397e-07,
+      "loss": 3.7494,
+      "step": 46433
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.010113780025284e-07,
+      "loss": 1.4467,
+      "step": 46434
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.009639696586599e-07,
+      "loss": 2.3149,
+      "step": 46435
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.009165613147915e-07,
+      "loss": 2.6004,
+      "step": 46436
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.008691529709228e-07,
+      "loss": 2.1277,
+      "step": 46437
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.008217446270544e-07,
+      "loss": 0.2092,
+      "step": 46438
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.007743362831859e-07,
+      "loss": 2.2196,
+      "step": 46439
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.007269279393173e-07,
+      "loss": 1.4565,
+      "step": 46440
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.006795195954489e-07,
+      "loss": 1.5315,
+      "step": 46441
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.006321112515804e-07,
+      "loss": 1.7847,
+      "step": 46442
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.005847029077117e-07,
+      "loss": 2.6402,
+      "step": 46443
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.005372945638432e-07,
+      "loss": 1.3904,
+      "step": 46444
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.004898862199746e-07,
+      "loss": 2.2922,
+      "step": 46445
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.004424778761062e-07,
+      "loss": 2.693,
+      "step": 46446
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.003950695322377e-07,
+      "loss": 1.6876,
+      "step": 46447
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.003476611883691e-07,
+      "loss": 1.9468,
+      "step": 46448
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.003002528445007e-07,
+      "loss": 2.9138,
+      "step": 46449
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.002528445006322e-07,
+      "loss": 2.3786,
+      "step": 46450
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.002054361567636e-07,
+      "loss": 2.5843,
+      "step": 46451
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 8.001580278128951e-07,
+      "loss": 1.623,
+      "step": 46452
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 8.001106194690267e-07,
+      "loss": 1.6844,
+      "step": 46453
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 8.00063211125158e-07,
+      "loss": 2.4179,
+      "step": 46454
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 8.000158027812895e-07,
+      "loss": 2.7625,
+      "step": 46455
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.999683944374209e-07,
+      "loss": 0.9727,
+      "step": 46456
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.999209860935524e-07,
+      "loss": 1.2041,
+      "step": 46457
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.99873577749684e-07,
+      "loss": 2.1985,
+      "step": 46458
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.998261694058154e-07,
+      "loss": 2.4067,
+      "step": 46459
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.997787610619469e-07,
+      "loss": 1.7136,
+      "step": 46460
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.997313527180785e-07,
+      "loss": 1.608,
+      "step": 46461
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.996839443742099e-07,
+      "loss": 2.0109,
+      "step": 46462
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.996365360303414e-07,
+      "loss": 3.3867,
+      "step": 46463
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.995891276864729e-07,
+      "loss": 1.1994,
+      "step": 46464
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.995417193426042e-07,
+      "loss": 1.2476,
+      "step": 46465
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.994943109987358e-07,
+      "loss": 2.4899,
+      "step": 46466
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.994469026548672e-07,
+      "loss": 1.6605,
+      "step": 46467
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.993994943109987e-07,
+      "loss": 2.6683,
+      "step": 46468
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.993520859671303e-07,
+      "loss": 0.7645,
+      "step": 46469
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.993046776232616e-07,
+      "loss": 0.8225,
+      "step": 46470
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.992572692793932e-07,
+      "loss": 1.2155,
+      "step": 46471
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.992098609355247e-07,
+      "loss": 2.4646,
+      "step": 46472
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.991624525916561e-07,
+      "loss": 2.2889,
+      "step": 46473
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.991150442477877e-07,
+      "loss": 3.5024,
+      "step": 46474
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.990676359039192e-07,
+      "loss": 1.3001,
+      "step": 46475
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.990202275600506e-07,
+      "loss": 2.8162,
+      "step": 46476
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.98972819216182e-07,
+      "loss": 0.8664,
+      "step": 46477
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.989254108723134e-07,
+      "loss": 1.7976,
+      "step": 46478
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.98878002528445e-07,
+      "loss": 2.2956,
+      "step": 46479
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.988305941845765e-07,
+      "loss": 2.1063,
+      "step": 46480
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.987831858407079e-07,
+      "loss": 1.3355,
+      "step": 46481
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.987357774968395e-07,
+      "loss": 1.645,
+      "step": 46482
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.98688369152971e-07,
+      "loss": 2.1748,
+      "step": 46483
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.986409608091024e-07,
+      "loss": 1.5418,
+      "step": 46484
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.985935524652339e-07,
+      "loss": 3.4087,
+      "step": 46485
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.985461441213655e-07,
+      "loss": 2.9276,
+      "step": 46486
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.984987357774969e-07,
+      "loss": 1.7188,
+      "step": 46487
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.984513274336284e-07,
+      "loss": 2.3932,
+      "step": 46488
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.984039190897597e-07,
+      "loss": 2.3179,
+      "step": 46489
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.983565107458912e-07,
+      "loss": 2.7004,
+      "step": 46490
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.983091024020228e-07,
+      "loss": 1.9158,
+      "step": 46491
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.982616940581542e-07,
+      "loss": 2.5376,
+      "step": 46492
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.982142857142857e-07,
+      "loss": 2.0231,
+      "step": 46493
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.981668773704173e-07,
+      "loss": 3.2089,
+      "step": 46494
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.981194690265487e-07,
+      "loss": 1.3652,
+      "step": 46495
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.980720606826802e-07,
+      "loss": 3.1161,
+      "step": 46496
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.980246523388117e-07,
+      "loss": 2.2581,
+      "step": 46497
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.979772439949431e-07,
+      "loss": 2.1775,
+      "step": 46498
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.979298356510747e-07,
+      "loss": 1.3825,
+      "step": 46499
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.97882427307206e-07,
+      "loss": 1.4369,
+      "step": 46500
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.978350189633375e-07,
+      "loss": 2.551,
+      "step": 46501
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.977876106194691e-07,
+      "loss": 1.3808,
+      "step": 46502
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.977402022756004e-07,
+      "loss": 1.3732,
+      "step": 46503
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.97692793931732e-07,
+      "loss": 5.0815,
+      "step": 46504
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.976453855878635e-07,
+      "loss": 2.4256,
+      "step": 46505
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.975979772439949e-07,
+      "loss": 0.7933,
+      "step": 46506
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.975505689001265e-07,
+      "loss": 2.1469,
+      "step": 46507
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.97503160556258e-07,
+      "loss": 2.2814,
+      "step": 46508
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.974557522123894e-07,
+      "loss": 1.9046,
+      "step": 46509
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.974083438685209e-07,
+      "loss": 1.9043,
+      "step": 46510
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.973609355246523e-07,
+      "loss": 1.0497,
+      "step": 46511
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.973135271807838e-07,
+      "loss": 2.4743,
+      "step": 46512
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.972661188369153e-07,
+      "loss": 2.3971,
+      "step": 46513
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.972187104930467e-07,
+      "loss": 3.1865,
+      "step": 46514
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 7.971713021491782e-07,
+      "loss": 3.7123,
+      "step": 46515
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.971238938053098e-07,
+      "loss": 2.1722,
+      "step": 46516
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.970764854614412e-07,
+      "loss": 0.6893,
+      "step": 46517
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.970290771175727e-07,
+      "loss": 1.5292,
+      "step": 46518
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.969816687737043e-07,
+      "loss": 2.7255,
+      "step": 46519
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.969342604298357e-07,
+      "loss": 2.394,
+      "step": 46520
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.968868520859672e-07,
+      "loss": 1.1407,
+      "step": 46521
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.968394437420986e-07,
+      "loss": 1.4914,
+      "step": 46522
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.9679203539823e-07,
+      "loss": 2.8232,
+      "step": 46523
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.967446270543617e-07,
+      "loss": 3.3403,
+      "step": 46524
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.96697218710493e-07,
+      "loss": 1.4096,
+      "step": 46525
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.966498103666245e-07,
+      "loss": 1.0058,
+      "step": 46526
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.966024020227561e-07,
+      "loss": 2.5961,
+      "step": 46527
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.965549936788874e-07,
+      "loss": 1.3874,
+      "step": 46528
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.96507585335019e-07,
+      "loss": 3.2204,
+      "step": 46529
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.964601769911505e-07,
+      "loss": 1.5474,
+      "step": 46530
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.964127686472819e-07,
+      "loss": 0.9294,
+      "step": 46531
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.963653603034135e-07,
+      "loss": 0.928,
+      "step": 46532
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.963179519595449e-07,
+      "loss": 3.1039,
+      "step": 46533
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.962705436156764e-07,
+      "loss": 3.0791,
+      "step": 46534
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.962231352718079e-07,
+      "loss": 1.8418,
+      "step": 46535
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.961757269279392e-07,
+      "loss": 0.7245,
+      "step": 46536
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.961283185840708e-07,
+      "loss": 2.4992,
+      "step": 46537
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.960809102402023e-07,
+      "loss": 1.0287,
+      "step": 46538
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.960335018963337e-07,
+      "loss": 1.2734,
+      "step": 46539
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.959860935524653e-07,
+      "loss": 2.1719,
+      "step": 46540
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.959386852085968e-07,
+      "loss": 0.869,
+      "step": 46541
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.958912768647282e-07,
+      "loss": 3.2845,
+      "step": 46542
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.958438685208597e-07,
+      "loss": 1.465,
+      "step": 46543
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.957964601769911e-07,
+      "loss": 2.6083,
+      "step": 46544
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.957490518331227e-07,
+      "loss": 1.7317,
+      "step": 46545
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.957016434892542e-07,
+      "loss": 0.9073,
+      "step": 46546
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.956542351453855e-07,
+      "loss": 2.1589,
+      "step": 46547
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.95606826801517e-07,
+      "loss": 0.9251,
+      "step": 46548
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.955594184576486e-07,
+      "loss": 1.2354,
+      "step": 46549
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.9551201011378e-07,
+      "loss": 2.6531,
+      "step": 46550
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.954646017699115e-07,
+      "loss": 2.136,
+      "step": 46551
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.954171934260431e-07,
+      "loss": 1.6229,
+      "step": 46552
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.953697850821745e-07,
+      "loss": 0.6422,
+      "step": 46553
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.95322376738306e-07,
+      "loss": 2.2139,
+      "step": 46554
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.952749683944374e-07,
+      "loss": 2.847,
+      "step": 46555
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.952275600505689e-07,
+      "loss": 2.8169,
+      "step": 46556
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.951801517067005e-07,
+      "loss": 0.9465,
+      "step": 46557
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.951327433628319e-07,
+      "loss": 1.5126,
+      "step": 46558
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.950853350189633e-07,
+      "loss": 1.3048,
+      "step": 46559
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.950379266750949e-07,
+      "loss": 1.1777,
+      "step": 46560
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.949905183312262e-07,
+      "loss": 1.1166,
+      "step": 46561
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.949431099873578e-07,
+      "loss": 2.6678,
+      "step": 46562
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.948957016434893e-07,
+      "loss": 2.8551,
+      "step": 46563
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.948482932996207e-07,
+      "loss": 2.1607,
+      "step": 46564
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.948008849557523e-07,
+      "loss": 1.0051,
+      "step": 46565
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.947534766118837e-07,
+      "loss": 1.5613,
+      "step": 46566
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.947060682680152e-07,
+      "loss": 2.3912,
+      "step": 46567
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.946586599241467e-07,
+      "loss": 1.8579,
+      "step": 46568
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.946112515802781e-07,
+      "loss": 2.2793,
+      "step": 46569
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.945638432364097e-07,
+      "loss": 1.2174,
+      "step": 46570
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.945164348925411e-07,
+      "loss": 1.0158,
+      "step": 46571
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.944690265486725e-07,
+      "loss": 3.5618,
+      "step": 46572
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.944216182048041e-07,
+      "loss": 3.3099,
+      "step": 46573
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.943742098609356e-07,
+      "loss": 2.0734,
+      "step": 46574
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.94326801517067e-07,
+      "loss": 2.2677,
+      "step": 46575
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.942793931731985e-07,
+      "loss": 2.033,
+      "step": 46576
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.942319848293299e-07,
+      "loss": 1.9499,
+      "step": 46577
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.941845764854615e-07,
+      "loss": 2.1257,
+      "step": 46578
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 7.94137168141593e-07,
+      "loss": 2.2746,
+      "step": 46579
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.940897597977244e-07,
+      "loss": 1.5826,
+      "step": 46580
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.940423514538559e-07,
+      "loss": 2.457,
+      "step": 46581
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.939949431099875e-07,
+      "loss": 0.9653,
+      "step": 46582
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.939475347661188e-07,
+      "loss": 2.0937,
+      "step": 46583
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.939001264222503e-07,
+      "loss": 1.4578,
+      "step": 46584
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.938527180783819e-07,
+      "loss": 1.856,
+      "step": 46585
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.938053097345132e-07,
+      "loss": 2.4227,
+      "step": 46586
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.937579013906448e-07,
+      "loss": 3.7703,
+      "step": 46587
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.937104930467762e-07,
+      "loss": 3.1418,
+      "step": 46588
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.936630847029077e-07,
+      "loss": 2.1505,
+      "step": 46589
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.936156763590393e-07,
+      "loss": 3.012,
+      "step": 46590
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.935682680151707e-07,
+      "loss": 1.6157,
+      "step": 46591
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.935208596713022e-07,
+      "loss": 1.7183,
+      "step": 46592
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.934734513274337e-07,
+      "loss": 1.9398,
+      "step": 46593
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.93426042983565e-07,
+      "loss": 0.5203,
+      "step": 46594
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.933786346396966e-07,
+      "loss": 0.8428,
+      "step": 46595
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.933312262958281e-07,
+      "loss": 1.4188,
+      "step": 46596
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.932838179519595e-07,
+      "loss": 2.8283,
+      "step": 46597
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.932364096080911e-07,
+      "loss": 1.1796,
+      "step": 46598
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.931890012642224e-07,
+      "loss": 1.4192,
+      "step": 46599
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.93141592920354e-07,
+      "loss": 0.8705,
+      "step": 46600
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.930941845764855e-07,
+      "loss": 1.9525,
+      "step": 46601
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.930467762326169e-07,
+      "loss": 2.5493,
+      "step": 46602
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.929993678887485e-07,
+      "loss": 2.3845,
+      "step": 46603
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.9295195954488e-07,
+      "loss": 1.1503,
+      "step": 46604
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.929045512010114e-07,
+      "loss": 2.5511,
+      "step": 46605
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.928571428571428e-07,
+      "loss": 2.6095,
+      "step": 46606
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.928097345132744e-07,
+      "loss": 1.9211,
+      "step": 46607
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.927623261694058e-07,
+      "loss": 0.565,
+      "step": 46608
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.927149178255373e-07,
+      "loss": 3.3581,
+      "step": 46609
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.926675094816687e-07,
+      "loss": 1.1982,
+      "step": 46610
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.926201011378003e-07,
+      "loss": 1.9822,
+      "step": 46611
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.925726927939318e-07,
+      "loss": 1.3327,
+      "step": 46612
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.925252844500632e-07,
+      "loss": 1.2081,
+      "step": 46613
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.924778761061947e-07,
+      "loss": 0.5205,
+      "step": 46614
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.924304677623263e-07,
+      "loss": 0.9952,
+      "step": 46615
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.923830594184577e-07,
+      "loss": 2.0546,
+      "step": 46616
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.923356510745892e-07,
+      "loss": 2.2785,
+      "step": 46617
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.922882427307208e-07,
+      "loss": 2.5557,
+      "step": 46618
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.92240834386852e-07,
+      "loss": 1.7386,
+      "step": 46619
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.921934260429836e-07,
+      "loss": 0.9878,
+      "step": 46620
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.92146017699115e-07,
+      "loss": 2.0788,
+      "step": 46621
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.920986093552465e-07,
+      "loss": 1.818,
+      "step": 46622
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.920512010113781e-07,
+      "loss": 1.6762,
+      "step": 46623
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.920037926675095e-07,
+      "loss": 3.2128,
+      "step": 46624
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.91956384323641e-07,
+      "loss": 1.2246,
+      "step": 46625
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.919089759797725e-07,
+      "loss": 2.5715,
+      "step": 46626
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.918615676359039e-07,
+      "loss": 3.7434,
+      "step": 46627
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.918141592920355e-07,
+      "loss": 1.9503,
+      "step": 46628
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.91766750948167e-07,
+      "loss": 1.7261,
+      "step": 46629
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.917193426042983e-07,
+      "loss": 2.3524,
+      "step": 46630
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.916719342604299e-07,
+      "loss": 1.8527,
+      "step": 46631
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.916245259165612e-07,
+      "loss": 0.5359,
+      "step": 46632
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.915771175726928e-07,
+      "loss": 2.2429,
+      "step": 46633
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.915297092288243e-07,
+      "loss": 1.5241,
+      "step": 46634
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.914823008849557e-07,
+      "loss": 2.4587,
+      "step": 46635
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.914348925410873e-07,
+      "loss": 2.0232,
+      "step": 46636
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.913874841972188e-07,
+      "loss": 3.2794,
+      "step": 46637
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.913400758533502e-07,
+      "loss": 1.6604,
+      "step": 46638
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.912926675094817e-07,
+      "loss": 1.2205,
+      "step": 46639
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.912452591656133e-07,
+      "loss": 2.5788,
+      "step": 46640
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.911978508217446e-07,
+      "loss": 1.6949,
+      "step": 46641
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 7.911504424778761e-07,
+      "loss": 1.7063,
+      "step": 46642
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.911030341340075e-07,
+      "loss": 0.522,
+      "step": 46643
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.91055625790139e-07,
+      "loss": 3.6077,
+      "step": 46644
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.910082174462706e-07,
+      "loss": 2.0089,
+      "step": 46645
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.90960809102402e-07,
+      "loss": 1.1234,
+      "step": 46646
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.909134007585335e-07,
+      "loss": 2.3004,
+      "step": 46647
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.908659924146651e-07,
+      "loss": 1.7742,
+      "step": 46648
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.908185840707965e-07,
+      "loss": 2.2788,
+      "step": 46649
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.90771175726928e-07,
+      "loss": 1.7112,
+      "step": 46650
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.907237673830595e-07,
+      "loss": 1.5397,
+      "step": 46651
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.906763590391908e-07,
+      "loss": 1.4402,
+      "step": 46652
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.906289506953224e-07,
+      "loss": 1.4173,
+      "step": 46653
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.905815423514538e-07,
+      "loss": 1.8765,
+      "step": 46654
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.905341340075853e-07,
+      "loss": 1.2983,
+      "step": 46655
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.904867256637169e-07,
+      "loss": 2.6201,
+      "step": 46656
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.904393173198482e-07,
+      "loss": 2.6264,
+      "step": 46657
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.903919089759798e-07,
+      "loss": 2.7226,
+      "step": 46658
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.903445006321113e-07,
+      "loss": 1.3844,
+      "step": 46659
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.902970922882427e-07,
+      "loss": 2.0927,
+      "step": 46660
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.902496839443743e-07,
+      "loss": 1.2952,
+      "step": 46661
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.902022756005058e-07,
+      "loss": 1.4377,
+      "step": 46662
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.901548672566372e-07,
+      "loss": 0.8822,
+      "step": 46663
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.901074589127686e-07,
+      "loss": 3.241,
+      "step": 46664
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.900600505689e-07,
+      "loss": 1.0371,
+      "step": 46665
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.900126422250316e-07,
+      "loss": 1.7383,
+      "step": 46666
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.899652338811631e-07,
+      "loss": 2.4058,
+      "step": 46667
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.899178255372945e-07,
+      "loss": 1.6228,
+      "step": 46668
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.898704171934261e-07,
+      "loss": 1.7578,
+      "step": 46669
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.898230088495576e-07,
+      "loss": 1.7735,
+      "step": 46670
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.89775600505689e-07,
+      "loss": 3.4987,
+      "step": 46671
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.897281921618205e-07,
+      "loss": 1.0525,
+      "step": 46672
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.896807838179521e-07,
+      "loss": 2.6264,
+      "step": 46673
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.896333754740835e-07,
+      "loss": 1.8898,
+      "step": 46674
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.89585967130215e-07,
+      "loss": 0.7091,
+      "step": 46675
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.895385587863463e-07,
+      "loss": 1.8158,
+      "step": 46676
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.894911504424778e-07,
+      "loss": 0.9559,
+      "step": 46677
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.894437420986094e-07,
+      "loss": 1.7247,
+      "step": 46678
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.893963337547408e-07,
+      "loss": 1.4458,
+      "step": 46679
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.893489254108723e-07,
+      "loss": 2.0586,
+      "step": 46680
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.893015170670039e-07,
+      "loss": 2.092,
+      "step": 46681
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.892541087231353e-07,
+      "loss": 2.1628,
+      "step": 46682
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.892067003792668e-07,
+      "loss": 0.9902,
+      "step": 46683
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.891592920353983e-07,
+      "loss": 2.0127,
+      "step": 46684
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.891118836915297e-07,
+      "loss": 2.5635,
+      "step": 46685
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.890644753476613e-07,
+      "loss": 2.7328,
+      "step": 46686
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.890170670037927e-07,
+      "loss": 1.4045,
+      "step": 46687
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.889696586599241e-07,
+      "loss": 3.9791,
+      "step": 46688
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.889222503160557e-07,
+      "loss": 2.0633,
+      "step": 46689
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.88874841972187e-07,
+      "loss": 1.7844,
+      "step": 46690
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.888274336283186e-07,
+      "loss": 1.684,
+      "step": 46691
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.887800252844501e-07,
+      "loss": 2.0124,
+      "step": 46692
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.887326169405815e-07,
+      "loss": 1.95,
+      "step": 46693
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.886852085967131e-07,
+      "loss": 1.977,
+      "step": 46694
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.886378002528446e-07,
+      "loss": 1.7421,
+      "step": 46695
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.88590391908976e-07,
+      "loss": 2.2382,
+      "step": 46696
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.885429835651075e-07,
+      "loss": 2.4268,
+      "step": 46697
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.884955752212389e-07,
+      "loss": 0.8846,
+      "step": 46698
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.884481668773705e-07,
+      "loss": 1.9436,
+      "step": 46699
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.884007585335019e-07,
+      "loss": 1.8415,
+      "step": 46700
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.883533501896333e-07,
+      "loss": 1.6829,
+      "step": 46701
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.883059418457649e-07,
+      "loss": 2.0304,
+      "step": 46702
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.882585335018964e-07,
+      "loss": 3.0858,
+      "step": 46703
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.882111251580278e-07,
+      "loss": 3.2079,
+      "step": 46704
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 7.881637168141593e-07,
+      "loss": 2.0767,
+      "step": 46705
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.881163084702909e-07,
+      "loss": 2.4372,
+      "step": 46706
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.880689001264223e-07,
+      "loss": 2.4667,
+      "step": 46707
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.880214917825538e-07,
+      "loss": 1.8417,
+      "step": 46708
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.879740834386852e-07,
+      "loss": 3.0475,
+      "step": 46709
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.879266750948167e-07,
+      "loss": 1.904,
+      "step": 46710
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.878792667509483e-07,
+      "loss": 2.5924,
+      "step": 46711
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.878318584070796e-07,
+      "loss": 2.0939,
+      "step": 46712
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.877844500632111e-07,
+      "loss": 1.3445,
+      "step": 46713
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.877370417193427e-07,
+      "loss": 2.5625,
+      "step": 46714
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.87689633375474e-07,
+      "loss": 1.4777,
+      "step": 46715
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.876422250316056e-07,
+      "loss": 3.6188,
+      "step": 46716
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.875948166877371e-07,
+      "loss": 2.3548,
+      "step": 46717
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.875474083438685e-07,
+      "loss": 2.5693,
+      "step": 46718
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.875000000000001e-07,
+      "loss": 1.5723,
+      "step": 46719
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.874525916561315e-07,
+      "loss": 1.4146,
+      "step": 46720
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.87405183312263e-07,
+      "loss": 2.7469,
+      "step": 46721
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.873577749683945e-07,
+      "loss": 2.3508,
+      "step": 46722
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.873103666245258e-07,
+      "loss": 2.7388,
+      "step": 46723
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.872629582806574e-07,
+      "loss": 1.6483,
+      "step": 46724
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.872155499367889e-07,
+      "loss": 3.2656,
+      "step": 46725
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.871681415929203e-07,
+      "loss": 3.0857,
+      "step": 46726
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.871207332490519e-07,
+      "loss": 1.7172,
+      "step": 46727
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.870733249051834e-07,
+      "loss": 2.7635,
+      "step": 46728
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.870259165613148e-07,
+      "loss": 1.9025,
+      "step": 46729
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.869785082174463e-07,
+      "loss": 0.6106,
+      "step": 46730
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.869310998735777e-07,
+      "loss": 3.4116,
+      "step": 46731
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.868836915297093e-07,
+      "loss": 1.4181,
+      "step": 46732
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.868362831858408e-07,
+      "loss": 1.5267,
+      "step": 46733
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.867888748419721e-07,
+      "loss": 2.3729,
+      "step": 46734
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.867414664981036e-07,
+      "loss": 1.3762,
+      "step": 46735
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.866940581542352e-07,
+      "loss": 1.2069,
+      "step": 46736
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.866466498103666e-07,
+      "loss": 1.0854,
+      "step": 46737
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.865992414664981e-07,
+      "loss": 3.5456,
+      "step": 46738
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.865518331226297e-07,
+      "loss": 1.2813,
+      "step": 46739
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.865044247787611e-07,
+      "loss": 2.4316,
+      "step": 46740
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.864570164348926e-07,
+      "loss": 1.6469,
+      "step": 46741
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.86409608091024e-07,
+      "loss": 3.2082,
+      "step": 46742
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.863621997471555e-07,
+      "loss": 3.197,
+      "step": 46743
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.863147914032871e-07,
+      "loss": 1.9996,
+      "step": 46744
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.862673830594185e-07,
+      "loss": 1.6692,
+      "step": 46745
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.8621997471555e-07,
+      "loss": 2.4079,
+      "step": 46746
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.861725663716815e-07,
+      "loss": 2.7099,
+      "step": 46747
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.861251580278128e-07,
+      "loss": 2.3401,
+      "step": 46748
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.860777496839444e-07,
+      "loss": 2.8138,
+      "step": 46749
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.860303413400759e-07,
+      "loss": 3.7357,
+      "step": 46750
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.859829329962073e-07,
+      "loss": 1.4108,
+      "step": 46751
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.859355246523389e-07,
+      "loss": 1.8936,
+      "step": 46752
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.858881163084703e-07,
+      "loss": 1.3364,
+      "step": 46753
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.858407079646018e-07,
+      "loss": 3.2827,
+      "step": 46754
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.857932996207333e-07,
+      "loss": 1.4627,
+      "step": 46755
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.857458912768647e-07,
+      "loss": 2.2389,
+      "step": 46756
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.856984829329963e-07,
+      "loss": 0.4842,
+      "step": 46757
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.856510745891277e-07,
+      "loss": 1.4936,
+      "step": 46758
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.856036662452591e-07,
+      "loss": 2.5043,
+      "step": 46759
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.855562579013907e-07,
+      "loss": 2.6905,
+      "step": 46760
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.855088495575222e-07,
+      "loss": 2.1294,
+      "step": 46761
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.854614412136536e-07,
+      "loss": 1.7682,
+      "step": 46762
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.854140328697851e-07,
+      "loss": 1.5363,
+      "step": 46763
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.853666245259165e-07,
+      "loss": 1.0288,
+      "step": 46764
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.853192161820481e-07,
+      "loss": 3.5755,
+      "step": 46765
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.852718078381796e-07,
+      "loss": 2.1589,
+      "step": 46766
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.85224399494311e-07,
+      "loss": 2.0349,
+      "step": 46767
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.851769911504425e-07,
+      "loss": 2.8225,
+      "step": 46768
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 7.851295828065741e-07,
+      "loss": 1.4939,
+      "step": 46769
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.850821744627054e-07,
+      "loss": 1.6663,
+      "step": 46770
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.850347661188369e-07,
+      "loss": 1.9935,
+      "step": 46771
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.849873577749685e-07,
+      "loss": 2.5072,
+      "step": 46772
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.849399494310999e-07,
+      "loss": 1.9178,
+      "step": 46773
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.848925410872314e-07,
+      "loss": 2.0174,
+      "step": 46774
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.848451327433628e-07,
+      "loss": 2.4866,
+      "step": 46775
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.847977243994943e-07,
+      "loss": 2.6833,
+      "step": 46776
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.847503160556259e-07,
+      "loss": 1.7464,
+      "step": 46777
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.847029077117573e-07,
+      "loss": 0.7996,
+      "step": 46778
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.846554993678888e-07,
+      "loss": 2.104,
+      "step": 46779
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.846080910240204e-07,
+      "loss": 1.6204,
+      "step": 46780
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.845606826801516e-07,
+      "loss": 1.8839,
+      "step": 46781
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.845132743362832e-07,
+      "loss": 1.8887,
+      "step": 46782
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.844658659924147e-07,
+      "loss": 1.5583,
+      "step": 46783
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.844184576485461e-07,
+      "loss": 1.9001,
+      "step": 46784
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.843710493046777e-07,
+      "loss": 1.3216,
+      "step": 46785
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.84323640960809e-07,
+      "loss": 1.9503,
+      "step": 46786
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.842762326169406e-07,
+      "loss": 1.0283,
+      "step": 46787
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.842288242730721e-07,
+      "loss": 3.1968,
+      "step": 46788
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.841814159292035e-07,
+      "loss": 1.2157,
+      "step": 46789
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.841340075853351e-07,
+      "loss": 1.9991,
+      "step": 46790
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.840865992414666e-07,
+      "loss": 3.9831,
+      "step": 46791
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.84039190897598e-07,
+      "loss": 2.948,
+      "step": 46792
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.839917825537294e-07,
+      "loss": 1.6159,
+      "step": 46793
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.83944374209861e-07,
+      "loss": 2.8679,
+      "step": 46794
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.838969658659924e-07,
+      "loss": 3.0381,
+      "step": 46795
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.838495575221239e-07,
+      "loss": 1.853,
+      "step": 46796
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.838021491782553e-07,
+      "loss": 2.3419,
+      "step": 46797
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.837547408343869e-07,
+      "loss": 0.4085,
+      "step": 46798
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.837073324905184e-07,
+      "loss": 0.6448,
+      "step": 46799
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.836599241466498e-07,
+      "loss": 2.3066,
+      "step": 46800
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.836125158027813e-07,
+      "loss": 2.041,
+      "step": 46801
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.835651074589129e-07,
+      "loss": 2.8832,
+      "step": 46802
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.835176991150443e-07,
+      "loss": 1.7979,
+      "step": 46803
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.834702907711758e-07,
+      "loss": 2.3807,
+      "step": 46804
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.834228824273074e-07,
+      "loss": 1.3767,
+      "step": 46805
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.833754740834386e-07,
+      "loss": 2.5435,
+      "step": 46806
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.833280657395702e-07,
+      "loss": 1.7599,
+      "step": 46807
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.832806573957016e-07,
+      "loss": 1.2392,
+      "step": 46808
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.832332490518331e-07,
+      "loss": 3.0604,
+      "step": 46809
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.831858407079647e-07,
+      "loss": 1.3456,
+      "step": 46810
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.831384323640961e-07,
+      "loss": 1.0613,
+      "step": 46811
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.830910240202276e-07,
+      "loss": 1.8197,
+      "step": 46812
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.830436156763591e-07,
+      "loss": 1.8474,
+      "step": 46813
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.829962073324905e-07,
+      "loss": 1.5871,
+      "step": 46814
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.829487989886221e-07,
+      "loss": 2.2226,
+      "step": 46815
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.829013906447536e-07,
+      "loss": 1.269,
+      "step": 46816
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.828539823008849e-07,
+      "loss": 1.2716,
+      "step": 46817
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.828065739570165e-07,
+      "loss": 2.4434,
+      "step": 46818
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.827591656131478e-07,
+      "loss": 1.7306,
+      "step": 46819
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.827117572692794e-07,
+      "loss": 2.0162,
+      "step": 46820
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.826643489254109e-07,
+      "loss": 2.3333,
+      "step": 46821
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.826169405815423e-07,
+      "loss": 0.7673,
+      "step": 46822
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.825695322376739e-07,
+      "loss": 0.8485,
+      "step": 46823
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.825221238938054e-07,
+      "loss": 3.4587,
+      "step": 46824
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.824747155499368e-07,
+      "loss": 1.6564,
+      "step": 46825
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.824273072060683e-07,
+      "loss": 1.1737,
+      "step": 46826
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.823798988621999e-07,
+      "loss": 1.6893,
+      "step": 46827
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.823324905183312e-07,
+      "loss": 0.9354,
+      "step": 46828
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.822850821744627e-07,
+      "loss": 0.2411,
+      "step": 46829
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.822376738305941e-07,
+      "loss": 2.3135,
+      "step": 46830
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.821902654867257e-07,
+      "loss": 3.6406,
+      "step": 46831
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 7.821428571428572e-07,
+      "loss": 1.9045,
+      "step": 46832
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.820954487989886e-07,
+      "loss": 2.5059,
+      "step": 46833
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.820480404551201e-07,
+      "loss": 2.7868,
+      "step": 46834
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.820006321112517e-07,
+      "loss": 1.2823,
+      "step": 46835
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.819532237673831e-07,
+      "loss": 3.2915,
+      "step": 46836
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.819058154235146e-07,
+      "loss": 2.6822,
+      "step": 46837
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.818584070796462e-07,
+      "loss": 2.3178,
+      "step": 46838
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.818109987357775e-07,
+      "loss": 2.3501,
+      "step": 46839
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.81763590391909e-07,
+      "loss": 2.1245,
+      "step": 46840
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.817161820480404e-07,
+      "loss": 0.853,
+      "step": 46841
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.816687737041719e-07,
+      "loss": 2.1826,
+      "step": 46842
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.816213653603035e-07,
+      "loss": 2.3434,
+      "step": 46843
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.815739570164349e-07,
+      "loss": 0.7916,
+      "step": 46844
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.815265486725664e-07,
+      "loss": 3.1998,
+      "step": 46845
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.814791403286979e-07,
+      "loss": 2.0396,
+      "step": 46846
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.814317319848293e-07,
+      "loss": 2.0378,
+      "step": 46847
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.813843236409609e-07,
+      "loss": 1.4391,
+      "step": 46848
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.813369152970924e-07,
+      "loss": 2.5237,
+      "step": 46849
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.812895069532238e-07,
+      "loss": 2.9474,
+      "step": 46850
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.812420986093554e-07,
+      "loss": 2.6548,
+      "step": 46851
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.811946902654866e-07,
+      "loss": 3.7525,
+      "step": 46852
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.811472819216182e-07,
+      "loss": 2.0043,
+      "step": 46853
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.810998735777497e-07,
+      "loss": 2.7481,
+      "step": 46854
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.810524652338811e-07,
+      "loss": 3.4286,
+      "step": 46855
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.810050568900127e-07,
+      "loss": 0.3808,
+      "step": 46856
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.809576485461442e-07,
+      "loss": 3.1054,
+      "step": 46857
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.809102402022756e-07,
+      "loss": 2.3468,
+      "step": 46858
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.808628318584071e-07,
+      "loss": 1.4089,
+      "step": 46859
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.808154235145385e-07,
+      "loss": 0.7709,
+      "step": 46860
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.807680151706701e-07,
+      "loss": 3.1207,
+      "step": 46861
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.807206068268016e-07,
+      "loss": 1.1446,
+      "step": 46862
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.806731984829329e-07,
+      "loss": 0.5137,
+      "step": 46863
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.806257901390644e-07,
+      "loss": 1.6151,
+      "step": 46864
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.80578381795196e-07,
+      "loss": 2.2551,
+      "step": 46865
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.805309734513274e-07,
+      "loss": 1.3893,
+      "step": 46866
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.804835651074589e-07,
+      "loss": 0.9624,
+      "step": 46867
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.804361567635905e-07,
+      "loss": 1.4373,
+      "step": 46868
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.803887484197219e-07,
+      "loss": 0.6311,
+      "step": 46869
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.803413400758534e-07,
+      "loss": 2.5256,
+      "step": 46870
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.802939317319848e-07,
+      "loss": 1.772,
+      "step": 46871
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.802465233881163e-07,
+      "loss": 2.9826,
+      "step": 46872
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.801991150442479e-07,
+      "loss": 2.0601,
+      "step": 46873
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.801517067003793e-07,
+      "loss": 1.4141,
+      "step": 46874
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.801042983565107e-07,
+      "loss": 2.1999,
+      "step": 46875
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.800568900126423e-07,
+      "loss": 1.0491,
+      "step": 46876
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.800094816687736e-07,
+      "loss": 2.6581,
+      "step": 46877
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.799620733249052e-07,
+      "loss": 1.4037,
+      "step": 46878
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.799146649810367e-07,
+      "loss": 0.7645,
+      "step": 46879
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.798672566371681e-07,
+      "loss": 1.249,
+      "step": 46880
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.798198482932997e-07,
+      "loss": 1.3434,
+      "step": 46881
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.797724399494311e-07,
+      "loss": 2.6256,
+      "step": 46882
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.797250316055626e-07,
+      "loss": 0.5372,
+      "step": 46883
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.796776232616941e-07,
+      "loss": 2.8462,
+      "step": 46884
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.796302149178255e-07,
+      "loss": 2.0352,
+      "step": 46885
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.795828065739571e-07,
+      "loss": 2.0397,
+      "step": 46886
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.795353982300885e-07,
+      "loss": 1.3154,
+      "step": 46887
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.794879898862199e-07,
+      "loss": 2.6065,
+      "step": 46888
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.794405815423515e-07,
+      "loss": 2.03,
+      "step": 46889
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.79393173198483e-07,
+      "loss": 1.4526,
+      "step": 46890
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.793457648546144e-07,
+      "loss": 2.2067,
+      "step": 46891
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.792983565107459e-07,
+      "loss": 1.2727,
+      "step": 46892
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.792509481668773e-07,
+      "loss": 2.4616,
+      "step": 46893
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.792035398230089e-07,
+      "loss": 2.4182,
+      "step": 46894
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 7.791561314791404e-07,
+      "loss": 1.9725,
+      "step": 46895
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.791087231352718e-07,
+      "loss": 2.8953,
+      "step": 46896
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.790613147914033e-07,
+      "loss": 3.345,
+      "step": 46897
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.790139064475349e-07,
+      "loss": 2.4861,
+      "step": 46898
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.789664981036662e-07,
+      "loss": 2.4369,
+      "step": 46899
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.789190897597977e-07,
+      "loss": 2.4886,
+      "step": 46900
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.788716814159293e-07,
+      "loss": 1.4845,
+      "step": 46901
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.788242730720607e-07,
+      "loss": 1.5463,
+      "step": 46902
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.787768647281922e-07,
+      "loss": 0.8553,
+      "step": 46903
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.787294563843236e-07,
+      "loss": 2.9745,
+      "step": 46904
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.786820480404551e-07,
+      "loss": 2.1626,
+      "step": 46905
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.786346396965867e-07,
+      "loss": 0.2967,
+      "step": 46906
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.785872313527181e-07,
+      "loss": 0.6527,
+      "step": 46907
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.785398230088496e-07,
+      "loss": 2.9069,
+      "step": 46908
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.784924146649812e-07,
+      "loss": 1.1994,
+      "step": 46909
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.784450063211124e-07,
+      "loss": 1.3604,
+      "step": 46910
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.78397597977244e-07,
+      "loss": 3.2133,
+      "step": 46911
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.783501896333755e-07,
+      "loss": 1.1512,
+      "step": 46912
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.783027812895069e-07,
+      "loss": 0.7623,
+      "step": 46913
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.782553729456385e-07,
+      "loss": 1.2806,
+      "step": 46914
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.782079646017698e-07,
+      "loss": 1.1385,
+      "step": 46915
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.781605562579014e-07,
+      "loss": 1.3901,
+      "step": 46916
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.781131479140329e-07,
+      "loss": 2.1402,
+      "step": 46917
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.780657395701643e-07,
+      "loss": 3.1205,
+      "step": 46918
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.780183312262959e-07,
+      "loss": 1.3279,
+      "step": 46919
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.779709228824274e-07,
+      "loss": 2.8211,
+      "step": 46920
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.779235145385588e-07,
+      "loss": 1.4957,
+      "step": 46921
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.778761061946902e-07,
+      "loss": 2.8296,
+      "step": 46922
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.778286978508218e-07,
+      "loss": 3.0598,
+      "step": 46923
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.777812895069532e-07,
+      "loss": 1.5441,
+      "step": 46924
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.777338811630847e-07,
+      "loss": 2.5143,
+      "step": 46925
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.776864728192161e-07,
+      "loss": 1.6296,
+      "step": 46926
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.776390644753477e-07,
+      "loss": 2.716,
+      "step": 46927
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.775916561314792e-07,
+      "loss": 1.3642,
+      "step": 46928
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.775442477876106e-07,
+      "loss": 1.9972,
+      "step": 46929
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.774968394437421e-07,
+      "loss": 0.5818,
+      "step": 46930
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.774494310998737e-07,
+      "loss": 1.3219,
+      "step": 46931
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.774020227560051e-07,
+      "loss": 2.1419,
+      "step": 46932
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.773546144121366e-07,
+      "loss": 1.4779,
+      "step": 46933
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.773072060682681e-07,
+      "loss": 2.5596,
+      "step": 46934
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.772597977243994e-07,
+      "loss": 1.8821,
+      "step": 46935
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.77212389380531e-07,
+      "loss": 3.0431,
+      "step": 46936
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.771649810366624e-07,
+      "loss": 2.5669,
+      "step": 46937
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.771175726927939e-07,
+      "loss": 2.2879,
+      "step": 46938
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.770701643489255e-07,
+      "loss": 3.9446,
+      "step": 46939
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.770227560050569e-07,
+      "loss": 3.2769,
+      "step": 46940
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.769753476611884e-07,
+      "loss": 3.9341,
+      "step": 46941
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.769279393173199e-07,
+      "loss": 3.0663,
+      "step": 46942
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.768805309734513e-07,
+      "loss": 0.6868,
+      "step": 46943
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.768331226295829e-07,
+      "loss": 1.0693,
+      "step": 46944
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.767857142857144e-07,
+      "loss": 2.497,
+      "step": 46945
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.767383059418457e-07,
+      "loss": 2.8537,
+      "step": 46946
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.766908975979773e-07,
+      "loss": 1.709,
+      "step": 46947
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.766434892541086e-07,
+      "loss": 2.6073,
+      "step": 46948
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.765960809102402e-07,
+      "loss": 1.1687,
+      "step": 46949
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.765486725663717e-07,
+      "loss": 3.2806,
+      "step": 46950
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.765012642225031e-07,
+      "loss": 2.8673,
+      "step": 46951
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.764538558786347e-07,
+      "loss": 2.597,
+      "step": 46952
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.764064475347662e-07,
+      "loss": 1.6793,
+      "step": 46953
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.763590391908976e-07,
+      "loss": 1.4852,
+      "step": 46954
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.763116308470291e-07,
+      "loss": 1.8868,
+      "step": 46955
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.762642225031607e-07,
+      "loss": 2.2788,
+      "step": 46956
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.76216814159292e-07,
+      "loss": 1.4424,
+      "step": 46957
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.761694058154235e-07,
+      "loss": 2.4345,
+      "step": 46958
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 7.761219974715549e-07,
+      "loss": 1.7051,
+      "step": 46959
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.760745891276865e-07,
+      "loss": 2.4416,
+      "step": 46960
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.76027180783818e-07,
+      "loss": 1.9852,
+      "step": 46961
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.759797724399494e-07,
+      "loss": 2.3073,
+      "step": 46962
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.759323640960809e-07,
+      "loss": 2.2582,
+      "step": 46963
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.758849557522125e-07,
+      "loss": 1.3998,
+      "step": 46964
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.758375474083439e-07,
+      "loss": 1.0982,
+      "step": 46965
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.757901390644754e-07,
+      "loss": 1.2927,
+      "step": 46966
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.75742730720607e-07,
+      "loss": 1.57,
+      "step": 46967
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.756953223767382e-07,
+      "loss": 2.931,
+      "step": 46968
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.756479140328698e-07,
+      "loss": 1.7481,
+      "step": 46969
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.756005056890012e-07,
+      "loss": 1.8142,
+      "step": 46970
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.755530973451327e-07,
+      "loss": 2.2067,
+      "step": 46971
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.755056890012643e-07,
+      "loss": 2.2285,
+      "step": 46972
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.754582806573957e-07,
+      "loss": 1.6098,
+      "step": 46973
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.754108723135272e-07,
+      "loss": 1.0522,
+      "step": 46974
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.753634639696587e-07,
+      "loss": 1.8839,
+      "step": 46975
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.753160556257901e-07,
+      "loss": 2.3911,
+      "step": 46976
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.752686472819217e-07,
+      "loss": 1.5411,
+      "step": 46977
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.752212389380532e-07,
+      "loss": 1.2985,
+      "step": 46978
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.751738305941846e-07,
+      "loss": 2.5069,
+      "step": 46979
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.751264222503162e-07,
+      "loss": 1.4245,
+      "step": 46980
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.750790139064474e-07,
+      "loss": 0.9188,
+      "step": 46981
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.75031605562579e-07,
+      "loss": 3.814,
+      "step": 46982
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.749841972187105e-07,
+      "loss": 1.2288,
+      "step": 46983
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.749367888748419e-07,
+      "loss": 2.6529,
+      "step": 46984
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.748893805309735e-07,
+      "loss": 1.1264,
+      "step": 46985
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.74841972187105e-07,
+      "loss": 1.4619,
+      "step": 46986
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.747945638432364e-07,
+      "loss": 2.7214,
+      "step": 46987
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.747471554993679e-07,
+      "loss": 2.7246,
+      "step": 46988
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.746997471554995e-07,
+      "loss": 1.4413,
+      "step": 46989
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.746523388116309e-07,
+      "loss": 0.6515,
+      "step": 46990
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.746049304677624e-07,
+      "loss": 2.3813,
+      "step": 46991
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.745575221238937e-07,
+      "loss": 3.1521,
+      "step": 46992
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.745101137800252e-07,
+      "loss": 2.5154,
+      "step": 46993
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.744627054361568e-07,
+      "loss": 1.9464,
+      "step": 46994
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.744152970922882e-07,
+      "loss": 3.4102,
+      "step": 46995
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.743678887484197e-07,
+      "loss": 1.3579,
+      "step": 46996
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.743204804045513e-07,
+      "loss": 1.4634,
+      "step": 46997
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.742730720606827e-07,
+      "loss": 3.5934,
+      "step": 46998
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.742256637168142e-07,
+      "loss": 1.6319,
+      "step": 46999
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.741782553729457e-07,
+      "loss": 0.9008,
+      "step": 47000
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.741308470290771e-07,
+      "loss": 1.8071,
+      "step": 47001
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.740834386852087e-07,
+      "loss": 1.9687,
+      "step": 47002
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.7403603034134e-07,
+      "loss": 1.6483,
+      "step": 47003
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.739886219974715e-07,
+      "loss": 2.4004,
+      "step": 47004
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.739412136536031e-07,
+      "loss": 1.6335,
+      "step": 47005
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.738938053097344e-07,
+      "loss": 1.7119,
+      "step": 47006
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.73846396965866e-07,
+      "loss": 1.5317,
+      "step": 47007
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.737989886219975e-07,
+      "loss": 1.3111,
+      "step": 47008
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.737515802781289e-07,
+      "loss": 1.8776,
+      "step": 47009
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.737041719342605e-07,
+      "loss": 3.1006,
+      "step": 47010
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.73656763590392e-07,
+      "loss": 2.719,
+      "step": 47011
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.736093552465234e-07,
+      "loss": 1.7123,
+      "step": 47012
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.735619469026549e-07,
+      "loss": 1.1494,
+      "step": 47013
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.735145385587863e-07,
+      "loss": 1.6717,
+      "step": 47014
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.734671302149179e-07,
+      "loss": 1.459,
+      "step": 47015
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.734197218710493e-07,
+      "loss": 1.0938,
+      "step": 47016
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.733723135271807e-07,
+      "loss": 2.333,
+      "step": 47017
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.733249051833123e-07,
+      "loss": 0.6916,
+      "step": 47018
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.732774968394438e-07,
+      "loss": 1.8314,
+      "step": 47019
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.732300884955752e-07,
+      "loss": 1.7292,
+      "step": 47020
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.731826801517067e-07,
+      "loss": 2.3946,
+      "step": 47021
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 7.731352718078383e-07,
+      "loss": 1.8068,
+      "step": 47022
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.730878634639697e-07,
+      "loss": 1.8087,
+      "step": 47023
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.730404551201012e-07,
+      "loss": 2.6418,
+      "step": 47024
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.729930467762326e-07,
+      "loss": 1.6036,
+      "step": 47025
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.72945638432364e-07,
+      "loss": 1.6644,
+      "step": 47026
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.728982300884957e-07,
+      "loss": 1.6048,
+      "step": 47027
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.72850821744627e-07,
+      "loss": 1.9728,
+      "step": 47028
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.728034134007585e-07,
+      "loss": 1.9352,
+      "step": 47029
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.727560050568901e-07,
+      "loss": 1.8062,
+      "step": 47030
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.727085967130215e-07,
+      "loss": 3.1355,
+      "step": 47031
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.72661188369153e-07,
+      "loss": 2.1919,
+      "step": 47032
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.726137800252845e-07,
+      "loss": 1.9023,
+      "step": 47033
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.725663716814159e-07,
+      "loss": 0.5251,
+      "step": 47034
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.725189633375475e-07,
+      "loss": 2.6231,
+      "step": 47035
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.724715549936789e-07,
+      "loss": 1.8082,
+      "step": 47036
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.724241466498104e-07,
+      "loss": 2.3686,
+      "step": 47037
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.72376738305942e-07,
+      "loss": 0.8351,
+      "step": 47038
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.723293299620732e-07,
+      "loss": 2.1739,
+      "step": 47039
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.722819216182048e-07,
+      "loss": 1.8655,
+      "step": 47040
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.722345132743363e-07,
+      "loss": 1.4227,
+      "step": 47041
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.721871049304677e-07,
+      "loss": 2.8988,
+      "step": 47042
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.721396965865993e-07,
+      "loss": 2.846,
+      "step": 47043
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.720922882427308e-07,
+      "loss": 2.722,
+      "step": 47044
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.720448798988622e-07,
+      "loss": 1.7935,
+      "step": 47045
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.719974715549937e-07,
+      "loss": 0.9823,
+      "step": 47046
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.719500632111251e-07,
+      "loss": 2.3468,
+      "step": 47047
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.719026548672567e-07,
+      "loss": 1.6166,
+      "step": 47048
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.718552465233882e-07,
+      "loss": 2.9116,
+      "step": 47049
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.718078381795195e-07,
+      "loss": 2.0364,
+      "step": 47050
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.717604298356511e-07,
+      "loss": 1.626,
+      "step": 47051
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.717130214917826e-07,
+      "loss": 0.8446,
+      "step": 47052
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.71665613147914e-07,
+      "loss": 2.742,
+      "step": 47053
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.716182048040455e-07,
+      "loss": 2.0992,
+      "step": 47054
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.715707964601771e-07,
+      "loss": 1.0876,
+      "step": 47055
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.715233881163085e-07,
+      "loss": 0.9337,
+      "step": 47056
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.7147597977244e-07,
+      "loss": 1.118,
+      "step": 47057
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.714285714285714e-07,
+      "loss": 2.9071,
+      "step": 47058
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.713811630847029e-07,
+      "loss": 1.1505,
+      "step": 47059
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.713337547408345e-07,
+      "loss": 2.296,
+      "step": 47060
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.712863463969659e-07,
+      "loss": 1.2881,
+      "step": 47061
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.712389380530973e-07,
+      "loss": 1.3892,
+      "step": 47062
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.711915297092289e-07,
+      "loss": 2.1873,
+      "step": 47063
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.711441213653602e-07,
+      "loss": 2.0459,
+      "step": 47064
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.710967130214918e-07,
+      "loss": 2.5522,
+      "step": 47065
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.710493046776233e-07,
+      "loss": 1.7782,
+      "step": 47066
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.710018963337547e-07,
+      "loss": 1.9859,
+      "step": 47067
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.709544879898863e-07,
+      "loss": 2.3184,
+      "step": 47068
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.709070796460177e-07,
+      "loss": 2.2426,
+      "step": 47069
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.708596713021492e-07,
+      "loss": 2.9214,
+      "step": 47070
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.708122629582807e-07,
+      "loss": 0.5554,
+      "step": 47071
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.707648546144121e-07,
+      "loss": 3.197,
+      "step": 47072
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.707174462705437e-07,
+      "loss": 3.9035,
+      "step": 47073
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.706700379266751e-07,
+      "loss": 2.0723,
+      "step": 47074
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.706226295828065e-07,
+      "loss": 1.6972,
+      "step": 47075
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.705752212389381e-07,
+      "loss": 0.3722,
+      "step": 47076
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.705278128950696e-07,
+      "loss": 2.3623,
+      "step": 47077
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.70480404551201e-07,
+      "loss": 2.7643,
+      "step": 47078
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.704329962073325e-07,
+      "loss": 2.4915,
+      "step": 47079
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.703855878634639e-07,
+      "loss": 1.142,
+      "step": 47080
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.703381795195955e-07,
+      "loss": 2.42,
+      "step": 47081
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.70290771175727e-07,
+      "loss": 3.8894,
+      "step": 47082
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.702433628318584e-07,
+      "loss": 1.1878,
+      "step": 47083
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.701959544879899e-07,
+      "loss": 2.3725,
+      "step": 47084
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 7.701485461441215e-07,
+      "loss": 1.4244,
+      "step": 47085
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.701011378002528e-07,
+      "loss": 0.4945,
+      "step": 47086
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.700537294563843e-07,
+      "loss": 2.1084,
+      "step": 47087
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.700063211125159e-07,
+      "loss": 1.1204,
+      "step": 47088
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.699589127686473e-07,
+      "loss": 1.806,
+      "step": 47089
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.699115044247788e-07,
+      "loss": 2.9926,
+      "step": 47090
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.698640960809102e-07,
+      "loss": 3.216,
+      "step": 47091
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.698166877370417e-07,
+      "loss": 1.5046,
+      "step": 47092
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.697692793931733e-07,
+      "loss": 2.2482,
+      "step": 47093
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.697218710493047e-07,
+      "loss": 0.6589,
+      "step": 47094
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.696744627054362e-07,
+      "loss": 0.8071,
+      "step": 47095
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.696270543615678e-07,
+      "loss": 2.082,
+      "step": 47096
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.69579646017699e-07,
+      "loss": 0.6518,
+      "step": 47097
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.695322376738306e-07,
+      "loss": 1.7744,
+      "step": 47098
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.694848293299621e-07,
+      "loss": 1.0912,
+      "step": 47099
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.694374209860935e-07,
+      "loss": 0.4113,
+      "step": 47100
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.693900126422251e-07,
+      "loss": 2.31,
+      "step": 47101
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.693426042983565e-07,
+      "loss": 0.6465,
+      "step": 47102
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.69295195954488e-07,
+      "loss": 1.9612,
+      "step": 47103
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.692477876106195e-07,
+      "loss": 1.7363,
+      "step": 47104
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.692003792667509e-07,
+      "loss": 1.6113,
+      "step": 47105
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.691529709228825e-07,
+      "loss": 2.1729,
+      "step": 47106
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.69105562579014e-07,
+      "loss": 1.2393,
+      "step": 47107
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.690581542351454e-07,
+      "loss": 2.4754,
+      "step": 47108
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.69010745891277e-07,
+      "loss": 1.6766,
+      "step": 47109
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.689633375474084e-07,
+      "loss": 1.556,
+      "step": 47110
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.689159292035398e-07,
+      "loss": 2.3864,
+      "step": 47111
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.688685208596713e-07,
+      "loss": 2.0393,
+      "step": 47112
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.688211125158027e-07,
+      "loss": 0.5773,
+      "step": 47113
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.687737041719343e-07,
+      "loss": 3.1398,
+      "step": 47114
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.687262958280658e-07,
+      "loss": 3.2499,
+      "step": 47115
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.686788874841972e-07,
+      "loss": 0.6219,
+      "step": 47116
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.686314791403287e-07,
+      "loss": 1.9664,
+      "step": 47117
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.685840707964603e-07,
+      "loss": 2.043,
+      "step": 47118
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.685366624525917e-07,
+      "loss": 2.5708,
+      "step": 47119
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.684892541087232e-07,
+      "loss": 1.1982,
+      "step": 47120
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.684418457648548e-07,
+      "loss": 0.5702,
+      "step": 47121
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.683944374209861e-07,
+      "loss": 0.5364,
+      "step": 47122
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.683470290771176e-07,
+      "loss": 0.8926,
+      "step": 47123
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.68299620733249e-07,
+      "loss": 1.902,
+      "step": 47124
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.682522123893805e-07,
+      "loss": 1.8296,
+      "step": 47125
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.682048040455121e-07,
+      "loss": 1.1992,
+      "step": 47126
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.681573957016435e-07,
+      "loss": 1.1237,
+      "step": 47127
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.68109987357775e-07,
+      "loss": 3.036,
+      "step": 47128
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.680625790139066e-07,
+      "loss": 1.3782,
+      "step": 47129
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.680151706700379e-07,
+      "loss": 1.0347,
+      "step": 47130
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.679677623261695e-07,
+      "loss": 0.9213,
+      "step": 47131
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.67920353982301e-07,
+      "loss": 1.5795,
+      "step": 47132
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.678729456384323e-07,
+      "loss": 2.8178,
+      "step": 47133
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.678255372945639e-07,
+      "loss": 2.4105,
+      "step": 47134
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.677781289506952e-07,
+      "loss": 0.4795,
+      "step": 47135
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.677307206068268e-07,
+      "loss": 2.7113,
+      "step": 47136
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.676833122629583e-07,
+      "loss": 2.0716,
+      "step": 47137
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.676359039190897e-07,
+      "loss": 2.1435,
+      "step": 47138
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.675884955752213e-07,
+      "loss": 2.0413,
+      "step": 47139
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.675410872313528e-07,
+      "loss": 1.1845,
+      "step": 47140
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.674936788874842e-07,
+      "loss": 1.721,
+      "step": 47141
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.674462705436157e-07,
+      "loss": 0.2307,
+      "step": 47142
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.673988621997473e-07,
+      "loss": 2.4011,
+      "step": 47143
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.673514538558786e-07,
+      "loss": 1.4517,
+      "step": 47144
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.673040455120101e-07,
+      "loss": 1.6608,
+      "step": 47145
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.672566371681415e-07,
+      "loss": 3.6987,
+      "step": 47146
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.672092288242731e-07,
+      "loss": 2.5687,
+      "step": 47147
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.671618204804046e-07,
+      "loss": 1.7395,
+      "step": 47148
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 7.67114412136536e-07,
+      "loss": 1.6253,
+      "step": 47149
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.670670037926675e-07,
+      "loss": 0.7114,
+      "step": 47150
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.670195954487991e-07,
+      "loss": 2.0216,
+      "step": 47151
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.669721871049305e-07,
+      "loss": 2.2988,
+      "step": 47152
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.66924778761062e-07,
+      "loss": 1.3271,
+      "step": 47153
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.668773704171936e-07,
+      "loss": 2.0827,
+      "step": 47154
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.668299620733248e-07,
+      "loss": 2.0308,
+      "step": 47155
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.667825537294564e-07,
+      "loss": 2.0402,
+      "step": 47156
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.667351453855878e-07,
+      "loss": 1.7748,
+      "step": 47157
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.666877370417193e-07,
+      "loss": 2.5371,
+      "step": 47158
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.666403286978509e-07,
+      "loss": 3.0784,
+      "step": 47159
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.665929203539823e-07,
+      "loss": 1.6737,
+      "step": 47160
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.665455120101138e-07,
+      "loss": 2.4685,
+      "step": 47161
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.664981036662453e-07,
+      "loss": 0.8346,
+      "step": 47162
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.664506953223767e-07,
+      "loss": 0.7961,
+      "step": 47163
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.664032869785083e-07,
+      "loss": 2.3354,
+      "step": 47164
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.663558786346398e-07,
+      "loss": 1.8714,
+      "step": 47165
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.663084702907712e-07,
+      "loss": 2.5951,
+      "step": 47166
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.662610619469028e-07,
+      "loss": 2.209,
+      "step": 47167
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.66213653603034e-07,
+      "loss": 2.9097,
+      "step": 47168
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.661662452591656e-07,
+      "loss": 2.0563,
+      "step": 47169
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.661188369152971e-07,
+      "loss": 1.7269,
+      "step": 47170
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.660714285714285e-07,
+      "loss": 1.6491,
+      "step": 47171
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.660240202275601e-07,
+      "loss": 1.2874,
+      "step": 47172
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.659766118836916e-07,
+      "loss": 0.9357,
+      "step": 47173
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.65929203539823e-07,
+      "loss": 1.9473,
+      "step": 47174
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.658817951959545e-07,
+      "loss": 1.4636,
+      "step": 47175
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.658343868520861e-07,
+      "loss": 1.4926,
+      "step": 47176
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.657869785082175e-07,
+      "loss": 1.9374,
+      "step": 47177
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.65739570164349e-07,
+      "loss": 2.095,
+      "step": 47178
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.656921618204803e-07,
+      "loss": 1.9582,
+      "step": 47179
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.656447534766119e-07,
+      "loss": 1.607,
+      "step": 47180
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.655973451327434e-07,
+      "loss": 2.8733,
+      "step": 47181
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.655499367888748e-07,
+      "loss": 1.8405,
+      "step": 47182
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.655025284450063e-07,
+      "loss": 2.416,
+      "step": 47183
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.654551201011379e-07,
+      "loss": 1.3843,
+      "step": 47184
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.654077117572693e-07,
+      "loss": 2.7126,
+      "step": 47185
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.653603034134008e-07,
+      "loss": 2.0373,
+      "step": 47186
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.653128950695324e-07,
+      "loss": 1.21,
+      "step": 47187
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.652654867256637e-07,
+      "loss": 1.8858,
+      "step": 47188
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.652180783817953e-07,
+      "loss": 1.3229,
+      "step": 47189
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.651706700379267e-07,
+      "loss": 2.3508,
+      "step": 47190
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.651232616940581e-07,
+      "loss": 1.4247,
+      "step": 47191
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.650758533501897e-07,
+      "loss": 1.8369,
+      "step": 47192
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.650284450063211e-07,
+      "loss": 2.6507,
+      "step": 47193
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.649810366624526e-07,
+      "loss": 3.7844,
+      "step": 47194
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.649336283185841e-07,
+      "loss": 1.6254,
+      "step": 47195
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.648862199747155e-07,
+      "loss": 1.1833,
+      "step": 47196
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.648388116308471e-07,
+      "loss": 1.2924,
+      "step": 47197
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.647914032869786e-07,
+      "loss": 2.5636,
+      "step": 47198
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.6474399494311e-07,
+      "loss": 0.9722,
+      "step": 47199
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.646965865992416e-07,
+      "loss": 1.1649,
+      "step": 47200
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.646491782553729e-07,
+      "loss": 2.7543,
+      "step": 47201
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.646017699115045e-07,
+      "loss": 1.2383,
+      "step": 47202
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.645543615676359e-07,
+      "loss": 2.9037,
+      "step": 47203
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.645069532237673e-07,
+      "loss": 2.4791,
+      "step": 47204
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.644595448798989e-07,
+      "loss": 1.29,
+      "step": 47205
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.644121365360304e-07,
+      "loss": 1.9314,
+      "step": 47206
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.643647281921618e-07,
+      "loss": 0.9172,
+      "step": 47207
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.643173198482933e-07,
+      "loss": 2.1735,
+      "step": 47208
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.642699115044249e-07,
+      "loss": 0.7037,
+      "step": 47209
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.642225031605563e-07,
+      "loss": 0.6301,
+      "step": 47210
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.641750948166878e-07,
+      "loss": 2.3293,
+      "step": 47211
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 7.641276864728192e-07,
+      "loss": 2.8653,
+      "step": 47212
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.640802781289507e-07,
+      "loss": 1.9603,
+      "step": 47213
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.640328697850823e-07,
+      "loss": 1.6765,
+      "step": 47214
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.639854614412136e-07,
+      "loss": 1.3281,
+      "step": 47215
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.639380530973451e-07,
+      "loss": 2.8681,
+      "step": 47216
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.638906447534767e-07,
+      "loss": 3.5936,
+      "step": 47217
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.638432364096081e-07,
+      "loss": 2.6597,
+      "step": 47218
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.637958280657396e-07,
+      "loss": 1.8201,
+      "step": 47219
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.637484197218711e-07,
+      "loss": 1.72,
+      "step": 47220
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.637010113780025e-07,
+      "loss": 2.5479,
+      "step": 47221
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.636536030341341e-07,
+      "loss": 3.1428,
+      "step": 47222
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.636061946902655e-07,
+      "loss": 0.2855,
+      "step": 47223
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.63558786346397e-07,
+      "loss": 1.9864,
+      "step": 47224
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.635113780025286e-07,
+      "loss": 1.6288,
+      "step": 47225
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.634639696586598e-07,
+      "loss": 0.246,
+      "step": 47226
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.634165613147914e-07,
+      "loss": 0.6885,
+      "step": 47227
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.633691529709229e-07,
+      "loss": 1.5861,
+      "step": 47228
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.633217446270543e-07,
+      "loss": 1.4864,
+      "step": 47229
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.632743362831859e-07,
+      "loss": 4.2895,
+      "step": 47230
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.632269279393174e-07,
+      "loss": 2.5245,
+      "step": 47231
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.631795195954488e-07,
+      "loss": 1.4391,
+      "step": 47232
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.631321112515803e-07,
+      "loss": 1.2908,
+      "step": 47233
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.630847029077117e-07,
+      "loss": 1.4813,
+      "step": 47234
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.630372945638433e-07,
+      "loss": 2.8151,
+      "step": 47235
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.629898862199748e-07,
+      "loss": 2.6425,
+      "step": 47236
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.629424778761061e-07,
+      "loss": 3.1749,
+      "step": 47237
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.628950695322377e-07,
+      "loss": 1.4624,
+      "step": 47238
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.628476611883692e-07,
+      "loss": 1.0671,
+      "step": 47239
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.628002528445006e-07,
+      "loss": 1.1427,
+      "step": 47240
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.627528445006321e-07,
+      "loss": 1.0486,
+      "step": 47241
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.627054361567637e-07,
+      "loss": 2.4235,
+      "step": 47242
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.626580278128951e-07,
+      "loss": 2.7115,
+      "step": 47243
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.626106194690266e-07,
+      "loss": 2.084,
+      "step": 47244
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.62563211125158e-07,
+      "loss": 4.3626,
+      "step": 47245
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.625158027812895e-07,
+      "loss": 1.7348,
+      "step": 47246
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.624683944374211e-07,
+      "loss": 0.7731,
+      "step": 47247
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.624209860935525e-07,
+      "loss": 1.7752,
+      "step": 47248
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.62373577749684e-07,
+      "loss": 0.9479,
+      "step": 47249
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.623261694058155e-07,
+      "loss": 1.3306,
+      "step": 47250
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.622787610619469e-07,
+      "loss": 1.134,
+      "step": 47251
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.622313527180784e-07,
+      "loss": 1.1747,
+      "step": 47252
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.621839443742099e-07,
+      "loss": 3.6427,
+      "step": 47253
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.621365360303413e-07,
+      "loss": 3.0029,
+      "step": 47254
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.620891276864729e-07,
+      "loss": 1.9498,
+      "step": 47255
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.620417193426043e-07,
+      "loss": 0.7555,
+      "step": 47256
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.619943109987358e-07,
+      "loss": 1.8985,
+      "step": 47257
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.619469026548674e-07,
+      "loss": 2.2872,
+      "step": 47258
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.618994943109987e-07,
+      "loss": 2.1969,
+      "step": 47259
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.618520859671303e-07,
+      "loss": 3.3277,
+      "step": 47260
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.618046776232617e-07,
+      "loss": 2.2316,
+      "step": 47261
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.617572692793931e-07,
+      "loss": 3.214,
+      "step": 47262
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.617098609355247e-07,
+      "loss": 1.4604,
+      "step": 47263
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.616624525916562e-07,
+      "loss": 2.4901,
+      "step": 47264
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.616150442477876e-07,
+      "loss": 1.4471,
+      "step": 47265
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.615676359039191e-07,
+      "loss": 1.8007,
+      "step": 47266
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.615202275600505e-07,
+      "loss": 1.3222,
+      "step": 47267
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.614728192161821e-07,
+      "loss": 2.8018,
+      "step": 47268
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.614254108723136e-07,
+      "loss": 0.945,
+      "step": 47269
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.61378002528445e-07,
+      "loss": 2.1954,
+      "step": 47270
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.613305941845765e-07,
+      "loss": 2.0691,
+      "step": 47271
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.612831858407081e-07,
+      "loss": 1.6097,
+      "step": 47272
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.612357774968394e-07,
+      "loss": 1.7686,
+      "step": 47273
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.611883691529709e-07,
+      "loss": 2.4065,
+      "step": 47274
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 7.611409608091025e-07,
+      "loss": 1.7772,
+      "step": 47275
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.610935524652339e-07,
+      "loss": 3.0913,
+      "step": 47276
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.610461441213654e-07,
+      "loss": 2.2824,
+      "step": 47277
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.609987357774968e-07,
+      "loss": 0.9214,
+      "step": 47278
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.609513274336283e-07,
+      "loss": 3.3467,
+      "step": 47279
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.609039190897599e-07,
+      "loss": 0.2798,
+      "step": 47280
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.608565107458913e-07,
+      "loss": 4.2661,
+      "step": 47281
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.608091024020228e-07,
+      "loss": 1.8404,
+      "step": 47282
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.607616940581544e-07,
+      "loss": 1.8375,
+      "step": 47283
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.607142857142856e-07,
+      "loss": 2.8609,
+      "step": 47284
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.606668773704172e-07,
+      "loss": 1.9787,
+      "step": 47285
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.606194690265487e-07,
+      "loss": 2.4431,
+      "step": 47286
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.605720606826801e-07,
+      "loss": 2.7759,
+      "step": 47287
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.605246523388117e-07,
+      "loss": 0.3538,
+      "step": 47288
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.604772439949431e-07,
+      "loss": 1.5304,
+      "step": 47289
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.604298356510746e-07,
+      "loss": 1.7768,
+      "step": 47290
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.603824273072061e-07,
+      "loss": 2.5815,
+      "step": 47291
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.603350189633375e-07,
+      "loss": 1.2913,
+      "step": 47292
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.602876106194691e-07,
+      "loss": 1.8825,
+      "step": 47293
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.602402022756006e-07,
+      "loss": 2.1098,
+      "step": 47294
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.60192793931732e-07,
+      "loss": 1.318,
+      "step": 47295
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.601453855878636e-07,
+      "loss": 2.5382,
+      "step": 47296
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.600979772439948e-07,
+      "loss": 0.6381,
+      "step": 47297
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.600505689001264e-07,
+      "loss": 1.1933,
+      "step": 47298
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.600031605562579e-07,
+      "loss": 2.0376,
+      "step": 47299
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.599557522123893e-07,
+      "loss": 0.7736,
+      "step": 47300
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.599083438685209e-07,
+      "loss": 0.6617,
+      "step": 47301
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.598609355246524e-07,
+      "loss": 2.4807,
+      "step": 47302
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.598135271807838e-07,
+      "loss": 1.068,
+      "step": 47303
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.597661188369153e-07,
+      "loss": 1.7286,
+      "step": 47304
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.597187104930469e-07,
+      "loss": 1.9071,
+      "step": 47305
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.596713021491783e-07,
+      "loss": 1.5453,
+      "step": 47306
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.596238938053098e-07,
+      "loss": 2.4684,
+      "step": 47307
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.595764854614411e-07,
+      "loss": 0.8432,
+      "step": 47308
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.595290771175727e-07,
+      "loss": 2.7341,
+      "step": 47309
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.594816687737042e-07,
+      "loss": 2.0011,
+      "step": 47310
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.594342604298356e-07,
+      "loss": 2.3635,
+      "step": 47311
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.593868520859671e-07,
+      "loss": 3.0802,
+      "step": 47312
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.593394437420987e-07,
+      "loss": 1.0493,
+      "step": 47313
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.592920353982301e-07,
+      "loss": 2.3849,
+      "step": 47314
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.592446270543616e-07,
+      "loss": 1.0728,
+      "step": 47315
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.591972187104932e-07,
+      "loss": 0.9995,
+      "step": 47316
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.591498103666245e-07,
+      "loss": 1.6673,
+      "step": 47317
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.591024020227561e-07,
+      "loss": 2.035,
+      "step": 47318
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.590549936788874e-07,
+      "loss": 0.873,
+      "step": 47319
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.590075853350189e-07,
+      "loss": 1.7816,
+      "step": 47320
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.589601769911505e-07,
+      "loss": 2.964,
+      "step": 47321
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.589127686472819e-07,
+      "loss": 2.3506,
+      "step": 47322
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.588653603034134e-07,
+      "loss": 2.0869,
+      "step": 47323
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.588179519595449e-07,
+      "loss": 2.7483,
+      "step": 47324
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.587705436156763e-07,
+      "loss": 2.2784,
+      "step": 47325
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.587231352718079e-07,
+      "loss": 2.173,
+      "step": 47326
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.586757269279394e-07,
+      "loss": 1.801,
+      "step": 47327
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.586283185840708e-07,
+      "loss": 1.1134,
+      "step": 47328
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.585809102402024e-07,
+      "loss": 2.4936,
+      "step": 47329
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.585335018963337e-07,
+      "loss": 1.0711,
+      "step": 47330
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.584860935524652e-07,
+      "loss": 2.8093,
+      "step": 47331
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.584386852085967e-07,
+      "loss": 1.8896,
+      "step": 47332
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.583912768647281e-07,
+      "loss": 2.1958,
+      "step": 47333
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.583438685208597e-07,
+      "loss": 1.6017,
+      "step": 47334
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.582964601769912e-07,
+      "loss": 2.5552,
+      "step": 47335
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.582490518331226e-07,
+      "loss": 2.5232,
+      "step": 47336
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.582016434892541e-07,
+      "loss": 2.2399,
+      "step": 47337
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.581542351453857e-07,
+      "loss": 0.8726,
+      "step": 47338
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.581068268015171e-07,
+      "loss": 1.4742,
+      "step": 47339
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.580594184576486e-07,
+      "loss": 2.7946,
+      "step": 47340
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.5801201011378e-07,
+      "loss": 1.7271,
+      "step": 47341
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.579646017699115e-07,
+      "loss": 1.2135,
+      "step": 47342
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.57917193426043e-07,
+      "loss": 1.5487,
+      "step": 47343
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.578697850821744e-07,
+      "loss": 2.5473,
+      "step": 47344
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.578223767383059e-07,
+      "loss": 1.2055,
+      "step": 47345
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.577749683944375e-07,
+      "loss": 1.1002,
+      "step": 47346
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.577275600505689e-07,
+      "loss": 0.814,
+      "step": 47347
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.576801517067004e-07,
+      "loss": 2.3501,
+      "step": 47348
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.576327433628319e-07,
+      "loss": 2.0331,
+      "step": 47349
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.575853350189633e-07,
+      "loss": 1.0148,
+      "step": 47350
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.575379266750949e-07,
+      "loss": 1.3751,
+      "step": 47351
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.574905183312263e-07,
+      "loss": 1.9164,
+      "step": 47352
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.574431099873578e-07,
+      "loss": 2.4641,
+      "step": 47353
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.573957016434894e-07,
+      "loss": 0.8974,
+      "step": 47354
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.573482932996206e-07,
+      "loss": 2.2614,
+      "step": 47355
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.573008849557522e-07,
+      "loss": 1.4835,
+      "step": 47356
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.572534766118837e-07,
+      "loss": 1.9017,
+      "step": 47357
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.572060682680151e-07,
+      "loss": 2.3067,
+      "step": 47358
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.571586599241467e-07,
+      "loss": 2.0249,
+      "step": 47359
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.571112515802782e-07,
+      "loss": 1.9599,
+      "step": 47360
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.570638432364096e-07,
+      "loss": 1.3885,
+      "step": 47361
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.570164348925411e-07,
+      "loss": 1.1602,
+      "step": 47362
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.569690265486725e-07,
+      "loss": 1.086,
+      "step": 47363
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.569216182048041e-07,
+      "loss": 1.6643,
+      "step": 47364
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.568742098609356e-07,
+      "loss": 2.8743,
+      "step": 47365
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.568268015170669e-07,
+      "loss": 2.8828,
+      "step": 47366
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.567793931731985e-07,
+      "loss": 1.9489,
+      "step": 47367
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.5673198482933e-07,
+      "loss": 1.7006,
+      "step": 47368
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.566845764854614e-07,
+      "loss": 2.5068,
+      "step": 47369
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.566371681415929e-07,
+      "loss": 2.0214,
+      "step": 47370
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.565897597977245e-07,
+      "loss": 1.7285,
+      "step": 47371
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.565423514538559e-07,
+      "loss": 2.0372,
+      "step": 47372
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.564949431099874e-07,
+      "loss": 1.0232,
+      "step": 47373
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.564475347661188e-07,
+      "loss": 3.0984,
+      "step": 47374
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.564001264222503e-07,
+      "loss": 2.2047,
+      "step": 47375
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.563527180783819e-07,
+      "loss": 1.8173,
+      "step": 47376
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.563053097345133e-07,
+      "loss": 1.7543,
+      "step": 47377
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.562579013906447e-07,
+      "loss": 3.5713,
+      "step": 47378
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.562104930467763e-07,
+      "loss": 1.0103,
+      "step": 47379
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.561630847029077e-07,
+      "loss": 1.8945,
+      "step": 47380
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.561156763590392e-07,
+      "loss": 1.9835,
+      "step": 47381
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.560682680151707e-07,
+      "loss": 2.8813,
+      "step": 47382
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.560208596713021e-07,
+      "loss": 1.8644,
+      "step": 47383
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.559734513274337e-07,
+      "loss": 3.0997,
+      "step": 47384
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.559260429835651e-07,
+      "loss": 2.1765,
+      "step": 47385
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.558786346396966e-07,
+      "loss": 2.2168,
+      "step": 47386
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.558312262958282e-07,
+      "loss": 1.5162,
+      "step": 47387
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.557838179519595e-07,
+      "loss": 2.2834,
+      "step": 47388
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.557364096080911e-07,
+      "loss": 1.365,
+      "step": 47389
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.556890012642225e-07,
+      "loss": 1.5596,
+      "step": 47390
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.556415929203539e-07,
+      "loss": 2.1416,
+      "step": 47391
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.555941845764855e-07,
+      "loss": 2.9439,
+      "step": 47392
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.55546776232617e-07,
+      "loss": 1.2508,
+      "step": 47393
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.554993678887484e-07,
+      "loss": 3.3863,
+      "step": 47394
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.554519595448799e-07,
+      "loss": 2.5785,
+      "step": 47395
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.554045512010113e-07,
+      "loss": 3.0096,
+      "step": 47396
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.553571428571429e-07,
+      "loss": 2.6217,
+      "step": 47397
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.553097345132744e-07,
+      "loss": 2.2425,
+      "step": 47398
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.552623261694058e-07,
+      "loss": 2.3553,
+      "step": 47399
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.552149178255374e-07,
+      "loss": 1.8959,
+      "step": 47400
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.551675094816689e-07,
+      "loss": 2.0375,
+      "step": 47401
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 7.551201011378002e-07,
+      "loss": 2.0667,
+      "step": 47402
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.550726927939317e-07,
+      "loss": 2.4992,
+      "step": 47403
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.550252844500633e-07,
+      "loss": 1.7934,
+      "step": 47404
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.549778761061947e-07,
+      "loss": 2.3665,
+      "step": 47405
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.549304677623262e-07,
+      "loss": 1.2982,
+      "step": 47406
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.548830594184576e-07,
+      "loss": 0.4745,
+      "step": 47407
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.548356510745891e-07,
+      "loss": 1.3799,
+      "step": 47408
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.547882427307207e-07,
+      "loss": 2.811,
+      "step": 47409
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.547408343868521e-07,
+      "loss": 1.5088,
+      "step": 47410
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.546934260429836e-07,
+      "loss": 0.9026,
+      "step": 47411
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.546460176991152e-07,
+      "loss": 1.3187,
+      "step": 47412
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.545986093552464e-07,
+      "loss": 1.6502,
+      "step": 47413
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.54551201011378e-07,
+      "loss": 1.5006,
+      "step": 47414
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.545037926675095e-07,
+      "loss": 1.6448,
+      "step": 47415
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.544563843236409e-07,
+      "loss": 2.0586,
+      "step": 47416
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.544089759797725e-07,
+      "loss": 3.1875,
+      "step": 47417
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.543615676359039e-07,
+      "loss": 2.3723,
+      "step": 47418
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.543141592920354e-07,
+      "loss": 2.4525,
+      "step": 47419
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.542667509481669e-07,
+      "loss": 1.3161,
+      "step": 47420
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.542193426042983e-07,
+      "loss": 1.4074,
+      "step": 47421
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.541719342604299e-07,
+      "loss": 1.5536,
+      "step": 47422
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.541245259165614e-07,
+      "loss": 2.5074,
+      "step": 47423
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.540771175726928e-07,
+      "loss": 2.1182,
+      "step": 47424
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.540297092288243e-07,
+      "loss": 1.9313,
+      "step": 47425
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.539823008849558e-07,
+      "loss": 1.3048,
+      "step": 47426
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.539348925410872e-07,
+      "loss": 3.3282,
+      "step": 47427
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.538874841972187e-07,
+      "loss": 2.9642,
+      "step": 47428
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.538400758533501e-07,
+      "loss": 1.2277,
+      "step": 47429
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.537926675094817e-07,
+      "loss": 0.9061,
+      "step": 47430
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.537452591656132e-07,
+      "loss": 1.1728,
+      "step": 47431
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.536978508217446e-07,
+      "loss": 1.3929,
+      "step": 47432
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.536504424778761e-07,
+      "loss": 0.9958,
+      "step": 47433
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.536030341340077e-07,
+      "loss": 2.7724,
+      "step": 47434
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.535556257901391e-07,
+      "loss": 1.5615,
+      "step": 47435
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.535082174462706e-07,
+      "loss": 1.4577,
+      "step": 47436
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.534608091024021e-07,
+      "loss": 2.4788,
+      "step": 47437
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.534134007585335e-07,
+      "loss": 2.5521,
+      "step": 47438
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.53365992414665e-07,
+      "loss": 1.6737,
+      "step": 47439
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.533185840707964e-07,
+      "loss": 2.3466,
+      "step": 47440
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.532711757269279e-07,
+      "loss": 1.4929,
+      "step": 47441
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.532237673830595e-07,
+      "loss": 1.3832,
+      "step": 47442
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.531763590391909e-07,
+      "loss": 0.4774,
+      "step": 47443
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.531289506953224e-07,
+      "loss": 1.7353,
+      "step": 47444
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.53081542351454e-07,
+      "loss": 1.7178,
+      "step": 47445
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.530341340075853e-07,
+      "loss": 0.9782,
+      "step": 47446
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.529867256637169e-07,
+      "loss": 1.7487,
+      "step": 47447
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.529393173198484e-07,
+      "loss": 1.5264,
+      "step": 47448
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.528919089759797e-07,
+      "loss": 0.3563,
+      "step": 47449
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.528445006321113e-07,
+      "loss": 1.8345,
+      "step": 47450
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.527970922882427e-07,
+      "loss": 0.7881,
+      "step": 47451
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.527496839443742e-07,
+      "loss": 2.3723,
+      "step": 47452
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.527022756005057e-07,
+      "loss": 2.3848,
+      "step": 47453
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.526548672566371e-07,
+      "loss": 1.5706,
+      "step": 47454
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.526074589127687e-07,
+      "loss": 1.6606,
+      "step": 47455
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.525600505689002e-07,
+      "loss": 2.1666,
+      "step": 47456
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.525126422250316e-07,
+      "loss": 1.4647,
+      "step": 47457
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.524652338811632e-07,
+      "loss": 1.1796,
+      "step": 47458
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.524178255372947e-07,
+      "loss": 1.9233,
+      "step": 47459
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.52370417193426e-07,
+      "loss": 2.4956,
+      "step": 47460
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.523230088495575e-07,
+      "loss": 0.5905,
+      "step": 47461
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.522756005056889e-07,
+      "loss": 1.8623,
+      "step": 47462
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.522281921618205e-07,
+      "loss": 2.9057,
+      "step": 47463
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.52180783817952e-07,
+      "loss": 1.844,
+      "step": 47464
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 7.521333754740834e-07,
+      "loss": 2.349,
+      "step": 47465
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.520859671302149e-07,
+      "loss": 0.3569,
+      "step": 47466
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.520385587863465e-07,
+      "loss": 2.366,
+      "step": 47467
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.519911504424779e-07,
+      "loss": 2.8038,
+      "step": 47468
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.519437420986094e-07,
+      "loss": 3.0166,
+      "step": 47469
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.51896333754741e-07,
+      "loss": 2.1472,
+      "step": 47470
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.518489254108724e-07,
+      "loss": 1.4335,
+      "step": 47471
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.518015170670038e-07,
+      "loss": 2.8141,
+      "step": 47472
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.517541087231352e-07,
+      "loss": 0.8888,
+      "step": 47473
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.517067003792667e-07,
+      "loss": 1.8183,
+      "step": 47474
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.516592920353983e-07,
+      "loss": 1.9246,
+      "step": 47475
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.516118836915297e-07,
+      "loss": 2.1248,
+      "step": 47476
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.515644753476612e-07,
+      "loss": 0.3421,
+      "step": 47477
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.515170670037928e-07,
+      "loss": 2.2699,
+      "step": 47478
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.514696586599241e-07,
+      "loss": 3.7608,
+      "step": 47479
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.514222503160557e-07,
+      "loss": 1.9186,
+      "step": 47480
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.513748419721872e-07,
+      "loss": 1.5924,
+      "step": 47481
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.513274336283186e-07,
+      "loss": 1.7955,
+      "step": 47482
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.512800252844502e-07,
+      "loss": 2.1995,
+      "step": 47483
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.512326169405814e-07,
+      "loss": 1.4807,
+      "step": 47484
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.51185208596713e-07,
+      "loss": 1.4576,
+      "step": 47485
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.511378002528445e-07,
+      "loss": 1.5197,
+      "step": 47486
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.510903919089759e-07,
+      "loss": 2.0654,
+      "step": 47487
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.510429835651075e-07,
+      "loss": 2.0276,
+      "step": 47488
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.50995575221239e-07,
+      "loss": 1.0858,
+      "step": 47489
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.509481668773704e-07,
+      "loss": 2.1132,
+      "step": 47490
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.509007585335019e-07,
+      "loss": 1.3611,
+      "step": 47491
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.508533501896335e-07,
+      "loss": 1.4753,
+      "step": 47492
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.508059418457649e-07,
+      "loss": 0.5674,
+      "step": 47493
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.507585335018964e-07,
+      "loss": 1.8214,
+      "step": 47494
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.507111251580277e-07,
+      "loss": 5.2409,
+      "step": 47495
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.506637168141593e-07,
+      "loss": 3.3817,
+      "step": 47496
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.506163084702908e-07,
+      "loss": 1.6371,
+      "step": 47497
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.505689001264222e-07,
+      "loss": 1.1539,
+      "step": 47498
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.505214917825537e-07,
+      "loss": 0.9853,
+      "step": 47499
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.504740834386853e-07,
+      "loss": 1.5765,
+      "step": 47500
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.504266750948167e-07,
+      "loss": 2.0559,
+      "step": 47501
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.503792667509482e-07,
+      "loss": 3.5761,
+      "step": 47502
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.503318584070798e-07,
+      "loss": 1.7395,
+      "step": 47503
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.502844500632111e-07,
+      "loss": 1.1751,
+      "step": 47504
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.502370417193427e-07,
+      "loss": 2.6777,
+      "step": 47505
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.50189633375474e-07,
+      "loss": 1.5318,
+      "step": 47506
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.501422250316055e-07,
+      "loss": 1.748,
+      "step": 47507
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.500948166877371e-07,
+      "loss": 1.4984,
+      "step": 47508
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.500474083438685e-07,
+      "loss": 1.1389,
+      "step": 47509
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.5e-07,
+      "loss": 2.5095,
+      "step": 47510
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.499525916561315e-07,
+      "loss": 2.2005,
+      "step": 47511
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.49905183312263e-07,
+      "loss": 1.6675,
+      "step": 47512
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.498577749683945e-07,
+      "loss": 1.4284,
+      "step": 47513
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.498103666245259e-07,
+      "loss": 1.8841,
+      "step": 47514
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.497629582806574e-07,
+      "loss": 1.5123,
+      "step": 47515
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.49715549936789e-07,
+      "loss": 0.6683,
+      "step": 47516
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.496681415929204e-07,
+      "loss": 3.0074,
+      "step": 47517
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.496207332490519e-07,
+      "loss": 2.1585,
+      "step": 47518
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.495733249051833e-07,
+      "loss": 1.1274,
+      "step": 47519
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.495259165613148e-07,
+      "loss": 2.1056,
+      "step": 47520
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.494785082174463e-07,
+      "loss": 1.5919,
+      "step": 47521
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.494310998735777e-07,
+      "loss": 3.5464,
+      "step": 47522
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.493836915297093e-07,
+      "loss": 1.3176,
+      "step": 47523
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.493362831858407e-07,
+      "loss": 1.6032,
+      "step": 47524
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.492888748419722e-07,
+      "loss": 1.6969,
+      "step": 47525
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.492414664981037e-07,
+      "loss": 2.0606,
+      "step": 47526
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.491940581542352e-07,
+      "loss": 2.6146,
+      "step": 47527
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.491466498103667e-07,
+      "loss": 2.3587,
+      "step": 47528
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 7.490992414664982e-07,
+      "loss": 2.3438,
+      "step": 47529
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.490518331226295e-07,
+      "loss": 1.7309,
+      "step": 47530
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.490044247787611e-07,
+      "loss": 1.5877,
+      "step": 47531
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.489570164348925e-07,
+      "loss": 2.9809,
+      "step": 47532
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.48909608091024e-07,
+      "loss": 2.1109,
+      "step": 47533
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.488621997471555e-07,
+      "loss": 1.5773,
+      "step": 47534
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.48814791403287e-07,
+      "loss": 2.4076,
+      "step": 47535
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.487673830594185e-07,
+      "loss": 1.059,
+      "step": 47536
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.487199747155499e-07,
+      "loss": 2.4752,
+      "step": 47537
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.486725663716815e-07,
+      "loss": 2.4132,
+      "step": 47538
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.48625158027813e-07,
+      "loss": 2.0376,
+      "step": 47539
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.485777496839444e-07,
+      "loss": 2.6783,
+      "step": 47540
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.485303413400759e-07,
+      "loss": 1.433,
+      "step": 47541
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.484829329962073e-07,
+      "loss": 3.0575,
+      "step": 47542
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.484355246523388e-07,
+      "loss": 1.4158,
+      "step": 47543
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.483881163084703e-07,
+      "loss": 2.9779,
+      "step": 47544
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.483407079646018e-07,
+      "loss": 2.5703,
+      "step": 47545
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.482932996207333e-07,
+      "loss": 1.8272,
+      "step": 47546
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.482458912768647e-07,
+      "loss": 2.2348,
+      "step": 47547
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.481984829329962e-07,
+      "loss": 2.2008,
+      "step": 47548
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.481510745891277e-07,
+      "loss": 2.1507,
+      "step": 47549
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.481036662452592e-07,
+      "loss": 1.6386,
+      "step": 47550
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.480562579013907e-07,
+      "loss": 1.4222,
+      "step": 47551
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.480088495575221e-07,
+      "loss": 1.1809,
+      "step": 47552
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.479614412136537e-07,
+      "loss": 1.4989,
+      "step": 47553
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.479140328697851e-07,
+      "loss": 1.8293,
+      "step": 47554
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.478666245259165e-07,
+      "loss": 1.7277,
+      "step": 47555
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.478192161820481e-07,
+      "loss": 2.6268,
+      "step": 47556
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.477718078381795e-07,
+      "loss": 1.7575,
+      "step": 47557
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.47724399494311e-07,
+      "loss": 2.6951,
+      "step": 47558
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.476769911504425e-07,
+      "loss": 1.9,
+      "step": 47559
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.47629582806574e-07,
+      "loss": 2.3403,
+      "step": 47560
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.475821744627055e-07,
+      "loss": 1.2537,
+      "step": 47561
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.475347661188369e-07,
+      "loss": 2.5547,
+      "step": 47562
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.474873577749684e-07,
+      "loss": 0.6307,
+      "step": 47563
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.474399494311e-07,
+      "loss": 2.486,
+      "step": 47564
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.473925410872313e-07,
+      "loss": 1.7897,
+      "step": 47565
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.473451327433628e-07,
+      "loss": 0.3738,
+      "step": 47566
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.472977243994943e-07,
+      "loss": 1.8569,
+      "step": 47567
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.472503160556258e-07,
+      "loss": 1.6963,
+      "step": 47568
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.472029077117573e-07,
+      "loss": 2.3253,
+      "step": 47569
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.471554993678887e-07,
+      "loss": 3.4327,
+      "step": 47570
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.471080910240203e-07,
+      "loss": 2.6336,
+      "step": 47571
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.470606826801517e-07,
+      "loss": 2.9931,
+      "step": 47572
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.470132743362832e-07,
+      "loss": 2.9209,
+      "step": 47573
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.469658659924147e-07,
+      "loss": 2.8173,
+      "step": 47574
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.469184576485462e-07,
+      "loss": 1.3911,
+      "step": 47575
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.468710493046777e-07,
+      "loss": 0.9591,
+      "step": 47576
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.468236409608091e-07,
+      "loss": 1.464,
+      "step": 47577
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.467762326169406e-07,
+      "loss": 3.0434,
+      "step": 47578
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.467288242730721e-07,
+      "loss": 2.6115,
+      "step": 47579
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.466814159292035e-07,
+      "loss": 1.1974,
+      "step": 47580
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.46634007585335e-07,
+      "loss": 1.8522,
+      "step": 47581
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.465865992414665e-07,
+      "loss": 1.4389,
+      "step": 47582
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.46539190897598e-07,
+      "loss": 1.9315,
+      "step": 47583
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.464917825537295e-07,
+      "loss": 1.8556,
+      "step": 47584
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.464443742098609e-07,
+      "loss": 1.2979,
+      "step": 47585
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.463969658659925e-07,
+      "loss": 1.0284,
+      "step": 47586
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.46349557522124e-07,
+      "loss": 1.6713,
+      "step": 47587
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.463021491782553e-07,
+      "loss": 1.333,
+      "step": 47588
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.46254740834387e-07,
+      "loss": 1.3926,
+      "step": 47589
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.462073324905183e-07,
+      "loss": 1.8145,
+      "step": 47590
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.461599241466498e-07,
+      "loss": 1.179,
+      "step": 47591
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 7.461125158027813e-07,
+      "loss": 1.1581,
+      "step": 47592
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.460651074589128e-07,
+      "loss": 2.52,
+      "step": 47593
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.460176991150443e-07,
+      "loss": 2.7364,
+      "step": 47594
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.459702907711757e-07,
+      "loss": 0.6042,
+      "step": 47595
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.459228824273072e-07,
+      "loss": 2.9262,
+      "step": 47596
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.458754740834388e-07,
+      "loss": 2.7159,
+      "step": 47597
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.458280657395702e-07,
+      "loss": 1.0772,
+      "step": 47598
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.457806573957017e-07,
+      "loss": 1.0559,
+      "step": 47599
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.457332490518332e-07,
+      "loss": 2.4567,
+      "step": 47600
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.456858407079646e-07,
+      "loss": 1.3625,
+      "step": 47601
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.456384323640961e-07,
+      "loss": 1.8963,
+      "step": 47602
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.455910240202275e-07,
+      "loss": 3.5458,
+      "step": 47603
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.455436156763591e-07,
+      "loss": 2.3495,
+      "step": 47604
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.454962073324905e-07,
+      "loss": 2.069,
+      "step": 47605
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.45448798988622e-07,
+      "loss": 0.9377,
+      "step": 47606
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.454013906447535e-07,
+      "loss": 1.4304,
+      "step": 47607
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.45353982300885e-07,
+      "loss": 2.2383,
+      "step": 47608
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.453065739570165e-07,
+      "loss": 1.6927,
+      "step": 47609
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.45259165613148e-07,
+      "loss": 2.2415,
+      "step": 47610
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.452117572692795e-07,
+      "loss": 2.8604,
+      "step": 47611
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.45164348925411e-07,
+      "loss": 1.8165,
+      "step": 47612
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.451169405815423e-07,
+      "loss": 1.892,
+      "step": 47613
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.450695322376738e-07,
+      "loss": 2.0117,
+      "step": 47614
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.450221238938053e-07,
+      "loss": 3.2202,
+      "step": 47615
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.449747155499368e-07,
+      "loss": 1.5179,
+      "step": 47616
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.449273072060683e-07,
+      "loss": 1.4272,
+      "step": 47617
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.448798988621997e-07,
+      "loss": 1.6083,
+      "step": 47618
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.448324905183313e-07,
+      "loss": 1.1263,
+      "step": 47619
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.447850821744627e-07,
+      "loss": 2.1973,
+      "step": 47620
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.447376738305942e-07,
+      "loss": 1.8105,
+      "step": 47621
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.446902654867257e-07,
+      "loss": 0.4527,
+      "step": 47622
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.446428571428572e-07,
+      "loss": 3.24,
+      "step": 47623
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.445954487989886e-07,
+      "loss": 3.2038,
+      "step": 47624
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.445480404551201e-07,
+      "loss": 2.3975,
+      "step": 47625
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.445006321112516e-07,
+      "loss": 1.7449,
+      "step": 47626
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.444532237673831e-07,
+      "loss": 2.2014,
+      "step": 47627
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.444058154235145e-07,
+      "loss": 2.1722,
+      "step": 47628
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.44358407079646e-07,
+      "loss": 2.4138,
+      "step": 47629
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.443109987357775e-07,
+      "loss": 0.4062,
+      "step": 47630
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.44263590391909e-07,
+      "loss": 3.55,
+      "step": 47631
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.442161820480405e-07,
+      "loss": 2.4421,
+      "step": 47632
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.441687737041719e-07,
+      "loss": 2.3292,
+      "step": 47633
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.441213653603035e-07,
+      "loss": 1.6704,
+      "step": 47634
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.44073957016435e-07,
+      "loss": 1.5572,
+      "step": 47635
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.440265486725663e-07,
+      "loss": 2.4623,
+      "step": 47636
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.439791403286979e-07,
+      "loss": 1.6285,
+      "step": 47637
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.439317319848293e-07,
+      "loss": 2.6811,
+      "step": 47638
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.438843236409608e-07,
+      "loss": 1.3991,
+      "step": 47639
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.438369152970923e-07,
+      "loss": 1.7824,
+      "step": 47640
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.437895069532238e-07,
+      "loss": 2.1042,
+      "step": 47641
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.437420986093553e-07,
+      "loss": 1.5016,
+      "step": 47642
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.436946902654867e-07,
+      "loss": 1.9907,
+      "step": 47643
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.436472819216182e-07,
+      "loss": 1.5062,
+      "step": 47644
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.435998735777498e-07,
+      "loss": 2.4286,
+      "step": 47645
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.435524652338812e-07,
+      "loss": 1.394,
+      "step": 47646
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.435050568900126e-07,
+      "loss": 1.3106,
+      "step": 47647
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.434576485461441e-07,
+      "loss": 1.6119,
+      "step": 47648
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.434102402022756e-07,
+      "loss": 4.0815,
+      "step": 47649
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.433628318584071e-07,
+      "loss": 3.1107,
+      "step": 47650
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.433154235145385e-07,
+      "loss": 1.5164,
+      "step": 47651
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.432680151706701e-07,
+      "loss": 3.3331,
+      "step": 47652
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.432206068268015e-07,
+      "loss": 3.6487,
+      "step": 47653
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.43173198482933e-07,
+      "loss": 2.0033,
+      "step": 47654
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 7.431257901390645e-07,
+      "loss": 1.6779,
+      "step": 47655
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.43078381795196e-07,
+      "loss": 2.1805,
+      "step": 47656
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.430309734513275e-07,
+      "loss": 3.3872,
+      "step": 47657
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.42983565107459e-07,
+      "loss": 1.1786,
+      "step": 47658
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.429361567635904e-07,
+      "loss": 2.6532,
+      "step": 47659
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.428887484197219e-07,
+      "loss": 2.2775,
+      "step": 47660
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.428413400758533e-07,
+      "loss": 2.5785,
+      "step": 47661
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.427939317319848e-07,
+      "loss": 1.9948,
+      "step": 47662
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.427465233881163e-07,
+      "loss": 2.3029,
+      "step": 47663
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.426991150442478e-07,
+      "loss": 1.2959,
+      "step": 47664
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.426517067003793e-07,
+      "loss": 2.0246,
+      "step": 47665
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.426042983565107e-07,
+      "loss": 3.2459,
+      "step": 47666
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.425568900126423e-07,
+      "loss": 2.1931,
+      "step": 47667
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.425094816687738e-07,
+      "loss": 2.5886,
+      "step": 47668
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.424620733249052e-07,
+      "loss": 3.0687,
+      "step": 47669
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.424146649810368e-07,
+      "loss": 1.3756,
+      "step": 47670
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.423672566371681e-07,
+      "loss": 2.0254,
+      "step": 47671
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.423198482932996e-07,
+      "loss": 0.6448,
+      "step": 47672
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.422724399494311e-07,
+      "loss": 1.6969,
+      "step": 47673
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.422250316055626e-07,
+      "loss": 2.9843,
+      "step": 47674
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.421776232616941e-07,
+      "loss": 3.5612,
+      "step": 47675
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.421302149178255e-07,
+      "loss": 1.8732,
+      "step": 47676
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.42082806573957e-07,
+      "loss": 1.2372,
+      "step": 47677
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.420353982300886e-07,
+      "loss": 2.2774,
+      "step": 47678
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.4198798988622e-07,
+      "loss": 2.2901,
+      "step": 47679
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.419405815423515e-07,
+      "loss": 1.9448,
+      "step": 47680
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.41893173198483e-07,
+      "loss": 1.4504,
+      "step": 47681
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.418457648546144e-07,
+      "loss": 1.6016,
+      "step": 47682
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.417983565107459e-07,
+      "loss": 2.3236,
+      "step": 47683
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.417509481668773e-07,
+      "loss": 2.7668,
+      "step": 47684
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.417035398230089e-07,
+      "loss": 3.1139,
+      "step": 47685
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.416561314791403e-07,
+      "loss": 0.9537,
+      "step": 47686
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.416087231352718e-07,
+      "loss": 2.0308,
+      "step": 47687
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.415613147914033e-07,
+      "loss": 2.8105,
+      "step": 47688
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.415139064475348e-07,
+      "loss": 2.2106,
+      "step": 47689
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.414664981036663e-07,
+      "loss": 1.8945,
+      "step": 47690
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.414190897597977e-07,
+      "loss": 0.9897,
+      "step": 47691
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.413716814159293e-07,
+      "loss": 2.2635,
+      "step": 47692
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.413242730720608e-07,
+      "loss": 1.2518,
+      "step": 47693
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.412768647281921e-07,
+      "loss": 2.8733,
+      "step": 47694
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.412294563843236e-07,
+      "loss": 0.8631,
+      "step": 47695
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.411820480404551e-07,
+      "loss": 0.9753,
+      "step": 47696
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.411346396965866e-07,
+      "loss": 3.3526,
+      "step": 47697
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.410872313527181e-07,
+      "loss": 1.1122,
+      "step": 47698
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.410398230088495e-07,
+      "loss": 0.4927,
+      "step": 47699
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.409924146649811e-07,
+      "loss": 0.9592,
+      "step": 47700
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.409450063211125e-07,
+      "loss": 2.4681,
+      "step": 47701
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.40897597977244e-07,
+      "loss": 2.6478,
+      "step": 47702
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.408501896333756e-07,
+      "loss": 2.2169,
+      "step": 47703
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.40802781289507e-07,
+      "loss": 1.7701,
+      "step": 47704
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.407553729456385e-07,
+      "loss": 2.3483,
+      "step": 47705
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.407079646017699e-07,
+      "loss": 1.2871,
+      "step": 47706
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.406605562579014e-07,
+      "loss": 1.022,
+      "step": 47707
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.406131479140329e-07,
+      "loss": 1.3467,
+      "step": 47708
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.405657395701643e-07,
+      "loss": 0.8341,
+      "step": 47709
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.405183312262958e-07,
+      "loss": 1.9223,
+      "step": 47710
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.404709228824273e-07,
+      "loss": 2.5634,
+      "step": 47711
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.404235145385588e-07,
+      "loss": 1.2135,
+      "step": 47712
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.403761061946903e-07,
+      "loss": 1.9877,
+      "step": 47713
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.403286978508218e-07,
+      "loss": 2.943,
+      "step": 47714
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.402812895069533e-07,
+      "loss": 3.1228,
+      "step": 47715
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.402338811630848e-07,
+      "loss": 1.9182,
+      "step": 47716
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.401864728192161e-07,
+      "loss": 1.6353,
+      "step": 47717
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.401390644753477e-07,
+      "loss": 2.1314,
+      "step": 47718
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 7.400916561314791e-07,
+      "loss": 0.5731,
+      "step": 47719
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.400442477876106e-07,
+      "loss": 1.1794,
+      "step": 47720
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.399968394437421e-07,
+      "loss": 1.5917,
+      "step": 47721
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.399494310998736e-07,
+      "loss": 2.818,
+      "step": 47722
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.399020227560051e-07,
+      "loss": 2.6066,
+      "step": 47723
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.398546144121365e-07,
+      "loss": 2.1669,
+      "step": 47724
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.398072060682681e-07,
+      "loss": 1.5696,
+      "step": 47725
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.397597977243996e-07,
+      "loss": 2.3594,
+      "step": 47726
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.39712389380531e-07,
+      "loss": 2.818,
+      "step": 47727
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.396649810366625e-07,
+      "loss": 4.7917,
+      "step": 47728
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.396175726927939e-07,
+      "loss": 0.6701,
+      "step": 47729
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.395701643489254e-07,
+      "loss": 2.2926,
+      "step": 47730
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.395227560050569e-07,
+      "loss": 1.0002,
+      "step": 47731
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.394753476611883e-07,
+      "loss": 2.1229,
+      "step": 47732
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.394279393173199e-07,
+      "loss": 3.5426,
+      "step": 47733
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.393805309734513e-07,
+      "loss": 2.4041,
+      "step": 47734
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.393331226295828e-07,
+      "loss": 0.6829,
+      "step": 47735
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.392857142857144e-07,
+      "loss": 3.4643,
+      "step": 47736
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.392383059418458e-07,
+      "loss": 2.4892,
+      "step": 47737
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.391908975979773e-07,
+      "loss": 2.1231,
+      "step": 47738
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.391434892541088e-07,
+      "loss": 3.023,
+      "step": 47739
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.390960809102403e-07,
+      "loss": 1.591,
+      "step": 47740
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.390486725663717e-07,
+      "loss": 1.4084,
+      "step": 47741
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.390012642225031e-07,
+      "loss": 1.9716,
+      "step": 47742
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.389538558786346e-07,
+      "loss": 2.5837,
+      "step": 47743
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.389064475347661e-07,
+      "loss": 1.3014,
+      "step": 47744
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.388590391908976e-07,
+      "loss": 2.0336,
+      "step": 47745
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.388116308470291e-07,
+      "loss": 1.7599,
+      "step": 47746
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.387642225031606e-07,
+      "loss": 2.3913,
+      "step": 47747
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.387168141592921e-07,
+      "loss": 2.0101,
+      "step": 47748
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.386694058154236e-07,
+      "loss": 2.7249,
+      "step": 47749
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.38621997471555e-07,
+      "loss": 2.0492,
+      "step": 47750
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.385745891276866e-07,
+      "loss": 0.878,
+      "step": 47751
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.38527180783818e-07,
+      "loss": 2.3402,
+      "step": 47752
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.384797724399494e-07,
+      "loss": 4.4847,
+      "step": 47753
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.384323640960809e-07,
+      "loss": 1.5943,
+      "step": 47754
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.383849557522124e-07,
+      "loss": 0.8928,
+      "step": 47755
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.383375474083439e-07,
+      "loss": 1.0231,
+      "step": 47756
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.382901390644753e-07,
+      "loss": 2.3502,
+      "step": 47757
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.382427307206069e-07,
+      "loss": 2.8869,
+      "step": 47758
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.381953223767383e-07,
+      "loss": 1.9482,
+      "step": 47759
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.381479140328698e-07,
+      "loss": 2.9761,
+      "step": 47760
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.381005056890013e-07,
+      "loss": 1.3438,
+      "step": 47761
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.380530973451328e-07,
+      "loss": 2.9866,
+      "step": 47762
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.380056890012643e-07,
+      "loss": 2.2641,
+      "step": 47763
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.379582806573957e-07,
+      "loss": 2.3371,
+      "step": 47764
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.379108723135271e-07,
+      "loss": 3.1177,
+      "step": 47765
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.378634639696587e-07,
+      "loss": 2.8972,
+      "step": 47766
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.378160556257901e-07,
+      "loss": 1.2247,
+      "step": 47767
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.377686472819216e-07,
+      "loss": 1.3514,
+      "step": 47768
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.377212389380531e-07,
+      "loss": 1.4346,
+      "step": 47769
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.376738305941846e-07,
+      "loss": 1.3376,
+      "step": 47770
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.376264222503161e-07,
+      "loss": 1.4675,
+      "step": 47771
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.375790139064475e-07,
+      "loss": 1.5428,
+      "step": 47772
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.375316055625791e-07,
+      "loss": 0.9327,
+      "step": 47773
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.374841972187106e-07,
+      "loss": 2.4287,
+      "step": 47774
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.37436788874842e-07,
+      "loss": 1.9885,
+      "step": 47775
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.373893805309734e-07,
+      "loss": 0.5225,
+      "step": 47776
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.373419721871049e-07,
+      "loss": 2.285,
+      "step": 47777
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.372945638432364e-07,
+      "loss": 2.1301,
+      "step": 47778
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.372471554993679e-07,
+      "loss": 1.7178,
+      "step": 47779
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.371997471554994e-07,
+      "loss": 1.906,
+      "step": 47780
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.371523388116309e-07,
+      "loss": 1.3637,
+      "step": 47781
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 7.371049304677623e-07,
+      "loss": 1.8934,
+      "step": 47782
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.370575221238938e-07,
+      "loss": 2.1709,
+      "step": 47783
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.370101137800254e-07,
+      "loss": 0.5241,
+      "step": 47784
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.369627054361568e-07,
+      "loss": 2.252,
+      "step": 47785
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.369152970922883e-07,
+      "loss": 2.1999,
+      "step": 47786
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.368678887484198e-07,
+      "loss": 0.9631,
+      "step": 47787
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.368204804045512e-07,
+      "loss": 0.5441,
+      "step": 47788
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.367730720606827e-07,
+      "loss": 1.3693,
+      "step": 47789
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.367256637168141e-07,
+      "loss": 1.1778,
+      "step": 47790
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.366782553729457e-07,
+      "loss": 2.0835,
+      "step": 47791
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.366308470290771e-07,
+      "loss": 2.1418,
+      "step": 47792
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.365834386852086e-07,
+      "loss": 2.3014,
+      "step": 47793
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.365360303413401e-07,
+      "loss": 1.7302,
+      "step": 47794
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.364886219974716e-07,
+      "loss": 2.1852,
+      "step": 47795
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.364412136536031e-07,
+      "loss": 0.733,
+      "step": 47796
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.363938053097346e-07,
+      "loss": 1.9353,
+      "step": 47797
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.36346396965866e-07,
+      "loss": 2.1714,
+      "step": 47798
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.362989886219976e-07,
+      "loss": 2.28,
+      "step": 47799
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.362515802781289e-07,
+      "loss": 2.2786,
+      "step": 47800
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.362041719342604e-07,
+      "loss": 2.141,
+      "step": 47801
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.361567635903919e-07,
+      "loss": 2.444,
+      "step": 47802
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.361093552465234e-07,
+      "loss": 1.3163,
+      "step": 47803
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.360619469026549e-07,
+      "loss": 0.4865,
+      "step": 47804
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.360145385587863e-07,
+      "loss": 2.172,
+      "step": 47805
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.359671302149179e-07,
+      "loss": 1.6166,
+      "step": 47806
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.359197218710494e-07,
+      "loss": 2.7959,
+      "step": 47807
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.358723135271808e-07,
+      "loss": 1.048,
+      "step": 47808
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.358249051833123e-07,
+      "loss": 1.8549,
+      "step": 47809
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.357774968394438e-07,
+      "loss": 2.2222,
+      "step": 47810
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.357300884955752e-07,
+      "loss": 2.6168,
+      "step": 47811
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.356826801517067e-07,
+      "loss": 2.785,
+      "step": 47812
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.356352718078382e-07,
+      "loss": 2.2439,
+      "step": 47813
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.355878634639697e-07,
+      "loss": 0.4704,
+      "step": 47814
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.355404551201011e-07,
+      "loss": 1.5815,
+      "step": 47815
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.354930467762326e-07,
+      "loss": 0.2947,
+      "step": 47816
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.354456384323642e-07,
+      "loss": 2.1367,
+      "step": 47817
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.353982300884956e-07,
+      "loss": 0.6657,
+      "step": 47818
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.353508217446271e-07,
+      "loss": 1.7131,
+      "step": 47819
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.353034134007585e-07,
+      "loss": 1.6967,
+      "step": 47820
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.352560050568901e-07,
+      "loss": 1.4807,
+      "step": 47821
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.352085967130216e-07,
+      "loss": 2.8502,
+      "step": 47822
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.351611883691529e-07,
+      "loss": 2.4989,
+      "step": 47823
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.351137800252845e-07,
+      "loss": 2.7765,
+      "step": 47824
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.350663716814159e-07,
+      "loss": 3.8404,
+      "step": 47825
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.350189633375474e-07,
+      "loss": 2.94,
+      "step": 47826
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.349715549936789e-07,
+      "loss": 1.4405,
+      "step": 47827
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.349241466498104e-07,
+      "loss": 2.1273,
+      "step": 47828
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.348767383059419e-07,
+      "loss": 1.5435,
+      "step": 47829
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.348293299620733e-07,
+      "loss": 1.3919,
+      "step": 47830
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.347819216182048e-07,
+      "loss": 1.2587,
+      "step": 47831
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.347345132743364e-07,
+      "loss": 1.8646,
+      "step": 47832
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.346871049304678e-07,
+      "loss": 2.3365,
+      "step": 47833
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.346396965865992e-07,
+      "loss": 1.7649,
+      "step": 47834
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.345922882427307e-07,
+      "loss": 1.4692,
+      "step": 47835
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.345448798988622e-07,
+      "loss": 2.8468,
+      "step": 47836
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.344974715549937e-07,
+      "loss": 2.1027,
+      "step": 47837
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.344500632111251e-07,
+      "loss": 2.1029,
+      "step": 47838
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.344026548672567e-07,
+      "loss": 2.5501,
+      "step": 47839
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.343552465233881e-07,
+      "loss": 1.8038,
+      "step": 47840
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.343078381795196e-07,
+      "loss": 2.4095,
+      "step": 47841
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.342604298356511e-07,
+      "loss": 1.8468,
+      "step": 47842
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.342130214917826e-07,
+      "loss": 1.771,
+      "step": 47843
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.341656131479141e-07,
+      "loss": 2.3761,
+      "step": 47844
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 7.341182048040456e-07,
+      "loss": 1.6964,
+      "step": 47845
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.340707964601769e-07,
+      "loss": 1.8377,
+      "step": 47846
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.340233881163085e-07,
+      "loss": 1.8339,
+      "step": 47847
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.339759797724399e-07,
+      "loss": 3.3502,
+      "step": 47848
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.339285714285714e-07,
+      "loss": 1.4927,
+      "step": 47849
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.338811630847029e-07,
+      "loss": 1.113,
+      "step": 47850
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.338337547408344e-07,
+      "loss": 1.3199,
+      "step": 47851
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.337863463969659e-07,
+      "loss": 1.8343,
+      "step": 47852
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.337389380530973e-07,
+      "loss": 1.8596,
+      "step": 47853
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.336915297092289e-07,
+      "loss": 2.7838,
+      "step": 47854
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.336441213653604e-07,
+      "loss": 1.4877,
+      "step": 47855
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.335967130214918e-07,
+      "loss": 1.0513,
+      "step": 47856
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.335493046776233e-07,
+      "loss": 4.307,
+      "step": 47857
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.335018963337547e-07,
+      "loss": 2.9298,
+      "step": 47858
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.334544879898862e-07,
+      "loss": 1.2426,
+      "step": 47859
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.334070796460177e-07,
+      "loss": 2.7841,
+      "step": 47860
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.333596713021492e-07,
+      "loss": 1.1325,
+      "step": 47861
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.333122629582807e-07,
+      "loss": 3.8627,
+      "step": 47862
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.332648546144121e-07,
+      "loss": 2.0994,
+      "step": 47863
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.332174462705436e-07,
+      "loss": 2.4339,
+      "step": 47864
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.331700379266752e-07,
+      "loss": 2.517,
+      "step": 47865
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.331226295828066e-07,
+      "loss": 2.3023,
+      "step": 47866
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.330752212389381e-07,
+      "loss": 1.8313,
+      "step": 47867
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.330278128950696e-07,
+      "loss": 1.3343,
+      "step": 47868
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.32980404551201e-07,
+      "loss": 1.0831,
+      "step": 47869
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.329329962073325e-07,
+      "loss": 1.4592,
+      "step": 47870
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.328855878634639e-07,
+      "loss": 2.0015,
+      "step": 47871
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.328381795195955e-07,
+      "loss": 2.4444,
+      "step": 47872
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.327907711757269e-07,
+      "loss": 1.626,
+      "step": 47873
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.327433628318584e-07,
+      "loss": 1.96,
+      "step": 47874
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.326959544879899e-07,
+      "loss": 1.983,
+      "step": 47875
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.326485461441214e-07,
+      "loss": 2.3617,
+      "step": 47876
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.326011378002529e-07,
+      "loss": 1.3074,
+      "step": 47877
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.325537294563844e-07,
+      "loss": 2.1929,
+      "step": 47878
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.325063211125158e-07,
+      "loss": 2.4656,
+      "step": 47879
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.324589127686474e-07,
+      "loss": 2.4675,
+      "step": 47880
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.324115044247787e-07,
+      "loss": 2.4606,
+      "step": 47881
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.323640960809102e-07,
+      "loss": 1.0945,
+      "step": 47882
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.323166877370417e-07,
+      "loss": 1.3491,
+      "step": 47883
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.322692793931732e-07,
+      "loss": 2.3443,
+      "step": 47884
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.322218710493047e-07,
+      "loss": 1.673,
+      "step": 47885
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.321744627054361e-07,
+      "loss": 1.8843,
+      "step": 47886
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.321270543615677e-07,
+      "loss": 1.942,
+      "step": 47887
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.320796460176992e-07,
+      "loss": 1.0487,
+      "step": 47888
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.320322376738306e-07,
+      "loss": 1.716,
+      "step": 47889
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.319848293299621e-07,
+      "loss": 2.7355,
+      "step": 47890
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.319374209860936e-07,
+      "loss": 3.1378,
+      "step": 47891
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.318900126422251e-07,
+      "loss": 2.4592,
+      "step": 47892
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.318426042983565e-07,
+      "loss": 1.7863,
+      "step": 47893
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.31795195954488e-07,
+      "loss": 3.8518,
+      "step": 47894
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.317477876106195e-07,
+      "loss": 0.7869,
+      "step": 47895
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.317003792667509e-07,
+      "loss": 2.428,
+      "step": 47896
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.316529709228824e-07,
+      "loss": 2.9556,
+      "step": 47897
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.316055625790139e-07,
+      "loss": 2.6106,
+      "step": 47898
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.315581542351454e-07,
+      "loss": 0.6808,
+      "step": 47899
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.315107458912769e-07,
+      "loss": 2.8583,
+      "step": 47900
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.314633375474083e-07,
+      "loss": 2.7487,
+      "step": 47901
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.314159292035399e-07,
+      "loss": 3.0674,
+      "step": 47902
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.313685208596714e-07,
+      "loss": 1.8005,
+      "step": 47903
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.313211125158027e-07,
+      "loss": 1.9363,
+      "step": 47904
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.312737041719343e-07,
+      "loss": 1.3349,
+      "step": 47905
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.312262958280657e-07,
+      "loss": 1.6072,
+      "step": 47906
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.311788874841972e-07,
+      "loss": 2.075,
+      "step": 47907
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.311314791403287e-07,
+      "loss": 1.074,
+      "step": 47908
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 7.310840707964602e-07,
+      "loss": 0.8576,
+      "step": 47909
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.310366624525917e-07,
+      "loss": 1.3007,
+      "step": 47910
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.309892541087231e-07,
+      "loss": 0.9931,
+      "step": 47911
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.309418457648546e-07,
+      "loss": 2.7127,
+      "step": 47912
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.308944374209862e-07,
+      "loss": 2.6652,
+      "step": 47913
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.308470290771176e-07,
+      "loss": 2.5215,
+      "step": 47914
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.307996207332491e-07,
+      "loss": 1.1802,
+      "step": 47915
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.307522123893805e-07,
+      "loss": 0.8963,
+      "step": 47916
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.30704804045512e-07,
+      "loss": 3.3497,
+      "step": 47917
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.306573957016435e-07,
+      "loss": 2.2186,
+      "step": 47918
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.306099873577749e-07,
+      "loss": 1.8581,
+      "step": 47919
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.305625790139065e-07,
+      "loss": 1.0723,
+      "step": 47920
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.305151706700379e-07,
+      "loss": 1.7557,
+      "step": 47921
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.304677623261694e-07,
+      "loss": 2.1885,
+      "step": 47922
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.304203539823009e-07,
+      "loss": 2.3175,
+      "step": 47923
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.303729456384324e-07,
+      "loss": 1.2776,
+      "step": 47924
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.303255372945639e-07,
+      "loss": 0.8608,
+      "step": 47925
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.302781289506954e-07,
+      "loss": 1.0369,
+      "step": 47926
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.302307206068269e-07,
+      "loss": 1.6648,
+      "step": 47927
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.301833122629583e-07,
+      "loss": 1.9958,
+      "step": 47928
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.301359039190897e-07,
+      "loss": 1.6986,
+      "step": 47929
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.300884955752212e-07,
+      "loss": 1.238,
+      "step": 47930
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.300410872313527e-07,
+      "loss": 0.299,
+      "step": 47931
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.299936788874842e-07,
+      "loss": 1.6653,
+      "step": 47932
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.299462705436157e-07,
+      "loss": 1.5829,
+      "step": 47933
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.298988621997471e-07,
+      "loss": 1.1866,
+      "step": 47934
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.298514538558787e-07,
+      "loss": 2.3166,
+      "step": 47935
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.298040455120102e-07,
+      "loss": 1.7556,
+      "step": 47936
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.297566371681416e-07,
+      "loss": 0.2063,
+      "step": 47937
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.297092288242732e-07,
+      "loss": 1.1408,
+      "step": 47938
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.296618204804046e-07,
+      "loss": 2.7321,
+      "step": 47939
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.29614412136536e-07,
+      "loss": 1.3857,
+      "step": 47940
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.295670037926675e-07,
+      "loss": 2.5571,
+      "step": 47941
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.29519595448799e-07,
+      "loss": 3.0114,
+      "step": 47942
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.294721871049305e-07,
+      "loss": 0.8349,
+      "step": 47943
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.294247787610619e-07,
+      "loss": 2.5303,
+      "step": 47944
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.293773704171934e-07,
+      "loss": 3.3105,
+      "step": 47945
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.29329962073325e-07,
+      "loss": 2.6331,
+      "step": 47946
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.292825537294564e-07,
+      "loss": 0.5446,
+      "step": 47947
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.292351453855879e-07,
+      "loss": 2.77,
+      "step": 47948
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.291877370417194e-07,
+      "loss": 2.1474,
+      "step": 47949
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.291403286978509e-07,
+      "loss": 1.5561,
+      "step": 47950
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.290929203539824e-07,
+      "loss": 0.9561,
+      "step": 47951
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.290455120101137e-07,
+      "loss": 1.9873,
+      "step": 47952
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.289981036662453e-07,
+      "loss": 2.5912,
+      "step": 47953
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.289506953223767e-07,
+      "loss": 2.2753,
+      "step": 47954
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.289032869785082e-07,
+      "loss": 2.0811,
+      "step": 47955
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.288558786346397e-07,
+      "loss": 1.72,
+      "step": 47956
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.288084702907712e-07,
+      "loss": 2.5562,
+      "step": 47957
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.287610619469027e-07,
+      "loss": 2.2389,
+      "step": 47958
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.287136536030342e-07,
+      "loss": 2.0827,
+      "step": 47959
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.286662452591657e-07,
+      "loss": 1.0552,
+      "step": 47960
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.286188369152972e-07,
+      "loss": 2.6613,
+      "step": 47961
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.285714285714286e-07,
+      "loss": 2.6108,
+      "step": 47962
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.2852402022756e-07,
+      "loss": 2.1822,
+      "step": 47963
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.284766118836915e-07,
+      "loss": 2.0884,
+      "step": 47964
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.28429203539823e-07,
+      "loss": 0.1999,
+      "step": 47965
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.283817951959545e-07,
+      "loss": 0.6471,
+      "step": 47966
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.283343868520859e-07,
+      "loss": 1.9928,
+      "step": 47967
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.282869785082175e-07,
+      "loss": 0.917,
+      "step": 47968
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.282395701643489e-07,
+      "loss": 2.5007,
+      "step": 47969
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.281921618204804e-07,
+      "loss": 2.0553,
+      "step": 47970
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.28144753476612e-07,
+      "loss": 2.4774,
+      "step": 47971
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 7.280973451327434e-07,
+      "loss": 1.0319,
+      "step": 47972
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.280499367888749e-07,
+      "loss": 3.1312,
+      "step": 47973
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.280025284450064e-07,
+      "loss": 2.0145,
+      "step": 47974
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.279551201011378e-07,
+      "loss": 0.7914,
+      "step": 47975
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.279077117572693e-07,
+      "loss": 0.3559,
+      "step": 47976
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.278603034134007e-07,
+      "loss": 2.2807,
+      "step": 47977
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.278128950695322e-07,
+      "loss": 2.2903,
+      "step": 47978
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.277654867256637e-07,
+      "loss": 1.2853,
+      "step": 47979
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.277180783817952e-07,
+      "loss": 1.6836,
+      "step": 47980
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.276706700379267e-07,
+      "loss": 0.7101,
+      "step": 47981
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.276232616940582e-07,
+      "loss": 0.8598,
+      "step": 47982
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.275758533501897e-07,
+      "loss": 1.4665,
+      "step": 47983
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.275284450063212e-07,
+      "loss": 1.6084,
+      "step": 47984
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.274810366624526e-07,
+      "loss": 2.7549,
+      "step": 47985
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.274336283185842e-07,
+      "loss": 2.9044,
+      "step": 47986
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.273862199747155e-07,
+      "loss": 1.4596,
+      "step": 47987
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.27338811630847e-07,
+      "loss": 1.7209,
+      "step": 47988
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.272914032869785e-07,
+      "loss": 2.3649,
+      "step": 47989
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.2724399494311e-07,
+      "loss": 1.1371,
+      "step": 47990
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.271965865992415e-07,
+      "loss": 2.7373,
+      "step": 47991
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.271491782553729e-07,
+      "loss": 1.0967,
+      "step": 47992
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.271017699115045e-07,
+      "loss": 2.0696,
+      "step": 47993
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.27054361567636e-07,
+      "loss": 0.7566,
+      "step": 47994
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.270069532237674e-07,
+      "loss": 1.4103,
+      "step": 47995
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.269595448798989e-07,
+      "loss": 1.1029,
+      "step": 47996
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.269121365360304e-07,
+      "loss": 1.8321,
+      "step": 47997
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.268647281921618e-07,
+      "loss": 0.8299,
+      "step": 47998
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.268173198482933e-07,
+      "loss": 1.3634,
+      "step": 47999
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.267699115044247e-07,
+      "loss": 1.0627,
+      "step": 48000
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.267225031605563e-07,
+      "loss": 3.6164,
+      "step": 48001
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.266750948166877e-07,
+      "loss": 2.0096,
+      "step": 48002
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.266276864728192e-07,
+      "loss": 2.7256,
+      "step": 48003
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.265802781289508e-07,
+      "loss": 1.2198,
+      "step": 48004
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.265328697850822e-07,
+      "loss": 2.3238,
+      "step": 48005
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.264854614412137e-07,
+      "loss": 1.3628,
+      "step": 48006
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.264380530973452e-07,
+      "loss": 1.3016,
+      "step": 48007
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.263906447534767e-07,
+      "loss": 2.0841,
+      "step": 48008
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.263432364096082e-07,
+      "loss": 1.2253,
+      "step": 48009
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.262958280657395e-07,
+      "loss": 0.6145,
+      "step": 48010
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.26248419721871e-07,
+      "loss": 2.1614,
+      "step": 48011
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.262010113780025e-07,
+      "loss": 1.4937,
+      "step": 48012
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.26153603034134e-07,
+      "loss": 1.8844,
+      "step": 48013
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.261061946902655e-07,
+      "loss": 2.6203,
+      "step": 48014
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.26058786346397e-07,
+      "loss": 3.0418,
+      "step": 48015
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.260113780025285e-07,
+      "loss": 3.3771,
+      "step": 48016
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.2596396965866e-07,
+      "loss": 0.9658,
+      "step": 48017
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.259165613147914e-07,
+      "loss": 2.5381,
+      "step": 48018
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.25869152970923e-07,
+      "loss": 2.1003,
+      "step": 48019
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.258217446270544e-07,
+      "loss": 1.4769,
+      "step": 48020
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.257743362831859e-07,
+      "loss": 1.7675,
+      "step": 48021
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.257269279393173e-07,
+      "loss": 2.599,
+      "step": 48022
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.256795195954488e-07,
+      "loss": 0.6014,
+      "step": 48023
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.256321112515803e-07,
+      "loss": 0.8351,
+      "step": 48024
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.255847029077117e-07,
+      "loss": 1.7163,
+      "step": 48025
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.255372945638433e-07,
+      "loss": 1.8885,
+      "step": 48026
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.254898862199748e-07,
+      "loss": 2.6479,
+      "step": 48027
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.254424778761062e-07,
+      "loss": 1.7487,
+      "step": 48028
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.253950695322377e-07,
+      "loss": 1.8834,
+      "step": 48029
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.253476611883692e-07,
+      "loss": 1.2287,
+      "step": 48030
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.253002528445007e-07,
+      "loss": 1.9591,
+      "step": 48031
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.252528445006322e-07,
+      "loss": 2.1324,
+      "step": 48032
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.252054361567635e-07,
+      "loss": 1.1501,
+      "step": 48033
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.251580278128951e-07,
+      "loss": 1.246,
+      "step": 48034
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 7.251106194690265e-07,
+      "loss": 1.1408,
+      "step": 48035
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.25063211125158e-07,
+      "loss": 2.0983,
+      "step": 48036
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.250158027812895e-07,
+      "loss": 2.6381,
+      "step": 48037
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.24968394437421e-07,
+      "loss": 2.7104,
+      "step": 48038
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.249209860935525e-07,
+      "loss": 1.3716,
+      "step": 48039
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.248735777496839e-07,
+      "loss": 3.4166,
+      "step": 48040
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.248261694058155e-07,
+      "loss": 2.8458,
+      "step": 48041
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.24778761061947e-07,
+      "loss": 1.2324,
+      "step": 48042
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.247313527180784e-07,
+      "loss": 3.9044,
+      "step": 48043
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.246839443742099e-07,
+      "loss": 1.9203,
+      "step": 48044
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.246365360303413e-07,
+      "loss": 2.3967,
+      "step": 48045
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.245891276864728e-07,
+      "loss": 2.4484,
+      "step": 48046
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.245417193426043e-07,
+      "loss": 1.7808,
+      "step": 48047
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.244943109987358e-07,
+      "loss": 1.5885,
+      "step": 48048
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.244469026548673e-07,
+      "loss": 2.7696,
+      "step": 48049
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.243994943109987e-07,
+      "loss": 1.6537,
+      "step": 48050
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.243520859671302e-07,
+      "loss": 1.4804,
+      "step": 48051
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.243046776232618e-07,
+      "loss": 2.519,
+      "step": 48052
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.242572692793932e-07,
+      "loss": 0.4943,
+      "step": 48053
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.242098609355247e-07,
+      "loss": 3.0573,
+      "step": 48054
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.241624525916562e-07,
+      "loss": 2.8421,
+      "step": 48055
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.241150442477877e-07,
+      "loss": 2.3067,
+      "step": 48056
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.240676359039191e-07,
+      "loss": 1.9055,
+      "step": 48057
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.240202275600505e-07,
+      "loss": 1.6676,
+      "step": 48058
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.239728192161821e-07,
+      "loss": 1.6255,
+      "step": 48059
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.239254108723135e-07,
+      "loss": 2.6154,
+      "step": 48060
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.23878002528445e-07,
+      "loss": 1.6626,
+      "step": 48061
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.238305941845765e-07,
+      "loss": 2.354,
+      "step": 48062
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.23783185840708e-07,
+      "loss": 0.8968,
+      "step": 48063
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.237357774968395e-07,
+      "loss": 1.8315,
+      "step": 48064
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.23688369152971e-07,
+      "loss": 0.817,
+      "step": 48065
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.236409608091024e-07,
+      "loss": 1.9976,
+      "step": 48066
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.23593552465234e-07,
+      "loss": 2.1142,
+      "step": 48067
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.235461441213653e-07,
+      "loss": 1.4477,
+      "step": 48068
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.234987357774968e-07,
+      "loss": 2.8224,
+      "step": 48069
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.234513274336283e-07,
+      "loss": 1.35,
+      "step": 48070
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.234039190897598e-07,
+      "loss": 1.088,
+      "step": 48071
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.233565107458913e-07,
+      "loss": 2.3976,
+      "step": 48072
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.233091024020227e-07,
+      "loss": 3.6625,
+      "step": 48073
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.232616940581543e-07,
+      "loss": 1.9528,
+      "step": 48074
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.232142857142858e-07,
+      "loss": 0.8903,
+      "step": 48075
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.231668773704172e-07,
+      "loss": 1.5063,
+      "step": 48076
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.231194690265487e-07,
+      "loss": 1.7739,
+      "step": 48077
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.230720606826802e-07,
+      "loss": 0.8705,
+      "step": 48078
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.230246523388117e-07,
+      "loss": 2.7266,
+      "step": 48079
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.229772439949431e-07,
+      "loss": 2.051,
+      "step": 48080
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.229298356510745e-07,
+      "loss": 1.9223,
+      "step": 48081
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.228824273072061e-07,
+      "loss": 1.6307,
+      "step": 48082
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.228350189633375e-07,
+      "loss": 2.0765,
+      "step": 48083
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.22787610619469e-07,
+      "loss": 3.4779,
+      "step": 48084
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.227402022756006e-07,
+      "loss": 2.6362,
+      "step": 48085
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.22692793931732e-07,
+      "loss": 1.1485,
+      "step": 48086
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.226453855878635e-07,
+      "loss": 1.7596,
+      "step": 48087
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.22597977243995e-07,
+      "loss": 3.0125,
+      "step": 48088
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.225505689001265e-07,
+      "loss": 1.3963,
+      "step": 48089
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.22503160556258e-07,
+      "loss": 2.3286,
+      "step": 48090
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.224557522123894e-07,
+      "loss": 1.4138,
+      "step": 48091
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.224083438685208e-07,
+      "loss": 1.3053,
+      "step": 48092
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.223609355246523e-07,
+      "loss": 1.6637,
+      "step": 48093
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.223135271807838e-07,
+      "loss": 2.4497,
+      "step": 48094
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.222661188369153e-07,
+      "loss": 1.8923,
+      "step": 48095
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.222187104930468e-07,
+      "loss": 1.0789,
+      "step": 48096
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.221713021491783e-07,
+      "loss": 1.5695,
+      "step": 48097
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.221238938053098e-07,
+      "loss": 2.0062,
+      "step": 48098
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 7.220764854614412e-07,
+      "loss": 1.2664,
+      "step": 48099
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.220290771175728e-07,
+      "loss": 3.4386,
+      "step": 48100
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.219816687737042e-07,
+      "loss": 0.9612,
+      "step": 48101
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.219342604298357e-07,
+      "loss": 2.5779,
+      "step": 48102
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.218868520859672e-07,
+      "loss": 2.5671,
+      "step": 48103
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.218394437420986e-07,
+      "loss": 1.4665,
+      "step": 48104
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.217920353982301e-07,
+      "loss": 1.8742,
+      "step": 48105
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.217446270543615e-07,
+      "loss": 2.6415,
+      "step": 48106
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.216972187104931e-07,
+      "loss": 3.7964,
+      "step": 48107
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.216498103666245e-07,
+      "loss": 2.16,
+      "step": 48108
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.21602402022756e-07,
+      "loss": 1.938,
+      "step": 48109
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.215549936788875e-07,
+      "loss": 1.5531,
+      "step": 48110
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.21507585335019e-07,
+      "loss": 2.1483,
+      "step": 48111
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.214601769911505e-07,
+      "loss": 1.0731,
+      "step": 48112
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.21412768647282e-07,
+      "loss": 3.1095,
+      "step": 48113
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.213653603034134e-07,
+      "loss": 2.2045,
+      "step": 48114
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.21317951959545e-07,
+      "loss": 2.1422,
+      "step": 48115
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.212705436156763e-07,
+      "loss": 2.7785,
+      "step": 48116
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.212231352718078e-07,
+      "loss": 1.6146,
+      "step": 48117
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.211757269279393e-07,
+      "loss": 2.164,
+      "step": 48118
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.211283185840708e-07,
+      "loss": 1.7501,
+      "step": 48119
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.210809102402023e-07,
+      "loss": 1.0648,
+      "step": 48120
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.210335018963337e-07,
+      "loss": 2.5703,
+      "step": 48121
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.209860935524653e-07,
+      "loss": 1.1848,
+      "step": 48122
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.209386852085968e-07,
+      "loss": 2.0546,
+      "step": 48123
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.208912768647282e-07,
+      "loss": 0.8075,
+      "step": 48124
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.208438685208597e-07,
+      "loss": 5.2314,
+      "step": 48125
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.207964601769912e-07,
+      "loss": 1.8686,
+      "step": 48126
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.207490518331226e-07,
+      "loss": 2.1889,
+      "step": 48127
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.207016434892541e-07,
+      "loss": 3.4267,
+      "step": 48128
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.206542351453856e-07,
+      "loss": 1.146,
+      "step": 48129
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.206068268015171e-07,
+      "loss": 2.137,
+      "step": 48130
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.205594184576485e-07,
+      "loss": 2.5659,
+      "step": 48131
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.2051201011378e-07,
+      "loss": 2.0346,
+      "step": 48132
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.204646017699116e-07,
+      "loss": 0.3793,
+      "step": 48133
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.20417193426043e-07,
+      "loss": 1.421,
+      "step": 48134
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.203697850821745e-07,
+      "loss": 1.7391,
+      "step": 48135
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.20322376738306e-07,
+      "loss": 1.4656,
+      "step": 48136
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.202749683944375e-07,
+      "loss": 1.5474,
+      "step": 48137
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.20227560050569e-07,
+      "loss": 1.1173,
+      "step": 48138
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.201801517067003e-07,
+      "loss": 2.264,
+      "step": 48139
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.201327433628319e-07,
+      "loss": 1.9854,
+      "step": 48140
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.200853350189633e-07,
+      "loss": 2.1478,
+      "step": 48141
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.200379266750948e-07,
+      "loss": 2.3002,
+      "step": 48142
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.199905183312263e-07,
+      "loss": 2.6898,
+      "step": 48143
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.199431099873578e-07,
+      "loss": 0.8599,
+      "step": 48144
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.198957016434893e-07,
+      "loss": 1.9509,
+      "step": 48145
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.198482932996208e-07,
+      "loss": 1.6672,
+      "step": 48146
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.198008849557522e-07,
+      "loss": 2.583,
+      "step": 48147
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.197534766118838e-07,
+      "loss": 2.4327,
+      "step": 48148
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.197060682680152e-07,
+      "loss": 1.3447,
+      "step": 48149
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.196586599241466e-07,
+      "loss": 1.3097,
+      "step": 48150
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.196112515802781e-07,
+      "loss": 1.8749,
+      "step": 48151
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.195638432364096e-07,
+      "loss": 2.9839,
+      "step": 48152
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.195164348925411e-07,
+      "loss": 4.0688,
+      "step": 48153
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.194690265486725e-07,
+      "loss": 4.195,
+      "step": 48154
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.194216182048041e-07,
+      "loss": 2.6591,
+      "step": 48155
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.193742098609356e-07,
+      "loss": 1.4545,
+      "step": 48156
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.19326801517067e-07,
+      "loss": 1.5358,
+      "step": 48157
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.192793931731985e-07,
+      "loss": 1.5818,
+      "step": 48158
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.1923198482933e-07,
+      "loss": 1.812,
+      "step": 48159
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.191845764854615e-07,
+      "loss": 1.5323,
+      "step": 48160
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.19137168141593e-07,
+      "loss": 1.6083,
+      "step": 48161
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 7.190897597977244e-07,
+      "loss": 1.5767,
+      "step": 48162
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.190423514538559e-07,
+      "loss": 1.6501,
+      "step": 48163
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.189949431099873e-07,
+      "loss": 2.3992,
+      "step": 48164
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.189475347661188e-07,
+      "loss": 1.4344,
+      "step": 48165
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.189001264222504e-07,
+      "loss": 0.7415,
+      "step": 48166
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.188527180783818e-07,
+      "loss": 1.2114,
+      "step": 48167
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.188053097345133e-07,
+      "loss": 2.1762,
+      "step": 48168
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.187579013906447e-07,
+      "loss": 2.3022,
+      "step": 48169
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.187104930467763e-07,
+      "loss": 0.9237,
+      "step": 48170
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.186630847029078e-07,
+      "loss": 2.6615,
+      "step": 48171
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.186156763590392e-07,
+      "loss": 1.8478,
+      "step": 48172
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.185682680151708e-07,
+      "loss": 1.5202,
+      "step": 48173
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.185208596713021e-07,
+      "loss": 2.2947,
+      "step": 48174
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.184734513274336e-07,
+      "loss": 1.5883,
+      "step": 48175
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.184260429835651e-07,
+      "loss": 2.2551,
+      "step": 48176
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.183786346396966e-07,
+      "loss": 2.5017,
+      "step": 48177
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.183312262958281e-07,
+      "loss": 2.507,
+      "step": 48178
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.182838179519595e-07,
+      "loss": 0.4969,
+      "step": 48179
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.18236409608091e-07,
+      "loss": 2.6288,
+      "step": 48180
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.181890012642226e-07,
+      "loss": 2.5712,
+      "step": 48181
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.18141592920354e-07,
+      "loss": 1.1577,
+      "step": 48182
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.180941845764855e-07,
+      "loss": 1.1396,
+      "step": 48183
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.18046776232617e-07,
+      "loss": 1.1729,
+      "step": 48184
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.179993678887485e-07,
+      "loss": 2.2575,
+      "step": 48185
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.179519595448799e-07,
+      "loss": 0.8094,
+      "step": 48186
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.179045512010113e-07,
+      "loss": 2.8447,
+      "step": 48187
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.178571428571429e-07,
+      "loss": 4.0652,
+      "step": 48188
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.178097345132743e-07,
+      "loss": 2.238,
+      "step": 48189
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.177623261694058e-07,
+      "loss": 2.215,
+      "step": 48190
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.177149178255373e-07,
+      "loss": 1.4707,
+      "step": 48191
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.176675094816688e-07,
+      "loss": 1.3887,
+      "step": 48192
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.176201011378003e-07,
+      "loss": 2.5446,
+      "step": 48193
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.175726927939318e-07,
+      "loss": 1.7619,
+      "step": 48194
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.175252844500633e-07,
+      "loss": 1.3534,
+      "step": 48195
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.174778761061948e-07,
+      "loss": 2.9466,
+      "step": 48196
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.174304677623261e-07,
+      "loss": 1.3147,
+      "step": 48197
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.173830594184576e-07,
+      "loss": 2.529,
+      "step": 48198
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.173356510745891e-07,
+      "loss": 2.4693,
+      "step": 48199
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.172882427307206e-07,
+      "loss": 2.5536,
+      "step": 48200
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.172408343868521e-07,
+      "loss": 3.9769,
+      "step": 48201
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.171934260429835e-07,
+      "loss": 2.2486,
+      "step": 48202
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.171460176991151e-07,
+      "loss": 1.9476,
+      "step": 48203
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.170986093552466e-07,
+      "loss": 1.2157,
+      "step": 48204
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.17051201011378e-07,
+      "loss": 2.6518,
+      "step": 48205
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.170037926675096e-07,
+      "loss": 2.1151,
+      "step": 48206
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.16956384323641e-07,
+      "loss": 2.3204,
+      "step": 48207
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.169089759797725e-07,
+      "loss": 2.2605,
+      "step": 48208
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.168615676359039e-07,
+      "loss": 2.5513,
+      "step": 48209
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.168141592920354e-07,
+      "loss": 1.5894,
+      "step": 48210
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.167667509481669e-07,
+      "loss": 3.41,
+      "step": 48211
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.167193426042983e-07,
+      "loss": 1.9954,
+      "step": 48212
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.166719342604298e-07,
+      "loss": 4.0669,
+      "step": 48213
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.166245259165614e-07,
+      "loss": 3.5164,
+      "step": 48214
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.165771175726928e-07,
+      "loss": 2.3509,
+      "step": 48215
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.165297092288243e-07,
+      "loss": 0.6146,
+      "step": 48216
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.164823008849558e-07,
+      "loss": 2.189,
+      "step": 48217
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.164348925410873e-07,
+      "loss": 1.9379,
+      "step": 48218
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.163874841972188e-07,
+      "loss": 3.298,
+      "step": 48219
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.163400758533501e-07,
+      "loss": 1.8859,
+      "step": 48220
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.162926675094817e-07,
+      "loss": 2.8982,
+      "step": 48221
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.162452591656131e-07,
+      "loss": 2.4415,
+      "step": 48222
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.161978508217446e-07,
+      "loss": 2.2296,
+      "step": 48223
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.161504424778761e-07,
+      "loss": 1.9503,
+      "step": 48224
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 7.161030341340076e-07,
+      "loss": 1.5078,
+      "step": 48225
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.160556257901391e-07,
+      "loss": 2.1009,
+      "step": 48226
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.160082174462706e-07,
+      "loss": 1.3154,
+      "step": 48227
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.159608091024021e-07,
+      "loss": 2.0377,
+      "step": 48228
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.159134007585336e-07,
+      "loss": 2.667,
+      "step": 48229
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.15865992414665e-07,
+      "loss": 1.1221,
+      "step": 48230
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.158185840707965e-07,
+      "loss": 2.7543,
+      "step": 48231
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.15771175726928e-07,
+      "loss": 1.1485,
+      "step": 48232
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.157237673830594e-07,
+      "loss": 1.617,
+      "step": 48233
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.156763590391909e-07,
+      "loss": 1.4185,
+      "step": 48234
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.156289506953223e-07,
+      "loss": 1.4923,
+      "step": 48235
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.155815423514539e-07,
+      "loss": 3.7931,
+      "step": 48236
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.155341340075854e-07,
+      "loss": 1.4446,
+      "step": 48237
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.154867256637168e-07,
+      "loss": 2.1588,
+      "step": 48238
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.154393173198484e-07,
+      "loss": 1.1767,
+      "step": 48239
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.153919089759798e-07,
+      "loss": 1.5222,
+      "step": 48240
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.153445006321113e-07,
+      "loss": 1.0467,
+      "step": 48241
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.152970922882428e-07,
+      "loss": 3.156,
+      "step": 48242
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.152496839443743e-07,
+      "loss": 0.3108,
+      "step": 48243
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.152022756005057e-07,
+      "loss": 2.7085,
+      "step": 48244
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.151548672566371e-07,
+      "loss": 2.0728,
+      "step": 48245
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.151074589127686e-07,
+      "loss": 1.4302,
+      "step": 48246
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.150600505689001e-07,
+      "loss": 1.7208,
+      "step": 48247
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.150126422250316e-07,
+      "loss": 2.3585,
+      "step": 48248
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.149652338811631e-07,
+      "loss": 3.6917,
+      "step": 48249
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.149178255372946e-07,
+      "loss": 2.2412,
+      "step": 48250
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.148704171934261e-07,
+      "loss": 2.8408,
+      "step": 48251
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.148230088495576e-07,
+      "loss": 1.736,
+      "step": 48252
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.14775600505689e-07,
+      "loss": 2.3533,
+      "step": 48253
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.147281921618206e-07,
+      "loss": 1.256,
+      "step": 48254
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.14680783817952e-07,
+      "loss": 1.4929,
+      "step": 48255
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.146333754740834e-07,
+      "loss": 1.3959,
+      "step": 48256
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.145859671302149e-07,
+      "loss": 2.5896,
+      "step": 48257
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.145385587863464e-07,
+      "loss": 1.9855,
+      "step": 48258
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.144911504424779e-07,
+      "loss": 3.6618,
+      "step": 48259
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.144437420986093e-07,
+      "loss": 2.1955,
+      "step": 48260
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.143963337547409e-07,
+      "loss": 2.8694,
+      "step": 48261
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.143489254108724e-07,
+      "loss": 0.9154,
+      "step": 48262
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.143015170670038e-07,
+      "loss": 1.3971,
+      "step": 48263
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.142541087231353e-07,
+      "loss": 3.7965,
+      "step": 48264
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.142067003792668e-07,
+      "loss": 2.2477,
+      "step": 48265
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.141592920353983e-07,
+      "loss": 2.0422,
+      "step": 48266
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.141118836915297e-07,
+      "loss": 0.9479,
+      "step": 48267
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.140644753476611e-07,
+      "loss": 1.5728,
+      "step": 48268
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.140170670037927e-07,
+      "loss": 1.4448,
+      "step": 48269
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.139696586599241e-07,
+      "loss": 2.9755,
+      "step": 48270
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.139222503160556e-07,
+      "loss": 2.6883,
+      "step": 48271
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.138748419721872e-07,
+      "loss": 3.6303,
+      "step": 48272
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.138274336283186e-07,
+      "loss": 1.4765,
+      "step": 48273
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.137800252844501e-07,
+      "loss": 1.8955,
+      "step": 48274
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.137326169405816e-07,
+      "loss": 1.8126,
+      "step": 48275
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.136852085967131e-07,
+      "loss": 2.6621,
+      "step": 48276
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.136378002528446e-07,
+      "loss": 1.4839,
+      "step": 48277
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.13590391908976e-07,
+      "loss": 0.8354,
+      "step": 48278
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.135429835651074e-07,
+      "loss": 1.4514,
+      "step": 48279
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.134955752212389e-07,
+      "loss": 3.3691,
+      "step": 48280
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.134481668773704e-07,
+      "loss": 0.8872,
+      "step": 48281
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.134007585335019e-07,
+      "loss": 1.9893,
+      "step": 48282
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.133533501896333e-07,
+      "loss": 2.248,
+      "step": 48283
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.133059418457649e-07,
+      "loss": 0.5467,
+      "step": 48284
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.132585335018964e-07,
+      "loss": 2.4853,
+      "step": 48285
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.132111251580278e-07,
+      "loss": 2.8943,
+      "step": 48286
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.131637168141594e-07,
+      "loss": 2.6349,
+      "step": 48287
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.131163084702908e-07,
+      "loss": 3.153,
+      "step": 48288
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 7.130689001264223e-07,
+      "loss": 2.0851,
+      "step": 48289
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.130214917825538e-07,
+      "loss": 2.755,
+      "step": 48290
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.129740834386852e-07,
+      "loss": 1.1064,
+      "step": 48291
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.129266750948167e-07,
+      "loss": 3.4296,
+      "step": 48292
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.128792667509481e-07,
+      "loss": 1.9513,
+      "step": 48293
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.128318584070796e-07,
+      "loss": 1.742,
+      "step": 48294
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.127844500632112e-07,
+      "loss": 1.7992,
+      "step": 48295
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.127370417193426e-07,
+      "loss": 2.112,
+      "step": 48296
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.126896333754741e-07,
+      "loss": 1.9623,
+      "step": 48297
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.126422250316056e-07,
+      "loss": 1.4055,
+      "step": 48298
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.125948166877371e-07,
+      "loss": 1.5022,
+      "step": 48299
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.125474083438686e-07,
+      "loss": 1.2541,
+      "step": 48300
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.125e-07,
+      "loss": 2.4738,
+      "step": 48301
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.124525916561316e-07,
+      "loss": 1.8683,
+      "step": 48302
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.124051833122629e-07,
+      "loss": 1.7046,
+      "step": 48303
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.123577749683944e-07,
+      "loss": 1.316,
+      "step": 48304
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.123103666245259e-07,
+      "loss": 2.6721,
+      "step": 48305
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.122629582806574e-07,
+      "loss": 3.6361,
+      "step": 48306
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.122155499367889e-07,
+      "loss": 1.9871,
+      "step": 48307
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.121681415929203e-07,
+      "loss": 3.3431,
+      "step": 48308
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.121207332490519e-07,
+      "loss": 1.109,
+      "step": 48309
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.120733249051834e-07,
+      "loss": 2.1407,
+      "step": 48310
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.120259165613148e-07,
+      "loss": 2.2106,
+      "step": 48311
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.119785082174463e-07,
+      "loss": 3.0029,
+      "step": 48312
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.119310998735778e-07,
+      "loss": 1.7101,
+      "step": 48313
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.118836915297092e-07,
+      "loss": 2.8587,
+      "step": 48314
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.118362831858407e-07,
+      "loss": 2.8648,
+      "step": 48315
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.117888748419721e-07,
+      "loss": 2.7952,
+      "step": 48316
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.117414664981037e-07,
+      "loss": 1.6948,
+      "step": 48317
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.116940581542351e-07,
+      "loss": 2.0221,
+      "step": 48318
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.116466498103666e-07,
+      "loss": 2.0062,
+      "step": 48319
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.115992414664982e-07,
+      "loss": 1.2127,
+      "step": 48320
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.115518331226296e-07,
+      "loss": 1.171,
+      "step": 48321
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.115044247787611e-07,
+      "loss": 2.6469,
+      "step": 48322
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.114570164348926e-07,
+      "loss": 1.2636,
+      "step": 48323
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.114096080910241e-07,
+      "loss": 3.2455,
+      "step": 48324
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.113621997471556e-07,
+      "loss": 2.598,
+      "step": 48325
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.113147914032869e-07,
+      "loss": 2.5723,
+      "step": 48326
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.112673830594184e-07,
+      "loss": 2.9217,
+      "step": 48327
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.112199747155499e-07,
+      "loss": 1.7064,
+      "step": 48328
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.111725663716814e-07,
+      "loss": 1.2202,
+      "step": 48329
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.111251580278129e-07,
+      "loss": 1.7841,
+      "step": 48330
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.110777496839444e-07,
+      "loss": 1.9058,
+      "step": 48331
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.110303413400759e-07,
+      "loss": 2.4995,
+      "step": 48332
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.109829329962074e-07,
+      "loss": 1.8521,
+      "step": 48333
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.109355246523388e-07,
+      "loss": 1.9118,
+      "step": 48334
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.108881163084704e-07,
+      "loss": 3.2875,
+      "step": 48335
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.108407079646018e-07,
+      "loss": 2.0537,
+      "step": 48336
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.107932996207332e-07,
+      "loss": 3.0001,
+      "step": 48337
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.107458912768647e-07,
+      "loss": 0.6489,
+      "step": 48338
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.106984829329962e-07,
+      "loss": 0.7091,
+      "step": 48339
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.106510745891277e-07,
+      "loss": 1.6429,
+      "step": 48340
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.106036662452591e-07,
+      "loss": 1.4994,
+      "step": 48341
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.105562579013907e-07,
+      "loss": 1.1567,
+      "step": 48342
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.105088495575222e-07,
+      "loss": 1.3315,
+      "step": 48343
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.104614412136536e-07,
+      "loss": 2.5742,
+      "step": 48344
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.104140328697851e-07,
+      "loss": 1.5889,
+      "step": 48345
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.103666245259166e-07,
+      "loss": 1.5225,
+      "step": 48346
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.103192161820481e-07,
+      "loss": 3.6327,
+      "step": 48347
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.102718078381796e-07,
+      "loss": 2.1325,
+      "step": 48348
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.102243994943109e-07,
+      "loss": 1.6285,
+      "step": 48349
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.101769911504425e-07,
+      "loss": 2.0488,
+      "step": 48350
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.101295828065739e-07,
+      "loss": 1.731,
+      "step": 48351
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 7.100821744627054e-07,
+      "loss": 2.0345,
+      "step": 48352
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.10034766118837e-07,
+      "loss": 1.5812,
+      "step": 48353
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.099873577749684e-07,
+      "loss": 2.1507,
+      "step": 48354
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.099399494310999e-07,
+      "loss": 2.564,
+      "step": 48355
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.098925410872314e-07,
+      "loss": 1.2211,
+      "step": 48356
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.098451327433629e-07,
+      "loss": 2.1674,
+      "step": 48357
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.097977243994944e-07,
+      "loss": 1.9465,
+      "step": 48358
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.097503160556258e-07,
+      "loss": 1.1702,
+      "step": 48359
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.097029077117573e-07,
+      "loss": 1.3548,
+      "step": 48360
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.096554993678887e-07,
+      "loss": 1.5314,
+      "step": 48361
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.096080910240202e-07,
+      "loss": 3.038,
+      "step": 48362
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.095606826801517e-07,
+      "loss": 1.7371,
+      "step": 48363
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.095132743362832e-07,
+      "loss": 2.1398,
+      "step": 48364
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.094658659924147e-07,
+      "loss": 1.1153,
+      "step": 48365
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.094184576485462e-07,
+      "loss": 1.7209,
+      "step": 48366
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.093710493046776e-07,
+      "loss": 2.7242,
+      "step": 48367
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.093236409608092e-07,
+      "loss": 2.2569,
+      "step": 48368
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.092762326169406e-07,
+      "loss": 1.5541,
+      "step": 48369
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.092288242730721e-07,
+      "loss": 2.1422,
+      "step": 48370
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.091814159292036e-07,
+      "loss": 1.7854,
+      "step": 48371
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.09134007585335e-07,
+      "loss": 2.2859,
+      "step": 48372
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.090865992414665e-07,
+      "loss": 1.4124,
+      "step": 48373
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.090391908975979e-07,
+      "loss": 1.11,
+      "step": 48374
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.089917825537295e-07,
+      "loss": 1.954,
+      "step": 48375
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.08944374209861e-07,
+      "loss": 1.05,
+      "step": 48376
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.088969658659924e-07,
+      "loss": 1.2092,
+      "step": 48377
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.088495575221239e-07,
+      "loss": 1.8479,
+      "step": 48378
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.088021491782554e-07,
+      "loss": 2.5334,
+      "step": 48379
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.087547408343869e-07,
+      "loss": 3.0591,
+      "step": 48380
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.087073324905184e-07,
+      "loss": 1.9762,
+      "step": 48381
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.086599241466498e-07,
+      "loss": 1.5697,
+      "step": 48382
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.086125158027814e-07,
+      "loss": 1.0519,
+      "step": 48383
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.085651074589127e-07,
+      "loss": 1.0571,
+      "step": 48384
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.085176991150442e-07,
+      "loss": 2.3575,
+      "step": 48385
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.084702907711757e-07,
+      "loss": 0.5319,
+      "step": 48386
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.084228824273072e-07,
+      "loss": 2.5567,
+      "step": 48387
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.083754740834387e-07,
+      "loss": 1.8305,
+      "step": 48388
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.083280657395701e-07,
+      "loss": 1.2241,
+      "step": 48389
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.082806573957017e-07,
+      "loss": 1.4148,
+      "step": 48390
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.082332490518332e-07,
+      "loss": 2.708,
+      "step": 48391
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.081858407079646e-07,
+      "loss": 2.9498,
+      "step": 48392
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.081384323640961e-07,
+      "loss": 1.455,
+      "step": 48393
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.080910240202276e-07,
+      "loss": 0.9226,
+      "step": 48394
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.080436156763591e-07,
+      "loss": 1.974,
+      "step": 48395
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.079962073324905e-07,
+      "loss": 1.3077,
+      "step": 48396
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.07948798988622e-07,
+      "loss": 1.9296,
+      "step": 48397
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.079013906447535e-07,
+      "loss": 1.6708,
+      "step": 48398
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.078539823008849e-07,
+      "loss": 1.6754,
+      "step": 48399
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.078065739570164e-07,
+      "loss": 1.3128,
+      "step": 48400
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.07759165613148e-07,
+      "loss": 0.7223,
+      "step": 48401
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.077117572692794e-07,
+      "loss": 1.268,
+      "step": 48402
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.076643489254109e-07,
+      "loss": 1.7301,
+      "step": 48403
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.076169405815424e-07,
+      "loss": 2.2252,
+      "step": 48404
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.075695322376739e-07,
+      "loss": 1.4695,
+      "step": 48405
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.075221238938054e-07,
+      "loss": 2.5374,
+      "step": 48406
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.074747155499367e-07,
+      "loss": 3.1791,
+      "step": 48407
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.074273072060683e-07,
+      "loss": 3.1836,
+      "step": 48408
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.073798988621997e-07,
+      "loss": 1.9149,
+      "step": 48409
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.073324905183312e-07,
+      "loss": 1.26,
+      "step": 48410
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.072850821744627e-07,
+      "loss": 4.9389,
+      "step": 48411
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.072376738305942e-07,
+      "loss": 1.6819,
+      "step": 48412
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.071902654867257e-07,
+      "loss": 2.4099,
+      "step": 48413
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.071428571428572e-07,
+      "loss": 1.9205,
+      "step": 48414
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 7.070954487989886e-07,
+      "loss": 1.5431,
+      "step": 48415
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.070480404551202e-07,
+      "loss": 1.5234,
+      "step": 48416
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.070006321112516e-07,
+      "loss": 1.0594,
+      "step": 48417
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.069532237673831e-07,
+      "loss": 4.9219,
+      "step": 48418
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.069058154235145e-07,
+      "loss": 1.8367,
+      "step": 48419
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.06858407079646e-07,
+      "loss": 1.7644,
+      "step": 48420
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.068109987357775e-07,
+      "loss": 2.2975,
+      "step": 48421
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.067635903919089e-07,
+      "loss": 2.9745,
+      "step": 48422
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.067161820480405e-07,
+      "loss": 1.5713,
+      "step": 48423
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.06668773704172e-07,
+      "loss": 1.4638,
+      "step": 48424
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.066213653603034e-07,
+      "loss": 1.8754,
+      "step": 48425
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.065739570164349e-07,
+      "loss": 2.0794,
+      "step": 48426
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.065265486725664e-07,
+      "loss": 2.1009,
+      "step": 48427
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.064791403286979e-07,
+      "loss": 3.0334,
+      "step": 48428
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.064317319848294e-07,
+      "loss": 2.9022,
+      "step": 48429
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.063843236409609e-07,
+      "loss": 1.5495,
+      "step": 48430
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.063369152970923e-07,
+      "loss": 2.6425,
+      "step": 48431
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.062895069532237e-07,
+      "loss": 2.8445,
+      "step": 48432
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.062420986093552e-07,
+      "loss": 1.4029,
+      "step": 48433
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.061946902654868e-07,
+      "loss": 1.4442,
+      "step": 48434
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.061472819216182e-07,
+      "loss": 3.7132,
+      "step": 48435
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.060998735777497e-07,
+      "loss": 2.9002,
+      "step": 48436
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.060524652338812e-07,
+      "loss": 3.2249,
+      "step": 48437
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.060050568900127e-07,
+      "loss": 2.9051,
+      "step": 48438
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.059576485461442e-07,
+      "loss": 2.2092,
+      "step": 48439
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.059102402022756e-07,
+      "loss": 2.1653,
+      "step": 48440
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.058628318584072e-07,
+      "loss": 1.7799,
+      "step": 48441
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.058154235145386e-07,
+      "loss": 1.2412,
+      "step": 48442
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.0576801517067e-07,
+      "loss": 0.9386,
+      "step": 48443
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.057206068268015e-07,
+      "loss": 0.4147,
+      "step": 48444
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.05673198482933e-07,
+      "loss": 2.9789,
+      "step": 48445
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.056257901390645e-07,
+      "loss": 1.3847,
+      "step": 48446
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.055783817951959e-07,
+      "loss": 1.0594,
+      "step": 48447
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.055309734513274e-07,
+      "loss": 2.4251,
+      "step": 48448
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.05483565107459e-07,
+      "loss": 2.4006,
+      "step": 48449
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.054361567635904e-07,
+      "loss": 2.4131,
+      "step": 48450
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.053887484197219e-07,
+      "loss": 2.3627,
+      "step": 48451
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.053413400758534e-07,
+      "loss": 0.8662,
+      "step": 48452
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.052939317319849e-07,
+      "loss": 1.7077,
+      "step": 48453
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.052465233881164e-07,
+      "loss": 2.1005,
+      "step": 48454
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.051991150442477e-07,
+      "loss": 2.077,
+      "step": 48455
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.051517067003793e-07,
+      "loss": 3.8604,
+      "step": 48456
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.051042983565107e-07,
+      "loss": 1.6286,
+      "step": 48457
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.050568900126422e-07,
+      "loss": 2.7711,
+      "step": 48458
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.050094816687737e-07,
+      "loss": 2.6985,
+      "step": 48459
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.049620733249052e-07,
+      "loss": 2.9485,
+      "step": 48460
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.049146649810367e-07,
+      "loss": 0.8945,
+      "step": 48461
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.048672566371682e-07,
+      "loss": 0.8537,
+      "step": 48462
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.048198482932997e-07,
+      "loss": 3.0743,
+      "step": 48463
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.047724399494312e-07,
+      "loss": 1.561,
+      "step": 48464
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.047250316055626e-07,
+      "loss": 1.8931,
+      "step": 48465
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.04677623261694e-07,
+      "loss": 2.5469,
+      "step": 48466
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.046302149178255e-07,
+      "loss": 2.8003,
+      "step": 48467
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.04582806573957e-07,
+      "loss": 2.2763,
+      "step": 48468
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.045353982300885e-07,
+      "loss": 1.0325,
+      "step": 48469
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.044879898862199e-07,
+      "loss": 2.2433,
+      "step": 48470
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.044405815423515e-07,
+      "loss": 4.8189,
+      "step": 48471
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.04393173198483e-07,
+      "loss": 2.6039,
+      "step": 48472
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.043457648546144e-07,
+      "loss": 1.9666,
+      "step": 48473
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.04298356510746e-07,
+      "loss": 1.3834,
+      "step": 48474
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.042509481668774e-07,
+      "loss": 0.6014,
+      "step": 48475
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.042035398230089e-07,
+      "loss": 2.1574,
+      "step": 48476
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.041561314791404e-07,
+      "loss": 2.2317,
+      "step": 48477
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.041087231352718e-07,
+      "loss": 2.6957,
+      "step": 48478
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 7.040613147914033e-07,
+      "loss": 0.5036,
+      "step": 48479
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.040139064475347e-07,
+      "loss": 1.7046,
+      "step": 48480
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.039664981036662e-07,
+      "loss": 1.8531,
+      "step": 48481
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.039190897597978e-07,
+      "loss": 2.4218,
+      "step": 48482
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.038716814159292e-07,
+      "loss": 2.0244,
+      "step": 48483
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.038242730720607e-07,
+      "loss": 2.707,
+      "step": 48484
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.037768647281922e-07,
+      "loss": 3.0189,
+      "step": 48485
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.037294563843237e-07,
+      "loss": 3.0014,
+      "step": 48486
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.036820480404552e-07,
+      "loss": 2.6879,
+      "step": 48487
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.036346396965866e-07,
+      "loss": 2.0216,
+      "step": 48488
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.035872313527182e-07,
+      "loss": 1.1416,
+      "step": 48489
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.035398230088495e-07,
+      "loss": 1.4221,
+      "step": 48490
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.03492414664981e-07,
+      "loss": 2.0295,
+      "step": 48491
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.034450063211125e-07,
+      "loss": 0.9992,
+      "step": 48492
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.03397597977244e-07,
+      "loss": 0.9328,
+      "step": 48493
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.033501896333755e-07,
+      "loss": 1.6576,
+      "step": 48494
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.03302781289507e-07,
+      "loss": 0.8722,
+      "step": 48495
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.032553729456385e-07,
+      "loss": 1.9368,
+      "step": 48496
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.0320796460177e-07,
+      "loss": 2.4567,
+      "step": 48497
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.031605562579014e-07,
+      "loss": 1.0661,
+      "step": 48498
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.031131479140329e-07,
+      "loss": 0.7314,
+      "step": 48499
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.030657395701644e-07,
+      "loss": 1.5392,
+      "step": 48500
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.030183312262958e-07,
+      "loss": 1.6723,
+      "step": 48501
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.029709228824273e-07,
+      "loss": 1.7261,
+      "step": 48502
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.029235145385587e-07,
+      "loss": 0.6601,
+      "step": 48503
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.028761061946903e-07,
+      "loss": 2.5833,
+      "step": 48504
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.028286978508218e-07,
+      "loss": 1.3944,
+      "step": 48505
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.027812895069532e-07,
+      "loss": 3.0381,
+      "step": 48506
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.027338811630847e-07,
+      "loss": 1.0942,
+      "step": 48507
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.026864728192162e-07,
+      "loss": 1.8976,
+      "step": 48508
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.026390644753477e-07,
+      "loss": 2.0436,
+      "step": 48509
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.025916561314792e-07,
+      "loss": 1.4607,
+      "step": 48510
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.025442477876107e-07,
+      "loss": 2.9284,
+      "step": 48511
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.024968394437422e-07,
+      "loss": 1.4597,
+      "step": 48512
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.024494310998735e-07,
+      "loss": 2.1446,
+      "step": 48513
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.02402022756005e-07,
+      "loss": 1.4114,
+      "step": 48514
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.023546144121366e-07,
+      "loss": 1.5575,
+      "step": 48515
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.02307206068268e-07,
+      "loss": 2.745,
+      "step": 48516
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.022597977243995e-07,
+      "loss": 1.3917,
+      "step": 48517
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.022123893805309e-07,
+      "loss": 2.0137,
+      "step": 48518
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.021649810366625e-07,
+      "loss": 2.236,
+      "step": 48519
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.02117572692794e-07,
+      "loss": 1.7619,
+      "step": 48520
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.020701643489254e-07,
+      "loss": 1.2147,
+      "step": 48521
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.02022756005057e-07,
+      "loss": 1.1733,
+      "step": 48522
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.019753476611884e-07,
+      "loss": 1.6482,
+      "step": 48523
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.019279393173199e-07,
+      "loss": 0.6676,
+      "step": 48524
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.018805309734513e-07,
+      "loss": 1.8991,
+      "step": 48525
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.018331226295828e-07,
+      "loss": 4.9152,
+      "step": 48526
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.017857142857143e-07,
+      "loss": 1.4937,
+      "step": 48527
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.017383059418457e-07,
+      "loss": 1.0289,
+      "step": 48528
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.016908975979772e-07,
+      "loss": 3.8292,
+      "step": 48529
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.016434892541088e-07,
+      "loss": 0.9732,
+      "step": 48530
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.015960809102402e-07,
+      "loss": 1.4062,
+      "step": 48531
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.015486725663717e-07,
+      "loss": 1.5603,
+      "step": 48532
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.015012642225032e-07,
+      "loss": 1.6247,
+      "step": 48533
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.014538558786347e-07,
+      "loss": 0.9959,
+      "step": 48534
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.014064475347662e-07,
+      "loss": 3.6646,
+      "step": 48535
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.013590391908975e-07,
+      "loss": 2.8484,
+      "step": 48536
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.013116308470291e-07,
+      "loss": 1.4346,
+      "step": 48537
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.012642225031605e-07,
+      "loss": 2.6508,
+      "step": 48538
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.01216814159292e-07,
+      "loss": 2.1341,
+      "step": 48539
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.011694058154235e-07,
+      "loss": 1.4446,
+      "step": 48540
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.01121997471555e-07,
+      "loss": 2.1702,
+      "step": 48541
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 7.010745891276865e-07,
+      "loss": 1.7799,
+      "step": 48542
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.01027180783818e-07,
+      "loss": 2.1803,
+      "step": 48543
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.009797724399495e-07,
+      "loss": 0.7571,
+      "step": 48544
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.00932364096081e-07,
+      "loss": 1.1024,
+      "step": 48545
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.008849557522124e-07,
+      "loss": 2.3337,
+      "step": 48546
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.008375474083439e-07,
+      "loss": 1.5514,
+      "step": 48547
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.007901390644753e-07,
+      "loss": 2.0866,
+      "step": 48548
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.007427307206068e-07,
+      "loss": 2.9571,
+      "step": 48549
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.006953223767383e-07,
+      "loss": 0.9346,
+      "step": 48550
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.006479140328697e-07,
+      "loss": 0.9127,
+      "step": 48551
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.006005056890013e-07,
+      "loss": 1.5946,
+      "step": 48552
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.005530973451328e-07,
+      "loss": 2.2947,
+      "step": 48553
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.005056890012642e-07,
+      "loss": 2.6833,
+      "step": 48554
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.004582806573958e-07,
+      "loss": 2.7737,
+      "step": 48555
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.004108723135272e-07,
+      "loss": 2.0713,
+      "step": 48556
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.003634639696587e-07,
+      "loss": 2.7529,
+      "step": 48557
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.003160556257902e-07,
+      "loss": 2.1557,
+      "step": 48558
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.002686472819217e-07,
+      "loss": 3.4242,
+      "step": 48559
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.002212389380531e-07,
+      "loss": 2.2356,
+      "step": 48560
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.001738305941845e-07,
+      "loss": 2.0716,
+      "step": 48561
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.00126422250316e-07,
+      "loss": 1.9521,
+      "step": 48562
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.000790139064476e-07,
+      "loss": 1.4554,
+      "step": 48563
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 7.00031605562579e-07,
+      "loss": 2.5595,
+      "step": 48564
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.999841972187105e-07,
+      "loss": 2.0209,
+      "step": 48565
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.99936788874842e-07,
+      "loss": 0.99,
+      "step": 48566
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.998893805309735e-07,
+      "loss": 3.9619,
+      "step": 48567
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.99841972187105e-07,
+      "loss": 2.1669,
+      "step": 48568
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.997945638432364e-07,
+      "loss": 2.9503,
+      "step": 48569
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.99747155499368e-07,
+      "loss": 1.228,
+      "step": 48570
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.996997471554993e-07,
+      "loss": 0.7798,
+      "step": 48571
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.996523388116308e-07,
+      "loss": 1.552,
+      "step": 48572
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.996049304677623e-07,
+      "loss": 1.6666,
+      "step": 48573
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.995575221238938e-07,
+      "loss": 2.3196,
+      "step": 48574
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.995101137800253e-07,
+      "loss": 3.3614,
+      "step": 48575
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.994627054361568e-07,
+      "loss": 2.5714,
+      "step": 48576
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.994152970922883e-07,
+      "loss": 2.7077,
+      "step": 48577
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.993678887484198e-07,
+      "loss": 3.3469,
+      "step": 48578
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.993204804045512e-07,
+      "loss": 1.7815,
+      "step": 48579
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.992730720606827e-07,
+      "loss": 0.7805,
+      "step": 48580
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.992256637168142e-07,
+      "loss": 2.6666,
+      "step": 48581
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.991782553729457e-07,
+      "loss": 3.0434,
+      "step": 48582
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.991308470290771e-07,
+      "loss": 0.7008,
+      "step": 48583
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.990834386852085e-07,
+      "loss": 3.8602,
+      "step": 48584
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.990360303413401e-07,
+      "loss": 2.6936,
+      "step": 48585
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.989886219974715e-07,
+      "loss": 1.7284,
+      "step": 48586
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.98941213653603e-07,
+      "loss": 2.0532,
+      "step": 48587
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.988938053097346e-07,
+      "loss": 2.6634,
+      "step": 48588
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.98846396965866e-07,
+      "loss": 2.1649,
+      "step": 48589
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.987989886219975e-07,
+      "loss": 1.55,
+      "step": 48590
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.98751580278129e-07,
+      "loss": 2.0452,
+      "step": 48591
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.987041719342605e-07,
+      "loss": 1.8834,
+      "step": 48592
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.98656763590392e-07,
+      "loss": 2.2987,
+      "step": 48593
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.986093552465234e-07,
+      "loss": 1.4987,
+      "step": 48594
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.985619469026548e-07,
+      "loss": 1.7367,
+      "step": 48595
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.985145385587863e-07,
+      "loss": 2.3254,
+      "step": 48596
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.984671302149178e-07,
+      "loss": 2.1145,
+      "step": 48597
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.984197218710493e-07,
+      "loss": 1.7705,
+      "step": 48598
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.983723135271808e-07,
+      "loss": 1.533,
+      "step": 48599
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.983249051833123e-07,
+      "loss": 2.9274,
+      "step": 48600
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.982774968394438e-07,
+      "loss": 3.3083,
+      "step": 48601
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.982300884955752e-07,
+      "loss": 2.4355,
+      "step": 48602
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.981826801517068e-07,
+      "loss": 1.9025,
+      "step": 48603
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.981352718078382e-07,
+      "loss": 0.5326,
+      "step": 48604
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 6.980878634639697e-07,
+      "loss": 2.4559,
+      "step": 48605
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.980404551201012e-07,
+      "loss": 1.4494,
+      "step": 48606
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.979930467762326e-07,
+      "loss": 1.6866,
+      "step": 48607
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.979456384323641e-07,
+      "loss": 2.145,
+      "step": 48608
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.978982300884955e-07,
+      "loss": 2.0863,
+      "step": 48609
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.978508217446271e-07,
+      "loss": 2.5928,
+      "step": 48610
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.978034134007586e-07,
+      "loss": 1.3891,
+      "step": 48611
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.9775600505689e-07,
+      "loss": 1.8294,
+      "step": 48612
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.977085967130215e-07,
+      "loss": 3.8943,
+      "step": 48613
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.97661188369153e-07,
+      "loss": 2.2527,
+      "step": 48614
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.976137800252845e-07,
+      "loss": 0.6752,
+      "step": 48615
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.97566371681416e-07,
+      "loss": 1.3477,
+      "step": 48616
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.975189633375474e-07,
+      "loss": 1.2723,
+      "step": 48617
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.97471554993679e-07,
+      "loss": 0.9575,
+      "step": 48618
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.974241466498103e-07,
+      "loss": 1.7737,
+      "step": 48619
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.973767383059418e-07,
+      "loss": 3.6402,
+      "step": 48620
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.973293299620734e-07,
+      "loss": 0.3684,
+      "step": 48621
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.972819216182048e-07,
+      "loss": 1.1386,
+      "step": 48622
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.972345132743363e-07,
+      "loss": 1.8587,
+      "step": 48623
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.971871049304678e-07,
+      "loss": 2.2314,
+      "step": 48624
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.971396965865993e-07,
+      "loss": 0.6771,
+      "step": 48625
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.970922882427308e-07,
+      "loss": 2.7767,
+      "step": 48626
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.970448798988622e-07,
+      "loss": 0.5349,
+      "step": 48627
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.969974715549937e-07,
+      "loss": 1.7257,
+      "step": 48628
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.969500632111252e-07,
+      "loss": 0.6165,
+      "step": 48629
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.969026548672566e-07,
+      "loss": 2.0533,
+      "step": 48630
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.968552465233881e-07,
+      "loss": 2.6224,
+      "step": 48631
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.968078381795196e-07,
+      "loss": 2.3389,
+      "step": 48632
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.967604298356511e-07,
+      "loss": 2.6955,
+      "step": 48633
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.967130214917826e-07,
+      "loss": 3.4355,
+      "step": 48634
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.96665613147914e-07,
+      "loss": 3.4367,
+      "step": 48635
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.966182048040456e-07,
+      "loss": 2.0345,
+      "step": 48636
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.96570796460177e-07,
+      "loss": 1.0194,
+      "step": 48637
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.965233881163085e-07,
+      "loss": 1.7249,
+      "step": 48638
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.9647597977244e-07,
+      "loss": 1.729,
+      "step": 48639
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.964285714285715e-07,
+      "loss": 1.9415,
+      "step": 48640
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.96381163084703e-07,
+      "loss": 0.9144,
+      "step": 48641
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.963337547408343e-07,
+      "loss": 1.2501,
+      "step": 48642
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.962863463969659e-07,
+      "loss": 1.8772,
+      "step": 48643
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.962389380530974e-07,
+      "loss": 1.4722,
+      "step": 48644
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.961915297092288e-07,
+      "loss": 2.5984,
+      "step": 48645
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.961441213653603e-07,
+      "loss": 1.5953,
+      "step": 48646
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.960967130214918e-07,
+      "loss": 1.0154,
+      "step": 48647
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.960493046776233e-07,
+      "loss": 1.5331,
+      "step": 48648
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.960018963337548e-07,
+      "loss": 1.4329,
+      "step": 48649
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.959544879898862e-07,
+      "loss": 1.0792,
+      "step": 48650
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.959070796460178e-07,
+      "loss": 0.8398,
+      "step": 48651
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.958596713021492e-07,
+      "loss": 2.5604,
+      "step": 48652
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.958122629582806e-07,
+      "loss": 1.5057,
+      "step": 48653
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.957648546144122e-07,
+      "loss": 1.5181,
+      "step": 48654
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.957174462705436e-07,
+      "loss": 1.0642,
+      "step": 48655
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.956700379266751e-07,
+      "loss": 0.6925,
+      "step": 48656
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.956226295828065e-07,
+      "loss": 1.7739,
+      "step": 48657
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.955752212389381e-07,
+      "loss": 1.5215,
+      "step": 48658
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.955278128950696e-07,
+      "loss": 1.0108,
+      "step": 48659
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.95480404551201e-07,
+      "loss": 1.7581,
+      "step": 48660
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.954329962073325e-07,
+      "loss": 2.8277,
+      "step": 48661
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.95385587863464e-07,
+      "loss": 1.9173,
+      "step": 48662
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.953381795195955e-07,
+      "loss": 1.4562,
+      "step": 48663
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.95290771175727e-07,
+      "loss": 3.1925,
+      "step": 48664
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.952433628318584e-07,
+      "loss": 2.233,
+      "step": 48665
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.951959544879899e-07,
+      "loss": 2.8607,
+      "step": 48666
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.951485461441213e-07,
+      "loss": 0.3066,
+      "step": 48667
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.951011378002528e-07,
+      "loss": 1.551,
+      "step": 48668
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 6.950537294563844e-07,
+      "loss": 2.2728,
+      "step": 48669
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.950063211125158e-07,
+      "loss": 1.2736,
+      "step": 48670
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.949589127686473e-07,
+      "loss": 2.9994,
+      "step": 48671
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.949115044247788e-07,
+      "loss": 1.1078,
+      "step": 48672
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.948640960809103e-07,
+      "loss": 2.5538,
+      "step": 48673
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.948166877370418e-07,
+      "loss": 0.7831,
+      "step": 48674
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.947692793931732e-07,
+      "loss": 3.1059,
+      "step": 48675
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.947218710493048e-07,
+      "loss": 1.3613,
+      "step": 48676
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.946744627054361e-07,
+      "loss": 1.7817,
+      "step": 48677
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.946270543615676e-07,
+      "loss": 2.2761,
+      "step": 48678
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.945796460176991e-07,
+      "loss": 3.4245,
+      "step": 48679
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.945322376738306e-07,
+      "loss": 3.7372,
+      "step": 48680
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.944848293299621e-07,
+      "loss": 2.4117,
+      "step": 48681
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.944374209860936e-07,
+      "loss": 0.9761,
+      "step": 48682
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.94390012642225e-07,
+      "loss": 2.3443,
+      "step": 48683
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.943426042983566e-07,
+      "loss": 1.3304,
+      "step": 48684
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.94295195954488e-07,
+      "loss": 1.8417,
+      "step": 48685
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.942477876106195e-07,
+      "loss": 1.786,
+      "step": 48686
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.94200379266751e-07,
+      "loss": 1.8046,
+      "step": 48687
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.941529709228825e-07,
+      "loss": 2.0104,
+      "step": 48688
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.941055625790139e-07,
+      "loss": 0.2513,
+      "step": 48689
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.940581542351453e-07,
+      "loss": 0.9267,
+      "step": 48690
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.940107458912769e-07,
+      "loss": 1.4257,
+      "step": 48691
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.939633375474084e-07,
+      "loss": 1.6854,
+      "step": 48692
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.939159292035398e-07,
+      "loss": 1.1259,
+      "step": 48693
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.938685208596713e-07,
+      "loss": 0.6615,
+      "step": 48694
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.938211125158028e-07,
+      "loss": 3.2755,
+      "step": 48695
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.937737041719343e-07,
+      "loss": 1.2999,
+      "step": 48696
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.937262958280658e-07,
+      "loss": 0.2068,
+      "step": 48697
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.936788874841973e-07,
+      "loss": 2.377,
+      "step": 48698
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.936314791403288e-07,
+      "loss": 0.3883,
+      "step": 48699
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.935840707964601e-07,
+      "loss": 3.2283,
+      "step": 48700
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.935366624525916e-07,
+      "loss": 2.7964,
+      "step": 48701
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.934892541087232e-07,
+      "loss": 1.7387,
+      "step": 48702
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.934418457648546e-07,
+      "loss": 0.4941,
+      "step": 48703
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.933944374209861e-07,
+      "loss": 2.3959,
+      "step": 48704
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.933470290771176e-07,
+      "loss": 2.3651,
+      "step": 48705
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.932996207332491e-07,
+      "loss": 3.0657,
+      "step": 48706
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.932522123893806e-07,
+      "loss": 2.189,
+      "step": 48707
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.93204804045512e-07,
+      "loss": 2.8379,
+      "step": 48708
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.931573957016436e-07,
+      "loss": 1.3511,
+      "step": 48709
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.93109987357775e-07,
+      "loss": 2.4289,
+      "step": 48710
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.930625790139065e-07,
+      "loss": 2.8604,
+      "step": 48711
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.930151706700379e-07,
+      "loss": 2.6849,
+      "step": 48712
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.929677623261694e-07,
+      "loss": 1.6545,
+      "step": 48713
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.929203539823009e-07,
+      "loss": 2.2926,
+      "step": 48714
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.928729456384324e-07,
+      "loss": 1.9186,
+      "step": 48715
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.928255372945638e-07,
+      "loss": 1.5649,
+      "step": 48716
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.927781289506954e-07,
+      "loss": 2.2484,
+      "step": 48717
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.927307206068268e-07,
+      "loss": 1.2874,
+      "step": 48718
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.926833122629583e-07,
+      "loss": 1.8652,
+      "step": 48719
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.926359039190898e-07,
+      "loss": 0.9054,
+      "step": 48720
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.925884955752213e-07,
+      "loss": 1.458,
+      "step": 48721
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.925410872313528e-07,
+      "loss": 4.0982,
+      "step": 48722
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.924936788874841e-07,
+      "loss": 3.4959,
+      "step": 48723
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.924462705436157e-07,
+      "loss": 2.0816,
+      "step": 48724
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.923988621997472e-07,
+      "loss": 1.5841,
+      "step": 48725
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.923514538558786e-07,
+      "loss": 1.8769,
+      "step": 48726
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.923040455120101e-07,
+      "loss": 2.5308,
+      "step": 48727
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.922566371681416e-07,
+      "loss": 3.3053,
+      "step": 48728
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.922092288242731e-07,
+      "loss": 1.8211,
+      "step": 48729
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.921618204804046e-07,
+      "loss": 1.8494,
+      "step": 48730
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.92114412136536e-07,
+      "loss": 1.2901,
+      "step": 48731
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 6.920670037926676e-07,
+      "loss": 2.4182,
+      "step": 48732
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.92019595448799e-07,
+      "loss": 1.2107,
+      "step": 48733
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.919721871049305e-07,
+      "loss": 1.8379,
+      "step": 48734
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.91924778761062e-07,
+      "loss": 1.2276,
+      "step": 48735
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.918773704171934e-07,
+      "loss": 2.4774,
+      "step": 48736
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.918299620733249e-07,
+      "loss": 2.2562,
+      "step": 48737
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.917825537294563e-07,
+      "loss": 1.111,
+      "step": 48738
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.917351453855879e-07,
+      "loss": 2.5044,
+      "step": 48739
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.916877370417194e-07,
+      "loss": 0.9523,
+      "step": 48740
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.916403286978508e-07,
+      "loss": 2.7141,
+      "step": 48741
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.915929203539823e-07,
+      "loss": 1.6263,
+      "step": 48742
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.915455120101138e-07,
+      "loss": 2.0031,
+      "step": 48743
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.914981036662453e-07,
+      "loss": 1.8671,
+      "step": 48744
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.914506953223768e-07,
+      "loss": 0.7738,
+      "step": 48745
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.914032869785083e-07,
+      "loss": 2.6982,
+      "step": 48746
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.913558786346397e-07,
+      "loss": 2.0273,
+      "step": 48747
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.913084702907711e-07,
+      "loss": 2.151,
+      "step": 48748
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.912610619469026e-07,
+      "loss": 3.5874,
+      "step": 48749
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.912136536030342e-07,
+      "loss": 1.0283,
+      "step": 48750
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.911662452591656e-07,
+      "loss": 2.8573,
+      "step": 48751
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.911188369152971e-07,
+      "loss": 3.1112,
+      "step": 48752
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.910714285714286e-07,
+      "loss": 1.8044,
+      "step": 48753
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.910240202275601e-07,
+      "loss": 3.694,
+      "step": 48754
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.909766118836916e-07,
+      "loss": 1.8359,
+      "step": 48755
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.90929203539823e-07,
+      "loss": 2.7663,
+      "step": 48756
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.908817951959546e-07,
+      "loss": 2.9253,
+      "step": 48757
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.90834386852086e-07,
+      "loss": 2.8817,
+      "step": 48758
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.907869785082174e-07,
+      "loss": 2.3981,
+      "step": 48759
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.907395701643489e-07,
+      "loss": 1.614,
+      "step": 48760
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.906921618204804e-07,
+      "loss": 1.1778,
+      "step": 48761
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.906447534766119e-07,
+      "loss": 0.9321,
+      "step": 48762
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.905973451327434e-07,
+      "loss": 2.4544,
+      "step": 48763
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.905499367888748e-07,
+      "loss": 1.5485,
+      "step": 48764
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.905025284450064e-07,
+      "loss": 0.9593,
+      "step": 48765
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.904551201011378e-07,
+      "loss": 1.6269,
+      "step": 48766
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.904077117572693e-07,
+      "loss": 1.2562,
+      "step": 48767
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.903603034134008e-07,
+      "loss": 1.4295,
+      "step": 48768
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.903128950695323e-07,
+      "loss": 1.1076,
+      "step": 48769
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.902654867256638e-07,
+      "loss": 2.4065,
+      "step": 48770
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.902180783817951e-07,
+      "loss": 1.7268,
+      "step": 48771
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.901706700379267e-07,
+      "loss": 3.3586,
+      "step": 48772
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.901232616940582e-07,
+      "loss": 3.1381,
+      "step": 48773
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.900758533501896e-07,
+      "loss": 3.4864,
+      "step": 48774
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.900284450063211e-07,
+      "loss": 2.7008,
+      "step": 48775
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.899810366624526e-07,
+      "loss": 2.6912,
+      "step": 48776
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.899336283185841e-07,
+      "loss": 1.8221,
+      "step": 48777
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.898862199747156e-07,
+      "loss": 2.6068,
+      "step": 48778
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.898388116308471e-07,
+      "loss": 0.9226,
+      "step": 48779
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.897914032869786e-07,
+      "loss": 2.1021,
+      "step": 48780
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.8974399494311e-07,
+      "loss": 1.9584,
+      "step": 48781
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.896965865992414e-07,
+      "loss": 0.4484,
+      "step": 48782
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.89649178255373e-07,
+      "loss": 1.5149,
+      "step": 48783
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.896017699115044e-07,
+      "loss": 1.9625,
+      "step": 48784
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.895543615676359e-07,
+      "loss": 1.8809,
+      "step": 48785
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.895069532237674e-07,
+      "loss": 2.3389,
+      "step": 48786
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.894595448798989e-07,
+      "loss": 2.5915,
+      "step": 48787
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.894121365360304e-07,
+      "loss": 1.4609,
+      "step": 48788
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.893647281921618e-07,
+      "loss": 2.1564,
+      "step": 48789
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.893173198482934e-07,
+      "loss": 2.1842,
+      "step": 48790
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.892699115044248e-07,
+      "loss": 2.7849,
+      "step": 48791
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.892225031605563e-07,
+      "loss": 2.0641,
+      "step": 48792
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.891750948166878e-07,
+      "loss": 1.9648,
+      "step": 48793
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.891276864728192e-07,
+      "loss": 2.207,
+      "step": 48794
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 6.890802781289507e-07,
+      "loss": 3.6461,
+      "step": 48795
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.890328697850821e-07,
+      "loss": 2.3819,
+      "step": 48796
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.889854614412136e-07,
+      "loss": 1.6841,
+      "step": 48797
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.889380530973452e-07,
+      "loss": 2.677,
+      "step": 48798
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.888906447534766e-07,
+      "loss": 1.8929,
+      "step": 48799
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.888432364096081e-07,
+      "loss": 1.4713,
+      "step": 48800
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.887958280657396e-07,
+      "loss": 1.0752,
+      "step": 48801
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.887484197218711e-07,
+      "loss": 1.8243,
+      "step": 48802
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.887010113780026e-07,
+      "loss": 2.3488,
+      "step": 48803
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.88653603034134e-07,
+      "loss": 1.4307,
+      "step": 48804
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.886061946902656e-07,
+      "loss": 0.9263,
+      "step": 48805
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.885587863463969e-07,
+      "loss": 2.1757,
+      "step": 48806
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.885113780025284e-07,
+      "loss": 3.0031,
+      "step": 48807
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.884639696586599e-07,
+      "loss": 1.5967,
+      "step": 48808
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.884165613147914e-07,
+      "loss": 1.3749,
+      "step": 48809
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.883691529709229e-07,
+      "loss": 2.092,
+      "step": 48810
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.883217446270544e-07,
+      "loss": 0.6502,
+      "step": 48811
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.882743362831859e-07,
+      "loss": 0.6125,
+      "step": 48812
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.882269279393174e-07,
+      "loss": 1.1635,
+      "step": 48813
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.881795195954488e-07,
+      "loss": 1.5441,
+      "step": 48814
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.881321112515803e-07,
+      "loss": 1.1318,
+      "step": 48815
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.880847029077118e-07,
+      "loss": 3.7135,
+      "step": 48816
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.880372945638432e-07,
+      "loss": 2.4461,
+      "step": 48817
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.879898862199747e-07,
+      "loss": 2.4822,
+      "step": 48818
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.879424778761061e-07,
+      "loss": 3.3968,
+      "step": 48819
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.878950695322377e-07,
+      "loss": 1.2241,
+      "step": 48820
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.878476611883692e-07,
+      "loss": 1.664,
+      "step": 48821
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.878002528445006e-07,
+      "loss": 1.5078,
+      "step": 48822
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.877528445006322e-07,
+      "loss": 2.3629,
+      "step": 48823
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.877054361567636e-07,
+      "loss": 4.3121,
+      "step": 48824
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.876580278128951e-07,
+      "loss": 2.1567,
+      "step": 48825
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.876106194690266e-07,
+      "loss": 2.2653,
+      "step": 48826
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.875632111251581e-07,
+      "loss": 1.7223,
+      "step": 48827
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.875158027812896e-07,
+      "loss": 1.8269,
+      "step": 48828
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.874683944374209e-07,
+      "loss": 2.1866,
+      "step": 48829
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.874209860935524e-07,
+      "loss": 2.4764,
+      "step": 48830
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.87373577749684e-07,
+      "loss": 1.94,
+      "step": 48831
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.873261694058154e-07,
+      "loss": 2.6021,
+      "step": 48832
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.872787610619469e-07,
+      "loss": 2.8406,
+      "step": 48833
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.872313527180784e-07,
+      "loss": 1.4537,
+      "step": 48834
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.871839443742099e-07,
+      "loss": 2.0248,
+      "step": 48835
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.871365360303414e-07,
+      "loss": 0.7651,
+      "step": 48836
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.870891276864728e-07,
+      "loss": 1.403,
+      "step": 48837
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.870417193426044e-07,
+      "loss": 2.1874,
+      "step": 48838
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.869943109987358e-07,
+      "loss": 1.4158,
+      "step": 48839
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.869469026548672e-07,
+      "loss": 1.9222,
+      "step": 48840
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.868994943109987e-07,
+      "loss": 1.684,
+      "step": 48841
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.868520859671302e-07,
+      "loss": 1.0063,
+      "step": 48842
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.868046776232617e-07,
+      "loss": 1.2292,
+      "step": 48843
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.867572692793932e-07,
+      "loss": 2.769,
+      "step": 48844
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.867098609355247e-07,
+      "loss": 2.592,
+      "step": 48845
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.866624525916562e-07,
+      "loss": 0.8344,
+      "step": 48846
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.866150442477876e-07,
+      "loss": 2.3599,
+      "step": 48847
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.865676359039191e-07,
+      "loss": 1.8609,
+      "step": 48848
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.865202275600506e-07,
+      "loss": 1.7759,
+      "step": 48849
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.864728192161821e-07,
+      "loss": 0.7194,
+      "step": 48850
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.864254108723136e-07,
+      "loss": 3.3719,
+      "step": 48851
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.863780025284449e-07,
+      "loss": 1.9685,
+      "step": 48852
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.863305941845765e-07,
+      "loss": 2.5313,
+      "step": 48853
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.86283185840708e-07,
+      "loss": 2.7678,
+      "step": 48854
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.862357774968394e-07,
+      "loss": 1.8351,
+      "step": 48855
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.86188369152971e-07,
+      "loss": 4.2808,
+      "step": 48856
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.861409608091024e-07,
+      "loss": 3.5707,
+      "step": 48857
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.860935524652339e-07,
+      "loss": 2.277,
+      "step": 48858
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 6.860461441213654e-07,
+      "loss": 0.7251,
+      "step": 48859
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.859987357774969e-07,
+      "loss": 1.9817,
+      "step": 48860
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.859513274336284e-07,
+      "loss": 2.0671,
+      "step": 48861
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.859039190897598e-07,
+      "loss": 1.3412,
+      "step": 48862
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.858565107458913e-07,
+      "loss": 1.074,
+      "step": 48863
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.858091024020229e-07,
+      "loss": 2.2979,
+      "step": 48864
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.857616940581542e-07,
+      "loss": 1.8072,
+      "step": 48865
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.857142857142857e-07,
+      "loss": 3.4843,
+      "step": 48866
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.856668773704172e-07,
+      "loss": 1.773,
+      "step": 48867
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.856194690265487e-07,
+      "loss": 1.0983,
+      "step": 48868
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.855720606826802e-07,
+      "loss": 1.6136,
+      "step": 48869
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.855246523388116e-07,
+      "loss": 3.1912,
+      "step": 48870
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.854772439949432e-07,
+      "loss": 1.7027,
+      "step": 48871
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.854298356510746e-07,
+      "loss": 1.9776,
+      "step": 48872
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.853824273072061e-07,
+      "loss": 0.5289,
+      "step": 48873
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.853350189633376e-07,
+      "loss": 2.3116,
+      "step": 48874
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.85287610619469e-07,
+      "loss": 0.4653,
+      "step": 48875
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.852402022756005e-07,
+      "loss": 2.5728,
+      "step": 48876
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.851927939317319e-07,
+      "loss": 1.3296,
+      "step": 48877
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.851453855878635e-07,
+      "loss": 1.3936,
+      "step": 48878
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.85097977243995e-07,
+      "loss": 1.903,
+      "step": 48879
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.850505689001264e-07,
+      "loss": 1.571,
+      "step": 48880
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.850031605562579e-07,
+      "loss": 1.7779,
+      "step": 48881
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.849557522123894e-07,
+      "loss": 2.189,
+      "step": 48882
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.849083438685209e-07,
+      "loss": 1.8492,
+      "step": 48883
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.848609355246524e-07,
+      "loss": 1.8579,
+      "step": 48884
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.848135271807838e-07,
+      "loss": 2.1525,
+      "step": 48885
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.847661188369154e-07,
+      "loss": 1.021,
+      "step": 48886
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.847187104930467e-07,
+      "loss": 2.0132,
+      "step": 48887
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.846713021491782e-07,
+      "loss": 1.2455,
+      "step": 48888
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.846238938053098e-07,
+      "loss": 1.7837,
+      "step": 48889
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.845764854614412e-07,
+      "loss": 1.773,
+      "step": 48890
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.845290771175727e-07,
+      "loss": 1.0028,
+      "step": 48891
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.844816687737042e-07,
+      "loss": 1.1563,
+      "step": 48892
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.844342604298357e-07,
+      "loss": 0.7262,
+      "step": 48893
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.843868520859672e-07,
+      "loss": 1.8961,
+      "step": 48894
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.843394437420986e-07,
+      "loss": 0.9125,
+      "step": 48895
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.842920353982301e-07,
+      "loss": 2.6194,
+      "step": 48896
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.842446270543616e-07,
+      "loss": 3.2584,
+      "step": 48897
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.841972187104931e-07,
+      "loss": 1.7065,
+      "step": 48898
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.841498103666245e-07,
+      "loss": 2.4054,
+      "step": 48899
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.84102402022756e-07,
+      "loss": 1.2262,
+      "step": 48900
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.840549936788875e-07,
+      "loss": 1.7785,
+      "step": 48901
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.84007585335019e-07,
+      "loss": 1.166,
+      "step": 48902
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.839601769911504e-07,
+      "loss": 0.7014,
+      "step": 48903
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.83912768647282e-07,
+      "loss": 1.9453,
+      "step": 48904
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.838653603034134e-07,
+      "loss": 1.2128,
+      "step": 48905
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.838179519595449e-07,
+      "loss": 1.9695,
+      "step": 48906
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.837705436156764e-07,
+      "loss": 2.2728,
+      "step": 48907
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.837231352718079e-07,
+      "loss": 1.6301,
+      "step": 48908
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.836757269279394e-07,
+      "loss": 3.4211,
+      "step": 48909
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.836283185840707e-07,
+      "loss": 1.9239,
+      "step": 48910
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.835809102402023e-07,
+      "loss": 2.6591,
+      "step": 48911
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.835335018963338e-07,
+      "loss": 1.743,
+      "step": 48912
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.834860935524652e-07,
+      "loss": 1.4729,
+      "step": 48913
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.834386852085967e-07,
+      "loss": 2.3928,
+      "step": 48914
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.833912768647282e-07,
+      "loss": 0.5072,
+      "step": 48915
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.833438685208597e-07,
+      "loss": 0.9061,
+      "step": 48916
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.832964601769912e-07,
+      "loss": 2.6979,
+      "step": 48917
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.832490518331226e-07,
+      "loss": 3.8355,
+      "step": 48918
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.832016434892542e-07,
+      "loss": 1.2286,
+      "step": 48919
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.831542351453856e-07,
+      "loss": 1.4228,
+      "step": 48920
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.831068268015171e-07,
+      "loss": 1.0656,
+      "step": 48921
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 6.830594184576487e-07,
+      "loss": 0.8165,
+      "step": 48922
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.8301201011378e-07,
+      "loss": 2.5283,
+      "step": 48923
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.829646017699115e-07,
+      "loss": 1.1657,
+      "step": 48924
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.82917193426043e-07,
+      "loss": 1.7328,
+      "step": 48925
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.828697850821745e-07,
+      "loss": 2.0979,
+      "step": 48926
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.82822376738306e-07,
+      "loss": 3.1347,
+      "step": 48927
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.827749683944374e-07,
+      "loss": 1.0928,
+      "step": 48928
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.827275600505689e-07,
+      "loss": 1.2637,
+      "step": 48929
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.826801517067004e-07,
+      "loss": 1.4738,
+      "step": 48930
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.826327433628319e-07,
+      "loss": 1.178,
+      "step": 48931
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.825853350189634e-07,
+      "loss": 1.2978,
+      "step": 48932
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.825379266750949e-07,
+      "loss": 1.5628,
+      "step": 48933
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.824905183312263e-07,
+      "loss": 1.4819,
+      "step": 48934
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.824431099873577e-07,
+      "loss": 1.8795,
+      "step": 48935
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.823957016434892e-07,
+      "loss": 2.7915,
+      "step": 48936
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.823482932996208e-07,
+      "loss": 1.3654,
+      "step": 48937
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.823008849557522e-07,
+      "loss": 1.3597,
+      "step": 48938
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.822534766118837e-07,
+      "loss": 1.6967,
+      "step": 48939
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.822060682680152e-07,
+      "loss": 1.1612,
+      "step": 48940
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.821586599241467e-07,
+      "loss": 1.5365,
+      "step": 48941
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.821112515802782e-07,
+      "loss": 2.4711,
+      "step": 48942
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.820638432364096e-07,
+      "loss": 4.0201,
+      "step": 48943
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.820164348925411e-07,
+      "loss": 1.0538,
+      "step": 48944
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.819690265486726e-07,
+      "loss": 2.4996,
+      "step": 48945
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.81921618204804e-07,
+      "loss": 1.6353,
+      "step": 48946
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.818742098609355e-07,
+      "loss": 1.6262,
+      "step": 48947
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.81826801517067e-07,
+      "loss": 2.222,
+      "step": 48948
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.817793931731985e-07,
+      "loss": 2.4532,
+      "step": 48949
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.8173198482933e-07,
+      "loss": 2.389,
+      "step": 48950
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.816845764854614e-07,
+      "loss": 2.2046,
+      "step": 48951
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.81637168141593e-07,
+      "loss": 3.1613,
+      "step": 48952
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.815897597977244e-07,
+      "loss": 3.0267,
+      "step": 48953
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.815423514538559e-07,
+      "loss": 1.9693,
+      "step": 48954
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.814949431099874e-07,
+      "loss": 3.0404,
+      "step": 48955
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.814475347661189e-07,
+      "loss": 1.4052,
+      "step": 48956
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.814001264222504e-07,
+      "loss": 1.3723,
+      "step": 48957
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.813527180783817e-07,
+      "loss": 2.0016,
+      "step": 48958
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.813053097345133e-07,
+      "loss": 0.8506,
+      "step": 48959
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.812579013906448e-07,
+      "loss": 1.9229,
+      "step": 48960
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.812104930467762e-07,
+      "loss": 1.2967,
+      "step": 48961
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.811630847029077e-07,
+      "loss": 1.7369,
+      "step": 48962
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.811156763590392e-07,
+      "loss": 0.988,
+      "step": 48963
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.810682680151707e-07,
+      "loss": 1.5999,
+      "step": 48964
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.810208596713022e-07,
+      "loss": 2.2805,
+      "step": 48965
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.809734513274336e-07,
+      "loss": 1.7953,
+      "step": 48966
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.809260429835652e-07,
+      "loss": 1.4209,
+      "step": 48967
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.808786346396966e-07,
+      "loss": 1.5247,
+      "step": 48968
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.80831226295828e-07,
+      "loss": 1.156,
+      "step": 48969
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.807838179519596e-07,
+      "loss": 1.8229,
+      "step": 48970
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.80736409608091e-07,
+      "loss": 0.5224,
+      "step": 48971
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.806890012642225e-07,
+      "loss": 2.3177,
+      "step": 48972
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.80641592920354e-07,
+      "loss": 1.8928,
+      "step": 48973
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.805941845764855e-07,
+      "loss": 0.6071,
+      "step": 48974
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.80546776232617e-07,
+      "loss": 2.4487,
+      "step": 48975
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.804993678887484e-07,
+      "loss": 1.972,
+      "step": 48976
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.804519595448799e-07,
+      "loss": 1.7564,
+      "step": 48977
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.804045512010114e-07,
+      "loss": 1.7897,
+      "step": 48978
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.803571428571429e-07,
+      "loss": 1.7601,
+      "step": 48979
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.803097345132744e-07,
+      "loss": 1.3968,
+      "step": 48980
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.802623261694058e-07,
+      "loss": 2.2077,
+      "step": 48981
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.802149178255373e-07,
+      "loss": 1.287,
+      "step": 48982
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.801675094816688e-07,
+      "loss": 2.7929,
+      "step": 48983
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.801201011378002e-07,
+      "loss": 2.8485,
+      "step": 48984
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.800726927939318e-07,
+      "loss": 2.0585,
+      "step": 48985
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.800252844500632e-07,
+      "loss": 2.7963,
+      "step": 48986
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.799778761061947e-07,
+      "loss": 2.4828,
+      "step": 48987
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.799304677623262e-07,
+      "loss": 2.3896,
+      "step": 48988
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.798830594184577e-07,
+      "loss": 0.5304,
+      "step": 48989
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.798356510745892e-07,
+      "loss": 2.636,
+      "step": 48990
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.797882427307206e-07,
+      "loss": 1.0042,
+      "step": 48991
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.797408343868522e-07,
+      "loss": 2.3842,
+      "step": 48992
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.796934260429836e-07,
+      "loss": 3.2531,
+      "step": 48993
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.79646017699115e-07,
+      "loss": 1.5141,
+      "step": 48994
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.795986093552465e-07,
+      "loss": 2.1397,
+      "step": 48995
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.79551201011378e-07,
+      "loss": 2.7339,
+      "step": 48996
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.795037926675095e-07,
+      "loss": 0.9945,
+      "step": 48997
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.79456384323641e-07,
+      "loss": 1.5088,
+      "step": 48998
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.794089759797724e-07,
+      "loss": 1.5297,
+      "step": 48999
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.79361567635904e-07,
+      "loss": 3.1956,
+      "step": 49000
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.793141592920354e-07,
+      "loss": 2.1026,
+      "step": 49001
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.792667509481669e-07,
+      "loss": 1.5543,
+      "step": 49002
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.792193426042985e-07,
+      "loss": 3.2293,
+      "step": 49003
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.791719342604298e-07,
+      "loss": 1.8653,
+      "step": 49004
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.791245259165613e-07,
+      "loss": 1.714,
+      "step": 49005
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.790771175726927e-07,
+      "loss": 2.5297,
+      "step": 49006
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.790297092288243e-07,
+      "loss": 2.8284,
+      "step": 49007
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.789823008849558e-07,
+      "loss": 1.9623,
+      "step": 49008
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.789348925410872e-07,
+      "loss": 2.7906,
+      "step": 49009
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.788874841972187e-07,
+      "loss": 2.1702,
+      "step": 49010
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.788400758533502e-07,
+      "loss": 2.5514,
+      "step": 49011
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.787926675094817e-07,
+      "loss": 0.1656,
+      "step": 49012
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.787452591656132e-07,
+      "loss": 2.335,
+      "step": 49013
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.786978508217447e-07,
+      "loss": 1.7752,
+      "step": 49014
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.786504424778762e-07,
+      "loss": 2.064,
+      "step": 49015
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.786030341340075e-07,
+      "loss": 1.2701,
+      "step": 49016
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.78555625790139e-07,
+      "loss": 0.904,
+      "step": 49017
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.785082174462706e-07,
+      "loss": 1.7776,
+      "step": 49018
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.78460809102402e-07,
+      "loss": 1.367,
+      "step": 49019
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.784134007585335e-07,
+      "loss": 1.6221,
+      "step": 49020
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.78365992414665e-07,
+      "loss": 4.2224,
+      "step": 49021
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.783185840707965e-07,
+      "loss": 2.2537,
+      "step": 49022
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.78271175726928e-07,
+      "loss": 0.6242,
+      "step": 49023
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.782237673830594e-07,
+      "loss": 1.6478,
+      "step": 49024
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.78176359039191e-07,
+      "loss": 1.336,
+      "step": 49025
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.781289506953224e-07,
+      "loss": 0.5851,
+      "step": 49026
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.780815423514539e-07,
+      "loss": 2.5724,
+      "step": 49027
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.780341340075853e-07,
+      "loss": 2.1624,
+      "step": 49028
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.779867256637168e-07,
+      "loss": 2.4226,
+      "step": 49029
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.779393173198483e-07,
+      "loss": 1.6714,
+      "step": 49030
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.778919089759798e-07,
+      "loss": 1.749,
+      "step": 49031
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.778445006321112e-07,
+      "loss": 1.4671,
+      "step": 49032
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.777970922882428e-07,
+      "loss": 2.0625,
+      "step": 49033
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.777496839443742e-07,
+      "loss": 1.2626,
+      "step": 49034
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.777022756005057e-07,
+      "loss": 1.7624,
+      "step": 49035
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.776548672566372e-07,
+      "loss": 0.8987,
+      "step": 49036
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.776074589127687e-07,
+      "loss": 2.1711,
+      "step": 49037
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.775600505689002e-07,
+      "loss": 2.5574,
+      "step": 49038
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.775126422250315e-07,
+      "loss": 1.5536,
+      "step": 49039
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.774652338811631e-07,
+      "loss": 1.3431,
+      "step": 49040
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.774178255372946e-07,
+      "loss": 2.2933,
+      "step": 49041
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.77370417193426e-07,
+      "loss": 1.3688,
+      "step": 49042
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.773230088495575e-07,
+      "loss": 2.4942,
+      "step": 49043
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.77275600505689e-07,
+      "loss": 1.7231,
+      "step": 49044
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.772281921618205e-07,
+      "loss": 1.1064,
+      "step": 49045
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.77180783817952e-07,
+      "loss": 1.8873,
+      "step": 49046
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.771333754740835e-07,
+      "loss": 1.0962,
+      "step": 49047
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.77085967130215e-07,
+      "loss": 1.6086,
+      "step": 49048
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.770385587863464e-07,
+      "loss": 1.422,
+      "step": 49049
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.769911504424779e-07,
+      "loss": 2.152,
+      "step": 49050
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.769437420986095e-07,
+      "loss": 2.824,
+      "step": 49051
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.768963337547408e-07,
+      "loss": 1.3223,
+      "step": 49052
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.768489254108723e-07,
+      "loss": 2.4148,
+      "step": 49053
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.768015170670038e-07,
+      "loss": 2.6363,
+      "step": 49054
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.767541087231353e-07,
+      "loss": 2.259,
+      "step": 49055
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.767067003792668e-07,
+      "loss": 2.4691,
+      "step": 49056
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.766592920353982e-07,
+      "loss": 2.6015,
+      "step": 49057
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.766118836915298e-07,
+      "loss": 2.0073,
+      "step": 49058
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.765644753476612e-07,
+      "loss": 2.4258,
+      "step": 49059
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.765170670037927e-07,
+      "loss": 1.7999,
+      "step": 49060
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.764696586599242e-07,
+      "loss": 2.5214,
+      "step": 49061
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.764222503160557e-07,
+      "loss": 2.6948,
+      "step": 49062
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.763748419721871e-07,
+      "loss": 1.2077,
+      "step": 49063
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.763274336283186e-07,
+      "loss": 1.1544,
+      "step": 49064
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.7628002528445e-07,
+      "loss": 1.3807,
+      "step": 49065
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.762326169405816e-07,
+      "loss": 2.7334,
+      "step": 49066
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.76185208596713e-07,
+      "loss": 1.5404,
+      "step": 49067
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.761378002528445e-07,
+      "loss": 2.9328,
+      "step": 49068
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.76090391908976e-07,
+      "loss": 1.9371,
+      "step": 49069
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.760429835651075e-07,
+      "loss": 2.5104,
+      "step": 49070
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.75995575221239e-07,
+      "loss": 2.0019,
+      "step": 49071
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.759481668773704e-07,
+      "loss": 1.2942,
+      "step": 49072
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.75900758533502e-07,
+      "loss": 0.9134,
+      "step": 49073
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.758533501896333e-07,
+      "loss": 1.7334,
+      "step": 49074
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.758059418457648e-07,
+      "loss": 2.3132,
+      "step": 49075
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.757585335018963e-07,
+      "loss": 3.345,
+      "step": 49076
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.757111251580278e-07,
+      "loss": 2.1214,
+      "step": 49077
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.756637168141593e-07,
+      "loss": 2.4146,
+      "step": 49078
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.756163084702908e-07,
+      "loss": 2.3348,
+      "step": 49079
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.755689001264223e-07,
+      "loss": 0.9208,
+      "step": 49080
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.755214917825538e-07,
+      "loss": 1.6528,
+      "step": 49081
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.754740834386852e-07,
+      "loss": 1.6163,
+      "step": 49082
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.754266750948167e-07,
+      "loss": 2.7023,
+      "step": 49083
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.753792667509482e-07,
+      "loss": 2.1119,
+      "step": 49084
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.753318584070797e-07,
+      "loss": 2.4425,
+      "step": 49085
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.752844500632111e-07,
+      "loss": 2.9803,
+      "step": 49086
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.752370417193425e-07,
+      "loss": 0.767,
+      "step": 49087
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.751896333754741e-07,
+      "loss": 1.9956,
+      "step": 49088
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.751422250316056e-07,
+      "loss": 1.1407,
+      "step": 49089
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.75094816687737e-07,
+      "loss": 2.4533,
+      "step": 49090
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.750474083438686e-07,
+      "loss": 2.1856,
+      "step": 49091
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.75e-07,
+      "loss": 1.5563,
+      "step": 49092
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.749525916561315e-07,
+      "loss": 1.1184,
+      "step": 49093
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.74905183312263e-07,
+      "loss": 2.955,
+      "step": 49094
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.748577749683945e-07,
+      "loss": 0.5805,
+      "step": 49095
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.74810366624526e-07,
+      "loss": 1.5517,
+      "step": 49096
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.747629582806574e-07,
+      "loss": 1.0374,
+      "step": 49097
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.747155499367888e-07,
+      "loss": 2.5252,
+      "step": 49098
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.746681415929204e-07,
+      "loss": 1.0205,
+      "step": 49099
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.746207332490518e-07,
+      "loss": 1.254,
+      "step": 49100
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.745733249051833e-07,
+      "loss": 1.0644,
+      "step": 49101
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.745259165613148e-07,
+      "loss": 2.2754,
+      "step": 49102
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.744785082174463e-07,
+      "loss": 1.8919,
+      "step": 49103
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.744310998735778e-07,
+      "loss": 0.9388,
+      "step": 49104
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.743836915297092e-07,
+      "loss": 2.0713,
+      "step": 49105
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.743362831858408e-07,
+      "loss": 1.561,
+      "step": 49106
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.742888748419722e-07,
+      "loss": 1.0435,
+      "step": 49107
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.742414664981037e-07,
+      "loss": 2.1538,
+      "step": 49108
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.741940581542352e-07,
+      "loss": 2.6897,
+      "step": 49109
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.741466498103666e-07,
+      "loss": 1.9994,
+      "step": 49110
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.740992414664981e-07,
+      "loss": 3.0474,
+      "step": 49111
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.740518331226296e-07,
+      "loss": 1.6135,
+      "step": 49112
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.740044247787611e-07,
+      "loss": 2.5642,
+      "step": 49113
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.739570164348926e-07,
+      "loss": 1.3206,
+      "step": 49114
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.73909608091024e-07,
+      "loss": 1.2325,
+      "step": 49115
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.738621997471555e-07,
+      "loss": 1.7309,
+      "step": 49116
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.73814791403287e-07,
+      "loss": 1.4029,
+      "step": 49117
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.737673830594185e-07,
+      "loss": 0.9892,
+      "step": 49118
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.7371997471555e-07,
+      "loss": 1.725,
+      "step": 49119
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.736725663716814e-07,
+      "loss": 1.496,
+      "step": 49120
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.73625158027813e-07,
+      "loss": 0.6401,
+      "step": 49121
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.735777496839444e-07,
+      "loss": 1.6873,
+      "step": 49122
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.735303413400758e-07,
+      "loss": 0.7556,
+      "step": 49123
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.734829329962074e-07,
+      "loss": 1.0439,
+      "step": 49124
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.734355246523388e-07,
+      "loss": 2.6521,
+      "step": 49125
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.733881163084703e-07,
+      "loss": 2.4276,
+      "step": 49126
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.733407079646018e-07,
+      "loss": 3.6337,
+      "step": 49127
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.732932996207333e-07,
+      "loss": 1.9699,
+      "step": 49128
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.732458912768648e-07,
+      "loss": 1.4737,
+      "step": 49129
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.731984829329962e-07,
+      "loss": 1.4352,
+      "step": 49130
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.731510745891277e-07,
+      "loss": 1.7694,
+      "step": 49131
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.731036662452593e-07,
+      "loss": 2.0612,
+      "step": 49132
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.730562579013906e-07,
+      "loss": 1.5754,
+      "step": 49133
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.730088495575221e-07,
+      "loss": 2.2093,
+      "step": 49134
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.729614412136536e-07,
+      "loss": 1.6758,
+      "step": 49135
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.729140328697851e-07,
+      "loss": 1.3477,
+      "step": 49136
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.728666245259166e-07,
+      "loss": 1.0132,
+      "step": 49137
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.72819216182048e-07,
+      "loss": 3.158,
+      "step": 49138
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.727718078381796e-07,
+      "loss": 1.1449,
+      "step": 49139
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.72724399494311e-07,
+      "loss": 1.7413,
+      "step": 49140
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.726769911504425e-07,
+      "loss": 1.4721,
+      "step": 49141
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.72629582806574e-07,
+      "loss": 2.3585,
+      "step": 49142
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.725821744627055e-07,
+      "loss": 1.0193,
+      "step": 49143
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.72534766118837e-07,
+      "loss": 1.2977,
+      "step": 49144
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.724873577749683e-07,
+      "loss": 3.5077,
+      "step": 49145
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.724399494310999e-07,
+      "loss": 2.0986,
+      "step": 49146
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.723925410872314e-07,
+      "loss": 1.7793,
+      "step": 49147
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.723451327433628e-07,
+      "loss": 1.5298,
+      "step": 49148
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.722977243994943e-07,
+      "loss": 1.2616,
+      "step": 49149
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.722503160556258e-07,
+      "loss": 2.4838,
+      "step": 49150
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.722029077117573e-07,
+      "loss": 1.3683,
+      "step": 49151
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.721554993678888e-07,
+      "loss": 2.5027,
+      "step": 49152
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.721080910240202e-07,
+      "loss": 1.2586,
+      "step": 49153
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.720606826801518e-07,
+      "loss": 3.9147,
+      "step": 49154
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.720132743362832e-07,
+      "loss": 2.661,
+      "step": 49155
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.719658659924146e-07,
+      "loss": 2.3009,
+      "step": 49156
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.719184576485462e-07,
+      "loss": 0.9129,
+      "step": 49157
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.718710493046776e-07,
+      "loss": 1.3578,
+      "step": 49158
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.718236409608091e-07,
+      "loss": 0.8215,
+      "step": 49159
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.717762326169406e-07,
+      "loss": 2.207,
+      "step": 49160
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.717288242730721e-07,
+      "loss": 2.2024,
+      "step": 49161
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.716814159292036e-07,
+      "loss": 0.7789,
+      "step": 49162
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.71634007585335e-07,
+      "loss": 1.8573,
+      "step": 49163
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.715865992414665e-07,
+      "loss": 0.9598,
+      "step": 49164
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.71539190897598e-07,
+      "loss": 3.4829,
+      "step": 49165
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.714917825537295e-07,
+      "loss": 2.6815,
+      "step": 49166
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.71444374209861e-07,
+      "loss": 1.3535,
+      "step": 49167
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.713969658659923e-07,
+      "loss": 2.7074,
+      "step": 49168
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.713495575221239e-07,
+      "loss": 0.7824,
+      "step": 49169
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.713021491782554e-07,
+      "loss": 2.026,
+      "step": 49170
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.712547408343868e-07,
+      "loss": 2.5272,
+      "step": 49171
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.712073324905184e-07,
+      "loss": 1.4193,
+      "step": 49172
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.711599241466498e-07,
+      "loss": 1.5495,
+      "step": 49173
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.711125158027813e-07,
+      "loss": 2.5232,
+      "step": 49174
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.710651074589128e-07,
+      "loss": 1.5587,
+      "step": 49175
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.710176991150443e-07,
+      "loss": 1.716,
+      "step": 49176
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.709702907711758e-07,
+      "loss": 1.7528,
+      "step": 49177
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.709228824273072e-07,
+      "loss": 1.1192,
+      "step": 49178
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.708754740834387e-07,
+      "loss": 2.2658,
+      "step": 49179
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.708280657395702e-07,
+      "loss": 1.3263,
+      "step": 49180
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.707806573957016e-07,
+      "loss": 1.776,
+      "step": 49181
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.707332490518331e-07,
+      "loss": 1.7846,
+      "step": 49182
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.706858407079646e-07,
+      "loss": 2.9766,
+      "step": 49183
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.706384323640961e-07,
+      "loss": 1.3887,
+      "step": 49184
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.705910240202276e-07,
+      "loss": 2.6366,
+      "step": 49185
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.70543615676359e-07,
+      "loss": 1.029,
+      "step": 49186
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.704962073324906e-07,
+      "loss": 1.7088,
+      "step": 49187
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.70448798988622e-07,
+      "loss": 2.1334,
+      "step": 49188
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.704013906447535e-07,
+      "loss": 2.2555,
+      "step": 49189
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.70353982300885e-07,
+      "loss": 2.0644,
+      "step": 49190
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.703065739570165e-07,
+      "loss": 0.9271,
+      "step": 49191
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.702591656131479e-07,
+      "loss": 2.8725,
+      "step": 49192
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.702117572692794e-07,
+      "loss": 1.4455,
+      "step": 49193
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.701643489254109e-07,
+      "loss": 1.7005,
+      "step": 49194
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.701169405815424e-07,
+      "loss": 3.0305,
+      "step": 49195
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.700695322376738e-07,
+      "loss": 1.4759,
+      "step": 49196
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.700221238938053e-07,
+      "loss": 2.1061,
+      "step": 49197
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.699747155499368e-07,
+      "loss": 1.942,
+      "step": 49198
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.699273072060683e-07,
+      "loss": 2.4526,
+      "step": 49199
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.698798988621998e-07,
+      "loss": 0.6892,
+      "step": 49200
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.698324905183312e-07,
+      "loss": 1.4925,
+      "step": 49201
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.697850821744628e-07,
+      "loss": 2.2844,
+      "step": 49202
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.697376738305943e-07,
+      "loss": 0.4624,
+      "step": 49203
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.696902654867256e-07,
+      "loss": 1.9183,
+      "step": 49204
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.696428571428572e-07,
+      "loss": 1.7682,
+      "step": 49205
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.695954487989886e-07,
+      "loss": 0.5352,
+      "step": 49206
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.695480404551201e-07,
+      "loss": 2.4963,
+      "step": 49207
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.695006321112516e-07,
+      "loss": 0.5231,
+      "step": 49208
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.694532237673831e-07,
+      "loss": 0.4698,
+      "step": 49209
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.694058154235146e-07,
+      "loss": 1.5559,
+      "step": 49210
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.69358407079646e-07,
+      "loss": 1.3502,
+      "step": 49211
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.693109987357775e-07,
+      "loss": 1.9405,
+      "step": 49212
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.69263590391909e-07,
+      "loss": 1.4866,
+      "step": 49213
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.692161820480405e-07,
+      "loss": 1.3975,
+      "step": 49214
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.691687737041719e-07,
+      "loss": 0.8815,
+      "step": 49215
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.691213653603034e-07,
+      "loss": 1.9272,
+      "step": 49216
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.690739570164349e-07,
+      "loss": 0.9574,
+      "step": 49217
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.690265486725664e-07,
+      "loss": 1.4469,
+      "step": 49218
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.689791403286978e-07,
+      "loss": 1.9665,
+      "step": 49219
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.689317319848294e-07,
+      "loss": 2.2405,
+      "step": 49220
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.688843236409608e-07,
+      "loss": 2.9026,
+      "step": 49221
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.688369152970923e-07,
+      "loss": 2.4234,
+      "step": 49222
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.687895069532238e-07,
+      "loss": 1.5119,
+      "step": 49223
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.687420986093553e-07,
+      "loss": 0.4575,
+      "step": 49224
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.686946902654868e-07,
+      "loss": 1.6683,
+      "step": 49225
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.686472819216181e-07,
+      "loss": 1.2824,
+      "step": 49226
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.685998735777497e-07,
+      "loss": 2.0553,
+      "step": 49227
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.685524652338812e-07,
+      "loss": 2.8242,
+      "step": 49228
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.685050568900126e-07,
+      "loss": 2.341,
+      "step": 49229
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.684576485461441e-07,
+      "loss": 3.1655,
+      "step": 49230
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.684102402022756e-07,
+      "loss": 1.2772,
+      "step": 49231
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.683628318584071e-07,
+      "loss": 1.9515,
+      "step": 49232
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.683154235145386e-07,
+      "loss": 1.3242,
+      "step": 49233
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.6826801517067e-07,
+      "loss": 1.9653,
+      "step": 49234
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.682206068268016e-07,
+      "loss": 1.7863,
+      "step": 49235
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.68173198482933e-07,
+      "loss": 4.2614,
+      "step": 49236
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.681257901390645e-07,
+      "loss": 3.3966,
+      "step": 49237
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.680783817951961e-07,
+      "loss": 2.1794,
+      "step": 49238
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 6.680309734513274e-07,
+      "loss": 2.0215,
+      "step": 49239
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.679835651074589e-07,
+      "loss": 1.663,
+      "step": 49240
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.679361567635904e-07,
+      "loss": 2.5154,
+      "step": 49241
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.678887484197219e-07,
+      "loss": 1.5083,
+      "step": 49242
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.678413400758534e-07,
+      "loss": 1.431,
+      "step": 49243
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.677939317319848e-07,
+      "loss": 1.4867,
+      "step": 49244
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.677465233881163e-07,
+      "loss": 1.8321,
+      "step": 49245
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.676991150442478e-07,
+      "loss": 2.0605,
+      "step": 49246
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.676517067003793e-07,
+      "loss": 2.4779,
+      "step": 49247
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.676042983565108e-07,
+      "loss": 2.4407,
+      "step": 49248
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.675568900126423e-07,
+      "loss": 0.8901,
+      "step": 49249
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.675094816687737e-07,
+      "loss": 2.2091,
+      "step": 49250
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.674620733249052e-07,
+      "loss": 1.9252,
+      "step": 49251
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.674146649810366e-07,
+      "loss": 1.5419,
+      "step": 49252
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.673672566371682e-07,
+      "loss": 1.8982,
+      "step": 49253
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.673198482932996e-07,
+      "loss": 1.2834,
+      "step": 49254
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.672724399494311e-07,
+      "loss": 1.235,
+      "step": 49255
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.672250316055626e-07,
+      "loss": 2.388,
+      "step": 49256
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.671776232616941e-07,
+      "loss": 1.6368,
+      "step": 49257
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.671302149178256e-07,
+      "loss": 3.3925,
+      "step": 49258
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.67082806573957e-07,
+      "loss": 2.6175,
+      "step": 49259
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.670353982300886e-07,
+      "loss": 1.323,
+      "step": 49260
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.669879898862201e-07,
+      "loss": 1.0471,
+      "step": 49261
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.669405815423514e-07,
+      "loss": 1.9528,
+      "step": 49262
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.668931731984829e-07,
+      "loss": 2.2101,
+      "step": 49263
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.668457648546144e-07,
+      "loss": 1.8602,
+      "step": 49264
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.667983565107459e-07,
+      "loss": 2.1703,
+      "step": 49265
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.667509481668774e-07,
+      "loss": 1.7166,
+      "step": 49266
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.667035398230088e-07,
+      "loss": 1.2829,
+      "step": 49267
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.666561314791404e-07,
+      "loss": 4.2577,
+      "step": 49268
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.666087231352718e-07,
+      "loss": 1.0204,
+      "step": 49269
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.665613147914033e-07,
+      "loss": 1.9176,
+      "step": 49270
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.665139064475349e-07,
+      "loss": 1.6995,
+      "step": 49271
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.664664981036663e-07,
+      "loss": 3.289,
+      "step": 49272
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.664190897597978e-07,
+      "loss": 1.1332,
+      "step": 49273
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.663716814159292e-07,
+      "loss": 1.1338,
+      "step": 49274
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.663242730720607e-07,
+      "loss": 1.51,
+      "step": 49275
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.662768647281922e-07,
+      "loss": 2.4663,
+      "step": 49276
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.662294563843236e-07,
+      "loss": 1.6509,
+      "step": 49277
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.661820480404551e-07,
+      "loss": 2.2758,
+      "step": 49278
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.661346396965866e-07,
+      "loss": 1.993,
+      "step": 49279
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.660872313527181e-07,
+      "loss": 1.9116,
+      "step": 49280
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.660398230088496e-07,
+      "loss": 1.4274,
+      "step": 49281
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.659924146649811e-07,
+      "loss": 2.5235,
+      "step": 49282
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.659450063211126e-07,
+      "loss": 2.7312,
+      "step": 49283
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.65897597977244e-07,
+      "loss": 0.8423,
+      "step": 49284
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.658501896333754e-07,
+      "loss": 0.8567,
+      "step": 49285
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.65802781289507e-07,
+      "loss": 2.3334,
+      "step": 49286
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.657553729456384e-07,
+      "loss": 1.7382,
+      "step": 49287
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.657079646017699e-07,
+      "loss": 2.4248,
+      "step": 49288
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.656605562579014e-07,
+      "loss": 1.0751,
+      "step": 49289
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.656131479140329e-07,
+      "loss": 1.3973,
+      "step": 49290
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.655657395701644e-07,
+      "loss": 1.6231,
+      "step": 49291
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.655183312262958e-07,
+      "loss": 1.4783,
+      "step": 49292
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.654709228824274e-07,
+      "loss": 2.2127,
+      "step": 49293
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.654235145385588e-07,
+      "loss": 2.5798,
+      "step": 49294
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.653761061946903e-07,
+      "loss": 2.214,
+      "step": 49295
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.653286978508218e-07,
+      "loss": 1.2407,
+      "step": 49296
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.652812895069532e-07,
+      "loss": 1.5778,
+      "step": 49297
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.652338811630847e-07,
+      "loss": 1.9467,
+      "step": 49298
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.651864728192162e-07,
+      "loss": 0.5999,
+      "step": 49299
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.651390644753476e-07,
+      "loss": 2.6487,
+      "step": 49300
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.650916561314792e-07,
+      "loss": 3.4776,
+      "step": 49301
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 6.650442477876106e-07,
+      "loss": 2.5621,
+      "step": 49302
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.649968394437421e-07,
+      "loss": 2.4334,
+      "step": 49303
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.649494310998736e-07,
+      "loss": 2.441,
+      "step": 49304
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.649020227560051e-07,
+      "loss": 3.0559,
+      "step": 49305
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.648546144121366e-07,
+      "loss": 2.5734,
+      "step": 49306
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.64807206068268e-07,
+      "loss": 3.4874,
+      "step": 49307
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.647597977243996e-07,
+      "loss": 0.9935,
+      "step": 49308
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.64712389380531e-07,
+      "loss": 2.727,
+      "step": 49309
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.646649810366624e-07,
+      "loss": 1.4918,
+      "step": 49310
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.646175726927939e-07,
+      "loss": 1.2894,
+      "step": 49311
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.645701643489254e-07,
+      "loss": 2.3265,
+      "step": 49312
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.645227560050569e-07,
+      "loss": 1.8913,
+      "step": 49313
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.644753476611884e-07,
+      "loss": 2.5353,
+      "step": 49314
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.644279393173199e-07,
+      "loss": 2.3252,
+      "step": 49315
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.643805309734514e-07,
+      "loss": 1.8358,
+      "step": 49316
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.643331226295828e-07,
+      "loss": 1.5566,
+      "step": 49317
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.642857142857143e-07,
+      "loss": 1.5431,
+      "step": 49318
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.642383059418459e-07,
+      "loss": 0.9815,
+      "step": 49319
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.641908975979772e-07,
+      "loss": 1.7834,
+      "step": 49320
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.641434892541087e-07,
+      "loss": 1.4033,
+      "step": 49321
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.640960809102402e-07,
+      "loss": 2.2112,
+      "step": 49322
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.640486725663717e-07,
+      "loss": 0.4,
+      "step": 49323
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.640012642225032e-07,
+      "loss": 2.0493,
+      "step": 49324
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.639538558786346e-07,
+      "loss": 1.1714,
+      "step": 49325
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.639064475347662e-07,
+      "loss": 2.7064,
+      "step": 49326
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.638590391908976e-07,
+      "loss": 1.5691,
+      "step": 49327
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.638116308470291e-07,
+      "loss": 2.8778,
+      "step": 49328
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.637642225031606e-07,
+      "loss": 0.4919,
+      "step": 49329
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.637168141592921e-07,
+      "loss": 1.1974,
+      "step": 49330
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.636694058154236e-07,
+      "loss": 2.3172,
+      "step": 49331
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.63621997471555e-07,
+      "loss": 2.3983,
+      "step": 49332
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.635745891276864e-07,
+      "loss": 2.6085,
+      "step": 49333
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.63527180783818e-07,
+      "loss": 1.649,
+      "step": 49334
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.634797724399494e-07,
+      "loss": 1.3936,
+      "step": 49335
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.634323640960809e-07,
+      "loss": 1.3525,
+      "step": 49336
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.633849557522124e-07,
+      "loss": 1.9105,
+      "step": 49337
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.633375474083439e-07,
+      "loss": 1.625,
+      "step": 49338
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.632901390644754e-07,
+      "loss": 2.5655,
+      "step": 49339
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.632427307206068e-07,
+      "loss": 2.0221,
+      "step": 49340
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.631953223767384e-07,
+      "loss": 1.8951,
+      "step": 49341
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.631479140328699e-07,
+      "loss": 0.9765,
+      "step": 49342
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.631005056890012e-07,
+      "loss": 1.9589,
+      "step": 49343
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.630530973451327e-07,
+      "loss": 1.1763,
+      "step": 49344
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.630056890012642e-07,
+      "loss": 2.0773,
+      "step": 49345
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.629582806573957e-07,
+      "loss": 2.031,
+      "step": 49346
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.629108723135272e-07,
+      "loss": 2.3101,
+      "step": 49347
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.628634639696587e-07,
+      "loss": 1.9849,
+      "step": 49348
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.628160556257902e-07,
+      "loss": 2.1425,
+      "step": 49349
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.627686472819216e-07,
+      "loss": 2.1818,
+      "step": 49350
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.627212389380531e-07,
+      "loss": 1.1761,
+      "step": 49351
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.626738305941846e-07,
+      "loss": 3.0439,
+      "step": 49352
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.626264222503161e-07,
+      "loss": 2.3184,
+      "step": 49353
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.625790139064476e-07,
+      "loss": 0.813,
+      "step": 49354
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.625316055625789e-07,
+      "loss": 2.1403,
+      "step": 49355
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.624841972187105e-07,
+      "loss": 0.7909,
+      "step": 49356
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.62436788874842e-07,
+      "loss": 3.1755,
+      "step": 49357
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.623893805309734e-07,
+      "loss": 0.8065,
+      "step": 49358
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.62341972187105e-07,
+      "loss": 2.6775,
+      "step": 49359
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.622945638432364e-07,
+      "loss": 3.9421,
+      "step": 49360
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.622471554993679e-07,
+      "loss": 1.8139,
+      "step": 49361
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.621997471554994e-07,
+      "loss": 1.7383,
+      "step": 49362
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.621523388116309e-07,
+      "loss": 2.1672,
+      "step": 49363
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.621049304677624e-07,
+      "loss": 2.855,
+      "step": 49364
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 6.620575221238938e-07,
+      "loss": 1.0147,
+      "step": 49365
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.620101137800253e-07,
+      "loss": 1.3949,
+      "step": 49366
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.619627054361569e-07,
+      "loss": 2.5732,
+      "step": 49367
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.619152970922882e-07,
+      "loss": 1.8597,
+      "step": 49368
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.618678887484197e-07,
+      "loss": 0.4308,
+      "step": 49369
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.618204804045512e-07,
+      "loss": 3.5845,
+      "step": 49370
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.617730720606827e-07,
+      "loss": 2.2195,
+      "step": 49371
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.617256637168142e-07,
+      "loss": 1.1867,
+      "step": 49372
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.616782553729456e-07,
+      "loss": 1.1229,
+      "step": 49373
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.616308470290772e-07,
+      "loss": 1.713,
+      "step": 49374
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.615834386852086e-07,
+      "loss": 1.5579,
+      "step": 49375
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.615360303413401e-07,
+      "loss": 1.0734,
+      "step": 49376
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.614886219974716e-07,
+      "loss": 1.8221,
+      "step": 49377
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.61441213653603e-07,
+      "loss": 1.805,
+      "step": 49378
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.613938053097345e-07,
+      "loss": 2.1946,
+      "step": 49379
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.61346396965866e-07,
+      "loss": 2.7258,
+      "step": 49380
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.612989886219974e-07,
+      "loss": 1.9563,
+      "step": 49381
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.61251580278129e-07,
+      "loss": 1.883,
+      "step": 49382
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.612041719342604e-07,
+      "loss": 3.3787,
+      "step": 49383
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.611567635903919e-07,
+      "loss": 2.3073,
+      "step": 49384
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.611093552465234e-07,
+      "loss": 0.8576,
+      "step": 49385
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.610619469026549e-07,
+      "loss": 1.268,
+      "step": 49386
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.610145385587864e-07,
+      "loss": 1.5068,
+      "step": 49387
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.609671302149178e-07,
+      "loss": 2.3884,
+      "step": 49388
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.609197218710494e-07,
+      "loss": 2.3498,
+      "step": 49389
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.608723135271809e-07,
+      "loss": 1.3099,
+      "step": 49390
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.608249051833122e-07,
+      "loss": 1.7298,
+      "step": 49391
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.607774968394437e-07,
+      "loss": 1.0685,
+      "step": 49392
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.607300884955752e-07,
+      "loss": 1.614,
+      "step": 49393
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.606826801517067e-07,
+      "loss": 2.5359,
+      "step": 49394
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.606352718078382e-07,
+      "loss": 1.7674,
+      "step": 49395
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.605878634639697e-07,
+      "loss": 0.9831,
+      "step": 49396
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.605404551201012e-07,
+      "loss": 3.4881,
+      "step": 49397
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.604930467762326e-07,
+      "loss": 1.8046,
+      "step": 49398
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.604456384323641e-07,
+      "loss": 2.2817,
+      "step": 49399
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.603982300884957e-07,
+      "loss": 2.1445,
+      "step": 49400
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.603508217446271e-07,
+      "loss": 2.181,
+      "step": 49401
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.603034134007585e-07,
+      "loss": 2.2697,
+      "step": 49402
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.6025600505689e-07,
+      "loss": 1.682,
+      "step": 49403
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.602085967130215e-07,
+      "loss": 1.3987,
+      "step": 49404
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.60161188369153e-07,
+      "loss": 2.1117,
+      "step": 49405
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.601137800252844e-07,
+      "loss": 3.596,
+      "step": 49406
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.60066371681416e-07,
+      "loss": 1.2278,
+      "step": 49407
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.600189633375474e-07,
+      "loss": 0.6459,
+      "step": 49408
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.599715549936789e-07,
+      "loss": 2.5018,
+      "step": 49409
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.599241466498104e-07,
+      "loss": 3.1235,
+      "step": 49410
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.598767383059419e-07,
+      "loss": 3.1456,
+      "step": 49411
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.598293299620734e-07,
+      "loss": 1.422,
+      "step": 49412
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.597819216182049e-07,
+      "loss": 2.7583,
+      "step": 49413
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.597345132743362e-07,
+      "loss": 2.5589,
+      "step": 49414
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.596871049304678e-07,
+      "loss": 1.8316,
+      "step": 49415
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.596396965865992e-07,
+      "loss": 0.9545,
+      "step": 49416
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.595922882427307e-07,
+      "loss": 1.3594,
+      "step": 49417
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.595448798988622e-07,
+      "loss": 1.5272,
+      "step": 49418
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.594974715549937e-07,
+      "loss": 1.2896,
+      "step": 49419
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.594500632111252e-07,
+      "loss": 1.8563,
+      "step": 49420
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.594026548672566e-07,
+      "loss": 2.0838,
+      "step": 49421
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.593552465233882e-07,
+      "loss": 2.6341,
+      "step": 49422
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.593078381795196e-07,
+      "loss": 0.9299,
+      "step": 49423
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.592604298356511e-07,
+      "loss": 1.9472,
+      "step": 49424
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.592130214917825e-07,
+      "loss": 2.3863,
+      "step": 49425
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.59165613147914e-07,
+      "loss": 0.4656,
+      "step": 49426
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.591182048040455e-07,
+      "loss": 2.5692,
+      "step": 49427
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.59070796460177e-07,
+      "loss": 2.107,
+      "step": 49428
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 6.590233881163085e-07,
+      "loss": 1.5835,
+      "step": 49429
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.5897597977244e-07,
+      "loss": 1.3416,
+      "step": 49430
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.589285714285714e-07,
+      "loss": 2.6993,
+      "step": 49431
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.588811630847029e-07,
+      "loss": 1.232,
+      "step": 49432
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.588337547408344e-07,
+      "loss": 1.4804,
+      "step": 49433
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.587863463969659e-07,
+      "loss": 1.982,
+      "step": 49434
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.587389380530974e-07,
+      "loss": 1.8356,
+      "step": 49435
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.586915297092288e-07,
+      "loss": 2.0067,
+      "step": 49436
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.586441213653603e-07,
+      "loss": 1.9603,
+      "step": 49437
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.585967130214918e-07,
+      "loss": 2.3171,
+      "step": 49438
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.585493046776232e-07,
+      "loss": 1.6075,
+      "step": 49439
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.585018963337548e-07,
+      "loss": 1.9334,
+      "step": 49440
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.584544879898862e-07,
+      "loss": 1.7258,
+      "step": 49441
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.584070796460177e-07,
+      "loss": 3.3608,
+      "step": 49442
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.583596713021492e-07,
+      "loss": 2.4177,
+      "step": 49443
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.583122629582807e-07,
+      "loss": 2.2781,
+      "step": 49444
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.582648546144122e-07,
+      "loss": 2.5523,
+      "step": 49445
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.582174462705436e-07,
+      "loss": 4.0418,
+      "step": 49446
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.581700379266751e-07,
+      "loss": 1.8363,
+      "step": 49447
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.581226295828067e-07,
+      "loss": 1.1111,
+      "step": 49448
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.58075221238938e-07,
+      "loss": 1.1589,
+      "step": 49449
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.580278128950695e-07,
+      "loss": 1.3242,
+      "step": 49450
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.57980404551201e-07,
+      "loss": 2.3465,
+      "step": 49451
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.579329962073325e-07,
+      "loss": 1.8698,
+      "step": 49452
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.57885587863464e-07,
+      "loss": 1.4285,
+      "step": 49453
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.578381795195954e-07,
+      "loss": 2.7232,
+      "step": 49454
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.57790771175727e-07,
+      "loss": 1.1486,
+      "step": 49455
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.577433628318584e-07,
+      "loss": 2.6798,
+      "step": 49456
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.576959544879899e-07,
+      "loss": 3.8255,
+      "step": 49457
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.576485461441214e-07,
+      "loss": 2.7107,
+      "step": 49458
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.576011378002529e-07,
+      "loss": 1.7849,
+      "step": 49459
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.575537294563844e-07,
+      "loss": 3.1607,
+      "step": 49460
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.575063211125158e-07,
+      "loss": 1.5741,
+      "step": 49461
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.574589127686473e-07,
+      "loss": 0.661,
+      "step": 49462
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.574115044247788e-07,
+      "loss": 2.367,
+      "step": 49463
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.573640960809102e-07,
+      "loss": 2.1401,
+      "step": 49464
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.573166877370417e-07,
+      "loss": 4.1688,
+      "step": 49465
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.572692793931732e-07,
+      "loss": 3.0461,
+      "step": 49466
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.572218710493047e-07,
+      "loss": 1.7343,
+      "step": 49467
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.571744627054362e-07,
+      "loss": 0.2308,
+      "step": 49468
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.571270543615676e-07,
+      "loss": 1.7901,
+      "step": 49469
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.570796460176992e-07,
+      "loss": 1.8641,
+      "step": 49470
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.570322376738307e-07,
+      "loss": 2.1895,
+      "step": 49471
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.56984829329962e-07,
+      "loss": 1.2703,
+      "step": 49472
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.569374209860936e-07,
+      "loss": 0.9872,
+      "step": 49473
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.56890012642225e-07,
+      "loss": 2.3602,
+      "step": 49474
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.568426042983565e-07,
+      "loss": 2.8312,
+      "step": 49475
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.56795195954488e-07,
+      "loss": 2.0044,
+      "step": 49476
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.567477876106195e-07,
+      "loss": 1.422,
+      "step": 49477
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.56700379266751e-07,
+      "loss": 1.6271,
+      "step": 49478
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.566529709228824e-07,
+      "loss": 2.3373,
+      "step": 49479
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.566055625790139e-07,
+      "loss": 1.8464,
+      "step": 49480
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.565581542351455e-07,
+      "loss": 0.877,
+      "step": 49481
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.565107458912769e-07,
+      "loss": 1.7828,
+      "step": 49482
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.564633375474084e-07,
+      "loss": 4.295,
+      "step": 49483
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.564159292035398e-07,
+      "loss": 1.5378,
+      "step": 49484
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.563685208596713e-07,
+      "loss": 2.2529,
+      "step": 49485
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.563211125158028e-07,
+      "loss": 2.5482,
+      "step": 49486
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.562737041719342e-07,
+      "loss": 2.1377,
+      "step": 49487
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.562262958280658e-07,
+      "loss": 0.6285,
+      "step": 49488
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.561788874841972e-07,
+      "loss": 2.0865,
+      "step": 49489
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.561314791403287e-07,
+      "loss": 2.4973,
+      "step": 49490
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.560840707964602e-07,
+      "loss": 1.8809,
+      "step": 49491
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 6.560366624525917e-07,
+      "loss": 1.3808,
+      "step": 49492
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.559892541087232e-07,
+      "loss": 1.3244,
+      "step": 49493
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.559418457648546e-07,
+      "loss": 1.6143,
+      "step": 49494
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.558944374209862e-07,
+      "loss": 1.7697,
+      "step": 49495
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.558470290771176e-07,
+      "loss": 1.76,
+      "step": 49496
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.55799620733249e-07,
+      "loss": 2.524,
+      "step": 49497
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.557522123893805e-07,
+      "loss": 0.992,
+      "step": 49498
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.55704804045512e-07,
+      "loss": 0.807,
+      "step": 49499
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.556573957016435e-07,
+      "loss": 2.2079,
+      "step": 49500
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.55609987357775e-07,
+      "loss": 1.0467,
+      "step": 49501
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.555625790139064e-07,
+      "loss": 3.2562,
+      "step": 49502
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.55515170670038e-07,
+      "loss": 2.4077,
+      "step": 49503
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.554677623261694e-07,
+      "loss": 2.3598,
+      "step": 49504
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.554203539823009e-07,
+      "loss": 3.3859,
+      "step": 49505
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.553729456384325e-07,
+      "loss": 1.8,
+      "step": 49506
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.553255372945638e-07,
+      "loss": 2.0699,
+      "step": 49507
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.552781289506953e-07,
+      "loss": 2.0609,
+      "step": 49508
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.552307206068268e-07,
+      "loss": 2.3699,
+      "step": 49509
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.551833122629583e-07,
+      "loss": 2.3731,
+      "step": 49510
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.551359039190898e-07,
+      "loss": 2.5883,
+      "step": 49511
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.550884955752212e-07,
+      "loss": 1.6832,
+      "step": 49512
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.550410872313527e-07,
+      "loss": 2.0765,
+      "step": 49513
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.549936788874842e-07,
+      "loss": 1.4746,
+      "step": 49514
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.549462705436157e-07,
+      "loss": 2.8934,
+      "step": 49515
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.548988621997472e-07,
+      "loss": 1.5667,
+      "step": 49516
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.548514538558787e-07,
+      "loss": 1.4479,
+      "step": 49517
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.548040455120102e-07,
+      "loss": 2.8345,
+      "step": 49518
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.547566371681416e-07,
+      "loss": 0.3335,
+      "step": 49519
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.54709228824273e-07,
+      "loss": 2.9116,
+      "step": 49520
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.546618204804046e-07,
+      "loss": 2.732,
+      "step": 49521
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.54614412136536e-07,
+      "loss": 2.4798,
+      "step": 49522
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.545670037926675e-07,
+      "loss": 2.2839,
+      "step": 49523
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.54519595448799e-07,
+      "loss": 1.6538,
+      "step": 49524
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.544721871049305e-07,
+      "loss": 1.2312,
+      "step": 49525
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.54424778761062e-07,
+      "loss": 3.6133,
+      "step": 49526
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.543773704171934e-07,
+      "loss": 1.4987,
+      "step": 49527
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.54329962073325e-07,
+      "loss": 1.7574,
+      "step": 49528
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.542825537294565e-07,
+      "loss": 1.5969,
+      "step": 49529
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.542351453855879e-07,
+      "loss": 1.763,
+      "step": 49530
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.541877370417193e-07,
+      "loss": 1.3796,
+      "step": 49531
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.541403286978508e-07,
+      "loss": 1.9546,
+      "step": 49532
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.540929203539823e-07,
+      "loss": 1.6701,
+      "step": 49533
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.540455120101138e-07,
+      "loss": 2.9517,
+      "step": 49534
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.539981036662452e-07,
+      "loss": 1.4388,
+      "step": 49535
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.539506953223768e-07,
+      "loss": 1.1837,
+      "step": 49536
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.539032869785082e-07,
+      "loss": 0.5485,
+      "step": 49537
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.538558786346397e-07,
+      "loss": 1.282,
+      "step": 49538
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.538084702907713e-07,
+      "loss": 1.4368,
+      "step": 49539
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.537610619469027e-07,
+      "loss": 1.8625,
+      "step": 49540
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.537136536030342e-07,
+      "loss": 1.7435,
+      "step": 49541
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.536662452591657e-07,
+      "loss": 2.0992,
+      "step": 49542
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.536188369152971e-07,
+      "loss": 3.5819,
+      "step": 49543
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.535714285714286e-07,
+      "loss": 1.3613,
+      "step": 49544
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.5352402022756e-07,
+      "loss": 2.3913,
+      "step": 49545
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.534766118836915e-07,
+      "loss": 1.5853,
+      "step": 49546
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.53429203539823e-07,
+      "loss": 2.0768,
+      "step": 49547
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.533817951959545e-07,
+      "loss": 2.507,
+      "step": 49548
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.53334386852086e-07,
+      "loss": 1.5915,
+      "step": 49549
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.532869785082175e-07,
+      "loss": 1.513,
+      "step": 49550
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.53239570164349e-07,
+      "loss": 1.8,
+      "step": 49551
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.531921618204805e-07,
+      "loss": 2.4335,
+      "step": 49552
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.531447534766119e-07,
+      "loss": 2.9766,
+      "step": 49553
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.530973451327435e-07,
+      "loss": 2.0655,
+      "step": 49554
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 6.530499367888748e-07,
+      "loss": 2.3343,
+      "step": 49555
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.530025284450063e-07,
+      "loss": 1.0275,
+      "step": 49556
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.529551201011378e-07,
+      "loss": 1.3799,
+      "step": 49557
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.529077117572693e-07,
+      "loss": 3.0491,
+      "step": 49558
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.528603034134008e-07,
+      "loss": 0.9593,
+      "step": 49559
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.528128950695322e-07,
+      "loss": 2.4062,
+      "step": 49560
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.527654867256638e-07,
+      "loss": 2.9096,
+      "step": 49561
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.527180783817952e-07,
+      "loss": 0.6833,
+      "step": 49562
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.526706700379267e-07,
+      "loss": 2.0953,
+      "step": 49563
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.526232616940582e-07,
+      "loss": 0.8885,
+      "step": 49564
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.525758533501897e-07,
+      "loss": 1.6393,
+      "step": 49565
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.525284450063211e-07,
+      "loss": 1.7805,
+      "step": 49566
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.524810366624526e-07,
+      "loss": 1.5969,
+      "step": 49567
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.52433628318584e-07,
+      "loss": 0.9439,
+      "step": 49568
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.523862199747156e-07,
+      "loss": 3.8594,
+      "step": 49569
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.52338811630847e-07,
+      "loss": 1.6725,
+      "step": 49570
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.522914032869785e-07,
+      "loss": 0.6989,
+      "step": 49571
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.5224399494311e-07,
+      "loss": 2.0935,
+      "step": 49572
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.521965865992415e-07,
+      "loss": 1.7403,
+      "step": 49573
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.52149178255373e-07,
+      "loss": 2.3984,
+      "step": 49574
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.521017699115044e-07,
+      "loss": 1.4993,
+      "step": 49575
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.52054361567636e-07,
+      "loss": 1.1271,
+      "step": 49576
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.520069532237675e-07,
+      "loss": 1.9045,
+      "step": 49577
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.519595448798988e-07,
+      "loss": 1.3084,
+      "step": 49578
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.519121365360303e-07,
+      "loss": 1.1649,
+      "step": 49579
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.518647281921618e-07,
+      "loss": 0.5794,
+      "step": 49580
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.518173198482933e-07,
+      "loss": 2.2686,
+      "step": 49581
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.517699115044248e-07,
+      "loss": 0.9963,
+      "step": 49582
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.517225031605563e-07,
+      "loss": 1.3591,
+      "step": 49583
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.516750948166878e-07,
+      "loss": 2.5166,
+      "step": 49584
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.516276864728192e-07,
+      "loss": 0.894,
+      "step": 49585
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.515802781289507e-07,
+      "loss": 0.9936,
+      "step": 49586
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.515328697850823e-07,
+      "loss": 0.9403,
+      "step": 49587
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.514854614412137e-07,
+      "loss": 2.3333,
+      "step": 49588
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.514380530973451e-07,
+      "loss": 0.8614,
+      "step": 49589
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.513906447534766e-07,
+      "loss": 1.7429,
+      "step": 49590
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.513432364096081e-07,
+      "loss": 1.8363,
+      "step": 49591
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.512958280657396e-07,
+      "loss": 1.349,
+      "step": 49592
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.51248419721871e-07,
+      "loss": 1.0331,
+      "step": 49593
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.512010113780026e-07,
+      "loss": 1.7784,
+      "step": 49594
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.51153603034134e-07,
+      "loss": 1.0416,
+      "step": 49595
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.511061946902655e-07,
+      "loss": 2.3468,
+      "step": 49596
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.51058786346397e-07,
+      "loss": 0.9025,
+      "step": 49597
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.510113780025285e-07,
+      "loss": 1.6747,
+      "step": 49598
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.5096396965866e-07,
+      "loss": 1.7362,
+      "step": 49599
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.509165613147915e-07,
+      "loss": 2.4462,
+      "step": 49600
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.508691529709228e-07,
+      "loss": 2.1968,
+      "step": 49601
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.508217446270544e-07,
+      "loss": 2.2598,
+      "step": 49602
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.507743362831858e-07,
+      "loss": 3.5776,
+      "step": 49603
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.507269279393173e-07,
+      "loss": 1.939,
+      "step": 49604
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.506795195954488e-07,
+      "loss": 2.876,
+      "step": 49605
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.506321112515803e-07,
+      "loss": 3.7956,
+      "step": 49606
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.505847029077118e-07,
+      "loss": 0.8858,
+      "step": 49607
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.505372945638432e-07,
+      "loss": 3.364,
+      "step": 49608
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.504898862199748e-07,
+      "loss": 0.8172,
+      "step": 49609
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.504424778761063e-07,
+      "loss": 0.9854,
+      "step": 49610
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.503950695322377e-07,
+      "loss": 2.0324,
+      "step": 49611
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.503476611883692e-07,
+      "loss": 2.1416,
+      "step": 49612
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.503002528445006e-07,
+      "loss": 2.4159,
+      "step": 49613
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.502528445006321e-07,
+      "loss": 2.1342,
+      "step": 49614
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.502054361567636e-07,
+      "loss": 1.7085,
+      "step": 49615
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.50158027812895e-07,
+      "loss": 2.8057,
+      "step": 49616
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.501106194690266e-07,
+      "loss": 1.8592,
+      "step": 49617
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.50063211125158e-07,
+      "loss": 2.5092,
+      "step": 49618
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 6.500158027812895e-07,
+      "loss": 1.9865,
+      "step": 49619
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.499683944374211e-07,
+      "loss": 2.763,
+      "step": 49620
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.499209860935525e-07,
+      "loss": 1.9245,
+      "step": 49621
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.49873577749684e-07,
+      "loss": 2.5133,
+      "step": 49622
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.498261694058155e-07,
+      "loss": 0.652,
+      "step": 49623
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.49778761061947e-07,
+      "loss": 1.4612,
+      "step": 49624
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.497313527180784e-07,
+      "loss": 2.6542,
+      "step": 49625
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.496839443742098e-07,
+      "loss": 3.3069,
+      "step": 49626
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.496365360303413e-07,
+      "loss": 2.4491,
+      "step": 49627
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.495891276864728e-07,
+      "loss": 5.8182,
+      "step": 49628
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.495417193426043e-07,
+      "loss": 1.4429,
+      "step": 49629
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.494943109987358e-07,
+      "loss": 1.7226,
+      "step": 49630
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.494469026548673e-07,
+      "loss": 1.8406,
+      "step": 49631
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.493994943109988e-07,
+      "loss": 1.7189,
+      "step": 49632
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.493520859671302e-07,
+      "loss": 1.4005,
+      "step": 49633
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.493046776232617e-07,
+      "loss": 1.4037,
+      "step": 49634
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.492572692793933e-07,
+      "loss": 0.2168,
+      "step": 49635
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.492098609355246e-07,
+      "loss": 3.5243,
+      "step": 49636
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.491624525916561e-07,
+      "loss": 2.1217,
+      "step": 49637
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.491150442477876e-07,
+      "loss": 1.6458,
+      "step": 49638
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.490676359039191e-07,
+      "loss": 1.8964,
+      "step": 49639
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.490202275600506e-07,
+      "loss": 1.9911,
+      "step": 49640
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.48972819216182e-07,
+      "loss": 2.8845,
+      "step": 49641
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.489254108723136e-07,
+      "loss": 2.4632,
+      "step": 49642
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.48878002528445e-07,
+      "loss": 2.0598,
+      "step": 49643
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.488305941845765e-07,
+      "loss": 1.8741,
+      "step": 49644
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.48783185840708e-07,
+      "loss": 2.2384,
+      "step": 49645
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.487357774968395e-07,
+      "loss": 2.2271,
+      "step": 49646
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.48688369152971e-07,
+      "loss": 3.1043,
+      "step": 49647
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.486409608091024e-07,
+      "loss": 1.6273,
+      "step": 49648
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.485935524652338e-07,
+      "loss": 1.7002,
+      "step": 49649
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.485461441213654e-07,
+      "loss": 3.8991,
+      "step": 49650
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.484987357774968e-07,
+      "loss": 1.5222,
+      "step": 49651
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.484513274336283e-07,
+      "loss": 1.2368,
+      "step": 49652
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.484039190897598e-07,
+      "loss": 1.8544,
+      "step": 49653
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.483565107458913e-07,
+      "loss": 1.0409,
+      "step": 49654
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.483091024020228e-07,
+      "loss": 2.351,
+      "step": 49655
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.482616940581542e-07,
+      "loss": 0.9921,
+      "step": 49656
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.482142857142858e-07,
+      "loss": 1.2253,
+      "step": 49657
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.481668773704173e-07,
+      "loss": 2.8377,
+      "step": 49658
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.481194690265486e-07,
+      "loss": 1.5477,
+      "step": 49659
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.480720606826801e-07,
+      "loss": 2.4144,
+      "step": 49660
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.480246523388116e-07,
+      "loss": 1.8062,
+      "step": 49661
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.479772439949431e-07,
+      "loss": 2.4992,
+      "step": 49662
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.479298356510746e-07,
+      "loss": 1.373,
+      "step": 49663
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.478824273072061e-07,
+      "loss": 1.7168,
+      "step": 49664
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.478350189633376e-07,
+      "loss": 3.7039,
+      "step": 49665
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.47787610619469e-07,
+      "loss": 1.6943,
+      "step": 49666
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.477402022756005e-07,
+      "loss": 1.5393,
+      "step": 49667
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.476927939317321e-07,
+      "loss": 0.7977,
+      "step": 49668
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.476453855878635e-07,
+      "loss": 3.0175,
+      "step": 49669
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.47597977243995e-07,
+      "loss": 1.2285,
+      "step": 49670
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.475505689001264e-07,
+      "loss": 2.8963,
+      "step": 49671
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.475031605562579e-07,
+      "loss": 3.0787,
+      "step": 49672
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.474557522123894e-07,
+      "loss": 1.6933,
+      "step": 49673
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.474083438685208e-07,
+      "loss": 1.2015,
+      "step": 49674
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.473609355246524e-07,
+      "loss": 2.7812,
+      "step": 49675
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.473135271807838e-07,
+      "loss": 2.1523,
+      "step": 49676
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.472661188369153e-07,
+      "loss": 1.9905,
+      "step": 49677
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.472187104930468e-07,
+      "loss": 2.2528,
+      "step": 49678
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.471713021491783e-07,
+      "loss": 3.5354,
+      "step": 49679
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.471238938053098e-07,
+      "loss": 2.551,
+      "step": 49680
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.470764854614413e-07,
+      "loss": 0.6222,
+      "step": 49681
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 6.470290771175727e-07,
+      "loss": 1.9696,
+      "step": 49682
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.469816687737042e-07,
+      "loss": 1.2379,
+      "step": 49683
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.469342604298356e-07,
+      "loss": 1.0321,
+      "step": 49684
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.468868520859671e-07,
+      "loss": 1.0132,
+      "step": 49685
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.468394437420986e-07,
+      "loss": 1.9023,
+      "step": 49686
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.467920353982301e-07,
+      "loss": 2.3111,
+      "step": 49687
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.467446270543616e-07,
+      "loss": 1.1022,
+      "step": 49688
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.46697218710493e-07,
+      "loss": 0.5154,
+      "step": 49689
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.466498103666246e-07,
+      "loss": 2.4598,
+      "step": 49690
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.466024020227561e-07,
+      "loss": 2.0753,
+      "step": 49691
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.465549936788875e-07,
+      "loss": 2.0334,
+      "step": 49692
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.46507585335019e-07,
+      "loss": 1.6108,
+      "step": 49693
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.464601769911505e-07,
+      "loss": 1.7618,
+      "step": 49694
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.464127686472819e-07,
+      "loss": 1.9618,
+      "step": 49695
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.463653603034134e-07,
+      "loss": 1.3642,
+      "step": 49696
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.463179519595449e-07,
+      "loss": 2.0745,
+      "step": 49697
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.462705436156764e-07,
+      "loss": 2.2585,
+      "step": 49698
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.462231352718078e-07,
+      "loss": 0.7797,
+      "step": 49699
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.461757269279393e-07,
+      "loss": 2.4369,
+      "step": 49700
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.461283185840708e-07,
+      "loss": 0.6912,
+      "step": 49701
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.460809102402023e-07,
+      "loss": 1.771,
+      "step": 49702
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.460335018963338e-07,
+      "loss": 1.5601,
+      "step": 49703
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.459860935524652e-07,
+      "loss": 2.9756,
+      "step": 49704
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.459386852085968e-07,
+      "loss": 2.9776,
+      "step": 49705
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.458912768647283e-07,
+      "loss": 1.7955,
+      "step": 49706
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.458438685208596e-07,
+      "loss": 1.5132,
+      "step": 49707
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.457964601769912e-07,
+      "loss": 2.4317,
+      "step": 49708
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.457490518331226e-07,
+      "loss": 1.3768,
+      "step": 49709
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.457016434892541e-07,
+      "loss": 1.6034,
+      "step": 49710
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.456542351453856e-07,
+      "loss": 4.127,
+      "step": 49711
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.456068268015171e-07,
+      "loss": 2.9985,
+      "step": 49712
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.455594184576486e-07,
+      "loss": 2.0828,
+      "step": 49713
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.4551201011378e-07,
+      "loss": 4.042,
+      "step": 49714
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.454646017699115e-07,
+      "loss": 2.9088,
+      "step": 49715
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.454171934260431e-07,
+      "loss": 0.8357,
+      "step": 49716
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.453697850821745e-07,
+      "loss": 0.6106,
+      "step": 49717
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.453223767383059e-07,
+      "loss": 0.6745,
+      "step": 49718
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.452749683944374e-07,
+      "loss": 2.4902,
+      "step": 49719
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.452275600505689e-07,
+      "loss": 1.6911,
+      "step": 49720
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.451801517067004e-07,
+      "loss": 1.7865,
+      "step": 49721
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.451327433628318e-07,
+      "loss": 1.8456,
+      "step": 49722
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.450853350189634e-07,
+      "loss": 1.7423,
+      "step": 49723
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.450379266750948e-07,
+      "loss": 2.7876,
+      "step": 49724
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.449905183312263e-07,
+      "loss": 2.3902,
+      "step": 49725
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.449431099873578e-07,
+      "loss": 2.4798,
+      "step": 49726
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.448957016434893e-07,
+      "loss": 0.681,
+      "step": 49727
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.448482932996208e-07,
+      "loss": 0.7013,
+      "step": 49728
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.448008849557523e-07,
+      "loss": 1.339,
+      "step": 49729
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.447534766118837e-07,
+      "loss": 2.2367,
+      "step": 49730
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.447060682680152e-07,
+      "loss": 1.7862,
+      "step": 49731
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.446586599241466e-07,
+      "loss": 1.1557,
+      "step": 49732
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.446112515802781e-07,
+      "loss": 3.5887,
+      "step": 49733
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.445638432364096e-07,
+      "loss": 2.1624,
+      "step": 49734
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.445164348925411e-07,
+      "loss": 2.4151,
+      "step": 49735
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.444690265486726e-07,
+      "loss": 1.5977,
+      "step": 49736
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.44421618204804e-07,
+      "loss": 1.5311,
+      "step": 49737
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.443742098609356e-07,
+      "loss": 1.355,
+      "step": 49738
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.443268015170671e-07,
+      "loss": 1.6787,
+      "step": 49739
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.442793931731985e-07,
+      "loss": 2.6258,
+      "step": 49740
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.442319848293301e-07,
+      "loss": 2.0302,
+      "step": 49741
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.441845764854614e-07,
+      "loss": 1.4129,
+      "step": 49742
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.441371681415929e-07,
+      "loss": 2.3262,
+      "step": 49743
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.440897597977244e-07,
+      "loss": 2.2284,
+      "step": 49744
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 6.440423514538559e-07,
+      "loss": 1.2015,
+      "step": 49745
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.439949431099874e-07,
+      "loss": 1.1722,
+      "step": 49746
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.439475347661188e-07,
+      "loss": 1.7254,
+      "step": 49747
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.439001264222503e-07,
+      "loss": 2.5371,
+      "step": 49748
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.438527180783819e-07,
+      "loss": 2.0219,
+      "step": 49749
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.438053097345133e-07,
+      "loss": 1.6471,
+      "step": 49750
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.437579013906448e-07,
+      "loss": 0.495,
+      "step": 49751
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.437104930467763e-07,
+      "loss": 1.863,
+      "step": 49752
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.436630847029077e-07,
+      "loss": 1.7524,
+      "step": 49753
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.436156763590392e-07,
+      "loss": 1.0304,
+      "step": 49754
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.435682680151706e-07,
+      "loss": 2.0555,
+      "step": 49755
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.435208596713022e-07,
+      "loss": 1.4561,
+      "step": 49756
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.434734513274336e-07,
+      "loss": 3.2313,
+      "step": 49757
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.434260429835651e-07,
+      "loss": 3.4857,
+      "step": 49758
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.433786346396966e-07,
+      "loss": 2.8246,
+      "step": 49759
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.433312262958281e-07,
+      "loss": 2.5743,
+      "step": 49760
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.432838179519596e-07,
+      "loss": 1.9484,
+      "step": 49761
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.432364096080911e-07,
+      "loss": 1.6002,
+      "step": 49762
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.431890012642226e-07,
+      "loss": 1.5235,
+      "step": 49763
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.431415929203541e-07,
+      "loss": 2.318,
+      "step": 49764
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.430941845764854e-07,
+      "loss": 1.8602,
+      "step": 49765
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.430467762326169e-07,
+      "loss": 1.6756,
+      "step": 49766
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.429993678887484e-07,
+      "loss": 1.0253,
+      "step": 49767
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.429519595448799e-07,
+      "loss": 0.5818,
+      "step": 49768
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.429045512010114e-07,
+      "loss": 1.517,
+      "step": 49769
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.428571428571428e-07,
+      "loss": 1.602,
+      "step": 49770
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.428097345132744e-07,
+      "loss": 1.7819,
+      "step": 49771
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.427623261694058e-07,
+      "loss": 2.591,
+      "step": 49772
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.427149178255373e-07,
+      "loss": 2.0245,
+      "step": 49773
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.426675094816689e-07,
+      "loss": 2.8903,
+      "step": 49774
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.426201011378003e-07,
+      "loss": 1.8429,
+      "step": 49775
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.425726927939318e-07,
+      "loss": 0.9988,
+      "step": 49776
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.425252844500632e-07,
+      "loss": 1.5429,
+      "step": 49777
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.424778761061947e-07,
+      "loss": 2.6476,
+      "step": 49778
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.424304677623262e-07,
+      "loss": 2.4187,
+      "step": 49779
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.423830594184576e-07,
+      "loss": 1.5676,
+      "step": 49780
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.423356510745891e-07,
+      "loss": 2.3448,
+      "step": 49781
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.422882427307206e-07,
+      "loss": 1.9089,
+      "step": 49782
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.422408343868521e-07,
+      "loss": 1.9873,
+      "step": 49783
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.421934260429836e-07,
+      "loss": 1.4009,
+      "step": 49784
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.421460176991151e-07,
+      "loss": 1.4747,
+      "step": 49785
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.420986093552466e-07,
+      "loss": 1.981,
+      "step": 49786
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.420512010113781e-07,
+      "loss": 1.0323,
+      "step": 49787
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.420037926675094e-07,
+      "loss": 2.7615,
+      "step": 49788
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.41956384323641e-07,
+      "loss": 2.2416,
+      "step": 49789
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.419089759797724e-07,
+      "loss": 1.3252,
+      "step": 49790
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.418615676359039e-07,
+      "loss": 2.6104,
+      "step": 49791
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.418141592920354e-07,
+      "loss": 1.0152,
+      "step": 49792
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.417667509481669e-07,
+      "loss": 0.7014,
+      "step": 49793
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.417193426042984e-07,
+      "loss": 3.062,
+      "step": 49794
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.416719342604298e-07,
+      "loss": 1.4141,
+      "step": 49795
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.416245259165614e-07,
+      "loss": 2.0318,
+      "step": 49796
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.415771175726929e-07,
+      "loss": 1.1786,
+      "step": 49797
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.415297092288243e-07,
+      "loss": 0.9761,
+      "step": 49798
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.414823008849558e-07,
+      "loss": 3.341,
+      "step": 49799
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.414348925410872e-07,
+      "loss": 2.3748,
+      "step": 49800
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.413874841972187e-07,
+      "loss": 2.4182,
+      "step": 49801
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.413400758533502e-07,
+      "loss": 2.863,
+      "step": 49802
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.412926675094816e-07,
+      "loss": 1.3431,
+      "step": 49803
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.412452591656132e-07,
+      "loss": 2.4464,
+      "step": 49804
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.411978508217446e-07,
+      "loss": 1.0852,
+      "step": 49805
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.411504424778761e-07,
+      "loss": 2.7587,
+      "step": 49806
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.411030341340077e-07,
+      "loss": 1.5354,
+      "step": 49807
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.410556257901391e-07,
+      "loss": 2.8153,
+      "step": 49808
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 6.410082174462706e-07,
+      "loss": 2.7938,
+      "step": 49809
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.409608091024021e-07,
+      "loss": 1.9454,
+      "step": 49810
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.409134007585336e-07,
+      "loss": 3.3422,
+      "step": 49811
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.40865992414665e-07,
+      "loss": 1.9088,
+      "step": 49812
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.408185840707964e-07,
+      "loss": 1.8035,
+      "step": 49813
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.407711757269279e-07,
+      "loss": 2.7216,
+      "step": 49814
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.407237673830594e-07,
+      "loss": 1.5444,
+      "step": 49815
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.406763590391909e-07,
+      "loss": 1.0675,
+      "step": 49816
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.406289506953224e-07,
+      "loss": 2.3,
+      "step": 49817
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.405815423514538e-07,
+      "loss": 2.2243,
+      "step": 49818
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.405341340075854e-07,
+      "loss": 3.4632,
+      "step": 49819
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.404867256637169e-07,
+      "loss": 1.8014,
+      "step": 49820
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.404393173198483e-07,
+      "loss": 1.8303,
+      "step": 49821
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.403919089759799e-07,
+      "loss": 2.1059,
+      "step": 49822
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.403445006321112e-07,
+      "loss": 1.6865,
+      "step": 49823
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.402970922882427e-07,
+      "loss": 2.2131,
+      "step": 49824
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.402496839443742e-07,
+      "loss": 2.4409,
+      "step": 49825
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.402022756005057e-07,
+      "loss": 0.9634,
+      "step": 49826
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.401548672566372e-07,
+      "loss": 0.872,
+      "step": 49827
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.401074589127686e-07,
+      "loss": 2.0008,
+      "step": 49828
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.400600505689001e-07,
+      "loss": 1.4698,
+      "step": 49829
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.400126422250317e-07,
+      "loss": 4.4005,
+      "step": 49830
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.399652338811631e-07,
+      "loss": 2.4314,
+      "step": 49831
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.399178255372946e-07,
+      "loss": 1.3453,
+      "step": 49832
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.398704171934261e-07,
+      "loss": 3.5869,
+      "step": 49833
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.398230088495576e-07,
+      "loss": 3.1424,
+      "step": 49834
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.39775600505689e-07,
+      "loss": 2.4774,
+      "step": 49835
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.397281921618204e-07,
+      "loss": 1.2581,
+      "step": 49836
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.39680783817952e-07,
+      "loss": 1.1527,
+      "step": 49837
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.396333754740834e-07,
+      "loss": 0.6801,
+      "step": 49838
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.395859671302149e-07,
+      "loss": 2.2242,
+      "step": 49839
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.395385587863464e-07,
+      "loss": 1.1405,
+      "step": 49840
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.394911504424779e-07,
+      "loss": 0.9925,
+      "step": 49841
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.394437420986094e-07,
+      "loss": 2.3736,
+      "step": 49842
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.393963337547408e-07,
+      "loss": 2.0515,
+      "step": 49843
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.393489254108724e-07,
+      "loss": 0.8092,
+      "step": 49844
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.393015170670039e-07,
+      "loss": 1.4649,
+      "step": 49845
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.392541087231353e-07,
+      "loss": 2.7564,
+      "step": 49846
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.392067003792667e-07,
+      "loss": 2.3588,
+      "step": 49847
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.391592920353982e-07,
+      "loss": 2.5986,
+      "step": 49848
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.391118836915297e-07,
+      "loss": 1.6281,
+      "step": 49849
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.390644753476612e-07,
+      "loss": 1.081,
+      "step": 49850
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.390170670037926e-07,
+      "loss": 1.3506,
+      "step": 49851
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.389696586599242e-07,
+      "loss": 2.2912,
+      "step": 49852
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.389222503160556e-07,
+      "loss": 1.5677,
+      "step": 49853
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.388748419721871e-07,
+      "loss": 1.2671,
+      "step": 49854
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.388274336283187e-07,
+      "loss": 3.4371,
+      "step": 49855
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.387800252844501e-07,
+      "loss": 1.8229,
+      "step": 49856
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.387326169405816e-07,
+      "loss": 2.6862,
+      "step": 49857
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.38685208596713e-07,
+      "loss": 0.857,
+      "step": 49858
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.386378002528445e-07,
+      "loss": 1.0564,
+      "step": 49859
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.38590391908976e-07,
+      "loss": 2.9914,
+      "step": 49860
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.385429835651074e-07,
+      "loss": 1.9666,
+      "step": 49861
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.384955752212389e-07,
+      "loss": 2.4413,
+      "step": 49862
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.384481668773704e-07,
+      "loss": 1.7884,
+      "step": 49863
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.384007585335019e-07,
+      "loss": 1.2395,
+      "step": 49864
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.383533501896334e-07,
+      "loss": 1.0702,
+      "step": 49865
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.383059418457649e-07,
+      "loss": 2.9899,
+      "step": 49866
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.382585335018964e-07,
+      "loss": 2.4814,
+      "step": 49867
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.382111251580279e-07,
+      "loss": 0.9699,
+      "step": 49868
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.381637168141593e-07,
+      "loss": 1.8112,
+      "step": 49869
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.381163084702909e-07,
+      "loss": 3.6037,
+      "step": 49870
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.380689001264222e-07,
+      "loss": 2.3003,
+      "step": 49871
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 6.380214917825537e-07,
+      "loss": 3.2852,
+      "step": 49872
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.379740834386852e-07,
+      "loss": 1.674,
+      "step": 49873
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.379266750948167e-07,
+      "loss": 1.3613,
+      "step": 49874
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.378792667509482e-07,
+      "loss": 1.9706,
+      "step": 49875
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.378318584070796e-07,
+      "loss": 2.2642,
+      "step": 49876
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.377844500632112e-07,
+      "loss": 0.4818,
+      "step": 49877
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.377370417193427e-07,
+      "loss": 1.795,
+      "step": 49878
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.376896333754741e-07,
+      "loss": 2.0894,
+      "step": 49879
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.376422250316056e-07,
+      "loss": 0.9059,
+      "step": 49880
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.375948166877371e-07,
+      "loss": 2.9294,
+      "step": 49881
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.375474083438685e-07,
+      "loss": 1.6772,
+      "step": 49882
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.375e-07,
+      "loss": 1.4597,
+      "step": 49883
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.374525916561314e-07,
+      "loss": 3.2211,
+      "step": 49884
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.37405183312263e-07,
+      "loss": 1.4643,
+      "step": 49885
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.373577749683944e-07,
+      "loss": 2.5518,
+      "step": 49886
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.373103666245259e-07,
+      "loss": 0.9997,
+      "step": 49887
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.372629582806575e-07,
+      "loss": 0.7595,
+      "step": 49888
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.372155499367889e-07,
+      "loss": 1.9973,
+      "step": 49889
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.371681415929204e-07,
+      "loss": 2.4751,
+      "step": 49890
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.371207332490519e-07,
+      "loss": 1.5085,
+      "step": 49891
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.370733249051834e-07,
+      "loss": 1.1214,
+      "step": 49892
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.370259165613149e-07,
+      "loss": 3.3239,
+      "step": 49893
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.369785082174462e-07,
+      "loss": 2.7558,
+      "step": 49894
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.369310998735777e-07,
+      "loss": 1.9697,
+      "step": 49895
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.368836915297092e-07,
+      "loss": 1.7378,
+      "step": 49896
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.368362831858407e-07,
+      "loss": 2.0063,
+      "step": 49897
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.367888748419722e-07,
+      "loss": 0.3711,
+      "step": 49898
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.367414664981037e-07,
+      "loss": 2.5595,
+      "step": 49899
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.366940581542352e-07,
+      "loss": 1.2839,
+      "step": 49900
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.366466498103667e-07,
+      "loss": 2.588,
+      "step": 49901
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.365992414664981e-07,
+      "loss": 1.6814,
+      "step": 49902
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.365518331226297e-07,
+      "loss": 1.6459,
+      "step": 49903
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.365044247787611e-07,
+      "loss": 1.4894,
+      "step": 49904
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.364570164348925e-07,
+      "loss": 3.0305,
+      "step": 49905
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.36409608091024e-07,
+      "loss": 1.9853,
+      "step": 49906
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.363621997471555e-07,
+      "loss": 2.2065,
+      "step": 49907
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.36314791403287e-07,
+      "loss": 0.5563,
+      "step": 49908
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.362673830594184e-07,
+      "loss": 3.1616,
+      "step": 49909
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.3621997471555e-07,
+      "loss": 0.9834,
+      "step": 49910
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.361725663716814e-07,
+      "loss": 2.0616,
+      "step": 49911
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.361251580278129e-07,
+      "loss": 2.3861,
+      "step": 49912
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.360777496839444e-07,
+      "loss": 1.6996,
+      "step": 49913
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.360303413400759e-07,
+      "loss": 1.7466,
+      "step": 49914
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.359829329962074e-07,
+      "loss": 2.4324,
+      "step": 49915
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.359355246523389e-07,
+      "loss": 2.0773,
+      "step": 49916
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.358881163084702e-07,
+      "loss": 0.9025,
+      "step": 49917
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.358407079646018e-07,
+      "loss": 2.4028,
+      "step": 49918
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.357932996207332e-07,
+      "loss": 1.4288,
+      "step": 49919
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.357458912768647e-07,
+      "loss": 1.1065,
+      "step": 49920
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.356984829329962e-07,
+      "loss": 1.2925,
+      "step": 49921
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.356510745891277e-07,
+      "loss": 2.23,
+      "step": 49922
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.356036662452592e-07,
+      "loss": 3.6481,
+      "step": 49923
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.355562579013906e-07,
+      "loss": 3.9315,
+      "step": 49924
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.355088495575222e-07,
+      "loss": 2.1664,
+      "step": 49925
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.354614412136537e-07,
+      "loss": 1.0486,
+      "step": 49926
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.354140328697851e-07,
+      "loss": 2.7385,
+      "step": 49927
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.353666245259165e-07,
+      "loss": 1.3677,
+      "step": 49928
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.35319216182048e-07,
+      "loss": 2.6995,
+      "step": 49929
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.352718078381795e-07,
+      "loss": 2.1756,
+      "step": 49930
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.35224399494311e-07,
+      "loss": 1.7169,
+      "step": 49931
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.351769911504425e-07,
+      "loss": 2.1674,
+      "step": 49932
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.35129582806574e-07,
+      "loss": 1.8604,
+      "step": 49933
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.350821744627054e-07,
+      "loss": 1.4707,
+      "step": 49934
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 6.350347661188369e-07,
+      "loss": 1.6047,
+      "step": 49935
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.349873577749685e-07,
+      "loss": 1.4436,
+      "step": 49936
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.349399494310999e-07,
+      "loss": 2.283,
+      "step": 49937
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.348925410872314e-07,
+      "loss": 1.9109,
+      "step": 49938
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.348451327433629e-07,
+      "loss": 1.5478,
+      "step": 49939
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.347977243994944e-07,
+      "loss": 2.3286,
+      "step": 49940
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.347503160556258e-07,
+      "loss": 0.9874,
+      "step": 49941
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.347029077117572e-07,
+      "loss": 1.6811,
+      "step": 49942
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.346554993678888e-07,
+      "loss": 1.4921,
+      "step": 49943
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.346080910240202e-07,
+      "loss": 1.877,
+      "step": 49944
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.345606826801517e-07,
+      "loss": 1.8828,
+      "step": 49945
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.345132743362832e-07,
+      "loss": 2.57,
+      "step": 49946
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.344658659924147e-07,
+      "loss": 1.4082,
+      "step": 49947
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.344184576485462e-07,
+      "loss": 2.1394,
+      "step": 49948
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.343710493046777e-07,
+      "loss": 1.4654,
+      "step": 49949
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.343236409608091e-07,
+      "loss": 1.0915,
+      "step": 49950
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.342762326169407e-07,
+      "loss": 1.6562,
+      "step": 49951
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.34228824273072e-07,
+      "loss": 1.736,
+      "step": 49952
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.341814159292035e-07,
+      "loss": 1.2017,
+      "step": 49953
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.34134007585335e-07,
+      "loss": 1.8786,
+      "step": 49954
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.340865992414665e-07,
+      "loss": 1.3459,
+      "step": 49955
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.34039190897598e-07,
+      "loss": 2.6026,
+      "step": 49956
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.339917825537294e-07,
+      "loss": 3.0503,
+      "step": 49957
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.33944374209861e-07,
+      "loss": 0.8341,
+      "step": 49958
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.338969658659925e-07,
+      "loss": 2.9028,
+      "step": 49959
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.338495575221239e-07,
+      "loss": 1.5904,
+      "step": 49960
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.338021491782554e-07,
+      "loss": 2.5771,
+      "step": 49961
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.337547408343869e-07,
+      "loss": 1.4469,
+      "step": 49962
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.337073324905184e-07,
+      "loss": 1.3347,
+      "step": 49963
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.336599241466498e-07,
+      "loss": 1.3952,
+      "step": 49964
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.336125158027813e-07,
+      "loss": 1.5881,
+      "step": 49965
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.335651074589128e-07,
+      "loss": 2.7596,
+      "step": 49966
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.335176991150442e-07,
+      "loss": 3.9161,
+      "step": 49967
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.334702907711757e-07,
+      "loss": 0.9471,
+      "step": 49968
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.334228824273073e-07,
+      "loss": 1.0215,
+      "step": 49969
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.333754740834387e-07,
+      "loss": 2.3315,
+      "step": 49970
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.333280657395702e-07,
+      "loss": 1.9449,
+      "step": 49971
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.332806573957016e-07,
+      "loss": 2.3009,
+      "step": 49972
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.332332490518332e-07,
+      "loss": 0.6549,
+      "step": 49973
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.331858407079647e-07,
+      "loss": 1.2188,
+      "step": 49974
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.33138432364096e-07,
+      "loss": 1.3648,
+      "step": 49975
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.330910240202276e-07,
+      "loss": 2.1309,
+      "step": 49976
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.33043615676359e-07,
+      "loss": 2.1528,
+      "step": 49977
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.329962073324905e-07,
+      "loss": 2.3138,
+      "step": 49978
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.32948798988622e-07,
+      "loss": 1.8892,
+      "step": 49979
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.329013906447535e-07,
+      "loss": 2.6919,
+      "step": 49980
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.32853982300885e-07,
+      "loss": 2.6565,
+      "step": 49981
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.328065739570164e-07,
+      "loss": 0.4119,
+      "step": 49982
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.327591656131479e-07,
+      "loss": 2.4927,
+      "step": 49983
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.327117572692795e-07,
+      "loss": 1.1055,
+      "step": 49984
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.326643489254109e-07,
+      "loss": 2.3539,
+      "step": 49985
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.326169405815424e-07,
+      "loss": 1.6145,
+      "step": 49986
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.325695322376738e-07,
+      "loss": 1.764,
+      "step": 49987
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.325221238938053e-07,
+      "loss": 0.6197,
+      "step": 49988
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.324747155499368e-07,
+      "loss": 1.4411,
+      "step": 49989
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.324273072060682e-07,
+      "loss": 2.3299,
+      "step": 49990
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.323798988621998e-07,
+      "loss": 3.7794,
+      "step": 49991
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.323324905183312e-07,
+      "loss": 1.4781,
+      "step": 49992
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.322850821744627e-07,
+      "loss": 2.231,
+      "step": 49993
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.322376738305942e-07,
+      "loss": 1.6028,
+      "step": 49994
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.321902654867257e-07,
+      "loss": 1.7364,
+      "step": 49995
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.321428571428572e-07,
+      "loss": 1.6964,
+      "step": 49996
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.320954487989887e-07,
+      "loss": 1.776,
+      "step": 49997
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.320480404551202e-07,
+      "loss": 2.1887,
+      "step": 49998
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 6.320006321112516e-07,
+      "loss": 1.3065,
+      "step": 49999
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.31953223767383e-07,
+      "loss": 1.177,
+      "step": 50000
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.319058154235145e-07,
+      "loss": 1.4024,
+      "step": 50001
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.31858407079646e-07,
+      "loss": 0.5664,
+      "step": 50002
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.318109987357775e-07,
+      "loss": 1.2503,
+      "step": 50003
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.31763590391909e-07,
+      "loss": 1.9169,
+      "step": 50004
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.317161820480404e-07,
+      "loss": 2.4281,
+      "step": 50005
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.31668773704172e-07,
+      "loss": 1.5742,
+      "step": 50006
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.316213653603035e-07,
+      "loss": 2.6901,
+      "step": 50007
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.315739570164349e-07,
+      "loss": 1.8274,
+      "step": 50008
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.315265486725665e-07,
+      "loss": 2.0203,
+      "step": 50009
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.314791403286978e-07,
+      "loss": 1.4927,
+      "step": 50010
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.314317319848293e-07,
+      "loss": 2.8862,
+      "step": 50011
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.313843236409608e-07,
+      "loss": 3.2662,
+      "step": 50012
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.313369152970923e-07,
+      "loss": 1.4576,
+      "step": 50013
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.312895069532238e-07,
+      "loss": 2.1906,
+      "step": 50014
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.312420986093552e-07,
+      "loss": 3.8009,
+      "step": 50015
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.311946902654867e-07,
+      "loss": 0.7299,
+      "step": 50016
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.311472819216183e-07,
+      "loss": 1.9611,
+      "step": 50017
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.310998735777497e-07,
+      "loss": 1.5187,
+      "step": 50018
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.310524652338812e-07,
+      "loss": 1.4515,
+      "step": 50019
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.310050568900127e-07,
+      "loss": 1.0605,
+      "step": 50020
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.309576485461442e-07,
+      "loss": 1.6272,
+      "step": 50021
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.309102402022756e-07,
+      "loss": 1.8046,
+      "step": 50022
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.30862831858407e-07,
+      "loss": 1.2233,
+      "step": 50023
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.308154235145386e-07,
+      "loss": 2.8129,
+      "step": 50024
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.3076801517067e-07,
+      "loss": 2.5152,
+      "step": 50025
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.307206068268015e-07,
+      "loss": 2.8099,
+      "step": 50026
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.30673198482933e-07,
+      "loss": 2.0528,
+      "step": 50027
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.306257901390645e-07,
+      "loss": 3.0254,
+      "step": 50028
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.30578381795196e-07,
+      "loss": 1.1629,
+      "step": 50029
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.305309734513275e-07,
+      "loss": 3.535,
+      "step": 50030
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.30483565107459e-07,
+      "loss": 1.387,
+      "step": 50031
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.304361567635905e-07,
+      "loss": 0.7442,
+      "step": 50032
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.303887484197219e-07,
+      "loss": 2.6404,
+      "step": 50033
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.303413400758533e-07,
+      "loss": 1.5962,
+      "step": 50034
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.302939317319848e-07,
+      "loss": 2.5175,
+      "step": 50035
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.302465233881163e-07,
+      "loss": 1.1855,
+      "step": 50036
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.301991150442478e-07,
+      "loss": 1.3359,
+      "step": 50037
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.301517067003792e-07,
+      "loss": 1.4204,
+      "step": 50038
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.301042983565108e-07,
+      "loss": 2.3981,
+      "step": 50039
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.300568900126423e-07,
+      "loss": 2.0464,
+      "step": 50040
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.300094816687737e-07,
+      "loss": 3.6521,
+      "step": 50041
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.299620733249052e-07,
+      "loss": 1.4655,
+      "step": 50042
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.299146649810367e-07,
+      "loss": 2.7747,
+      "step": 50043
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.298672566371682e-07,
+      "loss": 3.7218,
+      "step": 50044
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.298198482932997e-07,
+      "loss": 0.7193,
+      "step": 50045
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.297724399494311e-07,
+      "loss": 1.4611,
+      "step": 50046
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.297250316055626e-07,
+      "loss": 2.772,
+      "step": 50047
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.29677623261694e-07,
+      "loss": 2.9398,
+      "step": 50048
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.296302149178255e-07,
+      "loss": 2.3315,
+      "step": 50049
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.29582806573957e-07,
+      "loss": 1.8779,
+      "step": 50050
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.295353982300885e-07,
+      "loss": 1.7041,
+      "step": 50051
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.2948798988622e-07,
+      "loss": 0.9188,
+      "step": 50052
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.294405815423514e-07,
+      "loss": 2.8613,
+      "step": 50053
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.29393173198483e-07,
+      "loss": 0.8542,
+      "step": 50054
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.293457648546145e-07,
+      "loss": 1.77,
+      "step": 50055
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.292983565107459e-07,
+      "loss": 0.7461,
+      "step": 50056
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.292509481668775e-07,
+      "loss": 1.808,
+      "step": 50057
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.292035398230088e-07,
+      "loss": 1.2454,
+      "step": 50058
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.291561314791403e-07,
+      "loss": 1.4221,
+      "step": 50059
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.291087231352718e-07,
+      "loss": 3.5043,
+      "step": 50060
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.290613147914033e-07,
+      "loss": 2.1481,
+      "step": 50061
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 6.290139064475348e-07,
+      "loss": 1.7647,
+      "step": 50062
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.289664981036662e-07,
+      "loss": 1.8577,
+      "step": 50063
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.289190897597977e-07,
+      "loss": 2.8603,
+      "step": 50064
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.288716814159293e-07,
+      "loss": 2.0214,
+      "step": 50065
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.288242730720607e-07,
+      "loss": 1.5049,
+      "step": 50066
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.287768647281922e-07,
+      "loss": 4.3026,
+      "step": 50067
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.287294563843237e-07,
+      "loss": 1.0468,
+      "step": 50068
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.286820480404551e-07,
+      "loss": 1.0478,
+      "step": 50069
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.286346396965866e-07,
+      "loss": 2.6921,
+      "step": 50070
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.28587231352718e-07,
+      "loss": 1.6894,
+      "step": 50071
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.285398230088496e-07,
+      "loss": 1.5447,
+      "step": 50072
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.28492414664981e-07,
+      "loss": 1.4061,
+      "step": 50073
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.284450063211125e-07,
+      "loss": 1.8526,
+      "step": 50074
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.28397597977244e-07,
+      "loss": 1.5298,
+      "step": 50075
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.283501896333755e-07,
+      "loss": 1.136,
+      "step": 50076
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.28302781289507e-07,
+      "loss": 0.6785,
+      "step": 50077
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.282553729456385e-07,
+      "loss": 1.9951,
+      "step": 50078
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.2820796460177e-07,
+      "loss": 1.6923,
+      "step": 50079
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.281605562579015e-07,
+      "loss": 1.6537,
+      "step": 50080
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.281131479140328e-07,
+      "loss": 1.6847,
+      "step": 50081
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.280657395701643e-07,
+      "loss": 1.6659,
+      "step": 50082
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.280183312262958e-07,
+      "loss": 2.2171,
+      "step": 50083
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.279709228824273e-07,
+      "loss": 1.762,
+      "step": 50084
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.279235145385588e-07,
+      "loss": 1.0059,
+      "step": 50085
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.278761061946902e-07,
+      "loss": 0.9444,
+      "step": 50086
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.278286978508218e-07,
+      "loss": 1.7572,
+      "step": 50087
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.277812895069533e-07,
+      "loss": 1.9588,
+      "step": 50088
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.277338811630847e-07,
+      "loss": 1.1685,
+      "step": 50089
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.276864728192163e-07,
+      "loss": 1.7115,
+      "step": 50090
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.276390644753477e-07,
+      "loss": 1.8735,
+      "step": 50091
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.275916561314791e-07,
+      "loss": 2.5996,
+      "step": 50092
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.275442477876106e-07,
+      "loss": 3.4438,
+      "step": 50093
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.274968394437421e-07,
+      "loss": 1.579,
+      "step": 50094
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.274494310998736e-07,
+      "loss": 1.2008,
+      "step": 50095
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.27402022756005e-07,
+      "loss": 0.913,
+      "step": 50096
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.273546144121365e-07,
+      "loss": 0.9458,
+      "step": 50097
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.273072060682681e-07,
+      "loss": 2.1068,
+      "step": 50098
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.272597977243995e-07,
+      "loss": 0.8358,
+      "step": 50099
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.27212389380531e-07,
+      "loss": 2.2411,
+      "step": 50100
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.271649810366625e-07,
+      "loss": 1.6391,
+      "step": 50101
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.27117572692794e-07,
+      "loss": 1.5064,
+      "step": 50102
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.270701643489255e-07,
+      "loss": 1.6275,
+      "step": 50103
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.270227560050568e-07,
+      "loss": 3.9549,
+      "step": 50104
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.269753476611884e-07,
+      "loss": 1.5127,
+      "step": 50105
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.269279393173198e-07,
+      "loss": 1.6438,
+      "step": 50106
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.268805309734513e-07,
+      "loss": 1.4042,
+      "step": 50107
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.268331226295828e-07,
+      "loss": 1.2347,
+      "step": 50108
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.267857142857143e-07,
+      "loss": 2.4263,
+      "step": 50109
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.267383059418458e-07,
+      "loss": 2.0698,
+      "step": 50110
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.266908975979772e-07,
+      "loss": 2.1994,
+      "step": 50111
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.266434892541088e-07,
+      "loss": 3.1705,
+      "step": 50112
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.265960809102403e-07,
+      "loss": 1.9139,
+      "step": 50113
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.265486725663717e-07,
+      "loss": 2.2974,
+      "step": 50114
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.265012642225032e-07,
+      "loss": 2.3187,
+      "step": 50115
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.264538558786346e-07,
+      "loss": 2.9474,
+      "step": 50116
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.264064475347661e-07,
+      "loss": 2.5484,
+      "step": 50117
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.263590391908976e-07,
+      "loss": 0.6473,
+      "step": 50118
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.26311630847029e-07,
+      "loss": 2.4969,
+      "step": 50119
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.262642225031606e-07,
+      "loss": 2.9383,
+      "step": 50120
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.26216814159292e-07,
+      "loss": 2.3788,
+      "step": 50121
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.261694058154235e-07,
+      "loss": 2.0708,
+      "step": 50122
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.261219974715551e-07,
+      "loss": 2.605,
+      "step": 50123
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.260745891276865e-07,
+      "loss": 1.5344,
+      "step": 50124
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 6.26027180783818e-07,
+      "loss": 1.6838,
+      "step": 50125
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.259797724399495e-07,
+      "loss": 1.6506,
+      "step": 50126
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.25932364096081e-07,
+      "loss": 1.1596,
+      "step": 50127
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.258849557522124e-07,
+      "loss": 1.4409,
+      "step": 50128
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.258375474083438e-07,
+      "loss": 0.7387,
+      "step": 50129
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.257901390644753e-07,
+      "loss": 2.4633,
+      "step": 50130
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.257427307206068e-07,
+      "loss": 1.9518,
+      "step": 50131
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.256953223767383e-07,
+      "loss": 2.7214,
+      "step": 50132
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.256479140328698e-07,
+      "loss": 4.7907,
+      "step": 50133
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.256005056890013e-07,
+      "loss": 2.0542,
+      "step": 50134
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.255530973451328e-07,
+      "loss": 1.7213,
+      "step": 50135
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.255056890012643e-07,
+      "loss": 4.0615,
+      "step": 50136
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.254582806573957e-07,
+      "loss": 2.9997,
+      "step": 50137
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.254108723135273e-07,
+      "loss": 2.8317,
+      "step": 50138
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.253634639696586e-07,
+      "loss": 0.9283,
+      "step": 50139
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.253160556257901e-07,
+      "loss": 1.7549,
+      "step": 50140
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.252686472819216e-07,
+      "loss": 1.7961,
+      "step": 50141
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.252212389380531e-07,
+      "loss": 1.525,
+      "step": 50142
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.251738305941846e-07,
+      "loss": 1.7603,
+      "step": 50143
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.25126422250316e-07,
+      "loss": 2.4991,
+      "step": 50144
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.250790139064476e-07,
+      "loss": 2.6972,
+      "step": 50145
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.250316055625791e-07,
+      "loss": 1.6078,
+      "step": 50146
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.249841972187105e-07,
+      "loss": 2.1819,
+      "step": 50147
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.24936788874842e-07,
+      "loss": 1.9465,
+      "step": 50148
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.248893805309735e-07,
+      "loss": 1.2454,
+      "step": 50149
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.24841972187105e-07,
+      "loss": 1.9007,
+      "step": 50150
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.247945638432364e-07,
+      "loss": 0.8797,
+      "step": 50151
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.247471554993678e-07,
+      "loss": 1.2632,
+      "step": 50152
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.246997471554994e-07,
+      "loss": 1.9561,
+      "step": 50153
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.246523388116308e-07,
+      "loss": 2.9444,
+      "step": 50154
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.246049304677623e-07,
+      "loss": 1.4581,
+      "step": 50155
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.245575221238939e-07,
+      "loss": 1.321,
+      "step": 50156
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.245101137800253e-07,
+      "loss": 0.7549,
+      "step": 50157
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.244627054361568e-07,
+      "loss": 1.9548,
+      "step": 50158
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.244152970922883e-07,
+      "loss": 2.463,
+      "step": 50159
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.243678887484198e-07,
+      "loss": 3.0688,
+      "step": 50160
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.243204804045513e-07,
+      "loss": 3.3357,
+      "step": 50161
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.242730720606826e-07,
+      "loss": 1.8777,
+      "step": 50162
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.242256637168141e-07,
+      "loss": 2.8871,
+      "step": 50163
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.241782553729456e-07,
+      "loss": 1.9072,
+      "step": 50164
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.241308470290771e-07,
+      "loss": 1.5777,
+      "step": 50165
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.240834386852086e-07,
+      "loss": 1.5567,
+      "step": 50166
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.240360303413401e-07,
+      "loss": 2.5136,
+      "step": 50167
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.239886219974716e-07,
+      "loss": 3.1331,
+      "step": 50168
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.239412136536031e-07,
+      "loss": 1.379,
+      "step": 50169
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.238938053097345e-07,
+      "loss": 1.2638,
+      "step": 50170
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.238463969658661e-07,
+      "loss": 1.8727,
+      "step": 50171
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.237989886219975e-07,
+      "loss": 2.6827,
+      "step": 50172
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.23751580278129e-07,
+      "loss": 1.5861,
+      "step": 50173
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.237041719342604e-07,
+      "loss": 1.7914,
+      "step": 50174
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.236567635903919e-07,
+      "loss": 2.791,
+      "step": 50175
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.236093552465234e-07,
+      "loss": 3.1747,
+      "step": 50176
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.235619469026548e-07,
+      "loss": 0.5209,
+      "step": 50177
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.235145385587864e-07,
+      "loss": 2.2935,
+      "step": 50178
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.234671302149179e-07,
+      "loss": 1.4158,
+      "step": 50179
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.234197218710493e-07,
+      "loss": 3.0088,
+      "step": 50180
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.233723135271808e-07,
+      "loss": 2.0782,
+      "step": 50181
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.233249051833123e-07,
+      "loss": 2.0669,
+      "step": 50182
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.232774968394438e-07,
+      "loss": 2.5371,
+      "step": 50183
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.232300884955753e-07,
+      "loss": 2.4991,
+      "step": 50184
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.231826801517067e-07,
+      "loss": 1.8665,
+      "step": 50185
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.231352718078382e-07,
+      "loss": 1.8949,
+      "step": 50186
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.230878634639696e-07,
+      "loss": 0.8171,
+      "step": 50187
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.230404551201011e-07,
+      "loss": 1.8202,
+      "step": 50188
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 6.229930467762326e-07,
+      "loss": 2.8351,
+      "step": 50189
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.229456384323641e-07,
+      "loss": 1.3736,
+      "step": 50190
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.228982300884956e-07,
+      "loss": 0.6812,
+      "step": 50191
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.22850821744627e-07,
+      "loss": 1.4835,
+      "step": 50192
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.228034134007586e-07,
+      "loss": 1.4861,
+      "step": 50193
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.227560050568901e-07,
+      "loss": 1.0645,
+      "step": 50194
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.227085967130215e-07,
+      "loss": 2.7695,
+      "step": 50195
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.22661188369153e-07,
+      "loss": 2.5311,
+      "step": 50196
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.226137800252845e-07,
+      "loss": 2.3849,
+      "step": 50197
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.225663716814159e-07,
+      "loss": 3.1884,
+      "step": 50198
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.225189633375474e-07,
+      "loss": 1.9729,
+      "step": 50199
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.224715549936789e-07,
+      "loss": 0.671,
+      "step": 50200
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.224241466498104e-07,
+      "loss": 2.039,
+      "step": 50201
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.223767383059418e-07,
+      "loss": 1.0201,
+      "step": 50202
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.223293299620733e-07,
+      "loss": 2.5889,
+      "step": 50203
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.222819216182049e-07,
+      "loss": 3.7262,
+      "step": 50204
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.222345132743363e-07,
+      "loss": 3.1766,
+      "step": 50205
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.221871049304678e-07,
+      "loss": 2.1518,
+      "step": 50206
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.221396965865993e-07,
+      "loss": 2.3834,
+      "step": 50207
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.220922882427308e-07,
+      "loss": 2.5565,
+      "step": 50208
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.220448798988623e-07,
+      "loss": 1.2684,
+      "step": 50209
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.219974715549936e-07,
+      "loss": 1.0126,
+      "step": 50210
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.219500632111252e-07,
+      "loss": 1.7366,
+      "step": 50211
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.219026548672566e-07,
+      "loss": 0.9128,
+      "step": 50212
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.218552465233881e-07,
+      "loss": 0.9475,
+      "step": 50213
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.218078381795196e-07,
+      "loss": 1.3495,
+      "step": 50214
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.217604298356511e-07,
+      "loss": 2.8355,
+      "step": 50215
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.217130214917826e-07,
+      "loss": 1.9988,
+      "step": 50216
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.216656131479141e-07,
+      "loss": 0.4441,
+      "step": 50217
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.216182048040455e-07,
+      "loss": 1.8716,
+      "step": 50218
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.215707964601771e-07,
+      "loss": 0.6381,
+      "step": 50219
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.215233881163085e-07,
+      "loss": 2.1965,
+      "step": 50220
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.214759797724399e-07,
+      "loss": 1.2184,
+      "step": 50221
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.214285714285714e-07,
+      "loss": 1.0838,
+      "step": 50222
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.213811630847029e-07,
+      "loss": 1.3064,
+      "step": 50223
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.213337547408344e-07,
+      "loss": 1.5423,
+      "step": 50224
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.212863463969658e-07,
+      "loss": 2.9162,
+      "step": 50225
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.212389380530974e-07,
+      "loss": 1.3471,
+      "step": 50226
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.211915297092289e-07,
+      "loss": 1.3498,
+      "step": 50227
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.211441213653603e-07,
+      "loss": 2.6831,
+      "step": 50228
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.210967130214918e-07,
+      "loss": 1.8377,
+      "step": 50229
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.210493046776233e-07,
+      "loss": 1.6104,
+      "step": 50230
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.210018963337548e-07,
+      "loss": 1.245,
+      "step": 50231
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.209544879898863e-07,
+      "loss": 2.3212,
+      "step": 50232
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.209070796460177e-07,
+      "loss": 0.8417,
+      "step": 50233
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.208596713021492e-07,
+      "loss": 1.1997,
+      "step": 50234
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.208122629582806e-07,
+      "loss": 1.8877,
+      "step": 50235
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.207648546144121e-07,
+      "loss": 2.4201,
+      "step": 50236
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.207174462705437e-07,
+      "loss": 1.3291,
+      "step": 50237
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.206700379266751e-07,
+      "loss": 2.2794,
+      "step": 50238
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.206226295828066e-07,
+      "loss": 1.3277,
+      "step": 50239
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.205752212389381e-07,
+      "loss": 1.6041,
+      "step": 50240
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.205278128950696e-07,
+      "loss": 1.4475,
+      "step": 50241
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.204804045512011e-07,
+      "loss": 2.062,
+      "step": 50242
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.204329962073325e-07,
+      "loss": 1.7243,
+      "step": 50243
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.203855878634641e-07,
+      "loss": 2.0601,
+      "step": 50244
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.203381795195954e-07,
+      "loss": 0.6774,
+      "step": 50245
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.202907711757269e-07,
+      "loss": 1.8096,
+      "step": 50246
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.202433628318584e-07,
+      "loss": 1.807,
+      "step": 50247
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.201959544879899e-07,
+      "loss": 3.1611,
+      "step": 50248
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.201485461441214e-07,
+      "loss": 0.9526,
+      "step": 50249
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.201011378002528e-07,
+      "loss": 2.3387,
+      "step": 50250
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.200537294563843e-07,
+      "loss": 1.7314,
+      "step": 50251
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 6.200063211125159e-07,
+      "loss": 1.7369,
+      "step": 50252
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.199589127686473e-07,
+      "loss": 1.7741,
+      "step": 50253
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.199115044247788e-07,
+      "loss": 1.9755,
+      "step": 50254
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.198640960809103e-07,
+      "loss": 1.8393,
+      "step": 50255
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.198166877370417e-07,
+      "loss": 1.9197,
+      "step": 50256
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.197692793931732e-07,
+      "loss": 0.3375,
+      "step": 50257
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.197218710493046e-07,
+      "loss": 1.304,
+      "step": 50258
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.196744627054362e-07,
+      "loss": 2.4178,
+      "step": 50259
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.196270543615676e-07,
+      "loss": 0.8538,
+      "step": 50260
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.195796460176991e-07,
+      "loss": 1.7233,
+      "step": 50261
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.195322376738306e-07,
+      "loss": 1.5898,
+      "step": 50262
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.194848293299621e-07,
+      "loss": 0.8959,
+      "step": 50263
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.194374209860936e-07,
+      "loss": 3.1203,
+      "step": 50264
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.193900126422251e-07,
+      "loss": 1.5547,
+      "step": 50265
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.193426042983565e-07,
+      "loss": 1.3672,
+      "step": 50266
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.192951959544881e-07,
+      "loss": 2.5168,
+      "step": 50267
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.192477876106194e-07,
+      "loss": 3.0566,
+      "step": 50268
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.192003792667509e-07,
+      "loss": 1.4045,
+      "step": 50269
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.191529709228824e-07,
+      "loss": 3.3961,
+      "step": 50270
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.191055625790139e-07,
+      "loss": 3.085,
+      "step": 50271
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.190581542351454e-07,
+      "loss": 2.5032,
+      "step": 50272
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.190107458912768e-07,
+      "loss": 4.1014,
+      "step": 50273
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.189633375474084e-07,
+      "loss": 0.6715,
+      "step": 50274
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.189159292035399e-07,
+      "loss": 1.8956,
+      "step": 50275
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.188685208596713e-07,
+      "loss": 1.9732,
+      "step": 50276
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.188211125158028e-07,
+      "loss": 2.034,
+      "step": 50277
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.187737041719343e-07,
+      "loss": 2.9816,
+      "step": 50278
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.187262958280658e-07,
+      "loss": 2.804,
+      "step": 50279
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.186788874841972e-07,
+      "loss": 1.0941,
+      "step": 50280
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.186314791403287e-07,
+      "loss": 2.5179,
+      "step": 50281
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.185840707964602e-07,
+      "loss": 0.3561,
+      "step": 50282
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.185366624525916e-07,
+      "loss": 1.7762,
+      "step": 50283
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.184892541087231e-07,
+      "loss": 2.7332,
+      "step": 50284
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.184418457648547e-07,
+      "loss": 1.7856,
+      "step": 50285
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.183944374209861e-07,
+      "loss": 2.3397,
+      "step": 50286
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.183470290771176e-07,
+      "loss": 2.3971,
+      "step": 50287
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.182996207332491e-07,
+      "loss": 2.2584,
+      "step": 50288
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.182522123893806e-07,
+      "loss": 1.7247,
+      "step": 50289
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.182048040455121e-07,
+      "loss": 0.9249,
+      "step": 50290
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.181573957016434e-07,
+      "loss": 2.6554,
+      "step": 50291
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.18109987357775e-07,
+      "loss": 2.0849,
+      "step": 50292
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.180625790139064e-07,
+      "loss": 1.0316,
+      "step": 50293
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.180151706700379e-07,
+      "loss": 1.2343,
+      "step": 50294
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.179677623261694e-07,
+      "loss": 1.1133,
+      "step": 50295
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.179203539823009e-07,
+      "loss": 0.8896,
+      "step": 50296
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.178729456384324e-07,
+      "loss": 1.5495,
+      "step": 50297
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.178255372945639e-07,
+      "loss": 1.6819,
+      "step": 50298
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.177781289506953e-07,
+      "loss": 1.809,
+      "step": 50299
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.177307206068269e-07,
+      "loss": 1.358,
+      "step": 50300
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.176833122629583e-07,
+      "loss": 2.9477,
+      "step": 50301
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.176359039190898e-07,
+      "loss": 1.519,
+      "step": 50302
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.175884955752212e-07,
+      "loss": 1.8701,
+      "step": 50303
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.175410872313527e-07,
+      "loss": 1.8231,
+      "step": 50304
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.174936788874842e-07,
+      "loss": 1.5137,
+      "step": 50305
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.174462705436156e-07,
+      "loss": 1.2727,
+      "step": 50306
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.173988621997472e-07,
+      "loss": 1.3813,
+      "step": 50307
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.173514538558787e-07,
+      "loss": 1.3381,
+      "step": 50308
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.173040455120101e-07,
+      "loss": 3.0555,
+      "step": 50309
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.172566371681416e-07,
+      "loss": 2.1032,
+      "step": 50310
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.172092288242731e-07,
+      "loss": 2.1402,
+      "step": 50311
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.171618204804046e-07,
+      "loss": 1.5388,
+      "step": 50312
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.171144121365361e-07,
+      "loss": 2.4403,
+      "step": 50313
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.170670037926676e-07,
+      "loss": 1.463,
+      "step": 50314
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 6.17019595448799e-07,
+      "loss": 1.2079,
+      "step": 50315
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.169721871049304e-07,
+      "loss": 2.4176,
+      "step": 50316
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.169247787610619e-07,
+      "loss": 1.9561,
+      "step": 50317
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.168773704171935e-07,
+      "loss": 2.6072,
+      "step": 50318
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.168299620733249e-07,
+      "loss": 3.153,
+      "step": 50319
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.167825537294564e-07,
+      "loss": 2.3758,
+      "step": 50320
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.167351453855878e-07,
+      "loss": 1.2593,
+      "step": 50321
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.166877370417194e-07,
+      "loss": 2.0293,
+      "step": 50322
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.166403286978509e-07,
+      "loss": 1.4541,
+      "step": 50323
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.165929203539823e-07,
+      "loss": 4.0894,
+      "step": 50324
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.165455120101139e-07,
+      "loss": 1.6965,
+      "step": 50325
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.164981036662452e-07,
+      "loss": 1.4876,
+      "step": 50326
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.164506953223767e-07,
+      "loss": 1.5211,
+      "step": 50327
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.164032869785082e-07,
+      "loss": 2.4994,
+      "step": 50328
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.163558786346397e-07,
+      "loss": 1.8608,
+      "step": 50329
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.163084702907712e-07,
+      "loss": 1.0194,
+      "step": 50330
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.162610619469026e-07,
+      "loss": 2.6917,
+      "step": 50331
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.162136536030341e-07,
+      "loss": 1.1119,
+      "step": 50332
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.161662452591657e-07,
+      "loss": 1.9016,
+      "step": 50333
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.161188369152971e-07,
+      "loss": 1.5144,
+      "step": 50334
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.160714285714286e-07,
+      "loss": 1.6292,
+      "step": 50335
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.160240202275601e-07,
+      "loss": 1.1742,
+      "step": 50336
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.159766118836916e-07,
+      "loss": 2.7353,
+      "step": 50337
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.15929203539823e-07,
+      "loss": 1.4474,
+      "step": 50338
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.158817951959544e-07,
+      "loss": 1.9423,
+      "step": 50339
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.15834386852086e-07,
+      "loss": 2.8005,
+      "step": 50340
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.157869785082174e-07,
+      "loss": 0.9896,
+      "step": 50341
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.157395701643489e-07,
+      "loss": 0.9837,
+      "step": 50342
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.156921618204804e-07,
+      "loss": 1.2134,
+      "step": 50343
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.156447534766119e-07,
+      "loss": 1.9275,
+      "step": 50344
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.155973451327434e-07,
+      "loss": 2.5098,
+      "step": 50345
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.155499367888749e-07,
+      "loss": 2.1607,
+      "step": 50346
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.155025284450064e-07,
+      "loss": 1.184,
+      "step": 50347
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.154551201011379e-07,
+      "loss": 0.3142,
+      "step": 50348
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.154077117572693e-07,
+      "loss": 1.4332,
+      "step": 50349
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.153603034134007e-07,
+      "loss": 2.6315,
+      "step": 50350
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.153128950695322e-07,
+      "loss": 1.1472,
+      "step": 50351
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.152654867256637e-07,
+      "loss": 1.9481,
+      "step": 50352
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.152180783817952e-07,
+      "loss": 1.1331,
+      "step": 50353
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.151706700379266e-07,
+      "loss": 2.1949,
+      "step": 50354
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.151232616940582e-07,
+      "loss": 0.9762,
+      "step": 50355
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.150758533501897e-07,
+      "loss": 1.9833,
+      "step": 50356
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.150284450063211e-07,
+      "loss": 1.9408,
+      "step": 50357
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.149810366624527e-07,
+      "loss": 1.6939,
+      "step": 50358
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.149336283185841e-07,
+      "loss": 4.5267,
+      "step": 50359
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.148862199747156e-07,
+      "loss": 1.8567,
+      "step": 50360
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.14838811630847e-07,
+      "loss": 2.3702,
+      "step": 50361
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.147914032869785e-07,
+      "loss": 1.5146,
+      "step": 50362
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.1474399494311e-07,
+      "loss": 2.8029,
+      "step": 50363
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.146965865992414e-07,
+      "loss": 1.9846,
+      "step": 50364
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.146491782553729e-07,
+      "loss": 0.7208,
+      "step": 50365
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.146017699115045e-07,
+      "loss": 2.305,
+      "step": 50366
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.145543615676359e-07,
+      "loss": 2.2927,
+      "step": 50367
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.145069532237674e-07,
+      "loss": 1.9903,
+      "step": 50368
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.144595448798989e-07,
+      "loss": 1.9728,
+      "step": 50369
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.144121365360304e-07,
+      "loss": 1.9072,
+      "step": 50370
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.143647281921619e-07,
+      "loss": 1.6902,
+      "step": 50371
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.143173198482933e-07,
+      "loss": 2.3957,
+      "step": 50372
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.142699115044249e-07,
+      "loss": 1.5479,
+      "step": 50373
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.142225031605562e-07,
+      "loss": 0.1159,
+      "step": 50374
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.141750948166877e-07,
+      "loss": 1.4894,
+      "step": 50375
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.141276864728192e-07,
+      "loss": 1.3138,
+      "step": 50376
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.140802781289507e-07,
+      "loss": 1.7602,
+      "step": 50377
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.140328697850822e-07,
+      "loss": 1.4499,
+      "step": 50378
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 6.139854614412137e-07,
+      "loss": 2.1616,
+      "step": 50379
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.139380530973452e-07,
+      "loss": 1.268,
+      "step": 50380
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.138906447534767e-07,
+      "loss": 2.5984,
+      "step": 50381
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.138432364096081e-07,
+      "loss": 1.9216,
+      "step": 50382
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.137958280657396e-07,
+      "loss": 0.9059,
+      "step": 50383
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.137484197218711e-07,
+      "loss": 1.7039,
+      "step": 50384
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.137010113780025e-07,
+      "loss": 1.4011,
+      "step": 50385
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.13653603034134e-07,
+      "loss": 1.456,
+      "step": 50386
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.136061946902654e-07,
+      "loss": 1.6028,
+      "step": 50387
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.13558786346397e-07,
+      "loss": 0.8287,
+      "step": 50388
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.135113780025285e-07,
+      "loss": 1.4866,
+      "step": 50389
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.134639696586599e-07,
+      "loss": 2.4482,
+      "step": 50390
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.134165613147915e-07,
+      "loss": 2.2125,
+      "step": 50391
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.133691529709229e-07,
+      "loss": 2.0113,
+      "step": 50392
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.133217446270544e-07,
+      "loss": 1.3955,
+      "step": 50393
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.132743362831859e-07,
+      "loss": 2.2396,
+      "step": 50394
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.132269279393174e-07,
+      "loss": 2.5884,
+      "step": 50395
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.131795195954489e-07,
+      "loss": 0.6647,
+      "step": 50396
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.131321112515802e-07,
+      "loss": 2.2034,
+      "step": 50397
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.130847029077117e-07,
+      "loss": 2.3072,
+      "step": 50398
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.130372945638432e-07,
+      "loss": 2.4056,
+      "step": 50399
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.129898862199747e-07,
+      "loss": 1.5877,
+      "step": 50400
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.129424778761062e-07,
+      "loss": 2.5732,
+      "step": 50401
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.128950695322377e-07,
+      "loss": 2.0603,
+      "step": 50402
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.128476611883692e-07,
+      "loss": 0.812,
+      "step": 50403
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.128002528445007e-07,
+      "loss": 2.4002,
+      "step": 50404
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.127528445006321e-07,
+      "loss": 2.8602,
+      "step": 50405
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.127054361567637e-07,
+      "loss": 2.6501,
+      "step": 50406
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.126580278128951e-07,
+      "loss": 2.3923,
+      "step": 50407
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.126106194690265e-07,
+      "loss": 3.1029,
+      "step": 50408
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.12563211125158e-07,
+      "loss": 1.9,
+      "step": 50409
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.125158027812895e-07,
+      "loss": 0.7658,
+      "step": 50410
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.12468394437421e-07,
+      "loss": 0.9794,
+      "step": 50411
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.124209860935524e-07,
+      "loss": 1.6153,
+      "step": 50412
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.12373577749684e-07,
+      "loss": 1.8112,
+      "step": 50413
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.123261694058155e-07,
+      "loss": 2.7069,
+      "step": 50414
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.122787610619469e-07,
+      "loss": 2.7455,
+      "step": 50415
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.122313527180784e-07,
+      "loss": 0.8119,
+      "step": 50416
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.121839443742099e-07,
+      "loss": 1.328,
+      "step": 50417
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.121365360303414e-07,
+      "loss": 2.2632,
+      "step": 50418
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.120891276864729e-07,
+      "loss": 1.8013,
+      "step": 50419
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.120417193426042e-07,
+      "loss": 2.8225,
+      "step": 50420
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.119943109987358e-07,
+      "loss": 2.035,
+      "step": 50421
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.119469026548672e-07,
+      "loss": 2.2358,
+      "step": 50422
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.118994943109987e-07,
+      "loss": 1.0458,
+      "step": 50423
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.118520859671303e-07,
+      "loss": 2.0637,
+      "step": 50424
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.118046776232617e-07,
+      "loss": 2.1746,
+      "step": 50425
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.117572692793932e-07,
+      "loss": 2.3713,
+      "step": 50426
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.117098609355247e-07,
+      "loss": 1.9444,
+      "step": 50427
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.116624525916562e-07,
+      "loss": 2.7802,
+      "step": 50428
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.116150442477877e-07,
+      "loss": 2.3946,
+      "step": 50429
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.115676359039191e-07,
+      "loss": 3.201,
+      "step": 50430
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.115202275600506e-07,
+      "loss": 1.9736,
+      "step": 50431
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.11472819216182e-07,
+      "loss": 1.3771,
+      "step": 50432
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.114254108723135e-07,
+      "loss": 3.1838,
+      "step": 50433
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.11378002528445e-07,
+      "loss": 2.4306,
+      "step": 50434
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.113305941845765e-07,
+      "loss": 0.5482,
+      "step": 50435
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.11283185840708e-07,
+      "loss": 3.4476,
+      "step": 50436
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.112357774968395e-07,
+      "loss": 1.2387,
+      "step": 50437
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.111883691529709e-07,
+      "loss": 1.7928,
+      "step": 50438
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.111409608091025e-07,
+      "loss": 0.9875,
+      "step": 50439
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.110935524652339e-07,
+      "loss": 0.8896,
+      "step": 50440
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.110461441213654e-07,
+      "loss": 2.1385,
+      "step": 50441
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 6.109987357774969e-07,
+      "loss": 3.1861,
+      "step": 50442
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.109513274336284e-07,
+      "loss": 2.1798,
+      "step": 50443
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.109039190897598e-07,
+      "loss": 2.2439,
+      "step": 50444
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.108565107458912e-07,
+      "loss": 0.7805,
+      "step": 50445
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.108091024020228e-07,
+      "loss": 0.919,
+      "step": 50446
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.107616940581543e-07,
+      "loss": 2.5175,
+      "step": 50447
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.107142857142857e-07,
+      "loss": 2.4198,
+      "step": 50448
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.106668773704172e-07,
+      "loss": 1.3005,
+      "step": 50449
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.106194690265487e-07,
+      "loss": 1.9515,
+      "step": 50450
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.105720606826802e-07,
+      "loss": 1.0848,
+      "step": 50451
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.105246523388117e-07,
+      "loss": 1.6572,
+      "step": 50452
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.104772439949431e-07,
+      "loss": 1.212,
+      "step": 50453
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.104298356510747e-07,
+      "loss": 0.6642,
+      "step": 50454
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.10382427307206e-07,
+      "loss": 2.1606,
+      "step": 50455
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.103350189633375e-07,
+      "loss": 1.7511,
+      "step": 50456
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.102876106194691e-07,
+      "loss": 1.6235,
+      "step": 50457
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.102402022756005e-07,
+      "loss": 0.9403,
+      "step": 50458
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.10192793931732e-07,
+      "loss": 2.0607,
+      "step": 50459
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.101453855878634e-07,
+      "loss": 2.1777,
+      "step": 50460
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.10097977243995e-07,
+      "loss": 1.5599,
+      "step": 50461
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.100505689001265e-07,
+      "loss": 1.4757,
+      "step": 50462
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.100031605562579e-07,
+      "loss": 2.4073,
+      "step": 50463
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.099557522123894e-07,
+      "loss": 2.6577,
+      "step": 50464
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.099083438685209e-07,
+      "loss": 3.0843,
+      "step": 50465
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.098609355246524e-07,
+      "loss": 2.2008,
+      "step": 50466
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.098135271807838e-07,
+      "loss": 2.0626,
+      "step": 50467
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.097661188369153e-07,
+      "loss": 1.682,
+      "step": 50468
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.097187104930468e-07,
+      "loss": 3.8586,
+      "step": 50469
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.096713021491782e-07,
+      "loss": 2.0455,
+      "step": 50470
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.096238938053097e-07,
+      "loss": 0.6531,
+      "step": 50471
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.095764854614413e-07,
+      "loss": 1.7191,
+      "step": 50472
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.095290771175727e-07,
+      "loss": 3.0761,
+      "step": 50473
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.094816687737042e-07,
+      "loss": 2.7773,
+      "step": 50474
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.094342604298357e-07,
+      "loss": 2.1987,
+      "step": 50475
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.093868520859672e-07,
+      "loss": 2.3955,
+      "step": 50476
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.093394437420987e-07,
+      "loss": 2.072,
+      "step": 50477
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.0929203539823e-07,
+      "loss": 2.909,
+      "step": 50478
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.092446270543615e-07,
+      "loss": 1.5275,
+      "step": 50479
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.09197218710493e-07,
+      "loss": 2.1822,
+      "step": 50480
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.091498103666245e-07,
+      "loss": 2.2497,
+      "step": 50481
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.09102402022756e-07,
+      "loss": 1.0731,
+      "step": 50482
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.090549936788875e-07,
+      "loss": 1.6031,
+      "step": 50483
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.09007585335019e-07,
+      "loss": 1.3869,
+      "step": 50484
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.089601769911505e-07,
+      "loss": 1.0028,
+      "step": 50485
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.089127686472819e-07,
+      "loss": 1.7042,
+      "step": 50486
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.088653603034135e-07,
+      "loss": 2.8485,
+      "step": 50487
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.088179519595449e-07,
+      "loss": 0.4976,
+      "step": 50488
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.087705436156764e-07,
+      "loss": 2.3816,
+      "step": 50489
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.087231352718078e-07,
+      "loss": 0.9161,
+      "step": 50490
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.086757269279393e-07,
+      "loss": 1.4382,
+      "step": 50491
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.086283185840708e-07,
+      "loss": 2.2207,
+      "step": 50492
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.085809102402022e-07,
+      "loss": 3.1541,
+      "step": 50493
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.085335018963338e-07,
+      "loss": 1.2105,
+      "step": 50494
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.084860935524653e-07,
+      "loss": 2.3272,
+      "step": 50495
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.084386852085967e-07,
+      "loss": 2.3019,
+      "step": 50496
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.083912768647282e-07,
+      "loss": 1.58,
+      "step": 50497
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.083438685208597e-07,
+      "loss": 3.377,
+      "step": 50498
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.082964601769912e-07,
+      "loss": 2.3054,
+      "step": 50499
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.082490518331227e-07,
+      "loss": 1.9405,
+      "step": 50500
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.08201643489254e-07,
+      "loss": 1.7172,
+      "step": 50501
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.081542351453856e-07,
+      "loss": 0.4481,
+      "step": 50502
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.08106826801517e-07,
+      "loss": 2.2502,
+      "step": 50503
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.080594184576485e-07,
+      "loss": 2.5333,
+      "step": 50504
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 6.080120101137801e-07,
+      "loss": 1.6304,
+      "step": 50505
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.079646017699115e-07,
+      "loss": 1.3161,
+      "step": 50506
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.07917193426043e-07,
+      "loss": 3.3672,
+      "step": 50507
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.078697850821745e-07,
+      "loss": 1.7809,
+      "step": 50508
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.07822376738306e-07,
+      "loss": 2.3887,
+      "step": 50509
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.077749683944375e-07,
+      "loss": 1.9095,
+      "step": 50510
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.077275600505689e-07,
+      "loss": 2.4681,
+      "step": 50511
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.076801517067004e-07,
+      "loss": 3.0906,
+      "step": 50512
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.076327433628319e-07,
+      "loss": 2.3929,
+      "step": 50513
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.075853350189633e-07,
+      "loss": 1.4199,
+      "step": 50514
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.075379266750948e-07,
+      "loss": 1.5818,
+      "step": 50515
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.074905183312263e-07,
+      "loss": 2.035,
+      "step": 50516
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.074431099873578e-07,
+      "loss": 1.2648,
+      "step": 50517
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.073957016434893e-07,
+      "loss": 2.4299,
+      "step": 50518
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.073482932996207e-07,
+      "loss": 1.7943,
+      "step": 50519
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.073008849557523e-07,
+      "loss": 1.2211,
+      "step": 50520
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.072534766118837e-07,
+      "loss": 1.2432,
+      "step": 50521
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.072060682680152e-07,
+      "loss": 2.3649,
+      "step": 50522
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.071586599241467e-07,
+      "loss": 2.2876,
+      "step": 50523
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.071112515802782e-07,
+      "loss": 0.8608,
+      "step": 50524
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.070638432364097e-07,
+      "loss": 1.9755,
+      "step": 50525
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.07016434892541e-07,
+      "loss": 1.2815,
+      "step": 50526
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.069690265486726e-07,
+      "loss": 3.0848,
+      "step": 50527
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.069216182048041e-07,
+      "loss": 1.5671,
+      "step": 50528
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.068742098609355e-07,
+      "loss": 2.0716,
+      "step": 50529
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.06826801517067e-07,
+      "loss": 2.0889,
+      "step": 50530
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.067793931731985e-07,
+      "loss": 3.1373,
+      "step": 50531
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.0673198482933e-07,
+      "loss": 2.5844,
+      "step": 50532
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.066845764854615e-07,
+      "loss": 1.4612,
+      "step": 50533
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.066371681415929e-07,
+      "loss": 1.1702,
+      "step": 50534
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.065897597977245e-07,
+      "loss": 1.499,
+      "step": 50535
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.065423514538559e-07,
+      "loss": 2.1863,
+      "step": 50536
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.064949431099873e-07,
+      "loss": 1.7735,
+      "step": 50537
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.064475347661188e-07,
+      "loss": 0.5707,
+      "step": 50538
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.064001264222503e-07,
+      "loss": 1.3872,
+      "step": 50539
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.063527180783818e-07,
+      "loss": 2.8122,
+      "step": 50540
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.063053097345132e-07,
+      "loss": 2.7009,
+      "step": 50541
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.062579013906448e-07,
+      "loss": 1.4028,
+      "step": 50542
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.062104930467763e-07,
+      "loss": 3.694,
+      "step": 50543
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.061630847029077e-07,
+      "loss": 1.4805,
+      "step": 50544
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.061156763590392e-07,
+      "loss": 2.6787,
+      "step": 50545
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.060682680151707e-07,
+      "loss": 1.6256,
+      "step": 50546
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.060208596713022e-07,
+      "loss": 3.4228,
+      "step": 50547
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.059734513274337e-07,
+      "loss": 1.0989,
+      "step": 50548
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.059260429835651e-07,
+      "loss": 1.8187,
+      "step": 50549
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.058786346396966e-07,
+      "loss": 2.735,
+      "step": 50550
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.05831226295828e-07,
+      "loss": 2.6094,
+      "step": 50551
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.057838179519595e-07,
+      "loss": 2.0684,
+      "step": 50552
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.057364096080911e-07,
+      "loss": 2.191,
+      "step": 50553
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.056890012642225e-07,
+      "loss": 1.3747,
+      "step": 50554
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.05641592920354e-07,
+      "loss": 2.8826,
+      "step": 50555
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.055941845764855e-07,
+      "loss": 2.8177,
+      "step": 50556
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.05546776232617e-07,
+      "loss": 3.2529,
+      "step": 50557
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.054993678887485e-07,
+      "loss": 2.4263,
+      "step": 50558
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.054519595448799e-07,
+      "loss": 2.3647,
+      "step": 50559
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.054045512010115e-07,
+      "loss": 2.2443,
+      "step": 50560
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.053571428571428e-07,
+      "loss": 1.8774,
+      "step": 50561
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.053097345132743e-07,
+      "loss": 2.7092,
+      "step": 50562
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.052623261694058e-07,
+      "loss": 2.3923,
+      "step": 50563
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.052149178255373e-07,
+      "loss": 1.1499,
+      "step": 50564
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.051675094816688e-07,
+      "loss": 1.6082,
+      "step": 50565
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.051201011378003e-07,
+      "loss": 1.7148,
+      "step": 50566
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.050726927939317e-07,
+      "loss": 1.894,
+      "step": 50567
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.050252844500633e-07,
+      "loss": 1.203,
+      "step": 50568
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.049778761061947e-07,
+      "loss": 3.7056,
+      "step": 50569
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.049304677623262e-07,
+      "loss": 1.3699,
+      "step": 50570
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.048830594184577e-07,
+      "loss": 1.367,
+      "step": 50571
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.048356510745891e-07,
+      "loss": 1.2779,
+      "step": 50572
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.047882427307206e-07,
+      "loss": 4.6883,
+      "step": 50573
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.04740834386852e-07,
+      "loss": 2.5561,
+      "step": 50574
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.046934260429836e-07,
+      "loss": 2.6562,
+      "step": 50575
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.046460176991151e-07,
+      "loss": 2.6774,
+      "step": 50576
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.045986093552465e-07,
+      "loss": 1.5381,
+      "step": 50577
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.04551201011378e-07,
+      "loss": 0.7222,
+      "step": 50578
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.045037926675095e-07,
+      "loss": 2.2955,
+      "step": 50579
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.04456384323641e-07,
+      "loss": 2.0903,
+      "step": 50580
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.044089759797725e-07,
+      "loss": 0.9584,
+      "step": 50581
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.04361567635904e-07,
+      "loss": 0.8659,
+      "step": 50582
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.043141592920355e-07,
+      "loss": 2.0461,
+      "step": 50583
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.042667509481668e-07,
+      "loss": 1.5606,
+      "step": 50584
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.042193426042983e-07,
+      "loss": 2.0846,
+      "step": 50585
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.041719342604299e-07,
+      "loss": 1.4715,
+      "step": 50586
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.041245259165613e-07,
+      "loss": 0.4753,
+      "step": 50587
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.040771175726928e-07,
+      "loss": 1.3854,
+      "step": 50588
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.040297092288243e-07,
+      "loss": 1.3558,
+      "step": 50589
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.039823008849558e-07,
+      "loss": 1.3953,
+      "step": 50590
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.039348925410873e-07,
+      "loss": 2.894,
+      "step": 50591
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.038874841972187e-07,
+      "loss": 1.6242,
+      "step": 50592
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.038400758533503e-07,
+      "loss": 1.509,
+      "step": 50593
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.037926675094817e-07,
+      "loss": 2.9052,
+      "step": 50594
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.037452591656131e-07,
+      "loss": 1.5836,
+      "step": 50595
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.036978508217446e-07,
+      "loss": 2.4077,
+      "step": 50596
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.036504424778761e-07,
+      "loss": 2.2957,
+      "step": 50597
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.036030341340076e-07,
+      "loss": 0.5924,
+      "step": 50598
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.03555625790139e-07,
+      "loss": 1.0122,
+      "step": 50599
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.035082174462705e-07,
+      "loss": 1.4514,
+      "step": 50600
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.034608091024021e-07,
+      "loss": 2.6837,
+      "step": 50601
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.034134007585335e-07,
+      "loss": 0.666,
+      "step": 50602
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.03365992414665e-07,
+      "loss": 2.1754,
+      "step": 50603
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.033185840707965e-07,
+      "loss": 2.5919,
+      "step": 50604
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.03271175726928e-07,
+      "loss": 2.432,
+      "step": 50605
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.032237673830595e-07,
+      "loss": 2.2719,
+      "step": 50606
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.031763590391908e-07,
+      "loss": 2.0231,
+      "step": 50607
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.031289506953224e-07,
+      "loss": 1.5307,
+      "step": 50608
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.030815423514538e-07,
+      "loss": 2.1349,
+      "step": 50609
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.030341340075853e-07,
+      "loss": 2.3779,
+      "step": 50610
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.029867256637168e-07,
+      "loss": 1.8524,
+      "step": 50611
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.029393173198483e-07,
+      "loss": 0.9852,
+      "step": 50612
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.028919089759798e-07,
+      "loss": 0.8891,
+      "step": 50613
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.028445006321113e-07,
+      "loss": 3.7024,
+      "step": 50614
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.027970922882428e-07,
+      "loss": 1.0298,
+      "step": 50615
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.027496839443743e-07,
+      "loss": 2.1653,
+      "step": 50616
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.027022756005057e-07,
+      "loss": 1.714,
+      "step": 50617
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.026548672566372e-07,
+      "loss": 2.3236,
+      "step": 50618
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.026074589127686e-07,
+      "loss": 2.6379,
+      "step": 50619
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.025600505689001e-07,
+      "loss": 0.4415,
+      "step": 50620
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.025126422250316e-07,
+      "loss": 2.4344,
+      "step": 50621
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.02465233881163e-07,
+      "loss": 1.9232,
+      "step": 50622
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.024178255372946e-07,
+      "loss": 2.8945,
+      "step": 50623
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.023704171934261e-07,
+      "loss": 1.9222,
+      "step": 50624
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.023230088495575e-07,
+      "loss": 1.3409,
+      "step": 50625
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.022756005056891e-07,
+      "loss": 0.4815,
+      "step": 50626
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.022281921618205e-07,
+      "loss": 3.1582,
+      "step": 50627
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.02180783817952e-07,
+      "loss": 2.4477,
+      "step": 50628
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.021333754740835e-07,
+      "loss": 2.6824,
+      "step": 50629
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.02085967130215e-07,
+      "loss": 1.7144,
+      "step": 50630
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.020385587863464e-07,
+      "loss": 2.152,
+      "step": 50631
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 6.019911504424778e-07,
+      "loss": 3.0958,
+      "step": 50632
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.019437420986093e-07,
+      "loss": 2.9743,
+      "step": 50633
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.018963337547409e-07,
+      "loss": 1.4673,
+      "step": 50634
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.018489254108723e-07,
+      "loss": 1.5953,
+      "step": 50635
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.018015170670038e-07,
+      "loss": 0.7288,
+      "step": 50636
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.017541087231353e-07,
+      "loss": 0.4902,
+      "step": 50637
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.017067003792668e-07,
+      "loss": 0.9411,
+      "step": 50638
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.016592920353983e-07,
+      "loss": 2.7063,
+      "step": 50639
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.016118836915297e-07,
+      "loss": 1.9368,
+      "step": 50640
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.015644753476613e-07,
+      "loss": 1.3992,
+      "step": 50641
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.015170670037926e-07,
+      "loss": 2.4462,
+      "step": 50642
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.014696586599241e-07,
+      "loss": 3.1294,
+      "step": 50643
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.014222503160556e-07,
+      "loss": 1.1155,
+      "step": 50644
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.013748419721871e-07,
+      "loss": 1.7053,
+      "step": 50645
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.013274336283186e-07,
+      "loss": 3.7042,
+      "step": 50646
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.012800252844501e-07,
+      "loss": 2.2098,
+      "step": 50647
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.012326169405816e-07,
+      "loss": 2.3688,
+      "step": 50648
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.011852085967131e-07,
+      "loss": 1.7749,
+      "step": 50649
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.011378002528445e-07,
+      "loss": 2.536,
+      "step": 50650
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.01090391908976e-07,
+      "loss": 1.3605,
+      "step": 50651
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.010429835651075e-07,
+      "loss": 3.8157,
+      "step": 50652
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.00995575221239e-07,
+      "loss": 2.0149,
+      "step": 50653
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.009481668773704e-07,
+      "loss": 2.1901,
+      "step": 50654
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.009007585335018e-07,
+      "loss": 1.9796,
+      "step": 50655
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.008533501896334e-07,
+      "loss": 0.8339,
+      "step": 50656
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.008059418457649e-07,
+      "loss": 0.644,
+      "step": 50657
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.007585335018963e-07,
+      "loss": 0.7822,
+      "step": 50658
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.007111251580279e-07,
+      "loss": 1.802,
+      "step": 50659
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.006637168141593e-07,
+      "loss": 3.1157,
+      "step": 50660
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.006163084702908e-07,
+      "loss": 0.3783,
+      "step": 50661
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.005689001264223e-07,
+      "loss": 1.3844,
+      "step": 50662
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.005214917825538e-07,
+      "loss": 2.9399,
+      "step": 50663
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.004740834386853e-07,
+      "loss": 0.6649,
+      "step": 50664
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.5655417406749556,
+      "eval_loss": 1.85030198097229,
+      "eval_runtime": 154.4819,
+      "eval_samples_per_second": 18.222,
+      "eval_steps_per_second": 4.557,
+      "step": 50664
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.004266750948166e-07,
+      "loss": 0.7027,
+      "step": 50665
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.003792667509481e-07,
+      "loss": 2.0223,
+      "step": 50666
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.003318584070797e-07,
+      "loss": 1.729,
+      "step": 50667
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.002844500632111e-07,
+      "loss": 2.6822,
+      "step": 50668
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.002370417193426e-07,
+      "loss": 1.4272,
+      "step": 50669
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.001896333754741e-07,
+      "loss": 0.5354,
+      "step": 50670
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.001422250316056e-07,
+      "loss": 1.5673,
+      "step": 50671
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.000948166877371e-07,
+      "loss": 1.9868,
+      "step": 50672
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.000474083438685e-07,
+      "loss": 1.2584,
+      "step": 50673
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 2.1008,
+      "step": 50674
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.999525916561315e-07,
+      "loss": 1.6489,
+      "step": 50675
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.99905183312263e-07,
+      "loss": 1.2514,
+      "step": 50676
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.998577749683944e-07,
+      "loss": 1.7996,
+      "step": 50677
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.998103666245259e-07,
+      "loss": 2.263,
+      "step": 50678
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.997629582806574e-07,
+      "loss": 0.7763,
+      "step": 50679
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.997155499367888e-07,
+      "loss": 2.412,
+      "step": 50680
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.996681415929204e-07,
+      "loss": 2.7091,
+      "step": 50681
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.996207332490519e-07,
+      "loss": 1.5544,
+      "step": 50682
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.995733249051833e-07,
+      "loss": 1.7125,
+      "step": 50683
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.995259165613148e-07,
+      "loss": 0.9595,
+      "step": 50684
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.994785082174463e-07,
+      "loss": 1.3702,
+      "step": 50685
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.994310998735778e-07,
+      "loss": 1.4949,
+      "step": 50686
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.993836915297093e-07,
+      "loss": 2.9396,
+      "step": 50687
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.993362831858407e-07,
+      "loss": 1.4145,
+      "step": 50688
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.992888748419722e-07,
+      "loss": 2.4799,
+      "step": 50689
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.992414664981036e-07,
+      "loss": 2.4228,
+      "step": 50690
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.991940581542351e-07,
+      "loss": 2.7495,
+      "step": 50691
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.991466498103667e-07,
+      "loss": 2.1575,
+      "step": 50692
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.990992414664981e-07,
+      "loss": 1.0231,
+      "step": 50693
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.990518331226296e-07,
+      "loss": 2.3787,
+      "step": 50694
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 5.990044247787611e-07,
+      "loss": 1.793,
+      "step": 50695
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.989570164348926e-07,
+      "loss": 2.5184,
+      "step": 50696
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.989096080910241e-07,
+      "loss": 1.8296,
+      "step": 50697
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.988621997471555e-07,
+      "loss": 2.1741,
+      "step": 50698
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.98814791403287e-07,
+      "loss": 1.9305,
+      "step": 50699
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.987673830594185e-07,
+      "loss": 2.5251,
+      "step": 50700
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.987199747155499e-07,
+      "loss": 1.5395,
+      "step": 50701
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.986725663716814e-07,
+      "loss": 1.7566,
+      "step": 50702
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.986251580278128e-07,
+      "loss": 2.5582,
+      "step": 50703
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.985777496839444e-07,
+      "loss": 0.393,
+      "step": 50704
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.985303413400759e-07,
+      "loss": 1.2666,
+      "step": 50705
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.984829329962073e-07,
+      "loss": 1.8598,
+      "step": 50706
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.984355246523389e-07,
+      "loss": 1.9635,
+      "step": 50707
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.983881163084703e-07,
+      "loss": 1.4559,
+      "step": 50708
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.983407079646018e-07,
+      "loss": 1.0305,
+      "step": 50709
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.982932996207333e-07,
+      "loss": 1.2377,
+      "step": 50710
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.982458912768648e-07,
+      "loss": 2.8961,
+      "step": 50711
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.981984829329963e-07,
+      "loss": 3.5734,
+      "step": 50712
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.981510745891276e-07,
+      "loss": 0.9497,
+      "step": 50713
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.981036662452591e-07,
+      "loss": 1.8055,
+      "step": 50714
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.980562579013907e-07,
+      "loss": 1.6879,
+      "step": 50715
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.980088495575221e-07,
+      "loss": 2.3761,
+      "step": 50716
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.979614412136536e-07,
+      "loss": 1.1637,
+      "step": 50717
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.979140328697851e-07,
+      "loss": 1.3981,
+      "step": 50718
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.978666245259166e-07,
+      "loss": 2.207,
+      "step": 50719
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.978192161820481e-07,
+      "loss": 2.1865,
+      "step": 50720
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.977718078381795e-07,
+      "loss": 2.213,
+      "step": 50721
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.977243994943111e-07,
+      "loss": 2.1225,
+      "step": 50722
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.976769911504425e-07,
+      "loss": 1.1665,
+      "step": 50723
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.976295828065739e-07,
+      "loss": 2.2708,
+      "step": 50724
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.975821744627054e-07,
+      "loss": 1.1498,
+      "step": 50725
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.975347661188369e-07,
+      "loss": 1.51,
+      "step": 50726
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.974873577749684e-07,
+      "loss": 2.5944,
+      "step": 50727
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.974399494310999e-07,
+      "loss": 2.3163,
+      "step": 50728
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.973925410872314e-07,
+      "loss": 1.0131,
+      "step": 50729
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.973451327433629e-07,
+      "loss": 1.1295,
+      "step": 50730
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.972977243994943e-07,
+      "loss": 0.8001,
+      "step": 50731
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.972503160556258e-07,
+      "loss": 2.3819,
+      "step": 50732
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.972029077117573e-07,
+      "loss": 0.8279,
+      "step": 50733
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.971554993678888e-07,
+      "loss": 2.8507,
+      "step": 50734
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.971080910240203e-07,
+      "loss": 1.0524,
+      "step": 50735
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.970606826801516e-07,
+      "loss": 1.0741,
+      "step": 50736
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.970132743362832e-07,
+      "loss": 1.5564,
+      "step": 50737
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.969658659924146e-07,
+      "loss": 2.3409,
+      "step": 50738
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.969184576485461e-07,
+      "loss": 2.8619,
+      "step": 50739
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.968710493046777e-07,
+      "loss": 1.2799,
+      "step": 50740
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.968236409608091e-07,
+      "loss": 2.5035,
+      "step": 50741
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.967762326169406e-07,
+      "loss": 1.2852,
+      "step": 50742
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.967288242730721e-07,
+      "loss": 0.6374,
+      "step": 50743
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.966814159292036e-07,
+      "loss": 2.3975,
+      "step": 50744
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.966340075853351e-07,
+      "loss": 2.4976,
+      "step": 50745
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.965865992414665e-07,
+      "loss": 1.7167,
+      "step": 50746
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.96539190897598e-07,
+      "loss": 2.4081,
+      "step": 50747
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.964917825537294e-07,
+      "loss": 2.3101,
+      "step": 50748
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.964443742098609e-07,
+      "loss": 2.7689,
+      "step": 50749
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.963969658659924e-07,
+      "loss": 2.227,
+      "step": 50750
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.963495575221239e-07,
+      "loss": 1.955,
+      "step": 50751
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.963021491782554e-07,
+      "loss": 0.9078,
+      "step": 50752
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.962547408343869e-07,
+      "loss": 2.444,
+      "step": 50753
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.962073324905183e-07,
+      "loss": 2.7681,
+      "step": 50754
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.961599241466499e-07,
+      "loss": 2.6476,
+      "step": 50755
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.961125158027813e-07,
+      "loss": 2.9829,
+      "step": 50756
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.960651074589128e-07,
+      "loss": 1.7098,
+      "step": 50757
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 5.960176991150443e-07,
+      "loss": 0.8952,
+      "step": 50758
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.959702907711757e-07,
+      "loss": 1.0983,
+      "step": 50759
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.959228824273072e-07,
+      "loss": 2.1568,
+      "step": 50760
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.958754740834386e-07,
+      "loss": 2.3022,
+      "step": 50761
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.958280657395702e-07,
+      "loss": 2.3542,
+      "step": 50762
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.957806573957017e-07,
+      "loss": 3.7787,
+      "step": 50763
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.957332490518331e-07,
+      "loss": 1.8684,
+      "step": 50764
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.956858407079646e-07,
+      "loss": 1.6565,
+      "step": 50765
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.956384323640961e-07,
+      "loss": 2.3979,
+      "step": 50766
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.955910240202276e-07,
+      "loss": 1.4726,
+      "step": 50767
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.955436156763591e-07,
+      "loss": 0.7423,
+      "step": 50768
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.954962073324905e-07,
+      "loss": 1.1921,
+      "step": 50769
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.954487989886221e-07,
+      "loss": 2.6034,
+      "step": 50770
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.954013906447534e-07,
+      "loss": 2.7776,
+      "step": 50771
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.953539823008849e-07,
+      "loss": 1.7669,
+      "step": 50772
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.953065739570165e-07,
+      "loss": 4.0289,
+      "step": 50773
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.952591656131479e-07,
+      "loss": 1.2601,
+      "step": 50774
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.952117572692794e-07,
+      "loss": 1.9634,
+      "step": 50775
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.951643489254109e-07,
+      "loss": 1.4799,
+      "step": 50776
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.951169405815424e-07,
+      "loss": 1.3059,
+      "step": 50777
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.950695322376739e-07,
+      "loss": 2.5271,
+      "step": 50778
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.950221238938053e-07,
+      "loss": 2.522,
+      "step": 50779
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.949747155499368e-07,
+      "loss": 1.1017,
+      "step": 50780
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.949273072060683e-07,
+      "loss": 2.0376,
+      "step": 50781
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.948798988621998e-07,
+      "loss": 1.9423,
+      "step": 50782
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.948324905183312e-07,
+      "loss": 1.584,
+      "step": 50783
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.947850821744627e-07,
+      "loss": 1.9139,
+      "step": 50784
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.947376738305942e-07,
+      "loss": 0.7927,
+      "step": 50785
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.946902654867257e-07,
+      "loss": 0.9648,
+      "step": 50786
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.946428571428571e-07,
+      "loss": 2.0179,
+      "step": 50787
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.945954487989887e-07,
+      "loss": 1.4627,
+      "step": 50788
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.945480404551201e-07,
+      "loss": 0.7927,
+      "step": 50789
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.945006321112516e-07,
+      "loss": 1.4235,
+      "step": 50790
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.944532237673831e-07,
+      "loss": 1.9972,
+      "step": 50791
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.944058154235146e-07,
+      "loss": 2.6379,
+      "step": 50792
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.943584070796461e-07,
+      "loss": 2.3759,
+      "step": 50793
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.943109987357774e-07,
+      "loss": 2.5358,
+      "step": 50794
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.94263590391909e-07,
+      "loss": 1.3809,
+      "step": 50795
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.942161820480405e-07,
+      "loss": 1.4601,
+      "step": 50796
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.941687737041719e-07,
+      "loss": 1.6468,
+      "step": 50797
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.941213653603034e-07,
+      "loss": 1.4418,
+      "step": 50798
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.940739570164349e-07,
+      "loss": 2.9009,
+      "step": 50799
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.940265486725664e-07,
+      "loss": 1.8107,
+      "step": 50800
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.939791403286979e-07,
+      "loss": 2.5981,
+      "step": 50801
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.939317319848293e-07,
+      "loss": 2.4659,
+      "step": 50802
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.938843236409609e-07,
+      "loss": 1.2247,
+      "step": 50803
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.938369152970923e-07,
+      "loss": 2.4981,
+      "step": 50804
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.937895069532238e-07,
+      "loss": 0.4767,
+      "step": 50805
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.937420986093554e-07,
+      "loss": 2.7388,
+      "step": 50806
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.936946902654867e-07,
+      "loss": 2.0409,
+      "step": 50807
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.936472819216182e-07,
+      "loss": 1.2052,
+      "step": 50808
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.935998735777496e-07,
+      "loss": 0.3851,
+      "step": 50809
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.935524652338812e-07,
+      "loss": 2.245,
+      "step": 50810
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.935050568900127e-07,
+      "loss": 2.2514,
+      "step": 50811
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.934576485461441e-07,
+      "loss": 2.0396,
+      "step": 50812
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.934102402022756e-07,
+      "loss": 1.5687,
+      "step": 50813
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.933628318584071e-07,
+      "loss": 3.4469,
+      "step": 50814
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.933154235145386e-07,
+      "loss": 0.6362,
+      "step": 50815
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.932680151706701e-07,
+      "loss": 0.5899,
+      "step": 50816
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.932206068268016e-07,
+      "loss": 2.5534,
+      "step": 50817
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.93173198482933e-07,
+      "loss": 1.2142,
+      "step": 50818
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.931257901390644e-07,
+      "loss": 0.5945,
+      "step": 50819
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.930783817951959e-07,
+      "loss": 2.1679,
+      "step": 50820
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.930309734513275e-07,
+      "loss": 1.7235,
+      "step": 50821
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 5.929835651074589e-07,
+      "loss": 1.8085,
+      "step": 50822
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.929361567635904e-07,
+      "loss": 1.2536,
+      "step": 50823
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.928887484197219e-07,
+      "loss": 1.9056,
+      "step": 50824
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.928413400758534e-07,
+      "loss": 1.8887,
+      "step": 50825
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.927939317319849e-07,
+      "loss": 3.849,
+      "step": 50826
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.927465233881163e-07,
+      "loss": 1.5649,
+      "step": 50827
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.926991150442479e-07,
+      "loss": 3.7682,
+      "step": 50828
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.926517067003792e-07,
+      "loss": 1.7741,
+      "step": 50829
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.926042983565107e-07,
+      "loss": 2.6736,
+      "step": 50830
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.925568900126422e-07,
+      "loss": 2.0889,
+      "step": 50831
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.925094816687737e-07,
+      "loss": 1.5858,
+      "step": 50832
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.924620733249052e-07,
+      "loss": 2.3135,
+      "step": 50833
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.924146649810367e-07,
+      "loss": 1.0997,
+      "step": 50834
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.923672566371681e-07,
+      "loss": 1.9894,
+      "step": 50835
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.923198482932997e-07,
+      "loss": 2.6165,
+      "step": 50836
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.922724399494311e-07,
+      "loss": 0.6116,
+      "step": 50837
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.922250316055626e-07,
+      "loss": 2.7561,
+      "step": 50838
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.921776232616941e-07,
+      "loss": 1.3829,
+      "step": 50839
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.921302149178256e-07,
+      "loss": 2.8871,
+      "step": 50840
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.92082806573957e-07,
+      "loss": 2.1954,
+      "step": 50841
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.920353982300884e-07,
+      "loss": 2.1165,
+      "step": 50842
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.9198798988622e-07,
+      "loss": 2.0546,
+      "step": 50843
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.919405815423515e-07,
+      "loss": 2.0061,
+      "step": 50844
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.918931731984829e-07,
+      "loss": 1.0103,
+      "step": 50845
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.918457648546144e-07,
+      "loss": 2.0531,
+      "step": 50846
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.917983565107459e-07,
+      "loss": 1.9287,
+      "step": 50847
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.917509481668774e-07,
+      "loss": 2.44,
+      "step": 50848
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.917035398230089e-07,
+      "loss": 1.5531,
+      "step": 50849
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.916561314791404e-07,
+      "loss": 3.1271,
+      "step": 50850
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.916087231352719e-07,
+      "loss": 0.8316,
+      "step": 50851
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.915613147914033e-07,
+      "loss": 2.66,
+      "step": 50852
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.915139064475347e-07,
+      "loss": 2.5927,
+      "step": 50853
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.914664981036663e-07,
+      "loss": 2.3296,
+      "step": 50854
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.914190897597977e-07,
+      "loss": 1.6031,
+      "step": 50855
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.913716814159292e-07,
+      "loss": 2.7444,
+      "step": 50856
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.913242730720607e-07,
+      "loss": 2.7915,
+      "step": 50857
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.912768647281922e-07,
+      "loss": 1.8422,
+      "step": 50858
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.912294563843237e-07,
+      "loss": 2.6599,
+      "step": 50859
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.911820480404551e-07,
+      "loss": 1.4037,
+      "step": 50860
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.911346396965867e-07,
+      "loss": 0.3518,
+      "step": 50861
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.910872313527181e-07,
+      "loss": 3.0195,
+      "step": 50862
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.910398230088496e-07,
+      "loss": 2.2714,
+      "step": 50863
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.90992414664981e-07,
+      "loss": 1.4914,
+      "step": 50864
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.909450063211125e-07,
+      "loss": 1.7611,
+      "step": 50865
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.90897597977244e-07,
+      "loss": 2.404,
+      "step": 50866
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.908501896333755e-07,
+      "loss": 1.5304,
+      "step": 50867
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.908027812895069e-07,
+      "loss": 1.9509,
+      "step": 50868
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.907553729456385e-07,
+      "loss": 1.5352,
+      "step": 50869
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.907079646017699e-07,
+      "loss": 2.5658,
+      "step": 50870
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.906605562579014e-07,
+      "loss": 0.6804,
+      "step": 50871
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.906131479140329e-07,
+      "loss": 2.0103,
+      "step": 50872
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.905657395701644e-07,
+      "loss": 3.5706,
+      "step": 50873
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.905183312262959e-07,
+      "loss": 1.6822,
+      "step": 50874
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.904709228824273e-07,
+      "loss": 1.426,
+      "step": 50875
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.904235145385589e-07,
+      "loss": 0.6643,
+      "step": 50876
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.903761061946902e-07,
+      "loss": 1.7797,
+      "step": 50877
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.903286978508217e-07,
+      "loss": 2.4477,
+      "step": 50878
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.902812895069532e-07,
+      "loss": 2.7645,
+      "step": 50879
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.902338811630847e-07,
+      "loss": 3.0107,
+      "step": 50880
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.901864728192162e-07,
+      "loss": 2.1032,
+      "step": 50881
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.901390644753477e-07,
+      "loss": 3.5471,
+      "step": 50882
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.900916561314792e-07,
+      "loss": 1.5965,
+      "step": 50883
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.900442477876107e-07,
+      "loss": 1.1628,
+      "step": 50884
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 5.899968394437421e-07,
+      "loss": 1.854,
+      "step": 50885
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.899494310998736e-07,
+      "loss": 1.8058,
+      "step": 50886
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.899020227560051e-07,
+      "loss": 1.8761,
+      "step": 50887
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.898546144121365e-07,
+      "loss": 2.8224,
+      "step": 50888
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.89807206068268e-07,
+      "loss": 1.4672,
+      "step": 50889
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.897597977243994e-07,
+      "loss": 1.8064,
+      "step": 50890
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.89712389380531e-07,
+      "loss": 0.87,
+      "step": 50891
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.896649810366625e-07,
+      "loss": 1.4642,
+      "step": 50892
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.896175726927939e-07,
+      "loss": 1.5012,
+      "step": 50893
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.895701643489255e-07,
+      "loss": 2.3858,
+      "step": 50894
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.895227560050569e-07,
+      "loss": 2.0474,
+      "step": 50895
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.894753476611884e-07,
+      "loss": 0.8527,
+      "step": 50896
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.894279393173199e-07,
+      "loss": 2.2158,
+      "step": 50897
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.893805309734514e-07,
+      "loss": 1.3243,
+      "step": 50898
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.893331226295829e-07,
+      "loss": 1.3587,
+      "step": 50899
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.892857142857142e-07,
+      "loss": 2.8649,
+      "step": 50900
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.892383059418457e-07,
+      "loss": 2.0817,
+      "step": 50901
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.891908975979773e-07,
+      "loss": 0.947,
+      "step": 50902
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.891434892541087e-07,
+      "loss": 2.2682,
+      "step": 50903
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.890960809102402e-07,
+      "loss": 0.8807,
+      "step": 50904
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.890486725663717e-07,
+      "loss": 1.0136,
+      "step": 50905
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.890012642225032e-07,
+      "loss": 2.4122,
+      "step": 50906
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.889538558786347e-07,
+      "loss": 1.2225,
+      "step": 50907
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.889064475347661e-07,
+      "loss": 2.1288,
+      "step": 50908
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.888590391908977e-07,
+      "loss": 1.5934,
+      "step": 50909
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.888116308470291e-07,
+      "loss": 0.2888,
+      "step": 50910
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.887642225031605e-07,
+      "loss": 2.8581,
+      "step": 50911
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.88716814159292e-07,
+      "loss": 2.1351,
+      "step": 50912
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.886694058154235e-07,
+      "loss": 1.8322,
+      "step": 50913
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.88621997471555e-07,
+      "loss": 1.4492,
+      "step": 50914
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.885745891276865e-07,
+      "loss": 1.8872,
+      "step": 50915
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.885271807838179e-07,
+      "loss": 2.3932,
+      "step": 50916
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.884797724399495e-07,
+      "loss": 1.8429,
+      "step": 50917
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.884323640960809e-07,
+      "loss": 1.0824,
+      "step": 50918
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.883849557522124e-07,
+      "loss": 3.7912,
+      "step": 50919
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.883375474083439e-07,
+      "loss": 2.547,
+      "step": 50920
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.882901390644754e-07,
+      "loss": 2.1622,
+      "step": 50921
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.882427307206069e-07,
+      "loss": 0.7062,
+      "step": 50922
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.881953223767382e-07,
+      "loss": 1.0045,
+      "step": 50923
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.881479140328698e-07,
+      "loss": 2.2998,
+      "step": 50924
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.881005056890013e-07,
+      "loss": 1.2545,
+      "step": 50925
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.880530973451327e-07,
+      "loss": 1.6245,
+      "step": 50926
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.880056890012642e-07,
+      "loss": 1.3057,
+      "step": 50927
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.879582806573957e-07,
+      "loss": 0.7462,
+      "step": 50928
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.879108723135272e-07,
+      "loss": 2.2621,
+      "step": 50929
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.878634639696587e-07,
+      "loss": 2.3368,
+      "step": 50930
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.878160556257902e-07,
+      "loss": 4.6851,
+      "step": 50931
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.877686472819217e-07,
+      "loss": 1.4479,
+      "step": 50932
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.877212389380531e-07,
+      "loss": 3.2104,
+      "step": 50933
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.876738305941846e-07,
+      "loss": 0.7748,
+      "step": 50934
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.876264222503161e-07,
+      "loss": 0.8936,
+      "step": 50935
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.875790139064475e-07,
+      "loss": 1.1866,
+      "step": 50936
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.87531605562579e-07,
+      "loss": 0.8854,
+      "step": 50937
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.874841972187105e-07,
+      "loss": 1.2599,
+      "step": 50938
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.87436788874842e-07,
+      "loss": 1.7082,
+      "step": 50939
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.873893805309735e-07,
+      "loss": 2.3476,
+      "step": 50940
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.873419721871049e-07,
+      "loss": 0.8276,
+      "step": 50941
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.872945638432365e-07,
+      "loss": 1.9188,
+      "step": 50942
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.872471554993679e-07,
+      "loss": 3.1084,
+      "step": 50943
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.871997471554994e-07,
+      "loss": 1.6867,
+      "step": 50944
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.871523388116309e-07,
+      "loss": 1.4411,
+      "step": 50945
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.871049304677624e-07,
+      "loss": 2.1243,
+      "step": 50946
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.870575221238938e-07,
+      "loss": 1.2791,
+      "step": 50947
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 5.870101137800252e-07,
+      "loss": 2.1658,
+      "step": 50948
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.869627054361567e-07,
+      "loss": 2.5945,
+      "step": 50949
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.869152970922883e-07,
+      "loss": 1.7001,
+      "step": 50950
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.868678887484197e-07,
+      "loss": 3.0828,
+      "step": 50951
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.868204804045512e-07,
+      "loss": 1.5967,
+      "step": 50952
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.867730720606827e-07,
+      "loss": 1.5641,
+      "step": 50953
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.867256637168142e-07,
+      "loss": 1.3126,
+      "step": 50954
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.866782553729457e-07,
+      "loss": 1.2708,
+      "step": 50955
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.866308470290771e-07,
+      "loss": 1.8857,
+      "step": 50956
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.865834386852087e-07,
+      "loss": 2.0642,
+      "step": 50957
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.8653603034134e-07,
+      "loss": 0.8578,
+      "step": 50958
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.864886219974715e-07,
+      "loss": 1.8951,
+      "step": 50959
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.86441213653603e-07,
+      "loss": 1.2123,
+      "step": 50960
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.863938053097345e-07,
+      "loss": 1.9729,
+      "step": 50961
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.86346396965866e-07,
+      "loss": 1.6562,
+      "step": 50962
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.862989886219975e-07,
+      "loss": 1.3289,
+      "step": 50963
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.86251580278129e-07,
+      "loss": 4.1874,
+      "step": 50964
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.862041719342605e-07,
+      "loss": 3.6792,
+      "step": 50965
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.861567635903919e-07,
+      "loss": 2.307,
+      "step": 50966
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.861093552465234e-07,
+      "loss": 2.5314,
+      "step": 50967
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.860619469026549e-07,
+      "loss": 1.9798,
+      "step": 50968
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.860145385587864e-07,
+      "loss": 2.0184,
+      "step": 50969
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.859671302149178e-07,
+      "loss": 1.9347,
+      "step": 50970
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.859197218710492e-07,
+      "loss": 1.6537,
+      "step": 50971
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.858723135271808e-07,
+      "loss": 2.8096,
+      "step": 50972
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.858249051833123e-07,
+      "loss": 0.4744,
+      "step": 50973
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.857774968394437e-07,
+      "loss": 0.4571,
+      "step": 50974
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.857300884955753e-07,
+      "loss": 2.2844,
+      "step": 50975
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.856826801517067e-07,
+      "loss": 2.1637,
+      "step": 50976
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.856352718078382e-07,
+      "loss": 1.1096,
+      "step": 50977
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.855878634639697e-07,
+      "loss": 2.4073,
+      "step": 50978
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.855404551201012e-07,
+      "loss": 2.1305,
+      "step": 50979
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.854930467762327e-07,
+      "loss": 1.4047,
+      "step": 50980
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.85445638432364e-07,
+      "loss": 0.645,
+      "step": 50981
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.853982300884955e-07,
+      "loss": 1.8988,
+      "step": 50982
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.853508217446271e-07,
+      "loss": 0.4866,
+      "step": 50983
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.853034134007585e-07,
+      "loss": 0.8149,
+      "step": 50984
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.8525600505689e-07,
+      "loss": 2.0994,
+      "step": 50985
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.852085967130215e-07,
+      "loss": 1.6469,
+      "step": 50986
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.85161188369153e-07,
+      "loss": 1.1046,
+      "step": 50987
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.851137800252845e-07,
+      "loss": 1.9957,
+      "step": 50988
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.850663716814159e-07,
+      "loss": 1.1345,
+      "step": 50989
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.850189633375475e-07,
+      "loss": 1.5035,
+      "step": 50990
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.849715549936789e-07,
+      "loss": 1.911,
+      "step": 50991
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.849241466498104e-07,
+      "loss": 1.7462,
+      "step": 50992
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.848767383059418e-07,
+      "loss": 2.5102,
+      "step": 50993
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.848293299620733e-07,
+      "loss": 3.7996,
+      "step": 50994
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.847819216182048e-07,
+      "loss": 2.5577,
+      "step": 50995
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.847345132743363e-07,
+      "loss": 2.2192,
+      "step": 50996
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.846871049304678e-07,
+      "loss": 1.2391,
+      "step": 50997
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.846396965865993e-07,
+      "loss": 1.4574,
+      "step": 50998
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.845922882427307e-07,
+      "loss": 1.5197,
+      "step": 50999
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.845448798988622e-07,
+      "loss": 2.0705,
+      "step": 51000
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.844974715549937e-07,
+      "loss": 2.0378,
+      "step": 51001
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.844500632111252e-07,
+      "loss": 1.614,
+      "step": 51002
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.844026548672567e-07,
+      "loss": 1.4424,
+      "step": 51003
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.84355246523388e-07,
+      "loss": 1.1542,
+      "step": 51004
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.843078381795196e-07,
+      "loss": 2.0301,
+      "step": 51005
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.842604298356511e-07,
+      "loss": 0.8497,
+      "step": 51006
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.842130214917825e-07,
+      "loss": 1.3049,
+      "step": 51007
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.841656131479141e-07,
+      "loss": 1.3257,
+      "step": 51008
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.841182048040455e-07,
+      "loss": 3.1531,
+      "step": 51009
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.84070796460177e-07,
+      "loss": 1.9483,
+      "step": 51010
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.840233881163085e-07,
+      "loss": 0.8763,
+      "step": 51011
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 5.8397597977244e-07,
+      "loss": 1.4243,
+      "step": 51012
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.839285714285715e-07,
+      "loss": 2.6243,
+      "step": 51013
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.838811630847029e-07,
+      "loss": 2.5653,
+      "step": 51014
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.838337547408344e-07,
+      "loss": 2.0469,
+      "step": 51015
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.837863463969659e-07,
+      "loss": 1.7709,
+      "step": 51016
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.837389380530973e-07,
+      "loss": 1.845,
+      "step": 51017
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.836915297092288e-07,
+      "loss": 3.5283,
+      "step": 51018
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.836441213653603e-07,
+      "loss": 1.8012,
+      "step": 51019
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.835967130214918e-07,
+      "loss": 2.182,
+      "step": 51020
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.835493046776233e-07,
+      "loss": 1.4218,
+      "step": 51021
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.835018963337547e-07,
+      "loss": 1.2668,
+      "step": 51022
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.834544879898863e-07,
+      "loss": 2.6744,
+      "step": 51023
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.834070796460177e-07,
+      "loss": 1.2424,
+      "step": 51024
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.833596713021492e-07,
+      "loss": 2.2573,
+      "step": 51025
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.833122629582807e-07,
+      "loss": 0.9564,
+      "step": 51026
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.832648546144122e-07,
+      "loss": 2.2912,
+      "step": 51027
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.832174462705437e-07,
+      "loss": 1.8682,
+      "step": 51028
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.83170037926675e-07,
+      "loss": 0.9426,
+      "step": 51029
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.831226295828066e-07,
+      "loss": 1.9727,
+      "step": 51030
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.830752212389381e-07,
+      "loss": 1.9039,
+      "step": 51031
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.830278128950695e-07,
+      "loss": 2.4957,
+      "step": 51032
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.82980404551201e-07,
+      "loss": 1.4856,
+      "step": 51033
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.829329962073325e-07,
+      "loss": 1.5198,
+      "step": 51034
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.82885587863464e-07,
+      "loss": 0.7621,
+      "step": 51035
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.828381795195955e-07,
+      "loss": 1.6818,
+      "step": 51036
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.827907711757269e-07,
+      "loss": 2.2122,
+      "step": 51037
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.827433628318585e-07,
+      "loss": 3.0443,
+      "step": 51038
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.826959544879899e-07,
+      "loss": 1.95,
+      "step": 51039
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.826485461441213e-07,
+      "loss": 2.2001,
+      "step": 51040
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.826011378002529e-07,
+      "loss": 1.2143,
+      "step": 51041
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.825537294563843e-07,
+      "loss": 2.9534,
+      "step": 51042
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.825063211125158e-07,
+      "loss": 2.5669,
+      "step": 51043
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.824589127686473e-07,
+      "loss": 2.879,
+      "step": 51044
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.824115044247788e-07,
+      "loss": 1.3145,
+      "step": 51045
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.823640960809103e-07,
+      "loss": 1.6547,
+      "step": 51046
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.823166877370417e-07,
+      "loss": 1.4528,
+      "step": 51047
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.822692793931732e-07,
+      "loss": 3.1494,
+      "step": 51048
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.822218710493047e-07,
+      "loss": 1.2406,
+      "step": 51049
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.821744627054362e-07,
+      "loss": 1.078,
+      "step": 51050
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.821270543615677e-07,
+      "loss": 2.5852,
+      "step": 51051
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.820796460176991e-07,
+      "loss": 2.3579,
+      "step": 51052
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.820322376738306e-07,
+      "loss": 1.6154,
+      "step": 51053
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.819848293299621e-07,
+      "loss": 2.3511,
+      "step": 51054
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.819374209860935e-07,
+      "loss": 1.8859,
+      "step": 51055
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.818900126422251e-07,
+      "loss": 2.0267,
+      "step": 51056
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.818426042983565e-07,
+      "loss": 3.5606,
+      "step": 51057
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.81795195954488e-07,
+      "loss": 2.6149,
+      "step": 51058
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.817477876106195e-07,
+      "loss": 2.1693,
+      "step": 51059
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.81700379266751e-07,
+      "loss": 2.135,
+      "step": 51060
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.816529709228825e-07,
+      "loss": 2.6052,
+      "step": 51061
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.816055625790139e-07,
+      "loss": 2.5193,
+      "step": 51062
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.815581542351455e-07,
+      "loss": 0.5984,
+      "step": 51063
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.815107458912769e-07,
+      "loss": 2.1197,
+      "step": 51064
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.814633375474083e-07,
+      "loss": 0.779,
+      "step": 51065
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.814159292035398e-07,
+      "loss": 1.6222,
+      "step": 51066
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.813685208596713e-07,
+      "loss": 3.7069,
+      "step": 51067
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.813211125158028e-07,
+      "loss": 1.5553,
+      "step": 51068
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.812737041719343e-07,
+      "loss": 1.5942,
+      "step": 51069
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.812262958280657e-07,
+      "loss": 4.3775,
+      "step": 51070
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.811788874841973e-07,
+      "loss": 1.1856,
+      "step": 51071
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.811314791403287e-07,
+      "loss": 2.2441,
+      "step": 51072
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.810840707964602e-07,
+      "loss": 1.2311,
+      "step": 51073
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.810366624525918e-07,
+      "loss": 3.8898,
+      "step": 51074
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 5.809892541087231e-07,
+      "loss": 2.8998,
+      "step": 51075
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.809418457648546e-07,
+      "loss": 0.531,
+      "step": 51076
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.808944374209861e-07,
+      "loss": 1.5163,
+      "step": 51077
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.808470290771176e-07,
+      "loss": 2.2334,
+      "step": 51078
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.807996207332491e-07,
+      "loss": 1.7461,
+      "step": 51079
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.807522123893805e-07,
+      "loss": 2.8653,
+      "step": 51080
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.80704804045512e-07,
+      "loss": 1.433,
+      "step": 51081
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.806573957016435e-07,
+      "loss": 3.3586,
+      "step": 51082
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.80609987357775e-07,
+      "loss": 1.6889,
+      "step": 51083
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.805625790139065e-07,
+      "loss": 1.6419,
+      "step": 51084
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.80515170670038e-07,
+      "loss": 3.7685,
+      "step": 51085
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.804677623261695e-07,
+      "loss": 2.6192,
+      "step": 51086
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.804203539823008e-07,
+      "loss": 2.026,
+      "step": 51087
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.803729456384323e-07,
+      "loss": 1.4825,
+      "step": 51088
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.803255372945639e-07,
+      "loss": 1.9204,
+      "step": 51089
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.802781289506953e-07,
+      "loss": 2.6388,
+      "step": 51090
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.802307206068268e-07,
+      "loss": 2.9019,
+      "step": 51091
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.801833122629583e-07,
+      "loss": 1.8829,
+      "step": 51092
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.801359039190898e-07,
+      "loss": 2.0348,
+      "step": 51093
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.800884955752213e-07,
+      "loss": 3.7691,
+      "step": 51094
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.800410872313527e-07,
+      "loss": 1.6349,
+      "step": 51095
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.799936788874843e-07,
+      "loss": 3.0834,
+      "step": 51096
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.799462705436157e-07,
+      "loss": 1.9866,
+      "step": 51097
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.798988621997472e-07,
+      "loss": 1.5392,
+      "step": 51098
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.798514538558786e-07,
+      "loss": 0.2387,
+      "step": 51099
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.798040455120101e-07,
+      "loss": 3.341,
+      "step": 51100
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.797566371681416e-07,
+      "loss": 3.3404,
+      "step": 51101
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.797092288242731e-07,
+      "loss": 1.5326,
+      "step": 51102
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.796618204804045e-07,
+      "loss": 2.646,
+      "step": 51103
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.796144121365361e-07,
+      "loss": 1.9581,
+      "step": 51104
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.795670037926675e-07,
+      "loss": 2.0051,
+      "step": 51105
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.79519595448799e-07,
+      "loss": 1.4457,
+      "step": 51106
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.794721871049305e-07,
+      "loss": 2.9416,
+      "step": 51107
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.79424778761062e-07,
+      "loss": 1.2657,
+      "step": 51108
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.793773704171935e-07,
+      "loss": 1.1351,
+      "step": 51109
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.793299620733248e-07,
+      "loss": 2.444,
+      "step": 51110
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.792825537294564e-07,
+      "loss": 1.7584,
+      "step": 51111
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.792351453855879e-07,
+      "loss": 1.5591,
+      "step": 51112
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.791877370417193e-07,
+      "loss": 1.683,
+      "step": 51113
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.791403286978508e-07,
+      "loss": 1.9137,
+      "step": 51114
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.790929203539823e-07,
+      "loss": 2.6567,
+      "step": 51115
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.790455120101138e-07,
+      "loss": 1.8894,
+      "step": 51116
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.789981036662453e-07,
+      "loss": 2.7354,
+      "step": 51117
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.789506953223768e-07,
+      "loss": 0.4823,
+      "step": 51118
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.789032869785083e-07,
+      "loss": 2.19,
+      "step": 51119
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.788558786346397e-07,
+      "loss": 1.4879,
+      "step": 51120
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.788084702907712e-07,
+      "loss": 0.5695,
+      "step": 51121
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.787610619469028e-07,
+      "loss": 0.8013,
+      "step": 51122
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.787136536030341e-07,
+      "loss": 1.5231,
+      "step": 51123
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.786662452591656e-07,
+      "loss": 2.2126,
+      "step": 51124
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.786188369152971e-07,
+      "loss": 1.9594,
+      "step": 51125
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.785714285714286e-07,
+      "loss": 1.7744,
+      "step": 51126
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.785240202275601e-07,
+      "loss": 2.9331,
+      "step": 51127
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.784766118836915e-07,
+      "loss": 1.4144,
+      "step": 51128
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.784292035398231e-07,
+      "loss": 2.0797,
+      "step": 51129
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.783817951959545e-07,
+      "loss": 2.7163,
+      "step": 51130
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.78334386852086e-07,
+      "loss": 2.3697,
+      "step": 51131
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.782869785082175e-07,
+      "loss": 2.3899,
+      "step": 51132
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.78239570164349e-07,
+      "loss": 1.0004,
+      "step": 51133
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.781921618204804e-07,
+      "loss": 1.4212,
+      "step": 51134
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.781447534766119e-07,
+      "loss": 3.8986,
+      "step": 51135
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.780973451327433e-07,
+      "loss": 3.2977,
+      "step": 51136
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.780499367888749e-07,
+      "loss": 1.5931,
+      "step": 51137
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 5.780025284450063e-07,
+      "loss": 1.5992,
+      "step": 51138
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.779551201011378e-07,
+      "loss": 1.125,
+      "step": 51139
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.779077117572693e-07,
+      "loss": 1.4078,
+      "step": 51140
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.778603034134008e-07,
+      "loss": 2.0895,
+      "step": 51141
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.778128950695323e-07,
+      "loss": 2.0009,
+      "step": 51142
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.777654867256637e-07,
+      "loss": 1.0224,
+      "step": 51143
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.777180783817953e-07,
+      "loss": 1.3602,
+      "step": 51144
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.776706700379268e-07,
+      "loss": 3.8544,
+      "step": 51145
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.776232616940581e-07,
+      "loss": 2.16,
+      "step": 51146
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.775758533501896e-07,
+      "loss": 2.2395,
+      "step": 51147
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.775284450063211e-07,
+      "loss": 1.6929,
+      "step": 51148
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.774810366624526e-07,
+      "loss": 2.2384,
+      "step": 51149
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.774336283185841e-07,
+      "loss": 0.9559,
+      "step": 51150
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.773862199747155e-07,
+      "loss": 0.8428,
+      "step": 51151
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.773388116308471e-07,
+      "loss": 1.8184,
+      "step": 51152
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.772914032869785e-07,
+      "loss": 3.062,
+      "step": 51153
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.7724399494311e-07,
+      "loss": 1.9967,
+      "step": 51154
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.771965865992416e-07,
+      "loss": 1.7193,
+      "step": 51155
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.77149178255373e-07,
+      "loss": 0.8745,
+      "step": 51156
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.771017699115044e-07,
+      "loss": 0.9522,
+      "step": 51157
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.770543615676358e-07,
+      "loss": 1.2481,
+      "step": 51158
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.770069532237674e-07,
+      "loss": 1.2167,
+      "step": 51159
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.769595448798989e-07,
+      "loss": 1.654,
+      "step": 51160
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.769121365360303e-07,
+      "loss": 1.6197,
+      "step": 51161
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.768647281921618e-07,
+      "loss": 2.1632,
+      "step": 51162
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.768173198482933e-07,
+      "loss": 2.6627,
+      "step": 51163
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.767699115044248e-07,
+      "loss": 2.1924,
+      "step": 51164
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.767225031605563e-07,
+      "loss": 2.6041,
+      "step": 51165
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.766750948166878e-07,
+      "loss": 2.8831,
+      "step": 51166
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.766276864728193e-07,
+      "loss": 1.8496,
+      "step": 51167
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.765802781289506e-07,
+      "loss": 2.0875,
+      "step": 51168
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.765328697850821e-07,
+      "loss": 0.7233,
+      "step": 51169
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.764854614412137e-07,
+      "loss": 0.7096,
+      "step": 51170
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.764380530973451e-07,
+      "loss": 2.0068,
+      "step": 51171
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.763906447534766e-07,
+      "loss": 1.8486,
+      "step": 51172
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.763432364096081e-07,
+      "loss": 2.0142,
+      "step": 51173
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.762958280657396e-07,
+      "loss": 1.5755,
+      "step": 51174
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.762484197218711e-07,
+      "loss": 2.0621,
+      "step": 51175
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.762010113780025e-07,
+      "loss": 3.1833,
+      "step": 51176
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.761536030341341e-07,
+      "loss": 2.5496,
+      "step": 51177
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.761061946902655e-07,
+      "loss": 2.5058,
+      "step": 51178
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.76058786346397e-07,
+      "loss": 0.733,
+      "step": 51179
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.760113780025284e-07,
+      "loss": 1.3847,
+      "step": 51180
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.759639696586599e-07,
+      "loss": 2.8389,
+      "step": 51181
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.759165613147914e-07,
+      "loss": 2.3465,
+      "step": 51182
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.758691529709229e-07,
+      "loss": 1.9881,
+      "step": 51183
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.758217446270543e-07,
+      "loss": 3.6955,
+      "step": 51184
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.757743362831859e-07,
+      "loss": 1.6351,
+      "step": 51185
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.757269279393173e-07,
+      "loss": 2.1896,
+      "step": 51186
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.756795195954488e-07,
+      "loss": 2.6613,
+      "step": 51187
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.756321112515803e-07,
+      "loss": 1.6375,
+      "step": 51188
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.755847029077118e-07,
+      "loss": 1.8493,
+      "step": 51189
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.755372945638433e-07,
+      "loss": 0.5469,
+      "step": 51190
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.754898862199747e-07,
+      "loss": 1.0138,
+      "step": 51191
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.754424778761062e-07,
+      "loss": 2.6885,
+      "step": 51192
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.753950695322377e-07,
+      "loss": 2.4539,
+      "step": 51193
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.753476611883691e-07,
+      "loss": 1.3963,
+      "step": 51194
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.753002528445006e-07,
+      "loss": 1.3359,
+      "step": 51195
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.752528445006321e-07,
+      "loss": 0.976,
+      "step": 51196
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.752054361567636e-07,
+      "loss": 1.364,
+      "step": 51197
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.751580278128951e-07,
+      "loss": 1.8057,
+      "step": 51198
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.751106194690266e-07,
+      "loss": 1.6966,
+      "step": 51199
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.750632111251581e-07,
+      "loss": 1.0898,
+      "step": 51200
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.750158027812895e-07,
+      "loss": 1.2175,
+      "step": 51201
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 5.74968394437421e-07,
+      "loss": 1.2741,
+      "step": 51202
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.749209860935526e-07,
+      "loss": 2.4697,
+      "step": 51203
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.748735777496839e-07,
+      "loss": 2.7027,
+      "step": 51204
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.748261694058154e-07,
+      "loss": 1.8497,
+      "step": 51205
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.747787610619469e-07,
+      "loss": 1.1754,
+      "step": 51206
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.747313527180784e-07,
+      "loss": 1.4469,
+      "step": 51207
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.746839443742099e-07,
+      "loss": 2.7125,
+      "step": 51208
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.746365360303413e-07,
+      "loss": 2.2567,
+      "step": 51209
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.745891276864729e-07,
+      "loss": 1.2311,
+      "step": 51210
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.745417193426043e-07,
+      "loss": 2.0486,
+      "step": 51211
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.744943109987358e-07,
+      "loss": 1.7561,
+      "step": 51212
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.744469026548673e-07,
+      "loss": 2.616,
+      "step": 51213
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.743994943109988e-07,
+      "loss": 3.2663,
+      "step": 51214
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.743520859671303e-07,
+      "loss": 2.3756,
+      "step": 51215
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.743046776232617e-07,
+      "loss": 1.7181,
+      "step": 51216
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.742572692793931e-07,
+      "loss": 0.8856,
+      "step": 51217
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.742098609355247e-07,
+      "loss": 2.5932,
+      "step": 51218
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.741624525916561e-07,
+      "loss": 1.2746,
+      "step": 51219
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.741150442477876e-07,
+      "loss": 1.3339,
+      "step": 51220
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.740676359039191e-07,
+      "loss": 1.8409,
+      "step": 51221
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.740202275600506e-07,
+      "loss": 4.0229,
+      "step": 51222
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.739728192161821e-07,
+      "loss": 1.8636,
+      "step": 51223
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.739254108723135e-07,
+      "loss": 1.3262,
+      "step": 51224
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.738780025284451e-07,
+      "loss": 2.1692,
+      "step": 51225
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.738305941845765e-07,
+      "loss": 0.4814,
+      "step": 51226
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.737831858407079e-07,
+      "loss": 1.2576,
+      "step": 51227
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.737357774968394e-07,
+      "loss": 2.3059,
+      "step": 51228
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.736883691529709e-07,
+      "loss": 1.6512,
+      "step": 51229
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.736409608091024e-07,
+      "loss": 2.8848,
+      "step": 51230
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.735935524652339e-07,
+      "loss": 4.2968,
+      "step": 51231
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.735461441213654e-07,
+      "loss": 1.0293,
+      "step": 51232
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.734987357774969e-07,
+      "loss": 1.0801,
+      "step": 51233
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.734513274336283e-07,
+      "loss": 0.9005,
+      "step": 51234
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.734039190897598e-07,
+      "loss": 0.3164,
+      "step": 51235
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.733565107458913e-07,
+      "loss": 2.4528,
+      "step": 51236
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.733091024020228e-07,
+      "loss": 1.8796,
+      "step": 51237
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.732616940581543e-07,
+      "loss": 2.2761,
+      "step": 51238
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.732142857142856e-07,
+      "loss": 3.0245,
+      "step": 51239
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.731668773704172e-07,
+      "loss": 1.9738,
+      "step": 51240
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.731194690265487e-07,
+      "loss": 3.1759,
+      "step": 51241
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.730720606826801e-07,
+      "loss": 1.6074,
+      "step": 51242
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.730246523388117e-07,
+      "loss": 1.2326,
+      "step": 51243
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.729772439949431e-07,
+      "loss": 1.8306,
+      "step": 51244
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.729298356510746e-07,
+      "loss": 3.0447,
+      "step": 51245
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.728824273072061e-07,
+      "loss": 2.2945,
+      "step": 51246
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.728350189633376e-07,
+      "loss": 2.322,
+      "step": 51247
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.727876106194691e-07,
+      "loss": 3.3518,
+      "step": 51248
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.727402022756005e-07,
+      "loss": 1.6101,
+      "step": 51249
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.72692793931732e-07,
+      "loss": 3.2344,
+      "step": 51250
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.726453855878635e-07,
+      "loss": 2.011,
+      "step": 51251
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.725979772439949e-07,
+      "loss": 2.6026,
+      "step": 51252
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.725505689001264e-07,
+      "loss": 2.6433,
+      "step": 51253
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.725031605562579e-07,
+      "loss": 0.3721,
+      "step": 51254
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.724557522123894e-07,
+      "loss": 2.4194,
+      "step": 51255
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.724083438685209e-07,
+      "loss": 2.0217,
+      "step": 51256
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.723609355246523e-07,
+      "loss": 1.3888,
+      "step": 51257
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.723135271807839e-07,
+      "loss": 1.4652,
+      "step": 51258
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.722661188369153e-07,
+      "loss": 1.5584,
+      "step": 51259
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.722187104930468e-07,
+      "loss": 2.277,
+      "step": 51260
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.721713021491783e-07,
+      "loss": 2.6846,
+      "step": 51261
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.721238938053097e-07,
+      "loss": 1.192,
+      "step": 51262
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.720764854614412e-07,
+      "loss": 1.3577,
+      "step": 51263
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.720290771175727e-07,
+      "loss": 1.6086,
+      "step": 51264
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 5.719816687737042e-07,
+      "loss": 3.1424,
+      "step": 51265
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.719342604298357e-07,
+      "loss": 0.9861,
+      "step": 51266
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.718868520859671e-07,
+      "loss": 1.759,
+      "step": 51267
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.718394437420986e-07,
+      "loss": 2.6655,
+      "step": 51268
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.717920353982301e-07,
+      "loss": 2.3903,
+      "step": 51269
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.717446270543616e-07,
+      "loss": 1.9295,
+      "step": 51270
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.716972187104931e-07,
+      "loss": 1.8137,
+      "step": 51271
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.716498103666245e-07,
+      "loss": 1.6858,
+      "step": 51272
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.716024020227561e-07,
+      "loss": 1.6237,
+      "step": 51273
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.715549936788875e-07,
+      "loss": 2.0035,
+      "step": 51274
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.715075853350189e-07,
+      "loss": 1.6979,
+      "step": 51275
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.714601769911505e-07,
+      "loss": 2.4209,
+      "step": 51276
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.714127686472819e-07,
+      "loss": 1.9392,
+      "step": 51277
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.713653603034134e-07,
+      "loss": 1.6762,
+      "step": 51278
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.713179519595449e-07,
+      "loss": 3.647,
+      "step": 51279
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.712705436156764e-07,
+      "loss": 1.3303,
+      "step": 51280
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.712231352718079e-07,
+      "loss": 0.8879,
+      "step": 51281
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.711757269279393e-07,
+      "loss": 1.1007,
+      "step": 51282
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.711283185840708e-07,
+      "loss": 1.8034,
+      "step": 51283
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.710809102402024e-07,
+      "loss": 1.3782,
+      "step": 51284
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.710335018963338e-07,
+      "loss": 1.1139,
+      "step": 51285
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.709860935524652e-07,
+      "loss": 2.3304,
+      "step": 51286
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.709386852085967e-07,
+      "loss": 1.4966,
+      "step": 51287
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.708912768647282e-07,
+      "loss": 1.9523,
+      "step": 51288
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.708438685208597e-07,
+      "loss": 0.6301,
+      "step": 51289
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.707964601769911e-07,
+      "loss": 1.9742,
+      "step": 51290
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.707490518331227e-07,
+      "loss": 1.6839,
+      "step": 51291
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.707016434892541e-07,
+      "loss": 2.4743,
+      "step": 51292
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.706542351453856e-07,
+      "loss": 2.1101,
+      "step": 51293
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.706068268015171e-07,
+      "loss": 3.4418,
+      "step": 51294
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.705594184576486e-07,
+      "loss": 2.5004,
+      "step": 51295
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.705120101137801e-07,
+      "loss": 1.8401,
+      "step": 51296
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.704646017699114e-07,
+      "loss": 2.9491,
+      "step": 51297
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.70417193426043e-07,
+      "loss": 0.9464,
+      "step": 51298
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.703697850821745e-07,
+      "loss": 2.0096,
+      "step": 51299
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.703223767383059e-07,
+      "loss": 1.1755,
+      "step": 51300
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.702749683944374e-07,
+      "loss": 5.4132,
+      "step": 51301
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.702275600505689e-07,
+      "loss": 2.2028,
+      "step": 51302
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.701801517067004e-07,
+      "loss": 1.3746,
+      "step": 51303
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.701327433628319e-07,
+      "loss": 2.6616,
+      "step": 51304
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.700853350189633e-07,
+      "loss": 2.017,
+      "step": 51305
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.700379266750949e-07,
+      "loss": 3.6324,
+      "step": 51306
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.699905183312263e-07,
+      "loss": 2.1147,
+      "step": 51307
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.699431099873578e-07,
+      "loss": 1.051,
+      "step": 51308
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.698957016434894e-07,
+      "loss": 3.621,
+      "step": 51309
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.698482932996207e-07,
+      "loss": 1.7232,
+      "step": 51310
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.698008849557522e-07,
+      "loss": 2.7329,
+      "step": 51311
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.697534766118837e-07,
+      "loss": 1.721,
+      "step": 51312
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.697060682680152e-07,
+      "loss": 0.367,
+      "step": 51313
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.696586599241467e-07,
+      "loss": 1.4568,
+      "step": 51314
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.696112515802781e-07,
+      "loss": 2.1473,
+      "step": 51315
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.695638432364096e-07,
+      "loss": 1.5216,
+      "step": 51316
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.695164348925411e-07,
+      "loss": 1.4944,
+      "step": 51317
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.694690265486726e-07,
+      "loss": 1.9628,
+      "step": 51318
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.694216182048041e-07,
+      "loss": 1.1604,
+      "step": 51319
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.693742098609356e-07,
+      "loss": 1.8963,
+      "step": 51320
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.69326801517067e-07,
+      "loss": 2.1844,
+      "step": 51321
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.692793931731985e-07,
+      "loss": 2.4601,
+      "step": 51322
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.692319848293299e-07,
+      "loss": 1.6705,
+      "step": 51323
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.691845764854615e-07,
+      "loss": 2.6318,
+      "step": 51324
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.691371681415929e-07,
+      "loss": 2.0251,
+      "step": 51325
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.690897597977244e-07,
+      "loss": 3.3276,
+      "step": 51326
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.690423514538559e-07,
+      "loss": 1.9906,
+      "step": 51327
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 5.689949431099874e-07,
+      "loss": 1.9018,
+      "step": 51328
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.689475347661189e-07,
+      "loss": 1.2757,
+      "step": 51329
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.689001264222503e-07,
+      "loss": 1.9817,
+      "step": 51330
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.688527180783819e-07,
+      "loss": 2.7428,
+      "step": 51331
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.688053097345134e-07,
+      "loss": 0.8044,
+      "step": 51332
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.687579013906447e-07,
+      "loss": 1.89,
+      "step": 51333
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.687104930467762e-07,
+      "loss": 1.346,
+      "step": 51334
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.686630847029077e-07,
+      "loss": 1.0776,
+      "step": 51335
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.686156763590392e-07,
+      "loss": 1.9082,
+      "step": 51336
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.685682680151707e-07,
+      "loss": 0.8517,
+      "step": 51337
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.685208596713021e-07,
+      "loss": 2.8917,
+      "step": 51338
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.684734513274337e-07,
+      "loss": 1.1235,
+      "step": 51339
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.684260429835651e-07,
+      "loss": 1.3042,
+      "step": 51340
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.683786346396966e-07,
+      "loss": 3.0511,
+      "step": 51341
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.683312262958282e-07,
+      "loss": 2.7753,
+      "step": 51342
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.682838179519596e-07,
+      "loss": 1.7683,
+      "step": 51343
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.68236409608091e-07,
+      "loss": 2.1876,
+      "step": 51344
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.681890012642225e-07,
+      "loss": 1.7991,
+      "step": 51345
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.68141592920354e-07,
+      "loss": 1.4782,
+      "step": 51346
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.680941845764855e-07,
+      "loss": 3.4861,
+      "step": 51347
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.680467762326169e-07,
+      "loss": 2.2127,
+      "step": 51348
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.679993678887484e-07,
+      "loss": 1.9641,
+      "step": 51349
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.679519595448799e-07,
+      "loss": 1.5124,
+      "step": 51350
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.679045512010114e-07,
+      "loss": 0.3041,
+      "step": 51351
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.678571428571429e-07,
+      "loss": 1.4021,
+      "step": 51352
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.678097345132744e-07,
+      "loss": 1.9286,
+      "step": 51353
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.677623261694059e-07,
+      "loss": 1.8021,
+      "step": 51354
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.677149178255374e-07,
+      "loss": 1.0004,
+      "step": 51355
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.676675094816687e-07,
+      "loss": 1.8959,
+      "step": 51356
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.676201011378003e-07,
+      "loss": 1.7107,
+      "step": 51357
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.675726927939317e-07,
+      "loss": 1.2628,
+      "step": 51358
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.675252844500632e-07,
+      "loss": 2.0174,
+      "step": 51359
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.674778761061947e-07,
+      "loss": 2.0741,
+      "step": 51360
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.674304677623262e-07,
+      "loss": 3.7061,
+      "step": 51361
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.673830594184577e-07,
+      "loss": 1.2025,
+      "step": 51362
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.673356510745891e-07,
+      "loss": 1.944,
+      "step": 51363
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.672882427307206e-07,
+      "loss": 2.8654,
+      "step": 51364
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.672408343868521e-07,
+      "loss": 2.2771,
+      "step": 51365
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.671934260429836e-07,
+      "loss": 0.9933,
+      "step": 51366
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.67146017699115e-07,
+      "loss": 2.7578,
+      "step": 51367
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.670986093552465e-07,
+      "loss": 0.7273,
+      "step": 51368
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.67051201011378e-07,
+      "loss": 2.866,
+      "step": 51369
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.670037926675095e-07,
+      "loss": 2.8783,
+      "step": 51370
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.669563843236409e-07,
+      "loss": 1.1924,
+      "step": 51371
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.669089759797725e-07,
+      "loss": 1.8654,
+      "step": 51372
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.668615676359039e-07,
+      "loss": 2.1373,
+      "step": 51373
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.668141592920354e-07,
+      "loss": 2.0866,
+      "step": 51374
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.667667509481669e-07,
+      "loss": 1.5813,
+      "step": 51375
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.667193426042984e-07,
+      "loss": 1.4455,
+      "step": 51376
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.666719342604299e-07,
+      "loss": 2.7348,
+      "step": 51377
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.666245259165613e-07,
+      "loss": 2.0014,
+      "step": 51378
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.665771175726929e-07,
+      "loss": 1.5157,
+      "step": 51379
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.665297092288243e-07,
+      "loss": 1.8572,
+      "step": 51380
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.664823008849557e-07,
+      "loss": 2.3703,
+      "step": 51381
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.664348925410872e-07,
+      "loss": 2.0488,
+      "step": 51382
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.663874841972187e-07,
+      "loss": 1.7806,
+      "step": 51383
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.663400758533502e-07,
+      "loss": 2.2238,
+      "step": 51384
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.662926675094817e-07,
+      "loss": 1.6811,
+      "step": 51385
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.662452591656131e-07,
+      "loss": 2.5094,
+      "step": 51386
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.661978508217447e-07,
+      "loss": 2.0949,
+      "step": 51387
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.661504424778761e-07,
+      "loss": 1.6246,
+      "step": 51388
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.661030341340076e-07,
+      "loss": 2.0063,
+      "step": 51389
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.660556257901392e-07,
+      "loss": 3.1996,
+      "step": 51390
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.660082174462705e-07,
+      "loss": 1.6041,
+      "step": 51391
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 5.65960809102402e-07,
+      "loss": 0.9836,
+      "step": 51392
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.659134007585335e-07,
+      "loss": 0.656,
+      "step": 51393
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.65865992414665e-07,
+      "loss": 1.8462,
+      "step": 51394
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.658185840707965e-07,
+      "loss": 1.7601,
+      "step": 51395
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.657711757269279e-07,
+      "loss": 2.2791,
+      "step": 51396
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.657237673830594e-07,
+      "loss": 1.0254,
+      "step": 51397
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.656763590391909e-07,
+      "loss": 1.3409,
+      "step": 51398
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.656289506953224e-07,
+      "loss": 2.1701,
+      "step": 51399
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.655815423514539e-07,
+      "loss": 1.5435,
+      "step": 51400
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.655341340075854e-07,
+      "loss": 0.4234,
+      "step": 51401
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.654867256637169e-07,
+      "loss": 1.1497,
+      "step": 51402
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.654393173198483e-07,
+      "loss": 1.3861,
+      "step": 51403
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.653919089759797e-07,
+      "loss": 2.2261,
+      "step": 51404
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.653445006321113e-07,
+      "loss": 0.2495,
+      "step": 51405
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.652970922882427e-07,
+      "loss": 2.4495,
+      "step": 51406
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.652496839443742e-07,
+      "loss": 2.9193,
+      "step": 51407
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.652022756005057e-07,
+      "loss": 2.3579,
+      "step": 51408
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.651548672566372e-07,
+      "loss": 1.8362,
+      "step": 51409
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.651074589127687e-07,
+      "loss": 1.3343,
+      "step": 51410
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.650600505689001e-07,
+      "loss": 2.5701,
+      "step": 51411
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.650126422250317e-07,
+      "loss": 2.4146,
+      "step": 51412
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.649652338811632e-07,
+      "loss": 2.7681,
+      "step": 51413
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.649178255372945e-07,
+      "loss": 1.2345,
+      "step": 51414
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.64870417193426e-07,
+      "loss": 2.4491,
+      "step": 51415
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.648230088495575e-07,
+      "loss": 2.2447,
+      "step": 51416
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.64775600505689e-07,
+      "loss": 3.4937,
+      "step": 51417
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.647281921618205e-07,
+      "loss": 2.1854,
+      "step": 51418
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.646807838179519e-07,
+      "loss": 0.9929,
+      "step": 51419
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.646333754740835e-07,
+      "loss": 1.4876,
+      "step": 51420
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.645859671302149e-07,
+      "loss": 1.2556,
+      "step": 51421
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.645385587863464e-07,
+      "loss": 1.256,
+      "step": 51422
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.64491150442478e-07,
+      "loss": 2.1981,
+      "step": 51423
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.644437420986094e-07,
+      "loss": 2.6799,
+      "step": 51424
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.643963337547409e-07,
+      "loss": 1.012,
+      "step": 51425
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.643489254108723e-07,
+      "loss": 0.4594,
+      "step": 51426
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.643015170670038e-07,
+      "loss": 1.3549,
+      "step": 51427
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.642541087231353e-07,
+      "loss": 2.5092,
+      "step": 51428
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.642067003792667e-07,
+      "loss": 1.3073,
+      "step": 51429
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.641592920353982e-07,
+      "loss": 0.6639,
+      "step": 51430
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.641118836915297e-07,
+      "loss": 2.4873,
+      "step": 51431
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.640644753476612e-07,
+      "loss": 2.0837,
+      "step": 51432
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.640170670037927e-07,
+      "loss": 2.3795,
+      "step": 51433
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.639696586599242e-07,
+      "loss": 1.5435,
+      "step": 51434
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.639222503160557e-07,
+      "loss": 2.2997,
+      "step": 51435
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.638748419721871e-07,
+      "loss": 1.3269,
+      "step": 51436
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.638274336283186e-07,
+      "loss": 1.4332,
+      "step": 51437
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.637800252844501e-07,
+      "loss": 2.3245,
+      "step": 51438
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.637326169405815e-07,
+      "loss": 1.6036,
+      "step": 51439
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.63685208596713e-07,
+      "loss": 2.6653,
+      "step": 51440
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.636378002528445e-07,
+      "loss": 1.9186,
+      "step": 51441
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.63590391908976e-07,
+      "loss": 0.3206,
+      "step": 51442
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.635429835651075e-07,
+      "loss": 3.4066,
+      "step": 51443
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.634955752212389e-07,
+      "loss": 3.2405,
+      "step": 51444
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.634481668773705e-07,
+      "loss": 1.5405,
+      "step": 51445
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.634007585335019e-07,
+      "loss": 1.8195,
+      "step": 51446
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.633533501896334e-07,
+      "loss": 1.0419,
+      "step": 51447
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.633059418457649e-07,
+      "loss": 2.4257,
+      "step": 51448
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.632585335018964e-07,
+      "loss": 0.7234,
+      "step": 51449
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.632111251580278e-07,
+      "loss": 0.7596,
+      "step": 51450
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.631637168141593e-07,
+      "loss": 2.3694,
+      "step": 51451
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.631163084702907e-07,
+      "loss": 1.949,
+      "step": 51452
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.630689001264223e-07,
+      "loss": 2.1032,
+      "step": 51453
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.630214917825537e-07,
+      "loss": 2.0298,
+      "step": 51454
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 5.629740834386852e-07,
+      "loss": 1.8764,
+      "step": 51455
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.629266750948167e-07,
+      "loss": 2.9866,
+      "step": 51456
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.628792667509482e-07,
+      "loss": 0.7804,
+      "step": 51457
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.628318584070797e-07,
+      "loss": 3.0946,
+      "step": 51458
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.627844500632111e-07,
+      "loss": 1.2086,
+      "step": 51459
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.627370417193427e-07,
+      "loss": 2.4236,
+      "step": 51460
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.626896333754742e-07,
+      "loss": 2.5496,
+      "step": 51461
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.626422250316055e-07,
+      "loss": 1.9521,
+      "step": 51462
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.62594816687737e-07,
+      "loss": 1.9549,
+      "step": 51463
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.625474083438685e-07,
+      "loss": 1.8385,
+      "step": 51464
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.625e-07,
+      "loss": 1.9505,
+      "step": 51465
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.624525916561315e-07,
+      "loss": 2.11,
+      "step": 51466
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.62405183312263e-07,
+      "loss": 0.7823,
+      "step": 51467
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.623577749683945e-07,
+      "loss": 2.4306,
+      "step": 51468
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.623103666245259e-07,
+      "loss": 2.5952,
+      "step": 51469
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.622629582806574e-07,
+      "loss": 2.1094,
+      "step": 51470
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.62215549936789e-07,
+      "loss": 0.4454,
+      "step": 51471
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.621681415929204e-07,
+      "loss": 1.122,
+      "step": 51472
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.621207332490518e-07,
+      "loss": 3.3617,
+      "step": 51473
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.620733249051833e-07,
+      "loss": 2.4066,
+      "step": 51474
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.620259165613148e-07,
+      "loss": 1.6824,
+      "step": 51475
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.619785082174463e-07,
+      "loss": 1.0638,
+      "step": 51476
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.619310998735777e-07,
+      "loss": 1.7153,
+      "step": 51477
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.618836915297093e-07,
+      "loss": 1.4927,
+      "step": 51478
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.618362831858407e-07,
+      "loss": 2.3984,
+      "step": 51479
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.617888748419722e-07,
+      "loss": 2.1479,
+      "step": 51480
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.617414664981037e-07,
+      "loss": 1.71,
+      "step": 51481
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.616940581542352e-07,
+      "loss": 3.2959,
+      "step": 51482
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.616466498103667e-07,
+      "loss": 0.6417,
+      "step": 51483
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.615992414664982e-07,
+      "loss": 2.5885,
+      "step": 51484
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.615518331226295e-07,
+      "loss": 1.9265,
+      "step": 51485
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.615044247787611e-07,
+      "loss": 1.2003,
+      "step": 51486
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.614570164348925e-07,
+      "loss": 2.257,
+      "step": 51487
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.61409608091024e-07,
+      "loss": 1.5866,
+      "step": 51488
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.613621997471555e-07,
+      "loss": 2.2813,
+      "step": 51489
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.61314791403287e-07,
+      "loss": 0.7606,
+      "step": 51490
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.612673830594185e-07,
+      "loss": 0.7533,
+      "step": 51491
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.612199747155499e-07,
+      "loss": 2.002,
+      "step": 51492
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.611725663716815e-07,
+      "loss": 2.0604,
+      "step": 51493
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.61125158027813e-07,
+      "loss": 2.4364,
+      "step": 51494
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.610777496839444e-07,
+      "loss": 1.9558,
+      "step": 51495
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.610303413400758e-07,
+      "loss": 2.2789,
+      "step": 51496
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.609829329962073e-07,
+      "loss": 2.4159,
+      "step": 51497
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.609355246523388e-07,
+      "loss": 1.9591,
+      "step": 51498
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.608881163084703e-07,
+      "loss": 1.9212,
+      "step": 51499
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.608407079646018e-07,
+      "loss": 3.0621,
+      "step": 51500
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.607932996207333e-07,
+      "loss": 1.2183,
+      "step": 51501
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.607458912768647e-07,
+      "loss": 1.3782,
+      "step": 51502
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.606984829329962e-07,
+      "loss": 2.2245,
+      "step": 51503
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.606510745891277e-07,
+      "loss": 2.1646,
+      "step": 51504
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.606036662452592e-07,
+      "loss": 1.2225,
+      "step": 51505
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.605562579013907e-07,
+      "loss": 1.3031,
+      "step": 51506
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.60508849557522e-07,
+      "loss": 2.1093,
+      "step": 51507
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.604614412136536e-07,
+      "loss": 1.2014,
+      "step": 51508
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.604140328697851e-07,
+      "loss": 2.8594,
+      "step": 51509
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.603666245259165e-07,
+      "loss": 3.7171,
+      "step": 51510
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.603192161820481e-07,
+      "loss": 1.4779,
+      "step": 51511
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.602718078381795e-07,
+      "loss": 2.1907,
+      "step": 51512
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.60224399494311e-07,
+      "loss": 2.5522,
+      "step": 51513
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.601769911504425e-07,
+      "loss": 2.6441,
+      "step": 51514
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.60129582806574e-07,
+      "loss": 2.1468,
+      "step": 51515
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.600821744627055e-07,
+      "loss": 3.295,
+      "step": 51516
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.600347661188369e-07,
+      "loss": 1.3316,
+      "step": 51517
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 5.599873577749684e-07,
+      "loss": 3.2308,
+      "step": 51518
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.599399494311e-07,
+      "loss": 0.8304,
+      "step": 51519
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.598925410872313e-07,
+      "loss": 1.1859,
+      "step": 51520
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.598451327433628e-07,
+      "loss": 0.3242,
+      "step": 51521
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.597977243994943e-07,
+      "loss": 1.7921,
+      "step": 51522
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.597503160556258e-07,
+      "loss": 2.5524,
+      "step": 51523
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.597029077117573e-07,
+      "loss": 1.8686,
+      "step": 51524
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.596554993678887e-07,
+      "loss": 2.1025,
+      "step": 51525
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.596080910240203e-07,
+      "loss": 2.6405,
+      "step": 51526
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.595606826801517e-07,
+      "loss": 1.1863,
+      "step": 51527
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.595132743362832e-07,
+      "loss": 2.1516,
+      "step": 51528
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.594658659924147e-07,
+      "loss": 2.8183,
+      "step": 51529
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.594184576485462e-07,
+      "loss": 1.4788,
+      "step": 51530
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.593710493046777e-07,
+      "loss": 2.8593,
+      "step": 51531
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.593236409608091e-07,
+      "loss": 2.1133,
+      "step": 51532
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.592762326169406e-07,
+      "loss": 1.7415,
+      "step": 51533
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.592288242730721e-07,
+      "loss": 0.7633,
+      "step": 51534
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.591814159292035e-07,
+      "loss": 1.529,
+      "step": 51535
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.59134007585335e-07,
+      "loss": 1.1305,
+      "step": 51536
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.590865992414665e-07,
+      "loss": 1.3886,
+      "step": 51537
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.59039190897598e-07,
+      "loss": 3.4893,
+      "step": 51538
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.589917825537295e-07,
+      "loss": 1.8156,
+      "step": 51539
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.589443742098609e-07,
+      "loss": 0.8882,
+      "step": 51540
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.588969658659925e-07,
+      "loss": 1.1829,
+      "step": 51541
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.58849557522124e-07,
+      "loss": 1.3062,
+      "step": 51542
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.588021491782553e-07,
+      "loss": 1.5417,
+      "step": 51543
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.587547408343869e-07,
+      "loss": 0.8172,
+      "step": 51544
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.587073324905183e-07,
+      "loss": 1.1638,
+      "step": 51545
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.586599241466498e-07,
+      "loss": 2.0638,
+      "step": 51546
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.586125158027813e-07,
+      "loss": 1.7142,
+      "step": 51547
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.585651074589128e-07,
+      "loss": 2.9712,
+      "step": 51548
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.585176991150443e-07,
+      "loss": 1.0721,
+      "step": 51549
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.584702907711757e-07,
+      "loss": 1.4963,
+      "step": 51550
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.584228824273072e-07,
+      "loss": 0.8411,
+      "step": 51551
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.583754740834388e-07,
+      "loss": 2.1644,
+      "step": 51552
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.583280657395702e-07,
+      "loss": 2.6787,
+      "step": 51553
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.582806573957017e-07,
+      "loss": 1.787,
+      "step": 51554
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.582332490518331e-07,
+      "loss": 3.8442,
+      "step": 51555
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.581858407079646e-07,
+      "loss": 0.3751,
+      "step": 51556
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.581384323640961e-07,
+      "loss": 1.2179,
+      "step": 51557
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.580910240202275e-07,
+      "loss": 1.4199,
+      "step": 51558
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.580436156763591e-07,
+      "loss": 3.0047,
+      "step": 51559
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.579962073324905e-07,
+      "loss": 4.2466,
+      "step": 51560
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.57948798988622e-07,
+      "loss": 0.9935,
+      "step": 51561
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.579013906447535e-07,
+      "loss": 2.6184,
+      "step": 51562
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.57853982300885e-07,
+      "loss": 2.1757,
+      "step": 51563
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.578065739570165e-07,
+      "loss": 2.4728,
+      "step": 51564
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.57759165613148e-07,
+      "loss": 2.3123,
+      "step": 51565
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.577117572692795e-07,
+      "loss": 2.9686,
+      "step": 51566
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.576643489254109e-07,
+      "loss": 2.0659,
+      "step": 51567
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.576169405815423e-07,
+      "loss": 2.7881,
+      "step": 51568
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.575695322376738e-07,
+      "loss": 2.8319,
+      "step": 51569
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.575221238938053e-07,
+      "loss": 1.9463,
+      "step": 51570
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.574747155499368e-07,
+      "loss": 1.7624,
+      "step": 51571
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.574273072060683e-07,
+      "loss": 3.1795,
+      "step": 51572
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.573798988621997e-07,
+      "loss": 2.4032,
+      "step": 51573
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.573324905183313e-07,
+      "loss": 1.4475,
+      "step": 51574
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.572850821744627e-07,
+      "loss": 1.8897,
+      "step": 51575
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.572376738305942e-07,
+      "loss": 1.0399,
+      "step": 51576
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.571902654867257e-07,
+      "loss": 1.7592,
+      "step": 51577
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.571428571428571e-07,
+      "loss": 2.5808,
+      "step": 51578
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.570954487989886e-07,
+      "loss": 1.7157,
+      "step": 51579
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.570480404551201e-07,
+      "loss": 1.2469,
+      "step": 51580
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.570006321112516e-07,
+      "loss": 1.8869,
+      "step": 51581
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 5.569532237673831e-07,
+      "loss": 1.0722,
+      "step": 51582
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.569058154235145e-07,
+      "loss": 1.5114,
+      "step": 51583
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.56858407079646e-07,
+      "loss": 1.7592,
+      "step": 51584
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.568109987357775e-07,
+      "loss": 2.3672,
+      "step": 51585
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.56763590391909e-07,
+      "loss": 1.7816,
+      "step": 51586
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.567161820480405e-07,
+      "loss": 3.444,
+      "step": 51587
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.566687737041719e-07,
+      "loss": 1.542,
+      "step": 51588
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.566213653603035e-07,
+      "loss": 0.615,
+      "step": 51589
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.56573957016435e-07,
+      "loss": 3.4475,
+      "step": 51590
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.565265486725663e-07,
+      "loss": 2.4953,
+      "step": 51591
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.564791403286979e-07,
+      "loss": 2.7011,
+      "step": 51592
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.564317319848293e-07,
+      "loss": 1.7611,
+      "step": 51593
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.563843236409608e-07,
+      "loss": 2.9384,
+      "step": 51594
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.563369152970923e-07,
+      "loss": 2.1222,
+      "step": 51595
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.562895069532238e-07,
+      "loss": 1.3102,
+      "step": 51596
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.562420986093553e-07,
+      "loss": 2.1493,
+      "step": 51597
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.561946902654867e-07,
+      "loss": 1.5094,
+      "step": 51598
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.561472819216182e-07,
+      "loss": 2.0431,
+      "step": 51599
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.560998735777498e-07,
+      "loss": 1.9581,
+      "step": 51600
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.560524652338812e-07,
+      "loss": 1.8035,
+      "step": 51601
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.560050568900126e-07,
+      "loss": 2.9923,
+      "step": 51602
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.559576485461441e-07,
+      "loss": 2.0219,
+      "step": 51603
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.559102402022756e-07,
+      "loss": 0.8111,
+      "step": 51604
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.558628318584071e-07,
+      "loss": 2.1575,
+      "step": 51605
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.558154235145385e-07,
+      "loss": 0.5612,
+      "step": 51606
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.557680151706701e-07,
+      "loss": 1.0304,
+      "step": 51607
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.557206068268015e-07,
+      "loss": 1.8912,
+      "step": 51608
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.55673198482933e-07,
+      "loss": 2.4586,
+      "step": 51609
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.556257901390645e-07,
+      "loss": 2.6261,
+      "step": 51610
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.55578381795196e-07,
+      "loss": 3.4243,
+      "step": 51611
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.555309734513275e-07,
+      "loss": 2.2915,
+      "step": 51612
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.55483565107459e-07,
+      "loss": 1.3973,
+      "step": 51613
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.554361567635904e-07,
+      "loss": 1.953,
+      "step": 51614
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.553887484197219e-07,
+      "loss": 3.4751,
+      "step": 51615
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.553413400758533e-07,
+      "loss": 2.9816,
+      "step": 51616
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.552939317319848e-07,
+      "loss": 1.9119,
+      "step": 51617
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.552465233881163e-07,
+      "loss": 1.8592,
+      "step": 51618
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.551991150442478e-07,
+      "loss": 1.5625,
+      "step": 51619
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.551517067003793e-07,
+      "loss": 1.1594,
+      "step": 51620
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.551042983565107e-07,
+      "loss": 1.3286,
+      "step": 51621
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.550568900126423e-07,
+      "loss": 2.0885,
+      "step": 51622
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.550094816687738e-07,
+      "loss": 2.9234,
+      "step": 51623
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.549620733249052e-07,
+      "loss": 2.532,
+      "step": 51624
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.549146649810368e-07,
+      "loss": 1.4627,
+      "step": 51625
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.548672566371681e-07,
+      "loss": 3.0791,
+      "step": 51626
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.548198482932996e-07,
+      "loss": 1.1564,
+      "step": 51627
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.547724399494311e-07,
+      "loss": 2.318,
+      "step": 51628
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.547250316055626e-07,
+      "loss": 1.0853,
+      "step": 51629
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.546776232616941e-07,
+      "loss": 1.4349,
+      "step": 51630
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.546302149178255e-07,
+      "loss": 1.627,
+      "step": 51631
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.54582806573957e-07,
+      "loss": 2.2074,
+      "step": 51632
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.545353982300886e-07,
+      "loss": 1.9901,
+      "step": 51633
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.5448798988622e-07,
+      "loss": 1.2568,
+      "step": 51634
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.544405815423515e-07,
+      "loss": 1.0322,
+      "step": 51635
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.54393173198483e-07,
+      "loss": 1.8026,
+      "step": 51636
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.543457648546144e-07,
+      "loss": 2.9789,
+      "step": 51637
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.542983565107459e-07,
+      "loss": 2.9961,
+      "step": 51638
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.542509481668773e-07,
+      "loss": 1.9308,
+      "step": 51639
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.542035398230089e-07,
+      "loss": 2.3725,
+      "step": 51640
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.541561314791403e-07,
+      "loss": 2.2227,
+      "step": 51641
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.541087231352718e-07,
+      "loss": 0.7484,
+      "step": 51642
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.540613147914033e-07,
+      "loss": 2.3762,
+      "step": 51643
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.540139064475348e-07,
+      "loss": 2.4789,
+      "step": 51644
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 5.539664981036663e-07,
+      "loss": 1.3101,
+      "step": 51645
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.539190897597977e-07,
+      "loss": 2.8607,
+      "step": 51646
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.538716814159293e-07,
+      "loss": 2.9891,
+      "step": 51647
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.538242730720608e-07,
+      "loss": 1.4506,
+      "step": 51648
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.537768647281921e-07,
+      "loss": 0.5187,
+      "step": 51649
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.537294563843236e-07,
+      "loss": 1.2542,
+      "step": 51650
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.536820480404551e-07,
+      "loss": 1.5444,
+      "step": 51651
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.536346396965866e-07,
+      "loss": 0.891,
+      "step": 51652
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.535872313527181e-07,
+      "loss": 0.1903,
+      "step": 51653
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.535398230088495e-07,
+      "loss": 0.7505,
+      "step": 51654
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.534924146649811e-07,
+      "loss": 1.7036,
+      "step": 51655
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.534450063211125e-07,
+      "loss": 1.2741,
+      "step": 51656
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.53397597977244e-07,
+      "loss": 2.6844,
+      "step": 51657
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.533501896333756e-07,
+      "loss": 1.7725,
+      "step": 51658
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.53302781289507e-07,
+      "loss": 2.6342,
+      "step": 51659
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.532553729456384e-07,
+      "loss": 2.8141,
+      "step": 51660
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.532079646017699e-07,
+      "loss": 0.7536,
+      "step": 51661
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.531605562579014e-07,
+      "loss": 0.4595,
+      "step": 51662
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.531131479140329e-07,
+      "loss": 1.4838,
+      "step": 51663
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.530657395701643e-07,
+      "loss": 1.2354,
+      "step": 51664
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.530183312262958e-07,
+      "loss": 1.921,
+      "step": 51665
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.529709228824273e-07,
+      "loss": 3.9875,
+      "step": 51666
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.529235145385588e-07,
+      "loss": 3.0894,
+      "step": 51667
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.528761061946903e-07,
+      "loss": 3.998,
+      "step": 51668
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.528286978508218e-07,
+      "loss": 1.962,
+      "step": 51669
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.527812895069533e-07,
+      "loss": 1.5336,
+      "step": 51670
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.527338811630848e-07,
+      "loss": 2.4587,
+      "step": 51671
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.526864728192161e-07,
+      "loss": 0.662,
+      "step": 51672
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.526390644753477e-07,
+      "loss": 1.5528,
+      "step": 51673
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.525916561314791e-07,
+      "loss": 2.0134,
+      "step": 51674
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.525442477876106e-07,
+      "loss": 2.9732,
+      "step": 51675
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.524968394437421e-07,
+      "loss": 1.1131,
+      "step": 51676
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.524494310998736e-07,
+      "loss": 1.3982,
+      "step": 51677
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.524020227560051e-07,
+      "loss": 1.6282,
+      "step": 51678
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.523546144121365e-07,
+      "loss": 2.4265,
+      "step": 51679
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.523072060682681e-07,
+      "loss": 1.9286,
+      "step": 51680
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.522597977243996e-07,
+      "loss": 2.735,
+      "step": 51681
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.52212389380531e-07,
+      "loss": 3.2422,
+      "step": 51682
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.521649810366625e-07,
+      "loss": 1.5697,
+      "step": 51683
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.521175726927939e-07,
+      "loss": 1.6781,
+      "step": 51684
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.520701643489254e-07,
+      "loss": 1.1193,
+      "step": 51685
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.520227560050569e-07,
+      "loss": 1.2581,
+      "step": 51686
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.519753476611883e-07,
+      "loss": 3.1108,
+      "step": 51687
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.519279393173199e-07,
+      "loss": 0.8816,
+      "step": 51688
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.518805309734513e-07,
+      "loss": 2.0982,
+      "step": 51689
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.518331226295828e-07,
+      "loss": 1.5352,
+      "step": 51690
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.517857142857144e-07,
+      "loss": 0.8668,
+      "step": 51691
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.517383059418458e-07,
+      "loss": 2.3119,
+      "step": 51692
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.516908975979773e-07,
+      "loss": 1.0155,
+      "step": 51693
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.516434892541088e-07,
+      "loss": 2.5963,
+      "step": 51694
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.515960809102403e-07,
+      "loss": 1.735,
+      "step": 51695
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.515486725663717e-07,
+      "loss": 1.4297,
+      "step": 51696
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.515012642225031e-07,
+      "loss": 3.1145,
+      "step": 51697
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.514538558786346e-07,
+      "loss": 1.3209,
+      "step": 51698
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.514064475347661e-07,
+      "loss": 1.5554,
+      "step": 51699
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.513590391908976e-07,
+      "loss": 2.0102,
+      "step": 51700
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.513116308470291e-07,
+      "loss": 1.2789,
+      "step": 51701
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.512642225031606e-07,
+      "loss": 0.7714,
+      "step": 51702
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.512168141592921e-07,
+      "loss": 1.7742,
+      "step": 51703
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.511694058154236e-07,
+      "loss": 1.4228,
+      "step": 51704
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.51121997471555e-07,
+      "loss": 2.3269,
+      "step": 51705
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.510745891276866e-07,
+      "loss": 2.9573,
+      "step": 51706
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.510271807838179e-07,
+      "loss": 1.5674,
+      "step": 51707
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 5.509797724399494e-07,
+      "loss": 0.8171,
+      "step": 51708
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.509323640960809e-07,
+      "loss": 2.3267,
+      "step": 51709
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.508849557522124e-07,
+      "loss": 1.7708,
+      "step": 51710
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.508375474083439e-07,
+      "loss": 1.0555,
+      "step": 51711
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.507901390644753e-07,
+      "loss": 1.2625,
+      "step": 51712
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.507427307206069e-07,
+      "loss": 2.2418,
+      "step": 51713
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.506953223767383e-07,
+      "loss": 2.0764,
+      "step": 51714
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.506479140328698e-07,
+      "loss": 2.3302,
+      "step": 51715
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.506005056890013e-07,
+      "loss": 2.2767,
+      "step": 51716
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.505530973451328e-07,
+      "loss": 2.2851,
+      "step": 51717
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.505056890012643e-07,
+      "loss": 1.7501,
+      "step": 51718
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.504582806573957e-07,
+      "loss": 1.4374,
+      "step": 51719
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.504108723135271e-07,
+      "loss": 1.0045,
+      "step": 51720
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.503634639696587e-07,
+      "loss": 1.3361,
+      "step": 51721
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.503160556257901e-07,
+      "loss": 1.6828,
+      "step": 51722
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.502686472819216e-07,
+      "loss": 1.3879,
+      "step": 51723
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.502212389380531e-07,
+      "loss": 0.7205,
+      "step": 51724
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.501738305941846e-07,
+      "loss": 2.1064,
+      "step": 51725
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.501264222503161e-07,
+      "loss": 2.6503,
+      "step": 51726
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.500790139064475e-07,
+      "loss": 3.7782,
+      "step": 51727
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.500316055625791e-07,
+      "loss": 2.3966,
+      "step": 51728
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.499841972187106e-07,
+      "loss": 2.1233,
+      "step": 51729
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.499367888748419e-07,
+      "loss": 2.1113,
+      "step": 51730
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.498893805309734e-07,
+      "loss": 1.2511,
+      "step": 51731
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.498419721871049e-07,
+      "loss": 1.6139,
+      "step": 51732
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.497945638432364e-07,
+      "loss": 0.6682,
+      "step": 51733
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.497471554993679e-07,
+      "loss": 2.2096,
+      "step": 51734
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.496997471554994e-07,
+      "loss": 2.8386,
+      "step": 51735
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.496523388116309e-07,
+      "loss": 2.4707,
+      "step": 51736
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.496049304677623e-07,
+      "loss": 2.7536,
+      "step": 51737
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.495575221238938e-07,
+      "loss": 3.5915,
+      "step": 51738
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.495101137800254e-07,
+      "loss": 0.7907,
+      "step": 51739
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.494627054361568e-07,
+      "loss": 1.758,
+      "step": 51740
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.494152970922883e-07,
+      "loss": 2.2011,
+      "step": 51741
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.493678887484197e-07,
+      "loss": 2.2332,
+      "step": 51742
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.493204804045512e-07,
+      "loss": 3.2124,
+      "step": 51743
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.492730720606827e-07,
+      "loss": 1.039,
+      "step": 51744
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.492256637168141e-07,
+      "loss": 2.8641,
+      "step": 51745
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.491782553729457e-07,
+      "loss": 2.7929,
+      "step": 51746
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.491308470290771e-07,
+      "loss": 0.1556,
+      "step": 51747
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.490834386852086e-07,
+      "loss": 2.5518,
+      "step": 51748
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.490360303413401e-07,
+      "loss": 1.4958,
+      "step": 51749
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.489886219974716e-07,
+      "loss": 1.5557,
+      "step": 51750
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.489412136536031e-07,
+      "loss": 4.4447,
+      "step": 51751
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.488938053097346e-07,
+      "loss": 3.1683,
+      "step": 51752
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.48846396965866e-07,
+      "loss": 2.4545,
+      "step": 51753
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.487989886219975e-07,
+      "loss": 0.9441,
+      "step": 51754
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.487515802781289e-07,
+      "loss": 2.2236,
+      "step": 51755
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.487041719342604e-07,
+      "loss": 1.5639,
+      "step": 51756
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.486567635903919e-07,
+      "loss": 1.2102,
+      "step": 51757
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.486093552465234e-07,
+      "loss": 1.8303,
+      "step": 51758
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.485619469026549e-07,
+      "loss": 2.7699,
+      "step": 51759
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.485145385587863e-07,
+      "loss": 2.7907,
+      "step": 51760
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.484671302149179e-07,
+      "loss": 1.7775,
+      "step": 51761
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.484197218710494e-07,
+      "loss": 2.7913,
+      "step": 51762
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.483723135271808e-07,
+      "loss": 2.0725,
+      "step": 51763
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.483249051833123e-07,
+      "loss": 2.1636,
+      "step": 51764
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.482774968394437e-07,
+      "loss": 1.297,
+      "step": 51765
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.482300884955752e-07,
+      "loss": 1.5963,
+      "step": 51766
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.481826801517067e-07,
+      "loss": 2.2785,
+      "step": 51767
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.481352718078382e-07,
+      "loss": 0.9939,
+      "step": 51768
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.480878634639697e-07,
+      "loss": 2.024,
+      "step": 51769
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.480404551201011e-07,
+      "loss": 1.4439,
+      "step": 51770
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.479930467762326e-07,
+      "loss": 2.8095,
+      "step": 51771
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 5.479456384323642e-07,
+      "loss": 1.8951,
+      "step": 51772
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.478982300884956e-07,
+      "loss": 1.2927,
+      "step": 51773
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.478508217446271e-07,
+      "loss": 1.5373,
+      "step": 51774
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.478034134007585e-07,
+      "loss": 0.4216,
+      "step": 51775
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.477560050568901e-07,
+      "loss": 1.3438,
+      "step": 51776
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.477085967130215e-07,
+      "loss": 1.5673,
+      "step": 51777
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.476611883691529e-07,
+      "loss": 2.5117,
+      "step": 51778
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.476137800252845e-07,
+      "loss": 2.5836,
+      "step": 51779
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.475663716814159e-07,
+      "loss": 1.4072,
+      "step": 51780
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.475189633375474e-07,
+      "loss": 1.6001,
+      "step": 51781
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.474715549936789e-07,
+      "loss": 0.9914,
+      "step": 51782
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.474241466498104e-07,
+      "loss": 0.9377,
+      "step": 51783
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.473767383059419e-07,
+      "loss": 1.9927,
+      "step": 51784
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.473293299620733e-07,
+      "loss": 2.4103,
+      "step": 51785
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.472819216182048e-07,
+      "loss": 3.1808,
+      "step": 51786
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.472345132743364e-07,
+      "loss": 2.2934,
+      "step": 51787
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.471871049304678e-07,
+      "loss": 1.9237,
+      "step": 51788
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.471396965865992e-07,
+      "loss": 1.8597,
+      "step": 51789
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.470922882427307e-07,
+      "loss": 3.3098,
+      "step": 51790
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.470448798988622e-07,
+      "loss": 2.1546,
+      "step": 51791
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.469974715549937e-07,
+      "loss": 1.298,
+      "step": 51792
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.469500632111251e-07,
+      "loss": 0.9698,
+      "step": 51793
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.469026548672567e-07,
+      "loss": 1.9752,
+      "step": 51794
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.468552465233881e-07,
+      "loss": 1.629,
+      "step": 51795
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.468078381795196e-07,
+      "loss": 1.8731,
+      "step": 51796
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.467604298356511e-07,
+      "loss": 2.3929,
+      "step": 51797
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.467130214917826e-07,
+      "loss": 1.7756,
+      "step": 51798
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.466656131479141e-07,
+      "loss": 1.0588,
+      "step": 51799
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.466182048040456e-07,
+      "loss": 1.9559,
+      "step": 51800
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.465707964601769e-07,
+      "loss": 1.824,
+      "step": 51801
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.465233881163085e-07,
+      "loss": 3.2867,
+      "step": 51802
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.464759797724399e-07,
+      "loss": 2.0231,
+      "step": 51803
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.464285714285714e-07,
+      "loss": 1.6577,
+      "step": 51804
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.463811630847029e-07,
+      "loss": 1.0589,
+      "step": 51805
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.463337547408344e-07,
+      "loss": 1.1033,
+      "step": 51806
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.462863463969659e-07,
+      "loss": 2.0598,
+      "step": 51807
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.462389380530973e-07,
+      "loss": 2.6494,
+      "step": 51808
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.461915297092289e-07,
+      "loss": 1.9072,
+      "step": 51809
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.461441213653604e-07,
+      "loss": 1.4758,
+      "step": 51810
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.460967130214918e-07,
+      "loss": 1.664,
+      "step": 51811
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.460493046776232e-07,
+      "loss": 1.3813,
+      "step": 51812
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.460018963337547e-07,
+      "loss": 2.0492,
+      "step": 51813
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.459544879898862e-07,
+      "loss": 2.7276,
+      "step": 51814
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.459070796460177e-07,
+      "loss": 1.6951,
+      "step": 51815
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.458596713021492e-07,
+      "loss": 2.9604,
+      "step": 51816
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.458122629582807e-07,
+      "loss": 1.5855,
+      "step": 51817
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.457648546144121e-07,
+      "loss": 2.5917,
+      "step": 51818
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.457174462705436e-07,
+      "loss": 2.2658,
+      "step": 51819
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.456700379266752e-07,
+      "loss": 2.1664,
+      "step": 51820
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.456226295828066e-07,
+      "loss": 1.6591,
+      "step": 51821
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.455752212389381e-07,
+      "loss": 0.3798,
+      "step": 51822
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.455278128950696e-07,
+      "loss": 2.2573,
+      "step": 51823
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.45480404551201e-07,
+      "loss": 1.1831,
+      "step": 51824
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.454329962073325e-07,
+      "loss": 1.9434,
+      "step": 51825
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.453855878634639e-07,
+      "loss": 1.0388,
+      "step": 51826
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.453381795195955e-07,
+      "loss": 1.7795,
+      "step": 51827
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.452907711757269e-07,
+      "loss": 2.4633,
+      "step": 51828
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.452433628318584e-07,
+      "loss": 1.7232,
+      "step": 51829
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.451959544879899e-07,
+      "loss": 2.0953,
+      "step": 51830
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.451485461441214e-07,
+      "loss": 2.1241,
+      "step": 51831
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.451011378002529e-07,
+      "loss": 0.8827,
+      "step": 51832
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.450537294563844e-07,
+      "loss": 1.6662,
+      "step": 51833
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.450063211125158e-07,
+      "loss": 0.7278,
+      "step": 51834
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 5.449589127686474e-07,
+      "loss": 1.1002,
+      "step": 51835
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.449115044247787e-07,
+      "loss": 2.2223,
+      "step": 51836
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.448640960809102e-07,
+      "loss": 1.1338,
+      "step": 51837
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.448166877370417e-07,
+      "loss": 1.5408,
+      "step": 51838
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.447692793931732e-07,
+      "loss": 1.8393,
+      "step": 51839
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.447218710493047e-07,
+      "loss": 3.0177,
+      "step": 51840
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.446744627054361e-07,
+      "loss": 1.9933,
+      "step": 51841
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.446270543615677e-07,
+      "loss": 1.5179,
+      "step": 51842
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.445796460176992e-07,
+      "loss": 2.0,
+      "step": 51843
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.445322376738306e-07,
+      "loss": 1.5055,
+      "step": 51844
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.444848293299621e-07,
+      "loss": 2.6683,
+      "step": 51845
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.444374209860936e-07,
+      "loss": 1.3454,
+      "step": 51846
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.44390012642225e-07,
+      "loss": 0.9267,
+      "step": 51847
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.443426042983565e-07,
+      "loss": 0.5398,
+      "step": 51848
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.44295195954488e-07,
+      "loss": 1.3209,
+      "step": 51849
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.442477876106195e-07,
+      "loss": 1.2208,
+      "step": 51850
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.442003792667509e-07,
+      "loss": 3.0032,
+      "step": 51851
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.441529709228824e-07,
+      "loss": 2.7681,
+      "step": 51852
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.441055625790139e-07,
+      "loss": 4.3359,
+      "step": 51853
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.440581542351454e-07,
+      "loss": 1.2235,
+      "step": 51854
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.440107458912769e-07,
+      "loss": 2.3441,
+      "step": 51855
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.439633375474083e-07,
+      "loss": 2.6305,
+      "step": 51856
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.439159292035399e-07,
+      "loss": 1.1572,
+      "step": 51857
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.438685208596714e-07,
+      "loss": 1.4451,
+      "step": 51858
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.438211125158027e-07,
+      "loss": 1.8676,
+      "step": 51859
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.437737041719343e-07,
+      "loss": 2.7323,
+      "step": 51860
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.437262958280657e-07,
+      "loss": 0.968,
+      "step": 51861
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.436788874841972e-07,
+      "loss": 2.2948,
+      "step": 51862
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.436314791403287e-07,
+      "loss": 1.4175,
+      "step": 51863
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.435840707964602e-07,
+      "loss": 2.2513,
+      "step": 51864
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.435366624525917e-07,
+      "loss": 2.9267,
+      "step": 51865
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.434892541087231e-07,
+      "loss": 1.9708,
+      "step": 51866
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.434418457648546e-07,
+      "loss": 3.4889,
+      "step": 51867
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.433944374209862e-07,
+      "loss": 1.5907,
+      "step": 51868
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.433470290771176e-07,
+      "loss": 1.3027,
+      "step": 51869
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.43299620733249e-07,
+      "loss": 2.2636,
+      "step": 51870
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.432522123893805e-07,
+      "loss": 0.5625,
+      "step": 51871
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.43204804045512e-07,
+      "loss": 1.6701,
+      "step": 51872
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.431573957016435e-07,
+      "loss": 1.3833,
+      "step": 51873
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.431099873577749e-07,
+      "loss": 1.1373,
+      "step": 51874
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.430625790139065e-07,
+      "loss": 0.6714,
+      "step": 51875
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.430151706700379e-07,
+      "loss": 2.3964,
+      "step": 51876
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.429677623261694e-07,
+      "loss": 1.9302,
+      "step": 51877
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.429203539823009e-07,
+      "loss": 0.6871,
+      "step": 51878
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.428729456384324e-07,
+      "loss": 2.4454,
+      "step": 51879
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.428255372945639e-07,
+      "loss": 3.8127,
+      "step": 51880
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.427781289506954e-07,
+      "loss": 1.2978,
+      "step": 51881
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.427307206068269e-07,
+      "loss": 1.0764,
+      "step": 51882
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.426833122629583e-07,
+      "loss": 1.1011,
+      "step": 51883
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.426359039190897e-07,
+      "loss": 1.8192,
+      "step": 51884
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.425884955752212e-07,
+      "loss": 1.9128,
+      "step": 51885
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.425410872313527e-07,
+      "loss": 1.5209,
+      "step": 51886
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.424936788874842e-07,
+      "loss": 1.274,
+      "step": 51887
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.424462705436157e-07,
+      "loss": 1.0618,
+      "step": 51888
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.423988621997471e-07,
+      "loss": 1.8669,
+      "step": 51889
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.423514538558787e-07,
+      "loss": 1.3677,
+      "step": 51890
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.423040455120102e-07,
+      "loss": 3.1839,
+      "step": 51891
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.422566371681416e-07,
+      "loss": 1.8592,
+      "step": 51892
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.422092288242732e-07,
+      "loss": 1.2048,
+      "step": 51893
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.421618204804045e-07,
+      "loss": 0.6772,
+      "step": 51894
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.42114412136536e-07,
+      "loss": 1.4191,
+      "step": 51895
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.420670037926675e-07,
+      "loss": 0.6394,
+      "step": 51896
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.42019595448799e-07,
+      "loss": 1.6822,
+      "step": 51897
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 5.419721871049305e-07,
+      "loss": 1.8336,
+      "step": 51898
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.419247787610619e-07,
+      "loss": 2.0232,
+      "step": 51899
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.418773704171934e-07,
+      "loss": 2.7239,
+      "step": 51900
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.41829962073325e-07,
+      "loss": 1.8449,
+      "step": 51901
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.417825537294564e-07,
+      "loss": 0.9111,
+      "step": 51902
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.417351453855879e-07,
+      "loss": 1.4099,
+      "step": 51903
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.416877370417194e-07,
+      "loss": 1.6634,
+      "step": 51904
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.416403286978509e-07,
+      "loss": 1.2607,
+      "step": 51905
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.415929203539823e-07,
+      "loss": 0.3642,
+      "step": 51906
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.415455120101137e-07,
+      "loss": 2.0119,
+      "step": 51907
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.414981036662453e-07,
+      "loss": 1.3726,
+      "step": 51908
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.414506953223767e-07,
+      "loss": 2.0268,
+      "step": 51909
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.414032869785082e-07,
+      "loss": 2.4344,
+      "step": 51910
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.413558786346397e-07,
+      "loss": 2.1403,
+      "step": 51911
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.413084702907712e-07,
+      "loss": 1.3304,
+      "step": 51912
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.412610619469027e-07,
+      "loss": 2.7361,
+      "step": 51913
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.412136536030342e-07,
+      "loss": 2.8056,
+      "step": 51914
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.411662452591657e-07,
+      "loss": 2.3446,
+      "step": 51915
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.411188369152972e-07,
+      "loss": 1.8298,
+      "step": 51916
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.410714285714285e-07,
+      "loss": 1.6587,
+      "step": 51917
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.4102402022756e-07,
+      "loss": 1.6998,
+      "step": 51918
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.409766118836915e-07,
+      "loss": 1.1935,
+      "step": 51919
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.40929203539823e-07,
+      "loss": 1.6996,
+      "step": 51920
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.408817951959545e-07,
+      "loss": 1.2979,
+      "step": 51921
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.408343868520859e-07,
+      "loss": 1.1525,
+      "step": 51922
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.407869785082175e-07,
+      "loss": 1.8215,
+      "step": 51923
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.407395701643489e-07,
+      "loss": 2.3756,
+      "step": 51924
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.406921618204804e-07,
+      "loss": 0.6067,
+      "step": 51925
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.40644753476612e-07,
+      "loss": 1.3264,
+      "step": 51926
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.405973451327434e-07,
+      "loss": 3.2582,
+      "step": 51927
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.405499367888749e-07,
+      "loss": 1.6355,
+      "step": 51928
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.405025284450063e-07,
+      "loss": 2.6865,
+      "step": 51929
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.404551201011378e-07,
+      "loss": 1.847,
+      "step": 51930
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.404077117572693e-07,
+      "loss": 2.215,
+      "step": 51931
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.403603034134007e-07,
+      "loss": 1.8538,
+      "step": 51932
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.403128950695322e-07,
+      "loss": 1.4478,
+      "step": 51933
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.402654867256637e-07,
+      "loss": 2.572,
+      "step": 51934
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.402180783817952e-07,
+      "loss": 2.1243,
+      "step": 51935
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.401706700379267e-07,
+      "loss": 2.645,
+      "step": 51936
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.401232616940582e-07,
+      "loss": 1.7002,
+      "step": 51937
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.400758533501897e-07,
+      "loss": 1.457,
+      "step": 51938
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.400284450063212e-07,
+      "loss": 1.8491,
+      "step": 51939
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.399810366624526e-07,
+      "loss": 1.1597,
+      "step": 51940
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.399336283185841e-07,
+      "loss": 1.271,
+      "step": 51941
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.398862199747155e-07,
+      "loss": 1.7536,
+      "step": 51942
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.39838811630847e-07,
+      "loss": 2.1489,
+      "step": 51943
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.397914032869785e-07,
+      "loss": 0.4075,
+      "step": 51944
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.3974399494311e-07,
+      "loss": 1.5108,
+      "step": 51945
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.396965865992415e-07,
+      "loss": 1.0047,
+      "step": 51946
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.396491782553729e-07,
+      "loss": 2.0187,
+      "step": 51947
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.396017699115045e-07,
+      "loss": 1.0164,
+      "step": 51948
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.39554361567636e-07,
+      "loss": 1.8845,
+      "step": 51949
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.395069532237674e-07,
+      "loss": 1.6696,
+      "step": 51950
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.394595448798989e-07,
+      "loss": 1.0922,
+      "step": 51951
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.394121365360304e-07,
+      "loss": 2.4847,
+      "step": 51952
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.393647281921618e-07,
+      "loss": 3.2209,
+      "step": 51953
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.393173198482933e-07,
+      "loss": 2.2454,
+      "step": 51954
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.392699115044247e-07,
+      "loss": 2.1312,
+      "step": 51955
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.392225031605563e-07,
+      "loss": 1.1956,
+      "step": 51956
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.391750948166877e-07,
+      "loss": 1.8822,
+      "step": 51957
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.391276864728192e-07,
+      "loss": 2.5848,
+      "step": 51958
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.390802781289508e-07,
+      "loss": 2.0199,
+      "step": 51959
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.390328697850822e-07,
+      "loss": 1.8801,
+      "step": 51960
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.389854614412137e-07,
+      "loss": 0.542,
+      "step": 51961
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 5.389380530973452e-07,
+      "loss": 2.3765,
+      "step": 51962
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.388906447534767e-07,
+      "loss": 1.1426,
+      "step": 51963
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.388432364096082e-07,
+      "loss": 1.6274,
+      "step": 51964
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.387958280657395e-07,
+      "loss": 3.755,
+      "step": 51965
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.38748419721871e-07,
+      "loss": 3.0015,
+      "step": 51966
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.387010113780025e-07,
+      "loss": 2.9805,
+      "step": 51967
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.38653603034134e-07,
+      "loss": 2.551,
+      "step": 51968
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.386061946902655e-07,
+      "loss": 1.9751,
+      "step": 51969
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.38558786346397e-07,
+      "loss": 2.466,
+      "step": 51970
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.385113780025285e-07,
+      "loss": 2.0891,
+      "step": 51971
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.3846396965866e-07,
+      "loss": 1.545,
+      "step": 51972
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.384165613147914e-07,
+      "loss": 1.6053,
+      "step": 51973
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.38369152970923e-07,
+      "loss": 1.4028,
+      "step": 51974
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.383217446270544e-07,
+      "loss": 1.5357,
+      "step": 51975
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.382743362831858e-07,
+      "loss": 1.8348,
+      "step": 51976
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.382269279393173e-07,
+      "loss": 1.8005,
+      "step": 51977
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.381795195954488e-07,
+      "loss": 1.6487,
+      "step": 51978
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.381321112515803e-07,
+      "loss": 2.08,
+      "step": 51979
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.380847029077117e-07,
+      "loss": 3.3856,
+      "step": 51980
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.380372945638433e-07,
+      "loss": 2.0149,
+      "step": 51981
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.379898862199748e-07,
+      "loss": 2.6155,
+      "step": 51982
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.379424778761062e-07,
+      "loss": 2.6123,
+      "step": 51983
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.378950695322377e-07,
+      "loss": 1.9613,
+      "step": 51984
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.378476611883692e-07,
+      "loss": 0.4924,
+      "step": 51985
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.378002528445007e-07,
+      "loss": 2.6152,
+      "step": 51986
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.377528445006322e-07,
+      "loss": 0.782,
+      "step": 51987
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.377054361567635e-07,
+      "loss": 1.2236,
+      "step": 51988
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.376580278128951e-07,
+      "loss": 0.8361,
+      "step": 51989
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.376106194690265e-07,
+      "loss": 1.1579,
+      "step": 51990
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.37563211125158e-07,
+      "loss": 1.3304,
+      "step": 51991
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.375158027812895e-07,
+      "loss": 1.9423,
+      "step": 51992
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.37468394437421e-07,
+      "loss": 2.2217,
+      "step": 51993
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.374209860935525e-07,
+      "loss": 1.4513,
+      "step": 51994
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.373735777496839e-07,
+      "loss": 2.7303,
+      "step": 51995
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.373261694058155e-07,
+      "loss": 4.7061,
+      "step": 51996
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.37278761061947e-07,
+      "loss": 0.7226,
+      "step": 51997
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.372313527180784e-07,
+      "loss": 2.9572,
+      "step": 51998
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.371839443742098e-07,
+      "loss": 2.8117,
+      "step": 51999
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.371365360303413e-07,
+      "loss": 1.7242,
+      "step": 52000
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.370891276864728e-07,
+      "loss": 1.4457,
+      "step": 52001
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.370417193426043e-07,
+      "loss": 1.6251,
+      "step": 52002
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.369943109987358e-07,
+      "loss": 1.6062,
+      "step": 52003
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.369469026548673e-07,
+      "loss": 1.5904,
+      "step": 52004
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.368994943109987e-07,
+      "loss": 2.106,
+      "step": 52005
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.368520859671302e-07,
+      "loss": 1.3872,
+      "step": 52006
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.368046776232618e-07,
+      "loss": 0.8404,
+      "step": 52007
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.367572692793932e-07,
+      "loss": 3.8303,
+      "step": 52008
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.367098609355247e-07,
+      "loss": 0.7824,
+      "step": 52009
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.366624525916562e-07,
+      "loss": 1.6409,
+      "step": 52010
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.366150442477876e-07,
+      "loss": 2.1044,
+      "step": 52011
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.365676359039191e-07,
+      "loss": 0.5139,
+      "step": 52012
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.365202275600505e-07,
+      "loss": 2.4521,
+      "step": 52013
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.364728192161821e-07,
+      "loss": 2.6461,
+      "step": 52014
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.364254108723135e-07,
+      "loss": 2.6655,
+      "step": 52015
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.36378002528445e-07,
+      "loss": 1.5591,
+      "step": 52016
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.363305941845765e-07,
+      "loss": 2.8131,
+      "step": 52017
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.36283185840708e-07,
+      "loss": 2.3508,
+      "step": 52018
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.362357774968395e-07,
+      "loss": 1.698,
+      "step": 52019
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.36188369152971e-07,
+      "loss": 2.4393,
+      "step": 52020
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.361409608091024e-07,
+      "loss": 1.1204,
+      "step": 52021
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.36093552465234e-07,
+      "loss": 2.8166,
+      "step": 52022
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.360461441213653e-07,
+      "loss": 3.7726,
+      "step": 52023
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.359987357774968e-07,
+      "loss": 3.0455,
+      "step": 52024
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 5.359513274336283e-07,
+      "loss": 1.6321,
+      "step": 52025
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.359039190897598e-07,
+      "loss": 1.3553,
+      "step": 52026
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.358565107458913e-07,
+      "loss": 2.6418,
+      "step": 52027
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.358091024020227e-07,
+      "loss": 1.7512,
+      "step": 52028
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.357616940581543e-07,
+      "loss": 2.9916,
+      "step": 52029
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.357142857142858e-07,
+      "loss": 1.2813,
+      "step": 52030
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.356668773704172e-07,
+      "loss": 2.0754,
+      "step": 52031
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.356194690265487e-07,
+      "loss": 1.9621,
+      "step": 52032
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.355720606826802e-07,
+      "loss": 2.9031,
+      "step": 52033
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.355246523388117e-07,
+      "loss": 1.7237,
+      "step": 52034
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.354772439949431e-07,
+      "loss": 3.2237,
+      "step": 52035
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.354298356510745e-07,
+      "loss": 0.9388,
+      "step": 52036
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.353824273072061e-07,
+      "loss": 2.0556,
+      "step": 52037
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.353350189633375e-07,
+      "loss": 3.1375,
+      "step": 52038
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.35287610619469e-07,
+      "loss": 1.666,
+      "step": 52039
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.352402022756006e-07,
+      "loss": 2.0132,
+      "step": 52040
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.35192793931732e-07,
+      "loss": 1.4099,
+      "step": 52041
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.351453855878635e-07,
+      "loss": 0.8467,
+      "step": 52042
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.35097977243995e-07,
+      "loss": 1.4722,
+      "step": 52043
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.350505689001265e-07,
+      "loss": 2.2602,
+      "step": 52044
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.35003160556258e-07,
+      "loss": 2.5639,
+      "step": 52045
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.349557522123893e-07,
+      "loss": 2.385,
+      "step": 52046
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.349083438685208e-07,
+      "loss": 1.2827,
+      "step": 52047
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.348609355246523e-07,
+      "loss": 1.8739,
+      "step": 52048
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.348135271807838e-07,
+      "loss": 2.5124,
+      "step": 52049
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.347661188369153e-07,
+      "loss": 0.7597,
+      "step": 52050
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.347187104930468e-07,
+      "loss": 1.7923,
+      "step": 52051
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.346713021491783e-07,
+      "loss": 2.8637,
+      "step": 52052
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.346238938053098e-07,
+      "loss": 2.6887,
+      "step": 52053
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.345764854614412e-07,
+      "loss": 1.4541,
+      "step": 52054
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.345290771175728e-07,
+      "loss": 2.6298,
+      "step": 52055
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.344816687737042e-07,
+      "loss": 2.1245,
+      "step": 52056
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.344342604298357e-07,
+      "loss": 2.292,
+      "step": 52057
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.343868520859671e-07,
+      "loss": 2.0794,
+      "step": 52058
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.343394437420986e-07,
+      "loss": 2.7599,
+      "step": 52059
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.342920353982301e-07,
+      "loss": 1.9001,
+      "step": 52060
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.342446270543615e-07,
+      "loss": 3.1961,
+      "step": 52061
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.341972187104931e-07,
+      "loss": 2.4165,
+      "step": 52062
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.341498103666245e-07,
+      "loss": 2.5414,
+      "step": 52063
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.34102402022756e-07,
+      "loss": 1.8924,
+      "step": 52064
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.340549936788875e-07,
+      "loss": 1.7204,
+      "step": 52065
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.34007585335019e-07,
+      "loss": 1.6694,
+      "step": 52066
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.339601769911505e-07,
+      "loss": 1.7626,
+      "step": 52067
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.33912768647282e-07,
+      "loss": 2.3366,
+      "step": 52068
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.338653603034133e-07,
+      "loss": 2.0096,
+      "step": 52069
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.338179519595449e-07,
+      "loss": 1.8201,
+      "step": 52070
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.337705436156763e-07,
+      "loss": 0.5845,
+      "step": 52071
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.337231352718078e-07,
+      "loss": 3.7598,
+      "step": 52072
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.336757269279393e-07,
+      "loss": 1.5075,
+      "step": 52073
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.336283185840708e-07,
+      "loss": 1.4499,
+      "step": 52074
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.335809102402023e-07,
+      "loss": 2.3432,
+      "step": 52075
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.335335018963337e-07,
+      "loss": 3.4357,
+      "step": 52076
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.334860935524653e-07,
+      "loss": 2.4325,
+      "step": 52077
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.334386852085968e-07,
+      "loss": 2.9172,
+      "step": 52078
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.333912768647282e-07,
+      "loss": 1.9064,
+      "step": 52079
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.333438685208597e-07,
+      "loss": 2.1023,
+      "step": 52080
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.332964601769911e-07,
+      "loss": 1.599,
+      "step": 52081
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.332490518331226e-07,
+      "loss": 2.7771,
+      "step": 52082
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.332016434892541e-07,
+      "loss": 1.841,
+      "step": 52083
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.331542351453856e-07,
+      "loss": 2.5675,
+      "step": 52084
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.331068268015171e-07,
+      "loss": 3.5711,
+      "step": 52085
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.330594184576485e-07,
+      "loss": 1.1633,
+      "step": 52086
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.3301201011378e-07,
+      "loss": 1.6647,
+      "step": 52087
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 5.329646017699116e-07,
+      "loss": 1.299,
+      "step": 52088
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.32917193426043e-07,
+      "loss": 1.4656,
+      "step": 52089
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.328697850821745e-07,
+      "loss": 2.8608,
+      "step": 52090
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.32822376738306e-07,
+      "loss": 1.3461,
+      "step": 52091
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.327749683944375e-07,
+      "loss": 2.277,
+      "step": 52092
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.32727560050569e-07,
+      "loss": 2.6564,
+      "step": 52093
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.326801517067003e-07,
+      "loss": 2.522,
+      "step": 52094
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.326327433628319e-07,
+      "loss": 1.6999,
+      "step": 52095
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.325853350189633e-07,
+      "loss": 2.3934,
+      "step": 52096
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.325379266750948e-07,
+      "loss": 2.4649,
+      "step": 52097
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.324905183312263e-07,
+      "loss": 1.4724,
+      "step": 52098
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.324431099873578e-07,
+      "loss": 1.1688,
+      "step": 52099
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.323957016434893e-07,
+      "loss": 2.2693,
+      "step": 52100
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.323482932996208e-07,
+      "loss": 0.8714,
+      "step": 52101
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.323008849557522e-07,
+      "loss": 2.1102,
+      "step": 52102
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.322534766118838e-07,
+      "loss": 2.9087,
+      "step": 52103
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.322060682680152e-07,
+      "loss": 1.9072,
+      "step": 52104
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.321586599241466e-07,
+      "loss": 2.6762,
+      "step": 52105
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.321112515802781e-07,
+      "loss": 1.4048,
+      "step": 52106
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.320638432364096e-07,
+      "loss": 1.7166,
+      "step": 52107
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.320164348925411e-07,
+      "loss": 1.8093,
+      "step": 52108
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.319690265486725e-07,
+      "loss": 1.6244,
+      "step": 52109
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.319216182048041e-07,
+      "loss": 2.0394,
+      "step": 52110
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.318742098609356e-07,
+      "loss": 1.2864,
+      "step": 52111
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.31826801517067e-07,
+      "loss": 2.1512,
+      "step": 52112
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.317793931731985e-07,
+      "loss": 1.1505,
+      "step": 52113
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.3173198482933e-07,
+      "loss": 1.1794,
+      "step": 52114
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.316845764854615e-07,
+      "loss": 1.985,
+      "step": 52115
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.31637168141593e-07,
+      "loss": 1.9145,
+      "step": 52116
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.315897597977244e-07,
+      "loss": 0.6385,
+      "step": 52117
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.315423514538559e-07,
+      "loss": 1.0167,
+      "step": 52118
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.314949431099873e-07,
+      "loss": 2.5776,
+      "step": 52119
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.314475347661188e-07,
+      "loss": 2.0578,
+      "step": 52120
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.314001264222504e-07,
+      "loss": 2.6898,
+      "step": 52121
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.313527180783818e-07,
+      "loss": 0.8174,
+      "step": 52122
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.313053097345133e-07,
+      "loss": 1.9963,
+      "step": 52123
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.312579013906447e-07,
+      "loss": 3.6531,
+      "step": 52124
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.312104930467763e-07,
+      "loss": 1.5467,
+      "step": 52125
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.311630847029078e-07,
+      "loss": 2.5413,
+      "step": 52126
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.311156763590392e-07,
+      "loss": 3.0686,
+      "step": 52127
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.310682680151708e-07,
+      "loss": 2.7111,
+      "step": 52128
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.310208596713021e-07,
+      "loss": 2.1053,
+      "step": 52129
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.309734513274336e-07,
+      "loss": 2.3028,
+      "step": 52130
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.309260429835651e-07,
+      "loss": 2.5911,
+      "step": 52131
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.308786346396966e-07,
+      "loss": 1.9574,
+      "step": 52132
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.308312262958281e-07,
+      "loss": 3.2083,
+      "step": 52133
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.307838179519595e-07,
+      "loss": 4.1508,
+      "step": 52134
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.30736409608091e-07,
+      "loss": 1.3024,
+      "step": 52135
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.306890012642226e-07,
+      "loss": 0.9867,
+      "step": 52136
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.30641592920354e-07,
+      "loss": 1.0701,
+      "step": 52137
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.305941845764855e-07,
+      "loss": 1.5875,
+      "step": 52138
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.30546776232617e-07,
+      "loss": 2.5941,
+      "step": 52139
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.304993678887484e-07,
+      "loss": 1.289,
+      "step": 52140
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.304519595448799e-07,
+      "loss": 1.5934,
+      "step": 52141
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.304045512010113e-07,
+      "loss": 2.1037,
+      "step": 52142
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.303571428571429e-07,
+      "loss": 1.7787,
+      "step": 52143
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.303097345132743e-07,
+      "loss": 1.3836,
+      "step": 52144
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.302623261694058e-07,
+      "loss": 0.8725,
+      "step": 52145
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.302149178255373e-07,
+      "loss": 1.491,
+      "step": 52146
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.301675094816688e-07,
+      "loss": 1.6269,
+      "step": 52147
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.301201011378003e-07,
+      "loss": 2.3179,
+      "step": 52148
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.300726927939318e-07,
+      "loss": 1.7678,
+      "step": 52149
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.300252844500633e-07,
+      "loss": 2.4948,
+      "step": 52150
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.299778761061948e-07,
+      "loss": 1.5846,
+      "step": 52151
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 5.299304677623261e-07,
+      "loss": 1.7944,
+      "step": 52152
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.298830594184576e-07,
+      "loss": 3.6376,
+      "step": 52153
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.298356510745891e-07,
+      "loss": 2.5948,
+      "step": 52154
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.297882427307206e-07,
+      "loss": 2.4106,
+      "step": 52155
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.297408343868521e-07,
+      "loss": 1.9675,
+      "step": 52156
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.296934260429835e-07,
+      "loss": 1.0675,
+      "step": 52157
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.296460176991151e-07,
+      "loss": 2.2294,
+      "step": 52158
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.295986093552466e-07,
+      "loss": 0.7549,
+      "step": 52159
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.29551201011378e-07,
+      "loss": 1.635,
+      "step": 52160
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.295037926675096e-07,
+      "loss": 2.5238,
+      "step": 52161
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.29456384323641e-07,
+      "loss": 2.4398,
+      "step": 52162
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.294089759797724e-07,
+      "loss": 1.5392,
+      "step": 52163
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.293615676359039e-07,
+      "loss": 2.0381,
+      "step": 52164
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.293141592920354e-07,
+      "loss": 1.1668,
+      "step": 52165
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.292667509481669e-07,
+      "loss": 1.7427,
+      "step": 52166
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.292193426042983e-07,
+      "loss": 2.1476,
+      "step": 52167
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.291719342604298e-07,
+      "loss": 3.212,
+      "step": 52168
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.291245259165614e-07,
+      "loss": 2.5168,
+      "step": 52169
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.290771175726928e-07,
+      "loss": 1.7384,
+      "step": 52170
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.290297092288243e-07,
+      "loss": 0.8785,
+      "step": 52171
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.289823008849558e-07,
+      "loss": 0.8883,
+      "step": 52172
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.289348925410873e-07,
+      "loss": 1.6281,
+      "step": 52173
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.288874841972188e-07,
+      "loss": 1.8564,
+      "step": 52174
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.288400758533501e-07,
+      "loss": 2.835,
+      "step": 52175
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.287926675094817e-07,
+      "loss": 0.2878,
+      "step": 52176
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.287452591656131e-07,
+      "loss": 2.0237,
+      "step": 52177
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.286978508217446e-07,
+      "loss": 1.9596,
+      "step": 52178
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.286504424778761e-07,
+      "loss": 1.6187,
+      "step": 52179
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.286030341340076e-07,
+      "loss": 2.7966,
+      "step": 52180
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.285556257901391e-07,
+      "loss": 3.0702,
+      "step": 52181
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.285082174462706e-07,
+      "loss": 0.5122,
+      "step": 52182
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.284608091024021e-07,
+      "loss": 2.6939,
+      "step": 52183
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.284134007585336e-07,
+      "loss": 1.9722,
+      "step": 52184
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.28365992414665e-07,
+      "loss": 2.9947,
+      "step": 52185
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.283185840707965e-07,
+      "loss": 1.677,
+      "step": 52186
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.282711757269279e-07,
+      "loss": 1.5431,
+      "step": 52187
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.282237673830594e-07,
+      "loss": 1.9281,
+      "step": 52188
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.281763590391909e-07,
+      "loss": 1.2797,
+      "step": 52189
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.281289506953223e-07,
+      "loss": 0.8254,
+      "step": 52190
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.280815423514539e-07,
+      "loss": 1.2188,
+      "step": 52191
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.280341340075854e-07,
+      "loss": 1.9821,
+      "step": 52192
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.279867256637168e-07,
+      "loss": 1.5072,
+      "step": 52193
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.279393173198484e-07,
+      "loss": 1.9338,
+      "step": 52194
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.278919089759798e-07,
+      "loss": 1.5417,
+      "step": 52195
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.278445006321113e-07,
+      "loss": 1.5859,
+      "step": 52196
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.277970922882428e-07,
+      "loss": 2.7232,
+      "step": 52197
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.277496839443743e-07,
+      "loss": 2.2628,
+      "step": 52198
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.277022756005057e-07,
+      "loss": 1.9625,
+      "step": 52199
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.276548672566371e-07,
+      "loss": 1.9557,
+      "step": 52200
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.276074589127686e-07,
+      "loss": 2.8595,
+      "step": 52201
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.275600505689001e-07,
+      "loss": 0.8347,
+      "step": 52202
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.275126422250316e-07,
+      "loss": 1.8687,
+      "step": 52203
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.274652338811631e-07,
+      "loss": 0.989,
+      "step": 52204
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.274178255372946e-07,
+      "loss": 1.8503,
+      "step": 52205
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.273704171934261e-07,
+      "loss": 1.6591,
+      "step": 52206
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.273230088495576e-07,
+      "loss": 1.0852,
+      "step": 52207
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.27275600505689e-07,
+      "loss": 1.2155,
+      "step": 52208
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.272281921618206e-07,
+      "loss": 2.0373,
+      "step": 52209
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.271807838179519e-07,
+      "loss": 1.7788,
+      "step": 52210
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.271333754740834e-07,
+      "loss": 3.695,
+      "step": 52211
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.270859671302149e-07,
+      "loss": 1.1581,
+      "step": 52212
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.270385587863464e-07,
+      "loss": 0.6604,
+      "step": 52213
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.269911504424779e-07,
+      "loss": 2.1112,
+      "step": 52214
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 5.269437420986093e-07,
+      "loss": 1.7561,
+      "step": 52215
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.268963337547409e-07,
+      "loss": 1.6657,
+      "step": 52216
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.268489254108724e-07,
+      "loss": 1.2997,
+      "step": 52217
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.268015170670038e-07,
+      "loss": 2.1668,
+      "step": 52218
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.267541087231353e-07,
+      "loss": 1.6527,
+      "step": 52219
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.267067003792668e-07,
+      "loss": 1.569,
+      "step": 52220
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.266592920353983e-07,
+      "loss": 3.1569,
+      "step": 52221
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.266118836915297e-07,
+      "loss": 2.781,
+      "step": 52222
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.265644753476611e-07,
+      "loss": 1.1581,
+      "step": 52223
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.265170670037927e-07,
+      "loss": 1.5122,
+      "step": 52224
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.264696586599241e-07,
+      "loss": 0.8364,
+      "step": 52225
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.264222503160556e-07,
+      "loss": 3.5737,
+      "step": 52226
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.263748419721872e-07,
+      "loss": 2.3588,
+      "step": 52227
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.263274336283186e-07,
+      "loss": 0.584,
+      "step": 52228
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.262800252844501e-07,
+      "loss": 2.9997,
+      "step": 52229
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.262326169405816e-07,
+      "loss": 2.9429,
+      "step": 52230
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.261852085967131e-07,
+      "loss": 3.0788,
+      "step": 52231
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.261378002528446e-07,
+      "loss": 1.1661,
+      "step": 52232
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.26090391908976e-07,
+      "loss": 1.7366,
+      "step": 52233
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.260429835651074e-07,
+      "loss": 2.0911,
+      "step": 52234
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.259955752212389e-07,
+      "loss": 1.4669,
+      "step": 52235
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.259481668773704e-07,
+      "loss": 2.4919,
+      "step": 52236
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.259007585335019e-07,
+      "loss": 1.453,
+      "step": 52237
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.258533501896333e-07,
+      "loss": 1.8658,
+      "step": 52238
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.258059418457649e-07,
+      "loss": 2.2211,
+      "step": 52239
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.257585335018964e-07,
+      "loss": 1.9829,
+      "step": 52240
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.257111251580278e-07,
+      "loss": 1.8072,
+      "step": 52241
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.256637168141594e-07,
+      "loss": 1.3564,
+      "step": 52242
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.256163084702908e-07,
+      "loss": 2.5004,
+      "step": 52243
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.255689001264223e-07,
+      "loss": 2.3759,
+      "step": 52244
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.255214917825537e-07,
+      "loss": 1.7776,
+      "step": 52245
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.254740834386852e-07,
+      "loss": 1.1488,
+      "step": 52246
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.254266750948167e-07,
+      "loss": 1.543,
+      "step": 52247
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.253792667509481e-07,
+      "loss": 1.3055,
+      "step": 52248
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.253318584070796e-07,
+      "loss": 2.8675,
+      "step": 52249
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.252844500632112e-07,
+      "loss": 0.6128,
+      "step": 52250
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.252370417193426e-07,
+      "loss": 2.0375,
+      "step": 52251
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.251896333754741e-07,
+      "loss": 1.0438,
+      "step": 52252
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.251422250316056e-07,
+      "loss": 1.3805,
+      "step": 52253
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.250948166877371e-07,
+      "loss": 1.8228,
+      "step": 52254
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.250474083438686e-07,
+      "loss": 0.9306,
+      "step": 52255
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.25e-07,
+      "loss": 1.3455,
+      "step": 52256
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.249525916561315e-07,
+      "loss": 2.5137,
+      "step": 52257
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.249051833122629e-07,
+      "loss": 1.4996,
+      "step": 52258
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.248577749683944e-07,
+      "loss": 1.4087,
+      "step": 52259
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.248103666245259e-07,
+      "loss": 2.5769,
+      "step": 52260
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.247629582806574e-07,
+      "loss": 1.5074,
+      "step": 52261
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.247155499367889e-07,
+      "loss": 0.5943,
+      "step": 52262
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.246681415929203e-07,
+      "loss": 1.1789,
+      "step": 52263
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.246207332490519e-07,
+      "loss": 1.2533,
+      "step": 52264
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.245733249051834e-07,
+      "loss": 3.5844,
+      "step": 52265
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.245259165613148e-07,
+      "loss": 1.9605,
+      "step": 52266
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.244785082174463e-07,
+      "loss": 2.3729,
+      "step": 52267
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.244310998735778e-07,
+      "loss": 1.6536,
+      "step": 52268
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.243836915297092e-07,
+      "loss": 1.3935,
+      "step": 52269
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.243362831858407e-07,
+      "loss": 1.4984,
+      "step": 52270
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.242888748419721e-07,
+      "loss": 4.2373,
+      "step": 52271
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.242414664981037e-07,
+      "loss": 1.3132,
+      "step": 52272
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.241940581542351e-07,
+      "loss": 2.9667,
+      "step": 52273
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.241466498103666e-07,
+      "loss": 1.0215,
+      "step": 52274
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.240992414664982e-07,
+      "loss": 2.6667,
+      "step": 52275
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.240518331226296e-07,
+      "loss": 2.3477,
+      "step": 52276
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.240044247787611e-07,
+      "loss": 1.7406,
+      "step": 52277
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 5.239570164348926e-07,
+      "loss": 2.0318,
+      "step": 52278
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.239096080910241e-07,
+      "loss": 1.4533,
+      "step": 52279
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.238621997471556e-07,
+      "loss": 1.5397,
+      "step": 52280
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.238147914032869e-07,
+      "loss": 1.8505,
+      "step": 52281
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.237673830594184e-07,
+      "loss": 3.778,
+      "step": 52282
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.237199747155499e-07,
+      "loss": 1.8239,
+      "step": 52283
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.236725663716814e-07,
+      "loss": 0.6124,
+      "step": 52284
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.236251580278129e-07,
+      "loss": 0.4141,
+      "step": 52285
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.235777496839444e-07,
+      "loss": 1.5402,
+      "step": 52286
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.235303413400759e-07,
+      "loss": 1.0972,
+      "step": 52287
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.234829329962074e-07,
+      "loss": 0.5779,
+      "step": 52288
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.234355246523388e-07,
+      "loss": 1.8538,
+      "step": 52289
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.233881163084704e-07,
+      "loss": 1.3188,
+      "step": 52290
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.233407079646018e-07,
+      "loss": 1.9995,
+      "step": 52291
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.232932996207332e-07,
+      "loss": 3.5259,
+      "step": 52292
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.232458912768647e-07,
+      "loss": 2.1032,
+      "step": 52293
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.231984829329962e-07,
+      "loss": 1.7427,
+      "step": 52294
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.231510745891277e-07,
+      "loss": 1.6976,
+      "step": 52295
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.231036662452591e-07,
+      "loss": 2.3571,
+      "step": 52296
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.230562579013907e-07,
+      "loss": 1.4508,
+      "step": 52297
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.230088495575222e-07,
+      "loss": 2.3462,
+      "step": 52298
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.229614412136536e-07,
+      "loss": 1.138,
+      "step": 52299
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.229140328697851e-07,
+      "loss": 2.4399,
+      "step": 52300
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.228666245259166e-07,
+      "loss": 2.738,
+      "step": 52301
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.228192161820481e-07,
+      "loss": 1.6784,
+      "step": 52302
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.227718078381796e-07,
+      "loss": 2.8692,
+      "step": 52303
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.227243994943109e-07,
+      "loss": 0.8525,
+      "step": 52304
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.226769911504425e-07,
+      "loss": 1.1124,
+      "step": 52305
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.226295828065739e-07,
+      "loss": 2.0725,
+      "step": 52306
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.225821744627054e-07,
+      "loss": 2.1391,
+      "step": 52307
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.22534766118837e-07,
+      "loss": 1.3327,
+      "step": 52308
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.224873577749684e-07,
+      "loss": 0.8202,
+      "step": 52309
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.224399494310999e-07,
+      "loss": 1.7355,
+      "step": 52310
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.223925410872314e-07,
+      "loss": 0.5265,
+      "step": 52311
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.223451327433629e-07,
+      "loss": 3.6248,
+      "step": 52312
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.222977243994944e-07,
+      "loss": 1.6675,
+      "step": 52313
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.222503160556258e-07,
+      "loss": 1.493,
+      "step": 52314
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.222029077117572e-07,
+      "loss": 1.7253,
+      "step": 52315
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.221554993678887e-07,
+      "loss": 2.3925,
+      "step": 52316
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.221080910240202e-07,
+      "loss": 1.0454,
+      "step": 52317
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.220606826801517e-07,
+      "loss": 1.7891,
+      "step": 52318
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.220132743362832e-07,
+      "loss": 1.66,
+      "step": 52319
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.219658659924147e-07,
+      "loss": 0.6824,
+      "step": 52320
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.219184576485462e-07,
+      "loss": 2.9709,
+      "step": 52321
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.218710493046776e-07,
+      "loss": 1.5641,
+      "step": 52322
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.218236409608092e-07,
+      "loss": 3.068,
+      "step": 52323
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.217762326169406e-07,
+      "loss": 1.6908,
+      "step": 52324
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.217288242730721e-07,
+      "loss": 0.8883,
+      "step": 52325
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.216814159292036e-07,
+      "loss": 1.7944,
+      "step": 52326
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.21634007585335e-07,
+      "loss": 1.5059,
+      "step": 52327
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.215865992414665e-07,
+      "loss": 2.7742,
+      "step": 52328
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.215391908975979e-07,
+      "loss": 1.6334,
+      "step": 52329
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.214917825537295e-07,
+      "loss": 1.4602,
+      "step": 52330
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.21444374209861e-07,
+      "loss": 1.0103,
+      "step": 52331
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.213969658659924e-07,
+      "loss": 2.7473,
+      "step": 52332
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.213495575221239e-07,
+      "loss": 1.0864,
+      "step": 52333
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.213021491782554e-07,
+      "loss": 3.073,
+      "step": 52334
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.212547408343869e-07,
+      "loss": 1.2562,
+      "step": 52335
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.212073324905184e-07,
+      "loss": 1.7953,
+      "step": 52336
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.211599241466498e-07,
+      "loss": 0.9537,
+      "step": 52337
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.211125158027814e-07,
+      "loss": 2.2721,
+      "step": 52338
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.210651074589127e-07,
+      "loss": 4.8874,
+      "step": 52339
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.210176991150442e-07,
+      "loss": 1.8122,
+      "step": 52340
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.209702907711757e-07,
+      "loss": 1.7552,
+      "step": 52341
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 5.209228824273072e-07,
+      "loss": 1.5744,
+      "step": 52342
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.208754740834387e-07,
+      "loss": 1.7545,
+      "step": 52343
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.208280657395701e-07,
+      "loss": 3.2763,
+      "step": 52344
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.207806573957017e-07,
+      "loss": 1.5941,
+      "step": 52345
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.207332490518332e-07,
+      "loss": 1.5505,
+      "step": 52346
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.206858407079646e-07,
+      "loss": 2.9711,
+      "step": 52347
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.206384323640961e-07,
+      "loss": 2.6408,
+      "step": 52348
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.205910240202276e-07,
+      "loss": 1.259,
+      "step": 52349
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.20543615676359e-07,
+      "loss": 1.6299,
+      "step": 52350
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.204962073324905e-07,
+      "loss": 1.7037,
+      "step": 52351
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.20448798988622e-07,
+      "loss": 2.0266,
+      "step": 52352
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.204013906447535e-07,
+      "loss": 1.8225,
+      "step": 52353
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.203539823008849e-07,
+      "loss": 1.912,
+      "step": 52354
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.203065739570164e-07,
+      "loss": 1.0191,
+      "step": 52355
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.20259165613148e-07,
+      "loss": 2.158,
+      "step": 52356
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.202117572692794e-07,
+      "loss": 2.187,
+      "step": 52357
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.201643489254109e-07,
+      "loss": 2.3893,
+      "step": 52358
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.201169405815424e-07,
+      "loss": 2.4936,
+      "step": 52359
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.200695322376739e-07,
+      "loss": 3.0032,
+      "step": 52360
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.200221238938054e-07,
+      "loss": 2.7914,
+      "step": 52361
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.199747155499367e-07,
+      "loss": 0.7438,
+      "step": 52362
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.199273072060683e-07,
+      "loss": 2.533,
+      "step": 52363
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.198798988621997e-07,
+      "loss": 1.6624,
+      "step": 52364
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.198324905183312e-07,
+      "loss": 2.0766,
+      "step": 52365
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.197850821744627e-07,
+      "loss": 2.5958,
+      "step": 52366
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.197376738305942e-07,
+      "loss": 2.6432,
+      "step": 52367
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.196902654867257e-07,
+      "loss": 2.1076,
+      "step": 52368
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.196428571428572e-07,
+      "loss": 0.9233,
+      "step": 52369
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.195954487989886e-07,
+      "loss": 2.4612,
+      "step": 52370
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.195480404551202e-07,
+      "loss": 1.6802,
+      "step": 52371
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.195006321112516e-07,
+      "loss": 1.7172,
+      "step": 52372
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.19453223767383e-07,
+      "loss": 2.069,
+      "step": 52373
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.194058154235145e-07,
+      "loss": 3.5634,
+      "step": 52374
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.19358407079646e-07,
+      "loss": 2.0975,
+      "step": 52375
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.193109987357775e-07,
+      "loss": 0.6858,
+      "step": 52376
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.192635903919089e-07,
+      "loss": 1.2293,
+      "step": 52377
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.192161820480405e-07,
+      "loss": 1.2144,
+      "step": 52378
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.19168773704172e-07,
+      "loss": 1.9944,
+      "step": 52379
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.191213653603034e-07,
+      "loss": 3.2157,
+      "step": 52380
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.190739570164349e-07,
+      "loss": 1.6235,
+      "step": 52381
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.190265486725664e-07,
+      "loss": 1.3851,
+      "step": 52382
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.189791403286979e-07,
+      "loss": 1.7392,
+      "step": 52383
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.189317319848294e-07,
+      "loss": 1.6904,
+      "step": 52384
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.188843236409609e-07,
+      "loss": 0.643,
+      "step": 52385
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.188369152970923e-07,
+      "loss": 2.1027,
+      "step": 52386
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.187895069532237e-07,
+      "loss": 2.4118,
+      "step": 52387
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.187420986093552e-07,
+      "loss": 2.2589,
+      "step": 52388
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.186946902654868e-07,
+      "loss": 3.6202,
+      "step": 52389
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.186472819216182e-07,
+      "loss": 2.3582,
+      "step": 52390
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.185998735777497e-07,
+      "loss": 1.9409,
+      "step": 52391
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.185524652338812e-07,
+      "loss": 1.1974,
+      "step": 52392
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.185050568900127e-07,
+      "loss": 1.1904,
+      "step": 52393
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.184576485461442e-07,
+      "loss": 2.6996,
+      "step": 52394
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.184102402022756e-07,
+      "loss": 1.2507,
+      "step": 52395
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.183628318584072e-07,
+      "loss": 1.9076,
+      "step": 52396
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.183154235145385e-07,
+      "loss": 0.9343,
+      "step": 52397
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.1826801517067e-07,
+      "loss": 1.2608,
+      "step": 52398
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.182206068268015e-07,
+      "loss": 1.2766,
+      "step": 52399
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.18173198482933e-07,
+      "loss": 1.8121,
+      "step": 52400
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.181257901390645e-07,
+      "loss": 1.6513,
+      "step": 52401
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.180783817951959e-07,
+      "loss": 2.7391,
+      "step": 52402
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.180309734513274e-07,
+      "loss": 2.4279,
+      "step": 52403
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.17983565107459e-07,
+      "loss": 0.8628,
+      "step": 52404
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 5.179361567635904e-07,
+      "loss": 1.7614,
+      "step": 52405
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.178887484197219e-07,
+      "loss": 2.3,
+      "step": 52406
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.178413400758534e-07,
+      "loss": 1.8463,
+      "step": 52407
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.177939317319849e-07,
+      "loss": 0.6011,
+      "step": 52408
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.177465233881163e-07,
+      "loss": 1.0665,
+      "step": 52409
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.176991150442477e-07,
+      "loss": 2.5129,
+      "step": 52410
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.176517067003793e-07,
+      "loss": 1.5672,
+      "step": 52411
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.176042983565107e-07,
+      "loss": 0.9119,
+      "step": 52412
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.175568900126422e-07,
+      "loss": 1.3193,
+      "step": 52413
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.175094816687737e-07,
+      "loss": 1.0081,
+      "step": 52414
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.174620733249052e-07,
+      "loss": 0.7933,
+      "step": 52415
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.174146649810367e-07,
+      "loss": 1.4347,
+      "step": 52416
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.173672566371682e-07,
+      "loss": 2.0497,
+      "step": 52417
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.173198482932997e-07,
+      "loss": 1.6046,
+      "step": 52418
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.172724399494312e-07,
+      "loss": 1.7475,
+      "step": 52419
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.172250316055625e-07,
+      "loss": 1.9282,
+      "step": 52420
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.17177623261694e-07,
+      "loss": 0.8276,
+      "step": 52421
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.171302149178255e-07,
+      "loss": 1.3634,
+      "step": 52422
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.17082806573957e-07,
+      "loss": 1.2691,
+      "step": 52423
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.170353982300885e-07,
+      "loss": 1.2763,
+      "step": 52424
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.169879898862199e-07,
+      "loss": 0.6323,
+      "step": 52425
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.169405815423515e-07,
+      "loss": 1.4605,
+      "step": 52426
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.16893173198483e-07,
+      "loss": 1.7227,
+      "step": 52427
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.168457648546144e-07,
+      "loss": 2.2347,
+      "step": 52428
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.16798356510746e-07,
+      "loss": 2.4187,
+      "step": 52429
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.167509481668774e-07,
+      "loss": 2.113,
+      "step": 52430
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.167035398230089e-07,
+      "loss": 1.5351,
+      "step": 52431
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.166561314791403e-07,
+      "loss": 2.4457,
+      "step": 52432
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.166087231352718e-07,
+      "loss": 1.4525,
+      "step": 52433
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.165613147914033e-07,
+      "loss": 1.1931,
+      "step": 52434
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.165139064475347e-07,
+      "loss": 1.1609,
+      "step": 52435
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.164664981036662e-07,
+      "loss": 2.6862,
+      "step": 52436
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.164190897597978e-07,
+      "loss": 0.9911,
+      "step": 52437
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.163716814159292e-07,
+      "loss": 1.4702,
+      "step": 52438
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.163242730720607e-07,
+      "loss": 1.5905,
+      "step": 52439
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.162768647281922e-07,
+      "loss": 1.9751,
+      "step": 52440
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.162294563843237e-07,
+      "loss": 1.3688,
+      "step": 52441
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.161820480404552e-07,
+      "loss": 3.409,
+      "step": 52442
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.161346396965866e-07,
+      "loss": 1.2085,
+      "step": 52443
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.160872313527181e-07,
+      "loss": 2.5752,
+      "step": 52444
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.160398230088495e-07,
+      "loss": 2.4854,
+      "step": 52445
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.15992414664981e-07,
+      "loss": 2.3049,
+      "step": 52446
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.159450063211125e-07,
+      "loss": 2.1849,
+      "step": 52447
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.15897597977244e-07,
+      "loss": 1.8087,
+      "step": 52448
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.158501896333755e-07,
+      "loss": 0.9842,
+      "step": 52449
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.15802781289507e-07,
+      "loss": 2.0718,
+      "step": 52450
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.157553729456385e-07,
+      "loss": 1.8496,
+      "step": 52451
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.1570796460177e-07,
+      "loss": 1.5192,
+      "step": 52452
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.156605562579014e-07,
+      "loss": 1.989,
+      "step": 52453
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.156131479140329e-07,
+      "loss": 0.682,
+      "step": 52454
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.155657395701644e-07,
+      "loss": 1.4458,
+      "step": 52455
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.155183312262958e-07,
+      "loss": 2.4272,
+      "step": 52456
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.154709228824273e-07,
+      "loss": 1.3597,
+      "step": 52457
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.154235145385587e-07,
+      "loss": 2.9992,
+      "step": 52458
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.153761061946903e-07,
+      "loss": 2.0077,
+      "step": 52459
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.153286978508218e-07,
+      "loss": 1.2334,
+      "step": 52460
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.152812895069532e-07,
+      "loss": 1.7978,
+      "step": 52461
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.152338811630847e-07,
+      "loss": 1.9332,
+      "step": 52462
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.151864728192162e-07,
+      "loss": 0.6742,
+      "step": 52463
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.151390644753477e-07,
+      "loss": 1.7976,
+      "step": 52464
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.150916561314792e-07,
+      "loss": 1.1217,
+      "step": 52465
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.150442477876107e-07,
+      "loss": 2.0246,
+      "step": 52466
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.149968394437422e-07,
+      "loss": 3.1098,
+      "step": 52467
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 5.149494310998735e-07,
+      "loss": 3.0179,
+      "step": 52468
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.14902022756005e-07,
+      "loss": 1.4697,
+      "step": 52469
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.148546144121366e-07,
+      "loss": 3.3858,
+      "step": 52470
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.14807206068268e-07,
+      "loss": 2.4179,
+      "step": 52471
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.147597977243995e-07,
+      "loss": 2.6714,
+      "step": 52472
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.147123893805309e-07,
+      "loss": 1.832,
+      "step": 52473
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.146649810366625e-07,
+      "loss": 0.7555,
+      "step": 52474
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.14617572692794e-07,
+      "loss": 1.4189,
+      "step": 52475
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.145701643489254e-07,
+      "loss": 1.7142,
+      "step": 52476
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.14522756005057e-07,
+      "loss": 0.9883,
+      "step": 52477
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.144753476611884e-07,
+      "loss": 1.8932,
+      "step": 52478
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.144279393173198e-07,
+      "loss": 2.6047,
+      "step": 52479
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.143805309734513e-07,
+      "loss": 1.8373,
+      "step": 52480
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.143331226295828e-07,
+      "loss": 1.5426,
+      "step": 52481
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.142857142857143e-07,
+      "loss": 2.335,
+      "step": 52482
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.142383059418457e-07,
+      "loss": 2.3186,
+      "step": 52483
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.141908975979772e-07,
+      "loss": 0.5516,
+      "step": 52484
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.141434892541088e-07,
+      "loss": 1.8673,
+      "step": 52485
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.140960809102402e-07,
+      "loss": 1.9522,
+      "step": 52486
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.140486725663717e-07,
+      "loss": 2.4999,
+      "step": 52487
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.140012642225032e-07,
+      "loss": 1.4792,
+      "step": 52488
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.139538558786347e-07,
+      "loss": 1.9235,
+      "step": 52489
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.139064475347662e-07,
+      "loss": 2.8759,
+      "step": 52490
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.138590391908975e-07,
+      "loss": 2.561,
+      "step": 52491
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.138116308470291e-07,
+      "loss": 2.158,
+      "step": 52492
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.137642225031605e-07,
+      "loss": 3.4793,
+      "step": 52493
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.13716814159292e-07,
+      "loss": 1.3028,
+      "step": 52494
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.136694058154235e-07,
+      "loss": 1.5096,
+      "step": 52495
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.13621997471555e-07,
+      "loss": 1.6316,
+      "step": 52496
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.135745891276865e-07,
+      "loss": 1.1177,
+      "step": 52497
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.13527180783818e-07,
+      "loss": 2.274,
+      "step": 52498
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.134797724399495e-07,
+      "loss": 2.3711,
+      "step": 52499
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.13432364096081e-07,
+      "loss": 1.2711,
+      "step": 52500
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.133849557522124e-07,
+      "loss": 1.6786,
+      "step": 52501
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.133375474083438e-07,
+      "loss": 1.2502,
+      "step": 52502
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.132901390644753e-07,
+      "loss": 2.6315,
+      "step": 52503
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.132427307206068e-07,
+      "loss": 2.8075,
+      "step": 52504
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.131953223767383e-07,
+      "loss": 2.054,
+      "step": 52505
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.131479140328697e-07,
+      "loss": 0.7869,
+      "step": 52506
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.131005056890013e-07,
+      "loss": 2.275,
+      "step": 52507
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.130530973451328e-07,
+      "loss": 2.0784,
+      "step": 52508
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.130056890012642e-07,
+      "loss": 2.5354,
+      "step": 52509
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.129582806573958e-07,
+      "loss": 2.2911,
+      "step": 52510
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.129108723135272e-07,
+      "loss": 2.6288,
+      "step": 52511
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.128634639696587e-07,
+      "loss": 1.4215,
+      "step": 52512
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.128160556257902e-07,
+      "loss": 0.7373,
+      "step": 52513
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.127686472819216e-07,
+      "loss": 2.1956,
+      "step": 52514
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.127212389380531e-07,
+      "loss": 1.1826,
+      "step": 52515
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.126738305941845e-07,
+      "loss": 2.1905,
+      "step": 52516
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.12626422250316e-07,
+      "loss": 1.6884,
+      "step": 52517
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.125790139064476e-07,
+      "loss": 2.1987,
+      "step": 52518
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.12531605562579e-07,
+      "loss": 1.6807,
+      "step": 52519
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.124841972187105e-07,
+      "loss": 1.4756,
+      "step": 52520
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.12436788874842e-07,
+      "loss": 2.0421,
+      "step": 52521
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.123893805309735e-07,
+      "loss": 2.484,
+      "step": 52522
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.12341972187105e-07,
+      "loss": 1.3812,
+      "step": 52523
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.122945638432364e-07,
+      "loss": 2.4966,
+      "step": 52524
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.12247155499368e-07,
+      "loss": 1.0176,
+      "step": 52525
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.121997471554993e-07,
+      "loss": 1.6545,
+      "step": 52526
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.121523388116308e-07,
+      "loss": 1.6212,
+      "step": 52527
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.121049304677623e-07,
+      "loss": 2.3,
+      "step": 52528
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.120575221238938e-07,
+      "loss": 2.6923,
+      "step": 52529
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.120101137800253e-07,
+      "loss": 2.2361,
+      "step": 52530
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.119627054361568e-07,
+      "loss": 2.9889,
+      "step": 52531
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 5.119152970922883e-07,
+      "loss": 2.6419,
+      "step": 52532
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.118678887484198e-07,
+      "loss": 3.6032,
+      "step": 52533
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.118204804045512e-07,
+      "loss": 2.4052,
+      "step": 52534
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.117730720606827e-07,
+      "loss": 2.0176,
+      "step": 52535
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.117256637168142e-07,
+      "loss": 1.8942,
+      "step": 52536
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.116782553729457e-07,
+      "loss": 1.7875,
+      "step": 52537
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.116308470290771e-07,
+      "loss": 1.3958,
+      "step": 52538
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.115834386852085e-07,
+      "loss": 1.1663,
+      "step": 52539
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.115360303413401e-07,
+      "loss": 1.4053,
+      "step": 52540
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.114886219974715e-07,
+      "loss": 1.7349,
+      "step": 52541
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.11441213653603e-07,
+      "loss": 1.4746,
+      "step": 52542
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.113938053097346e-07,
+      "loss": 0.9478,
+      "step": 52543
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.11346396965866e-07,
+      "loss": 1.5255,
+      "step": 52544
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.112989886219975e-07,
+      "loss": 2.0508,
+      "step": 52545
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.11251580278129e-07,
+      "loss": 1.9767,
+      "step": 52546
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.112041719342605e-07,
+      "loss": 2.4591,
+      "step": 52547
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.11156763590392e-07,
+      "loss": 1.612,
+      "step": 52548
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.111093552465233e-07,
+      "loss": 1.6051,
+      "step": 52549
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.110619469026548e-07,
+      "loss": 2.0478,
+      "step": 52550
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.110145385587863e-07,
+      "loss": 1.5479,
+      "step": 52551
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.109671302149178e-07,
+      "loss": 2.7602,
+      "step": 52552
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.109197218710493e-07,
+      "loss": 1.7999,
+      "step": 52553
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.108723135271808e-07,
+      "loss": 0.2747,
+      "step": 52554
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.108249051833123e-07,
+      "loss": 0.47,
+      "step": 52555
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.107774968394438e-07,
+      "loss": 1.7336,
+      "step": 52556
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.107300884955752e-07,
+      "loss": 1.5963,
+      "step": 52557
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.106826801517068e-07,
+      "loss": 2.9758,
+      "step": 52558
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.106352718078382e-07,
+      "loss": 2.079,
+      "step": 52559
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.105878634639697e-07,
+      "loss": 1.0059,
+      "step": 52560
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.105404551201011e-07,
+      "loss": 1.8918,
+      "step": 52561
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.104930467762326e-07,
+      "loss": 0.8901,
+      "step": 52562
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.104456384323641e-07,
+      "loss": 2.3434,
+      "step": 52563
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.103982300884955e-07,
+      "loss": 1.4143,
+      "step": 52564
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.103508217446271e-07,
+      "loss": 0.4549,
+      "step": 52565
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.103034134007586e-07,
+      "loss": 3.6965,
+      "step": 52566
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.1025600505689e-07,
+      "loss": 0.8302,
+      "step": 52567
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.102085967130215e-07,
+      "loss": 2.3081,
+      "step": 52568
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.10161188369153e-07,
+      "loss": 2.2245,
+      "step": 52569
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.101137800252845e-07,
+      "loss": 1.4128,
+      "step": 52570
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.10066371681416e-07,
+      "loss": 1.3644,
+      "step": 52571
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.100189633375473e-07,
+      "loss": 0.9726,
+      "step": 52572
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.099715549936789e-07,
+      "loss": 1.9898,
+      "step": 52573
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.099241466498103e-07,
+      "loss": 1.1008,
+      "step": 52574
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.098767383059418e-07,
+      "loss": 2.1931,
+      "step": 52575
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.098293299620734e-07,
+      "loss": 3.3714,
+      "step": 52576
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.097819216182048e-07,
+      "loss": 0.5035,
+      "step": 52577
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.097345132743363e-07,
+      "loss": 0.5832,
+      "step": 52578
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.096871049304678e-07,
+      "loss": 1.409,
+      "step": 52579
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.096396965865993e-07,
+      "loss": 3.1288,
+      "step": 52580
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.095922882427308e-07,
+      "loss": 1.7913,
+      "step": 52581
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.095448798988622e-07,
+      "loss": 2.4493,
+      "step": 52582
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.094974715549937e-07,
+      "loss": 2.7062,
+      "step": 52583
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.094500632111251e-07,
+      "loss": 1.7908,
+      "step": 52584
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.094026548672566e-07,
+      "loss": 1.9716,
+      "step": 52585
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.093552465233881e-07,
+      "loss": 1.7139,
+      "step": 52586
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.093078381795196e-07,
+      "loss": 1.3192,
+      "step": 52587
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.092604298356511e-07,
+      "loss": 1.2865,
+      "step": 52588
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.092130214917826e-07,
+      "loss": 2.1977,
+      "step": 52589
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.09165613147914e-07,
+      "loss": 2.2242,
+      "step": 52590
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.091182048040456e-07,
+      "loss": 1.031,
+      "step": 52591
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.09070796460177e-07,
+      "loss": 2.7277,
+      "step": 52592
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.090233881163085e-07,
+      "loss": 1.322,
+      "step": 52593
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.0897597977244e-07,
+      "loss": 2.2729,
+      "step": 52594
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 5.089285714285715e-07,
+      "loss": 2.3099,
+      "step": 52595
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.08881163084703e-07,
+      "loss": 2.86,
+      "step": 52596
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.088337547408343e-07,
+      "loss": 1.0844,
+      "step": 52597
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.087863463969659e-07,
+      "loss": 2.5139,
+      "step": 52598
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.087389380530974e-07,
+      "loss": 1.8353,
+      "step": 52599
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.086915297092288e-07,
+      "loss": 2.0671,
+      "step": 52600
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.086441213653603e-07,
+      "loss": 2.7382,
+      "step": 52601
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.085967130214918e-07,
+      "loss": 1.4449,
+      "step": 52602
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.085493046776233e-07,
+      "loss": 3.2347,
+      "step": 52603
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.085018963337548e-07,
+      "loss": 1.3487,
+      "step": 52604
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.084544879898862e-07,
+      "loss": 0.244,
+      "step": 52605
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.084070796460178e-07,
+      "loss": 1.9755,
+      "step": 52606
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.083596713021492e-07,
+      "loss": 1.9276,
+      "step": 52607
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.083122629582806e-07,
+      "loss": 1.3716,
+      "step": 52608
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.082648546144122e-07,
+      "loss": 2.3333,
+      "step": 52609
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.082174462705436e-07,
+      "loss": 1.7501,
+      "step": 52610
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.081700379266751e-07,
+      "loss": 2.2912,
+      "step": 52611
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.081226295828065e-07,
+      "loss": 1.3483,
+      "step": 52612
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.080752212389381e-07,
+      "loss": 2.5259,
+      "step": 52613
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.080278128950696e-07,
+      "loss": 1.9631,
+      "step": 52614
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.07980404551201e-07,
+      "loss": 0.9966,
+      "step": 52615
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.079329962073325e-07,
+      "loss": 1.7985,
+      "step": 52616
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.07885587863464e-07,
+      "loss": 3.6834,
+      "step": 52617
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.078381795195955e-07,
+      "loss": 2.645,
+      "step": 52618
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.07790771175727e-07,
+      "loss": 1.6961,
+      "step": 52619
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.077433628318584e-07,
+      "loss": 2.4038,
+      "step": 52620
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.076959544879899e-07,
+      "loss": 2.3265,
+      "step": 52621
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.076485461441213e-07,
+      "loss": 1.0972,
+      "step": 52622
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.076011378002528e-07,
+      "loss": 1.9727,
+      "step": 52623
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.075537294563844e-07,
+      "loss": 4.1025,
+      "step": 52624
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.075063211125158e-07,
+      "loss": 0.9513,
+      "step": 52625
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.074589127686473e-07,
+      "loss": 1.8016,
+      "step": 52626
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.074115044247788e-07,
+      "loss": 1.5097,
+      "step": 52627
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.073640960809103e-07,
+      "loss": 2.5926,
+      "step": 52628
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.073166877370418e-07,
+      "loss": 1.9005,
+      "step": 52629
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.072692793931732e-07,
+      "loss": 2.7864,
+      "step": 52630
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.072218710493048e-07,
+      "loss": 2.1625,
+      "step": 52631
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.071744627054361e-07,
+      "loss": 3.0173,
+      "step": 52632
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.071270543615676e-07,
+      "loss": 2.7971,
+      "step": 52633
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.070796460176991e-07,
+      "loss": 0.911,
+      "step": 52634
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.070322376738306e-07,
+      "loss": 4.1366,
+      "step": 52635
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.069848293299621e-07,
+      "loss": 1.1894,
+      "step": 52636
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.069374209860936e-07,
+      "loss": 1.8609,
+      "step": 52637
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.06890012642225e-07,
+      "loss": 1.8841,
+      "step": 52638
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.068426042983566e-07,
+      "loss": 2.3997,
+      "step": 52639
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.06795195954488e-07,
+      "loss": 1.6314,
+      "step": 52640
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.067477876106195e-07,
+      "loss": 2.9611,
+      "step": 52641
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.06700379266751e-07,
+      "loss": 1.5264,
+      "step": 52642
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.066529709228824e-07,
+      "loss": 1.3259,
+      "step": 52643
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.066055625790139e-07,
+      "loss": 0.8479,
+      "step": 52644
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.065581542351453e-07,
+      "loss": 1.1853,
+      "step": 52645
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.065107458912769e-07,
+      "loss": 3.5471,
+      "step": 52646
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.064633375474084e-07,
+      "loss": 2.5685,
+      "step": 52647
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.064159292035398e-07,
+      "loss": 2.1693,
+      "step": 52648
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.063685208596713e-07,
+      "loss": 0.8702,
+      "step": 52649
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.063211125158028e-07,
+      "loss": 1.4442,
+      "step": 52650
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.062737041719343e-07,
+      "loss": 0.5974,
+      "step": 52651
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.062262958280658e-07,
+      "loss": 1.0423,
+      "step": 52652
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.061788874841973e-07,
+      "loss": 2.3906,
+      "step": 52653
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.061314791403288e-07,
+      "loss": 0.565,
+      "step": 52654
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.060840707964601e-07,
+      "loss": 1.7567,
+      "step": 52655
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.060366624525916e-07,
+      "loss": 2.854,
+      "step": 52656
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.059892541087232e-07,
+      "loss": 1.169,
+      "step": 52657
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 5.059418457648546e-07,
+      "loss": 1.0439,
+      "step": 52658
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.058944374209861e-07,
+      "loss": 1.9058,
+      "step": 52659
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.058470290771176e-07,
+      "loss": 1.4294,
+      "step": 52660
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.057996207332491e-07,
+      "loss": 3.1067,
+      "step": 52661
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.057522123893806e-07,
+      "loss": 1.0921,
+      "step": 52662
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.05704804045512e-07,
+      "loss": 0.8039,
+      "step": 52663
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.056573957016436e-07,
+      "loss": 2.4567,
+      "step": 52664
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.05609987357775e-07,
+      "loss": 3.1782,
+      "step": 52665
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.055625790139064e-07,
+      "loss": 2.6431,
+      "step": 52666
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.055151706700379e-07,
+      "loss": 1.507,
+      "step": 52667
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.054677623261694e-07,
+      "loss": 1.4713,
+      "step": 52668
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.054203539823009e-07,
+      "loss": 1.7046,
+      "step": 52669
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.053729456384324e-07,
+      "loss": 1.9562,
+      "step": 52670
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.053255372945638e-07,
+      "loss": 1.8818,
+      "step": 52671
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.052781289506954e-07,
+      "loss": 1.0076,
+      "step": 52672
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.052307206068268e-07,
+      "loss": 1.7619,
+      "step": 52673
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.051833122629583e-07,
+      "loss": 3.12,
+      "step": 52674
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.051359039190898e-07,
+      "loss": 1.7468,
+      "step": 52675
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.050884955752213e-07,
+      "loss": 1.3517,
+      "step": 52676
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.050410872313528e-07,
+      "loss": 1.9037,
+      "step": 52677
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.049936788874841e-07,
+      "loss": 3.4943,
+      "step": 52678
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.049462705436157e-07,
+      "loss": 2.2218,
+      "step": 52679
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.048988621997472e-07,
+      "loss": 2.6813,
+      "step": 52680
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.048514538558786e-07,
+      "loss": 2.7316,
+      "step": 52681
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.048040455120101e-07,
+      "loss": 1.7893,
+      "step": 52682
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.047566371681416e-07,
+      "loss": 1.4419,
+      "step": 52683
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.047092288242731e-07,
+      "loss": 1.7852,
+      "step": 52684
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.046618204804046e-07,
+      "loss": 2.603,
+      "step": 52685
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.04614412136536e-07,
+      "loss": 2.1277,
+      "step": 52686
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.045670037926676e-07,
+      "loss": 0.5803,
+      "step": 52687
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.04519595448799e-07,
+      "loss": 1.105,
+      "step": 52688
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.044721871049305e-07,
+      "loss": 0.993,
+      "step": 52689
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.044247787610619e-07,
+      "loss": 2.0972,
+      "step": 52690
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.043773704171934e-07,
+      "loss": 1.1037,
+      "step": 52691
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.043299620733249e-07,
+      "loss": 0.9046,
+      "step": 52692
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.042825537294563e-07,
+      "loss": 2.5992,
+      "step": 52693
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.042351453855879e-07,
+      "loss": 2.9202,
+      "step": 52694
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.041877370417194e-07,
+      "loss": 3.8157,
+      "step": 52695
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.041403286978508e-07,
+      "loss": 1.6779,
+      "step": 52696
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.040929203539823e-07,
+      "loss": 1.3442,
+      "step": 52697
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.040455120101138e-07,
+      "loss": 2.3635,
+      "step": 52698
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.039981036662453e-07,
+      "loss": 1.6905,
+      "step": 52699
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.039506953223768e-07,
+      "loss": 1.973,
+      "step": 52700
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.039032869785083e-07,
+      "loss": 1.5852,
+      "step": 52701
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.038558786346397e-07,
+      "loss": 0.9019,
+      "step": 52702
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.038084702907711e-07,
+      "loss": 1.8384,
+      "step": 52703
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.037610619469026e-07,
+      "loss": 1.6126,
+      "step": 52704
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.037136536030342e-07,
+      "loss": 2.509,
+      "step": 52705
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.036662452591656e-07,
+      "loss": 1.3214,
+      "step": 52706
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.036188369152971e-07,
+      "loss": 1.3082,
+      "step": 52707
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.035714285714286e-07,
+      "loss": 1.3062,
+      "step": 52708
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.035240202275601e-07,
+      "loss": 1.1508,
+      "step": 52709
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.034766118836916e-07,
+      "loss": 2.5967,
+      "step": 52710
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.03429203539823e-07,
+      "loss": 0.5051,
+      "step": 52711
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.033817951959546e-07,
+      "loss": 1.1347,
+      "step": 52712
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.033343868520859e-07,
+      "loss": 1.0857,
+      "step": 52713
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.032869785082174e-07,
+      "loss": 1.9307,
+      "step": 52714
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.032395701643489e-07,
+      "loss": 2.5599,
+      "step": 52715
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.031921618204804e-07,
+      "loss": 3.4127,
+      "step": 52716
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.031447534766119e-07,
+      "loss": 2.561,
+      "step": 52717
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.030973451327434e-07,
+      "loss": 2.046,
+      "step": 52718
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.030499367888748e-07,
+      "loss": 1.8315,
+      "step": 52719
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.030025284450064e-07,
+      "loss": 1.2725,
+      "step": 52720
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.029551201011378e-07,
+      "loss": 2.0631,
+      "step": 52721
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 5.029077117572693e-07,
+      "loss": 2.7904,
+      "step": 52722
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.028603034134008e-07,
+      "loss": 1.3695,
+      "step": 52723
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.028128950695323e-07,
+      "loss": 1.5989,
+      "step": 52724
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.027654867256637e-07,
+      "loss": 2.5695,
+      "step": 52725
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.027180783817951e-07,
+      "loss": 1.6611,
+      "step": 52726
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.026706700379267e-07,
+      "loss": 2.3646,
+      "step": 52727
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.026232616940582e-07,
+      "loss": 3.1568,
+      "step": 52728
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.025758533501896e-07,
+      "loss": 1.1089,
+      "step": 52729
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.025284450063211e-07,
+      "loss": 1.6964,
+      "step": 52730
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.024810366624526e-07,
+      "loss": 1.4548,
+      "step": 52731
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.024336283185841e-07,
+      "loss": 0.6176,
+      "step": 52732
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.023862199747156e-07,
+      "loss": 1.8982,
+      "step": 52733
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.023388116308471e-07,
+      "loss": 1.9016,
+      "step": 52734
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.022914032869786e-07,
+      "loss": 3.138,
+      "step": 52735
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.0224399494311e-07,
+      "loss": 2.7335,
+      "step": 52736
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.021965865992414e-07,
+      "loss": 1.6597,
+      "step": 52737
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.02149178255373e-07,
+      "loss": 1.2951,
+      "step": 52738
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.021017699115044e-07,
+      "loss": 2.5163,
+      "step": 52739
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.020543615676359e-07,
+      "loss": 2.1324,
+      "step": 52740
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.020069532237674e-07,
+      "loss": 2.113,
+      "step": 52741
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.019595448798989e-07,
+      "loss": 1.888,
+      "step": 52742
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.019121365360304e-07,
+      "loss": 1.6449,
+      "step": 52743
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.018647281921618e-07,
+      "loss": 2.0009,
+      "step": 52744
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.018173198482934e-07,
+      "loss": 2.502,
+      "step": 52745
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.017699115044248e-07,
+      "loss": 1.6233,
+      "step": 52746
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.017225031605563e-07,
+      "loss": 1.3173,
+      "step": 52747
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.016750948166877e-07,
+      "loss": 2.2451,
+      "step": 52748
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.016276864728192e-07,
+      "loss": 1.3209,
+      "step": 52749
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.015802781289507e-07,
+      "loss": 0.9539,
+      "step": 52750
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.015328697850821e-07,
+      "loss": 3.7375,
+      "step": 52751
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.014854614412136e-07,
+      "loss": 1.1711,
+      "step": 52752
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.014380530973452e-07,
+      "loss": 2.2666,
+      "step": 52753
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.013906447534766e-07,
+      "loss": 1.1417,
+      "step": 52754
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.013432364096081e-07,
+      "loss": 2.6727,
+      "step": 52755
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.012958280657396e-07,
+      "loss": 1.7053,
+      "step": 52756
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.012484197218711e-07,
+      "loss": 1.2711,
+      "step": 52757
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.012010113780026e-07,
+      "loss": 1.8658,
+      "step": 52758
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.01153603034134e-07,
+      "loss": 1.1351,
+      "step": 52759
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.011061946902655e-07,
+      "loss": 1.7928,
+      "step": 52760
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.010587863463969e-07,
+      "loss": 2.7266,
+      "step": 52761
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.010113780025284e-07,
+      "loss": 3.0013,
+      "step": 52762
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.009639696586599e-07,
+      "loss": 1.848,
+      "step": 52763
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.009165613147914e-07,
+      "loss": 2.6188,
+      "step": 52764
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.008691529709229e-07,
+      "loss": 1.8359,
+      "step": 52765
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.008217446270544e-07,
+      "loss": 1.7683,
+      "step": 52766
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.007743362831859e-07,
+      "loss": 2.8901,
+      "step": 52767
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.007269279393174e-07,
+      "loss": 2.3936,
+      "step": 52768
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.006795195954488e-07,
+      "loss": 4.4396,
+      "step": 52769
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.006321112515803e-07,
+      "loss": 1.9569,
+      "step": 52770
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.005847029077118e-07,
+      "loss": 1.8729,
+      "step": 52771
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.005372945638432e-07,
+      "loss": 2.9708,
+      "step": 52772
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.004898862199747e-07,
+      "loss": 1.5356,
+      "step": 52773
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.004424778761061e-07,
+      "loss": 1.5029,
+      "step": 52774
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.003950695322377e-07,
+      "loss": 1.6588,
+      "step": 52775
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.003476611883692e-07,
+      "loss": 0.8336,
+      "step": 52776
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.003002528445006e-07,
+      "loss": 2.0112,
+      "step": 52777
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.002528445006322e-07,
+      "loss": 1.5363,
+      "step": 52778
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.002054361567636e-07,
+      "loss": 1.2121,
+      "step": 52779
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.001580278128951e-07,
+      "loss": 2.6132,
+      "step": 52780
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.001106194690266e-07,
+      "loss": 2.6836,
+      "step": 52781
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.000632111251581e-07,
+      "loss": 1.0872,
+      "step": 52782
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 5.000158027812896e-07,
+      "loss": 1.2729,
+      "step": 52783
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.999683944374209e-07,
+      "loss": 1.111,
+      "step": 52784
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.999209860935524e-07,
+      "loss": 1.5683,
+      "step": 52785
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.99873577749684e-07,
+      "loss": 0.9575,
+      "step": 52786
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.998261694058154e-07,
+      "loss": 1.4838,
+      "step": 52787
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.997787610619469e-07,
+      "loss": 1.2214,
+      "step": 52788
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.997313527180784e-07,
+      "loss": 2.1368,
+      "step": 52789
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.996839443742099e-07,
+      "loss": 1.7771,
+      "step": 52790
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.996365360303414e-07,
+      "loss": 1.8746,
+      "step": 52791
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.995891276864728e-07,
+      "loss": 1.8566,
+      "step": 52792
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.995417193426044e-07,
+      "loss": 1.9822,
+      "step": 52793
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.994943109987358e-07,
+      "loss": 2.0286,
+      "step": 52794
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.994469026548672e-07,
+      "loss": 3.2378,
+      "step": 52795
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.993994943109987e-07,
+      "loss": 2.3512,
+      "step": 52796
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.993520859671302e-07,
+      "loss": 1.626,
+      "step": 52797
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.993046776232617e-07,
+      "loss": 2.0679,
+      "step": 52798
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.992572692793932e-07,
+      "loss": 1.5908,
+      "step": 52799
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.992098609355247e-07,
+      "loss": 1.4267,
+      "step": 52800
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.991624525916562e-07,
+      "loss": 2.0506,
+      "step": 52801
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.991150442477876e-07,
+      "loss": 1.8848,
+      "step": 52802
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.990676359039191e-07,
+      "loss": 1.5478,
+      "step": 52803
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.990202275600506e-07,
+      "loss": 2.9089,
+      "step": 52804
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.989728192161821e-07,
+      "loss": 2.2202,
+      "step": 52805
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.989254108723136e-07,
+      "loss": 1.7933,
+      "step": 52806
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.988780025284449e-07,
+      "loss": 1.8361,
+      "step": 52807
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.988305941845765e-07,
+      "loss": 1.3493,
+      "step": 52808
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.98783185840708e-07,
+      "loss": 2.1644,
+      "step": 52809
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.987357774968394e-07,
+      "loss": 2.0946,
+      "step": 52810
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.98688369152971e-07,
+      "loss": 1.8655,
+      "step": 52811
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.986409608091024e-07,
+      "loss": 2.2865,
+      "step": 52812
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.985935524652339e-07,
+      "loss": 2.2157,
+      "step": 52813
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.985461441213654e-07,
+      "loss": 0.3899,
+      "step": 52814
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.984987357774969e-07,
+      "loss": 2.4805,
+      "step": 52815
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.984513274336284e-07,
+      "loss": 1.442,
+      "step": 52816
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.984039190897598e-07,
+      "loss": 1.1228,
+      "step": 52817
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.983565107458912e-07,
+      "loss": 2.9211,
+      "step": 52818
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.983091024020228e-07,
+      "loss": 2.7959,
+      "step": 52819
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.982616940581542e-07,
+      "loss": 1.2092,
+      "step": 52820
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.982142857142857e-07,
+      "loss": 2.32,
+      "step": 52821
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.981668773704172e-07,
+      "loss": 1.038,
+      "step": 52822
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.981194690265487e-07,
+      "loss": 1.2357,
+      "step": 52823
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.980720606826802e-07,
+      "loss": 2.8992,
+      "step": 52824
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.980246523388116e-07,
+      "loss": 1.3238,
+      "step": 52825
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.979772439949432e-07,
+      "loss": 2.2416,
+      "step": 52826
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.979298356510746e-07,
+      "loss": 1.8492,
+      "step": 52827
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.978824273072061e-07,
+      "loss": 3.3326,
+      "step": 52828
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.978350189633376e-07,
+      "loss": 2.5875,
+      "step": 52829
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.97787610619469e-07,
+      "loss": 3.5401,
+      "step": 52830
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.977402022756005e-07,
+      "loss": 2.0077,
+      "step": 52831
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.976927939317319e-07,
+      "loss": 1.4668,
+      "step": 52832
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.976453855878635e-07,
+      "loss": 1.9731,
+      "step": 52833
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.97597977243995e-07,
+      "loss": 2.0721,
+      "step": 52834
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.975505689001264e-07,
+      "loss": 1.3931,
+      "step": 52835
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.975031605562579e-07,
+      "loss": 1.5768,
+      "step": 52836
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.974557522123894e-07,
+      "loss": 2.2122,
+      "step": 52837
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.974083438685209e-07,
+      "loss": 0.9367,
+      "step": 52838
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.973609355246524e-07,
+      "loss": 1.4979,
+      "step": 52839
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.973135271807838e-07,
+      "loss": 0.4303,
+      "step": 52840
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.972661188369154e-07,
+      "loss": 2.4957,
+      "step": 52841
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.972187104930467e-07,
+      "loss": 1.0878,
+      "step": 52842
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.971713021491782e-07,
+      "loss": 0.9734,
+      "step": 52843
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.971238938053098e-07,
+      "loss": 1.3891,
+      "step": 52844
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.970764854614412e-07,
+      "loss": 0.7746,
+      "step": 52845
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.970290771175727e-07,
+      "loss": 1.4617,
+      "step": 52846
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.969816687737042e-07,
+      "loss": 0.6773,
+      "step": 52847
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.969342604298357e-07,
+      "loss": 2.1029,
+      "step": 52848
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.968868520859672e-07,
+      "loss": 1.5851,
+      "step": 52849
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.968394437420986e-07,
+      "loss": 2.3046,
+      "step": 52850
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.967920353982301e-07,
+      "loss": 2.4415,
+      "step": 52851
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.967446270543616e-07,
+      "loss": 2.6311,
+      "step": 52852
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.96697218710493e-07,
+      "loss": 2.8389,
+      "step": 52853
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.966498103666245e-07,
+      "loss": 1.0972,
+      "step": 52854
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.96602402022756e-07,
+      "loss": 1.9072,
+      "step": 52855
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.965549936788875e-07,
+      "loss": 2.9357,
+      "step": 52856
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.96507585335019e-07,
+      "loss": 1.8452,
+      "step": 52857
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.964601769911504e-07,
+      "loss": 1.6079,
+      "step": 52858
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.96412768647282e-07,
+      "loss": 2.4495,
+      "step": 52859
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.963653603034134e-07,
+      "loss": 3.1449,
+      "step": 52860
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.963179519595449e-07,
+      "loss": 0.4498,
+      "step": 52861
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.962705436156764e-07,
+      "loss": 2.9441,
+      "step": 52862
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.962231352718079e-07,
+      "loss": 2.7938,
+      "step": 52863
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.961757269279394e-07,
+      "loss": 3.2958,
+      "step": 52864
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.961283185840707e-07,
+      "loss": 1.874,
+      "step": 52865
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.960809102402023e-07,
+      "loss": 1.6325,
+      "step": 52866
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.960335018963338e-07,
+      "loss": 1.4494,
+      "step": 52867
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.959860935524652e-07,
+      "loss": 1.4408,
+      "step": 52868
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.959386852085967e-07,
+      "loss": 1.9438,
+      "step": 52869
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.958912768647282e-07,
+      "loss": 1.6465,
+      "step": 52870
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.958438685208597e-07,
+      "loss": 1.0984,
+      "step": 52871
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.957964601769912e-07,
+      "loss": 3.5023,
+      "step": 52872
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.957490518331226e-07,
+      "loss": 2.2472,
+      "step": 52873
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.957016434892542e-07,
+      "loss": 2.1819,
+      "step": 52874
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.956542351453856e-07,
+      "loss": 1.3907,
+      "step": 52875
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.95606826801517e-07,
+      "loss": 2.3316,
+      "step": 52876
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.955594184576487e-07,
+      "loss": 1.2585,
+      "step": 52877
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.9551201011378e-07,
+      "loss": 0.9222,
+      "step": 52878
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.954646017699115e-07,
+      "loss": 0.4544,
+      "step": 52879
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.95417193426043e-07,
+      "loss": 2.0151,
+      "step": 52880
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.953697850821745e-07,
+      "loss": 2.3042,
+      "step": 52881
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.95322376738306e-07,
+      "loss": 2.5232,
+      "step": 52882
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.952749683944374e-07,
+      "loss": 2.5049,
+      "step": 52883
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.952275600505689e-07,
+      "loss": 2.3358,
+      "step": 52884
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.951801517067004e-07,
+      "loss": 0.9764,
+      "step": 52885
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.951327433628319e-07,
+      "loss": 1.3979,
+      "step": 52886
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.950853350189634e-07,
+      "loss": 1.5669,
+      "step": 52887
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.950379266750949e-07,
+      "loss": 3.6099,
+      "step": 52888
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.949905183312263e-07,
+      "loss": 1.6977,
+      "step": 52889
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.949431099873577e-07,
+      "loss": 1.947,
+      "step": 52890
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.948957016434892e-07,
+      "loss": 1.3803,
+      "step": 52891
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.948482932996208e-07,
+      "loss": 1.8439,
+      "step": 52892
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.948008849557522e-07,
+      "loss": 2.4247,
+      "step": 52893
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.947534766118837e-07,
+      "loss": 2.2429,
+      "step": 52894
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.947060682680152e-07,
+      "loss": 2.4952,
+      "step": 52895
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.946586599241467e-07,
+      "loss": 2.0119,
+      "step": 52896
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.946112515802782e-07,
+      "loss": 1.6298,
+      "step": 52897
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.945638432364096e-07,
+      "loss": 1.1067,
+      "step": 52898
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.945164348925411e-07,
+      "loss": 0.898,
+      "step": 52899
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.944690265486725e-07,
+      "loss": 1.6864,
+      "step": 52900
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.94421618204804e-07,
+      "loss": 1.1355,
+      "step": 52901
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.943742098609355e-07,
+      "loss": 2.6861,
+      "step": 52902
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.94326801517067e-07,
+      "loss": 2.8541,
+      "step": 52903
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.942793931731985e-07,
+      "loss": 1.3272,
+      "step": 52904
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.9423198482933e-07,
+      "loss": 1.96,
+      "step": 52905
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.941845764854614e-07,
+      "loss": 1.7198,
+      "step": 52906
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.94137168141593e-07,
+      "loss": 1.1981,
+      "step": 52907
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.940897597977244e-07,
+      "loss": 3.3761,
+      "step": 52908
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.940423514538559e-07,
+      "loss": 2.2113,
+      "step": 52909
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.939949431099874e-07,
+      "loss": 1.3513,
+      "step": 52910
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.939475347661189e-07,
+      "loss": 2.6936,
+      "step": 52911
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.939001264222503e-07,
+      "loss": 0.8139,
+      "step": 52912
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.938527180783817e-07,
+      "loss": 2.363,
+      "step": 52913
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.938053097345133e-07,
+      "loss": 1.0941,
+      "step": 52914
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.937579013906448e-07,
+      "loss": 0.9194,
+      "step": 52915
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.937104930467762e-07,
+      "loss": 1.6928,
+      "step": 52916
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.936630847029077e-07,
+      "loss": 2.9682,
+      "step": 52917
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.936156763590392e-07,
+      "loss": 1.8821,
+      "step": 52918
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.935682680151707e-07,
+      "loss": 2.4144,
+      "step": 52919
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.935208596713022e-07,
+      "loss": 1.0856,
+      "step": 52920
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.934734513274336e-07,
+      "loss": 1.5223,
+      "step": 52921
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.934260429835652e-07,
+      "loss": 1.4142,
+      "step": 52922
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.933786346396965e-07,
+      "loss": 1.2684,
+      "step": 52923
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.93331226295828e-07,
+      "loss": 3.9677,
+      "step": 52924
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.932838179519596e-07,
+      "loss": 4.0813,
+      "step": 52925
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.93236409608091e-07,
+      "loss": 1.072,
+      "step": 52926
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.931890012642225e-07,
+      "loss": 2.075,
+      "step": 52927
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.93141592920354e-07,
+      "loss": 3.2171,
+      "step": 52928
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.930941845764855e-07,
+      "loss": 2.9642,
+      "step": 52929
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.93046776232617e-07,
+      "loss": 1.472,
+      "step": 52930
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.929993678887484e-07,
+      "loss": 1.8824,
+      "step": 52931
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.929519595448799e-07,
+      "loss": 2.2941,
+      "step": 52932
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.929045512010114e-07,
+      "loss": 1.8739,
+      "step": 52933
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.928571428571429e-07,
+      "loss": 2.0451,
+      "step": 52934
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.928097345132743e-07,
+      "loss": 1.546,
+      "step": 52935
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.927623261694058e-07,
+      "loss": 1.9736,
+      "step": 52936
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.927149178255373e-07,
+      "loss": 1.2715,
+      "step": 52937
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.926675094816688e-07,
+      "loss": 1.2009,
+      "step": 52938
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.926201011378002e-07,
+      "loss": 0.974,
+      "step": 52939
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.925726927939318e-07,
+      "loss": 1.2099,
+      "step": 52940
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.925252844500632e-07,
+      "loss": 2.0558,
+      "step": 52941
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.924778761061947e-07,
+      "loss": 1.9711,
+      "step": 52942
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.924304677623262e-07,
+      "loss": 2.689,
+      "step": 52943
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.923830594184577e-07,
+      "loss": 1.8475,
+      "step": 52944
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.923356510745892e-07,
+      "loss": 3.6289,
+      "step": 52945
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.922882427307206e-07,
+      "loss": 3.4241,
+      "step": 52946
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.922408343868521e-07,
+      "loss": 1.8077,
+      "step": 52947
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.921934260429836e-07,
+      "loss": 1.6986,
+      "step": 52948
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.92146017699115e-07,
+      "loss": 1.998,
+      "step": 52949
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.920986093552465e-07,
+      "loss": 1.7938,
+      "step": 52950
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.92051201011378e-07,
+      "loss": 2.2399,
+      "step": 52951
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.920037926675095e-07,
+      "loss": 1.9651,
+      "step": 52952
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.91956384323641e-07,
+      "loss": 1.5878,
+      "step": 52953
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.919089759797724e-07,
+      "loss": 2.1397,
+      "step": 52954
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.91861567635904e-07,
+      "loss": 1.941,
+      "step": 52955
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.918141592920354e-07,
+      "loss": 2.8999,
+      "step": 52956
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.917667509481669e-07,
+      "loss": 1.7088,
+      "step": 52957
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.917193426042985e-07,
+      "loss": 3.1903,
+      "step": 52958
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.916719342604298e-07,
+      "loss": 1.8769,
+      "step": 52959
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.916245259165613e-07,
+      "loss": 0.9394,
+      "step": 52960
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.915771175726927e-07,
+      "loss": 0.9989,
+      "step": 52961
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.915297092288243e-07,
+      "loss": 0.8243,
+      "step": 52962
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.914823008849558e-07,
+      "loss": 2.0461,
+      "step": 52963
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.914348925410872e-07,
+      "loss": 1.8426,
+      "step": 52964
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.913874841972187e-07,
+      "loss": 1.0251,
+      "step": 52965
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.913400758533502e-07,
+      "loss": 0.6123,
+      "step": 52966
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.912926675094817e-07,
+      "loss": 1.6565,
+      "step": 52967
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.912452591656132e-07,
+      "loss": 2.0581,
+      "step": 52968
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.911978508217447e-07,
+      "loss": 0.9607,
+      "step": 52969
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.911504424778762e-07,
+      "loss": 1.8507,
+      "step": 52970
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.911030341340075e-07,
+      "loss": 2.0892,
+      "step": 52971
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.91055625790139e-07,
+      "loss": 2.1663,
+      "step": 52972
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.910082174462706e-07,
+      "loss": 1.3568,
+      "step": 52973
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.90960809102402e-07,
+      "loss": 1.5909,
+      "step": 52974
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.909134007585335e-07,
+      "loss": 1.3609,
+      "step": 52975
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.90865992414665e-07,
+      "loss": 2.4241,
+      "step": 52976
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.908185840707965e-07,
+      "loss": 1.6739,
+      "step": 52977
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.90771175726928e-07,
+      "loss": 2.2019,
+      "step": 52978
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.907237673830594e-07,
+      "loss": 2.8561,
+      "step": 52979
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.90676359039191e-07,
+      "loss": 1.6353,
+      "step": 52980
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.906289506953224e-07,
+      "loss": 1.9307,
+      "step": 52981
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.905815423514538e-07,
+      "loss": 1.8444,
+      "step": 52982
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.905341340075853e-07,
+      "loss": 1.9472,
+      "step": 52983
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.904867256637168e-07,
+      "loss": 1.1281,
+      "step": 52984
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.904393173198483e-07,
+      "loss": 2.9148,
+      "step": 52985
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.903919089759798e-07,
+      "loss": 0.7861,
+      "step": 52986
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.903445006321112e-07,
+      "loss": 0.835,
+      "step": 52987
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.902970922882428e-07,
+      "loss": 0.7851,
+      "step": 52988
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.902496839443742e-07,
+      "loss": 2.1024,
+      "step": 52989
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.902022756005057e-07,
+      "loss": 1.6697,
+      "step": 52990
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.901548672566372e-07,
+      "loss": 0.3759,
+      "step": 52991
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.901074589127687e-07,
+      "loss": 1.1361,
+      "step": 52992
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.900600505689002e-07,
+      "loss": 1.5609,
+      "step": 52993
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.900126422250315e-07,
+      "loss": 1.6642,
+      "step": 52994
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.899652338811631e-07,
+      "loss": 2.933,
+      "step": 52995
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.899178255372946e-07,
+      "loss": 1.4194,
+      "step": 52996
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.89870417193426e-07,
+      "loss": 2.3793,
+      "step": 52997
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.898230088495575e-07,
+      "loss": 2.1137,
+      "step": 52998
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.89775600505689e-07,
+      "loss": 3.3635,
+      "step": 52999
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.897281921618205e-07,
+      "loss": 1.9921,
+      "step": 53000
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.89680783817952e-07,
+      "loss": 1.3559,
+      "step": 53001
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.896333754740835e-07,
+      "loss": 1.0344,
+      "step": 53002
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.89585967130215e-07,
+      "loss": 1.1857,
+      "step": 53003
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.895385587863464e-07,
+      "loss": 2.8984,
+      "step": 53004
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.894911504424778e-07,
+      "loss": 2.9953,
+      "step": 53005
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.894437420986094e-07,
+      "loss": 1.8888,
+      "step": 53006
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.893963337547408e-07,
+      "loss": 1.1632,
+      "step": 53007
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.893489254108723e-07,
+      "loss": 1.3268,
+      "step": 53008
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.893015170670038e-07,
+      "loss": 2.281,
+      "step": 53009
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.892541087231353e-07,
+      "loss": 1.4906,
+      "step": 53010
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.892067003792668e-07,
+      "loss": 1.3909,
+      "step": 53011
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.891592920353982e-07,
+      "loss": 3.2323,
+      "step": 53012
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.891118836915298e-07,
+      "loss": 2.253,
+      "step": 53013
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.890644753476612e-07,
+      "loss": 1.0613,
+      "step": 53014
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.890170670037927e-07,
+      "loss": 1.2362,
+      "step": 53015
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.889696586599242e-07,
+      "loss": 3.2355,
+      "step": 53016
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.889222503160556e-07,
+      "loss": 3.3064,
+      "step": 53017
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.888748419721871e-07,
+      "loss": 3.3382,
+      "step": 53018
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.888274336283186e-07,
+      "loss": 3.0237,
+      "step": 53019
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.8878002528445e-07,
+      "loss": 2.2572,
+      "step": 53020
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.887326169405816e-07,
+      "loss": 1.1201,
+      "step": 53021
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.88685208596713e-07,
+      "loss": 2.3987,
+      "step": 53022
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.886378002528445e-07,
+      "loss": 1.2603,
+      "step": 53023
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.88590391908976e-07,
+      "loss": 2.5154,
+      "step": 53024
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.885429835651075e-07,
+      "loss": 1.9356,
+      "step": 53025
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.88495575221239e-07,
+      "loss": 2.4029,
+      "step": 53026
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.884481668773704e-07,
+      "loss": 1.7367,
+      "step": 53027
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.88400758533502e-07,
+      "loss": 2.07,
+      "step": 53028
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.883533501896333e-07,
+      "loss": 2.2556,
+      "step": 53029
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.883059418457648e-07,
+      "loss": 1.5729,
+      "step": 53030
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.882585335018963e-07,
+      "loss": 1.731,
+      "step": 53031
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.882111251580278e-07,
+      "loss": 0.6247,
+      "step": 53032
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.881637168141593e-07,
+      "loss": 3.8236,
+      "step": 53033
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.881163084702908e-07,
+      "loss": 2.0342,
+      "step": 53034
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.880689001264223e-07,
+      "loss": 2.1775,
+      "step": 53035
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.880214917825538e-07,
+      "loss": 1.8134,
+      "step": 53036
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.879740834386852e-07,
+      "loss": 2.1729,
+      "step": 53037
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.879266750948167e-07,
+      "loss": 2.773,
+      "step": 53038
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.878792667509482e-07,
+      "loss": 3.1479,
+      "step": 53039
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.878318584070797e-07,
+      "loss": 0.9939,
+      "step": 53040
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.877844500632111e-07,
+      "loss": 1.1011,
+      "step": 53041
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.877370417193425e-07,
+      "loss": 2.518,
+      "step": 53042
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.876896333754741e-07,
+      "loss": 2.8365,
+      "step": 53043
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.876422250316056e-07,
+      "loss": 1.6015,
+      "step": 53044
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.87594816687737e-07,
+      "loss": 1.6813,
+      "step": 53045
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.875474083438686e-07,
+      "loss": 0.9181,
+      "step": 53046
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.875e-07,
+      "loss": 2.0367,
+      "step": 53047
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.874525916561315e-07,
+      "loss": 2.0416,
+      "step": 53048
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.87405183312263e-07,
+      "loss": 2.2797,
+      "step": 53049
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.873577749683945e-07,
+      "loss": 1.5995,
+      "step": 53050
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.87310366624526e-07,
+      "loss": 1.4735,
+      "step": 53051
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.872629582806573e-07,
+      "loss": 2.7488,
+      "step": 53052
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.872155499367888e-07,
+      "loss": 0.8593,
+      "step": 53053
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.871681415929204e-07,
+      "loss": 1.7039,
+      "step": 53054
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.871207332490518e-07,
+      "loss": 2.2473,
+      "step": 53055
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.870733249051833e-07,
+      "loss": 1.5716,
+      "step": 53056
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.870259165613148e-07,
+      "loss": 1.3671,
+      "step": 53057
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.869785082174463e-07,
+      "loss": 1.1028,
+      "step": 53058
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.869310998735778e-07,
+      "loss": 2.1771,
+      "step": 53059
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.868836915297092e-07,
+      "loss": 2.5872,
+      "step": 53060
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.868362831858408e-07,
+      "loss": 2.6769,
+      "step": 53061
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.867888748419722e-07,
+      "loss": 1.2569,
+      "step": 53062
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.867414664981037e-07,
+      "loss": 0.9351,
+      "step": 53063
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.866940581542351e-07,
+      "loss": 3.7281,
+      "step": 53064
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.866466498103666e-07,
+      "loss": 1.0111,
+      "step": 53065
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.865992414664981e-07,
+      "loss": 1.4412,
+      "step": 53066
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.865518331226296e-07,
+      "loss": 2.3849,
+      "step": 53067
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.865044247787611e-07,
+      "loss": 0.9541,
+      "step": 53068
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.864570164348926e-07,
+      "loss": 3.1101,
+      "step": 53069
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.86409608091024e-07,
+      "loss": 2.6773,
+      "step": 53070
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.863621997471555e-07,
+      "loss": 1.9911,
+      "step": 53071
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.86314791403287e-07,
+      "loss": 2.0359,
+      "step": 53072
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.862673830594185e-07,
+      "loss": 1.4666,
+      "step": 53073
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.8621997471555e-07,
+      "loss": 2.0529,
+      "step": 53074
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.861725663716813e-07,
+      "loss": 2.4642,
+      "step": 53075
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.861251580278129e-07,
+      "loss": 2.25,
+      "step": 53076
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.860777496839444e-07,
+      "loss": 1.3816,
+      "step": 53077
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.860303413400758e-07,
+      "loss": 0.6694,
+      "step": 53078
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.859829329962074e-07,
+      "loss": 2.0656,
+      "step": 53079
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.859355246523388e-07,
+      "loss": 0.8347,
+      "step": 53080
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.858881163084703e-07,
+      "loss": 1.6009,
+      "step": 53081
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.858407079646018e-07,
+      "loss": 3.2044,
+      "step": 53082
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.857932996207333e-07,
+      "loss": 1.3338,
+      "step": 53083
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.857458912768648e-07,
+      "loss": 1.6718,
+      "step": 53084
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.856984829329962e-07,
+      "loss": 2.5729,
+      "step": 53085
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.856510745891277e-07,
+      "loss": 2.028,
+      "step": 53086
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.856036662452593e-07,
+      "loss": 2.8031,
+      "step": 53087
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.855562579013906e-07,
+      "loss": 2.5821,
+      "step": 53088
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.855088495575221e-07,
+      "loss": 0.3878,
+      "step": 53089
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.854614412136536e-07,
+      "loss": 1.9878,
+      "step": 53090
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.854140328697851e-07,
+      "loss": 0.8647,
+      "step": 53091
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.853666245259166e-07,
+      "loss": 1.8152,
+      "step": 53092
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.85319216182048e-07,
+      "loss": 0.5476,
+      "step": 53093
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.852718078381796e-07,
+      "loss": 3.0346,
+      "step": 53094
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.85224399494311e-07,
+      "loss": 3.2642,
+      "step": 53095
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.851769911504425e-07,
+      "loss": 3.4961,
+      "step": 53096
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.85129582806574e-07,
+      "loss": 1.8596,
+      "step": 53097
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.850821744627055e-07,
+      "loss": 2.9007,
+      "step": 53098
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.85034766118837e-07,
+      "loss": 1.0041,
+      "step": 53099
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.849873577749683e-07,
+      "loss": 1.3982,
+      "step": 53100
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.849399494310999e-07,
+      "loss": 2.9101,
+      "step": 53101
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.848925410872314e-07,
+      "loss": 2.0063,
+      "step": 53102
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.848451327433628e-07,
+      "loss": 2.1443,
+      "step": 53103
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.847977243994943e-07,
+      "loss": 1.6945,
+      "step": 53104
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.847503160556258e-07,
+      "loss": 3.0588,
+      "step": 53105
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.847029077117573e-07,
+      "loss": 2.9622,
+      "step": 53106
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.846554993678888e-07,
+      "loss": 2.6589,
+      "step": 53107
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.846080910240202e-07,
+      "loss": 1.6154,
+      "step": 53108
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.845606826801518e-07,
+      "loss": 1.9946,
+      "step": 53109
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.845132743362832e-07,
+      "loss": 2.5832,
+      "step": 53110
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.844658659924146e-07,
+      "loss": 3.7042,
+      "step": 53111
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.844184576485462e-07,
+      "loss": 0.9223,
+      "step": 53112
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.843710493046776e-07,
+      "loss": 1.8546,
+      "step": 53113
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.843236409608091e-07,
+      "loss": 2.0443,
+      "step": 53114
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.842762326169406e-07,
+      "loss": 0.8727,
+      "step": 53115
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.842288242730721e-07,
+      "loss": 1.3376,
+      "step": 53116
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.841814159292036e-07,
+      "loss": 1.6311,
+      "step": 53117
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.84134007585335e-07,
+      "loss": 2.0383,
+      "step": 53118
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.840865992414665e-07,
+      "loss": 0.3411,
+      "step": 53119
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.84039190897598e-07,
+      "loss": 1.877,
+      "step": 53120
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.839917825537295e-07,
+      "loss": 1.2858,
+      "step": 53121
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.83944374209861e-07,
+      "loss": 1.8429,
+      "step": 53122
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.838969658659923e-07,
+      "loss": 0.4229,
+      "step": 53123
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.838495575221239e-07,
+      "loss": 1.1599,
+      "step": 53124
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.838021491782554e-07,
+      "loss": 1.0037,
+      "step": 53125
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.837547408343868e-07,
+      "loss": 1.9287,
+      "step": 53126
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.837073324905184e-07,
+      "loss": 2.2157,
+      "step": 53127
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.836599241466498e-07,
+      "loss": 2.6697,
+      "step": 53128
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.836125158027813e-07,
+      "loss": 2.0829,
+      "step": 53129
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.835651074589128e-07,
+      "loss": 1.5624,
+      "step": 53130
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.835176991150443e-07,
+      "loss": 1.7524,
+      "step": 53131
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.834702907711758e-07,
+      "loss": 2.6687,
+      "step": 53132
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.834228824273072e-07,
+      "loss": 2.0754,
+      "step": 53133
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.833754740834386e-07,
+      "loss": 2.4601,
+      "step": 53134
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.833280657395702e-07,
+      "loss": 2.03,
+      "step": 53135
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.832806573957016e-07,
+      "loss": 2.9933,
+      "step": 53136
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.832332490518331e-07,
+      "loss": 1.473,
+      "step": 53137
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.831858407079646e-07,
+      "loss": 1.5362,
+      "step": 53138
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.831384323640961e-07,
+      "loss": 4.007,
+      "step": 53139
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.830910240202276e-07,
+      "loss": 2.3269,
+      "step": 53140
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.83043615676359e-07,
+      "loss": 3.4218,
+      "step": 53141
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.829962073324906e-07,
+      "loss": 0.1997,
+      "step": 53142
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.82948798988622e-07,
+      "loss": 1.7033,
+      "step": 53143
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.829013906447535e-07,
+      "loss": 1.4393,
+      "step": 53144
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.82853982300885e-07,
+      "loss": 1.3675,
+      "step": 53145
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.828065739570164e-07,
+      "loss": 0.3848,
+      "step": 53146
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.827591656131479e-07,
+      "loss": 1.5106,
+      "step": 53147
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.827117572692794e-07,
+      "loss": 2.0226,
+      "step": 53148
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.826643489254109e-07,
+      "loss": 2.0201,
+      "step": 53149
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.826169405815424e-07,
+      "loss": 1.7023,
+      "step": 53150
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.825695322376738e-07,
+      "loss": 0.8073,
+      "step": 53151
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.825221238938053e-07,
+      "loss": 1.2878,
+      "step": 53152
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.824747155499368e-07,
+      "loss": 1.3988,
+      "step": 53153
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.824273072060683e-07,
+      "loss": 1.7706,
+      "step": 53154
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.823798988621998e-07,
+      "loss": 2.1753,
+      "step": 53155
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.823324905183312e-07,
+      "loss": 1.7866,
+      "step": 53156
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.822850821744628e-07,
+      "loss": 2.0582,
+      "step": 53157
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.822376738305942e-07,
+      "loss": 2.5126,
+      "step": 53158
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.821902654867256e-07,
+      "loss": 2.9353,
+      "step": 53159
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.821428571428572e-07,
+      "loss": 0.8013,
+      "step": 53160
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.820954487989886e-07,
+      "loss": 1.7592,
+      "step": 53161
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.820480404551201e-07,
+      "loss": 1.4191,
+      "step": 53162
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.820006321112516e-07,
+      "loss": 2.6905,
+      "step": 53163
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.819532237673831e-07,
+      "loss": 0.8622,
+      "step": 53164
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.819058154235146e-07,
+      "loss": 0.8075,
+      "step": 53165
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.81858407079646e-07,
+      "loss": 1.8118,
+      "step": 53166
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.818109987357775e-07,
+      "loss": 1.9246,
+      "step": 53167
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.81763590391909e-07,
+      "loss": 2.2704,
+      "step": 53168
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.817161820480404e-07,
+      "loss": 1.1501,
+      "step": 53169
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.816687737041719e-07,
+      "loss": 1.1778,
+      "step": 53170
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.816213653603034e-07,
+      "loss": 2.4156,
+      "step": 53171
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.815739570164349e-07,
+      "loss": 1.796,
+      "step": 53172
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.815265486725664e-07,
+      "loss": 1.3046,
+      "step": 53173
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.814791403286978e-07,
+      "loss": 1.659,
+      "step": 53174
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.814317319848294e-07,
+      "loss": 2.7358,
+      "step": 53175
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.813843236409608e-07,
+      "loss": 2.5923,
+      "step": 53176
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.813369152970923e-07,
+      "loss": 1.6734,
+      "step": 53177
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.812895069532238e-07,
+      "loss": 2.0882,
+      "step": 53178
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.812420986093553e-07,
+      "loss": 1.7846,
+      "step": 53179
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.811946902654868e-07,
+      "loss": 1.2766,
+      "step": 53180
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.811472819216181e-07,
+      "loss": 3.4089,
+      "step": 53181
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.810998735777497e-07,
+      "loss": 1.2625,
+      "step": 53182
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.810524652338812e-07,
+      "loss": 1.8457,
+      "step": 53183
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.810050568900126e-07,
+      "loss": 2.7125,
+      "step": 53184
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.809576485461441e-07,
+      "loss": 1.5669,
+      "step": 53185
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.809102402022756e-07,
+      "loss": 3.8281,
+      "step": 53186
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.808628318584071e-07,
+      "loss": 1.9774,
+      "step": 53187
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.808154235145386e-07,
+      "loss": 2.0371,
+      "step": 53188
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.8076801517067e-07,
+      "loss": 1.2937,
+      "step": 53189
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.807206068268016e-07,
+      "loss": 0.9845,
+      "step": 53190
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.80673198482933e-07,
+      "loss": 2.128,
+      "step": 53191
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.806257901390645e-07,
+      "loss": 1.9034,
+      "step": 53192
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.80578381795196e-07,
+      "loss": 2.7494,
+      "step": 53193
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.805309734513274e-07,
+      "loss": 1.7198,
+      "step": 53194
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.804835651074589e-07,
+      "loss": 1.7404,
+      "step": 53195
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.804361567635904e-07,
+      "loss": 1.9537,
+      "step": 53196
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.803887484197219e-07,
+      "loss": 1.3741,
+      "step": 53197
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.803413400758534e-07,
+      "loss": 2.3313,
+      "step": 53198
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.802939317319848e-07,
+      "loss": 2.2971,
+      "step": 53199
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.802465233881163e-07,
+      "loss": 3.2047,
+      "step": 53200
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.801991150442478e-07,
+      "loss": 1.8278,
+      "step": 53201
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.801517067003793e-07,
+      "loss": 2.503,
+      "step": 53202
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.801042983565108e-07,
+      "loss": 1.4263,
+      "step": 53203
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.800568900126423e-07,
+      "loss": 1.3749,
+      "step": 53204
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.800094816687737e-07,
+      "loss": 4.1836,
+      "step": 53205
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.799620733249052e-07,
+      "loss": 1.4675,
+      "step": 53206
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.799146649810366e-07,
+      "loss": 3.1249,
+      "step": 53207
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.798672566371682e-07,
+      "loss": 1.6749,
+      "step": 53208
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.798198482932996e-07,
+      "loss": 2.3477,
+      "step": 53209
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.797724399494311e-07,
+      "loss": 1.1611,
+      "step": 53210
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.797250316055626e-07,
+      "loss": 1.5761,
+      "step": 53211
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.796776232616941e-07,
+      "loss": 1.7818,
+      "step": 53212
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.796302149178256e-07,
+      "loss": 1.0351,
+      "step": 53213
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.79582806573957e-07,
+      "loss": 1.822,
+      "step": 53214
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.795353982300886e-07,
+      "loss": 3.0117,
+      "step": 53215
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.7948798988622e-07,
+      "loss": 1.4902,
+      "step": 53216
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.794405815423514e-07,
+      "loss": 2.3477,
+      "step": 53217
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.793931731984829e-07,
+      "loss": 1.6518,
+      "step": 53218
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.793457648546144e-07,
+      "loss": 2.2368,
+      "step": 53219
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.792983565107459e-07,
+      "loss": 1.8487,
+      "step": 53220
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.792509481668774e-07,
+      "loss": 1.1004,
+      "step": 53221
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.792035398230088e-07,
+      "loss": 1.0187,
+      "step": 53222
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.791561314791404e-07,
+      "loss": 2.1971,
+      "step": 53223
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.791087231352718e-07,
+      "loss": 1.6831,
+      "step": 53224
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.790613147914033e-07,
+      "loss": 1.4901,
+      "step": 53225
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.790139064475349e-07,
+      "loss": 2.0246,
+      "step": 53226
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.789664981036663e-07,
+      "loss": 1.4838,
+      "step": 53227
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.789190897597977e-07,
+      "loss": 1.3499,
+      "step": 53228
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.788716814159292e-07,
+      "loss": 1.8497,
+      "step": 53229
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.788242730720607e-07,
+      "loss": 2.6665,
+      "step": 53230
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.787768647281922e-07,
+      "loss": 2.731,
+      "step": 53231
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.787294563843236e-07,
+      "loss": 1.0076,
+      "step": 53232
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.786820480404551e-07,
+      "loss": 2.1765,
+      "step": 53233
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.786346396965866e-07,
+      "loss": 2.2933,
+      "step": 53234
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.785872313527181e-07,
+      "loss": 1.8199,
+      "step": 53235
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.785398230088496e-07,
+      "loss": 2.4908,
+      "step": 53236
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.784924146649811e-07,
+      "loss": 2.5752,
+      "step": 53237
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.784450063211126e-07,
+      "loss": 1.1211,
+      "step": 53238
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.78397597977244e-07,
+      "loss": 2.814,
+      "step": 53239
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.783501896333754e-07,
+      "loss": 0.7749,
+      "step": 53240
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.78302781289507e-07,
+      "loss": 3.0434,
+      "step": 53241
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.782553729456384e-07,
+      "loss": 1.4337,
+      "step": 53242
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.782079646017699e-07,
+      "loss": 1.2824,
+      "step": 53243
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.781605562579014e-07,
+      "loss": 2.7111,
+      "step": 53244
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.781131479140329e-07,
+      "loss": 2.7731,
+      "step": 53245
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.780657395701644e-07,
+      "loss": 1.9277,
+      "step": 53246
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.780183312262958e-07,
+      "loss": 1.4595,
+      "step": 53247
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.779709228824274e-07,
+      "loss": 2.5507,
+      "step": 53248
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.779235145385588e-07,
+      "loss": 1.689,
+      "step": 53249
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.778761061946903e-07,
+      "loss": 1.1975,
+      "step": 53250
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.778286978508217e-07,
+      "loss": 1.033,
+      "step": 53251
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.777812895069532e-07,
+      "loss": 2.2011,
+      "step": 53252
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.777338811630847e-07,
+      "loss": 2.1031,
+      "step": 53253
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.776864728192162e-07,
+      "loss": 0.9826,
+      "step": 53254
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.776390644753476e-07,
+      "loss": 1.533,
+      "step": 53255
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.775916561314792e-07,
+      "loss": 1.3369,
+      "step": 53256
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.775442477876106e-07,
+      "loss": 2.3951,
+      "step": 53257
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.774968394437421e-07,
+      "loss": 1.83,
+      "step": 53258
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.774494310998736e-07,
+      "loss": 1.7582,
+      "step": 53259
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.774020227560051e-07,
+      "loss": 2.1491,
+      "step": 53260
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.773546144121366e-07,
+      "loss": 1.409,
+      "step": 53261
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.77307206068268e-07,
+      "loss": 2.2939,
+      "step": 53262
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.772597977243995e-07,
+      "loss": 1.8944,
+      "step": 53263
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.77212389380531e-07,
+      "loss": 3.4485,
+      "step": 53264
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.771649810366624e-07,
+      "loss": 1.2753,
+      "step": 53265
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.771175726927939e-07,
+      "loss": 2.4453,
+      "step": 53266
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.770701643489254e-07,
+      "loss": 1.0006,
+      "step": 53267
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.770227560050569e-07,
+      "loss": 1.054,
+      "step": 53268
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.769753476611884e-07,
+      "loss": 2.2464,
+      "step": 53269
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.769279393173199e-07,
+      "loss": 1.2981,
+      "step": 53270
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.768805309734514e-07,
+      "loss": 1.1329,
+      "step": 53271
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.768331226295828e-07,
+      "loss": 1.9019,
+      "step": 53272
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7678571428571426e-07,
+      "loss": 2.7077,
+      "step": 53273
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.767383059418458e-07,
+      "loss": 2.4391,
+      "step": 53274
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7669089759797724e-07,
+      "loss": 0.658,
+      "step": 53275
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7664348925410873e-07,
+      "loss": 2.9974,
+      "step": 53276
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.765960809102402e-07,
+      "loss": 2.1282,
+      "step": 53277
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.765486725663717e-07,
+      "loss": 1.3742,
+      "step": 53278
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7650126422250316e-07,
+      "loss": 1.9474,
+      "step": 53279
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7645385587863465e-07,
+      "loss": 1.0627,
+      "step": 53280
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.764064475347662e-07,
+      "loss": 1.6718,
+      "step": 53281
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7635903919089764e-07,
+      "loss": 1.0861,
+      "step": 53282
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.763116308470291e-07,
+      "loss": 1.8555,
+      "step": 53283
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7626422250316057e-07,
+      "loss": 1.5257,
+      "step": 53284
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7621681415929206e-07,
+      "loss": 1.7342,
+      "step": 53285
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7616940581542355e-07,
+      "loss": 0.7035,
+      "step": 53286
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.76121997471555e-07,
+      "loss": 2.0342,
+      "step": 53287
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7607458912768643e-07,
+      "loss": 1.5588,
+      "step": 53288
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.76027180783818e-07,
+      "loss": 2.0313,
+      "step": 53289
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7597977243994947e-07,
+      "loss": 2.5161,
+      "step": 53290
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.759323640960809e-07,
+      "loss": 0.9911,
+      "step": 53291
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.7588495575221245e-07,
+      "loss": 1.2346,
+      "step": 53292
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.758375474083439e-07,
+      "loss": 2.1149,
+      "step": 53293
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7579013906447533e-07,
+      "loss": 2.1017,
+      "step": 53294
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.757427307206068e-07,
+      "loss": 1.6119,
+      "step": 53295
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7569532237673837e-07,
+      "loss": 3.0349,
+      "step": 53296
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.756479140328698e-07,
+      "loss": 2.857,
+      "step": 53297
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7560050568900125e-07,
+      "loss": 2.9345,
+      "step": 53298
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7555309734513274e-07,
+      "loss": 2.1033,
+      "step": 53299
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.755056890012643e-07,
+      "loss": 1.9647,
+      "step": 53300
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.754582806573957e-07,
+      "loss": 2.2784,
+      "step": 53301
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7541087231352716e-07,
+      "loss": 2.7092,
+      "step": 53302
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.753634639696587e-07,
+      "loss": 1.4398,
+      "step": 53303
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7531605562579015e-07,
+      "loss": 1.8818,
+      "step": 53304
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7526864728192164e-07,
+      "loss": 0.7664,
+      "step": 53305
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.752212389380531e-07,
+      "loss": 2.1252,
+      "step": 53306
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.751738305941846e-07,
+      "loss": 0.8092,
+      "step": 53307
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7512642225031606e-07,
+      "loss": 2.1134,
+      "step": 53308
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7507901390644755e-07,
+      "loss": 2.2987,
+      "step": 53309
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.75031605562579e-07,
+      "loss": 2.244,
+      "step": 53310
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7498419721871054e-07,
+      "loss": 2.927,
+      "step": 53311
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.74936788874842e-07,
+      "loss": 2.2893,
+      "step": 53312
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7488938053097347e-07,
+      "loss": 2.6402,
+      "step": 53313
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7484197218710496e-07,
+      "loss": 2.996,
+      "step": 53314
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7479456384323645e-07,
+      "loss": 1.3117,
+      "step": 53315
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.747471554993679e-07,
+      "loss": 2.6664,
+      "step": 53316
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7469974715549933e-07,
+      "loss": 0.9371,
+      "step": 53317
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.746523388116309e-07,
+      "loss": 2.8193,
+      "step": 53318
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7460493046776237e-07,
+      "loss": 1.6867,
+      "step": 53319
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.745575221238938e-07,
+      "loss": 3.3796,
+      "step": 53320
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7451011378002525e-07,
+      "loss": 2.8401,
+      "step": 53321
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.744627054361568e-07,
+      "loss": 2.1538,
+      "step": 53322
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7441529709228823e-07,
+      "loss": 0.8345,
+      "step": 53323
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.743678887484197e-07,
+      "loss": 2.6355,
+      "step": 53324
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7432048040455127e-07,
+      "loss": 2.8556,
+      "step": 53325
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.742730720606827e-07,
+      "loss": 1.1747,
+      "step": 53326
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7422566371681415e-07,
+      "loss": 0.364,
+      "step": 53327
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7417825537294564e-07,
+      "loss": 2.3682,
+      "step": 53328
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.741308470290772e-07,
+      "loss": 1.9916,
+      "step": 53329
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.740834386852086e-07,
+      "loss": 2.7478,
+      "step": 53330
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7403603034134006e-07,
+      "loss": 1.7178,
+      "step": 53331
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7398862199747156e-07,
+      "loss": 2.7264,
+      "step": 53332
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7394121365360305e-07,
+      "loss": 1.3533,
+      "step": 53333
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7389380530973454e-07,
+      "loss": 1.0935,
+      "step": 53334
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.73846396965866e-07,
+      "loss": 1.2309,
+      "step": 53335
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.737989886219974e-07,
+      "loss": 1.8365,
+      "step": 53336
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7375158027812896e-07,
+      "loss": 3.4461,
+      "step": 53337
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7370417193426046e-07,
+      "loss": 1.5789,
+      "step": 53338
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.736567635903919e-07,
+      "loss": 1.3571,
+      "step": 53339
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7360935524652344e-07,
+      "loss": 2.0155,
+      "step": 53340
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.735619469026549e-07,
+      "loss": 1.7098,
+      "step": 53341
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7351453855878637e-07,
+      "loss": 2.1627,
+      "step": 53342
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.734671302149178e-07,
+      "loss": 3.1158,
+      "step": 53343
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7341972187104936e-07,
+      "loss": 1.8091,
+      "step": 53344
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.733723135271808e-07,
+      "loss": 1.6596,
+      "step": 53345
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7332490518331223e-07,
+      "loss": 1.3022,
+      "step": 53346
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7327749683944373e-07,
+      "loss": 1.9912,
+      "step": 53347
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7323008849557527e-07,
+      "loss": 1.3226,
+      "step": 53348
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.731826801517067e-07,
+      "loss": 1.7655,
+      "step": 53349
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7313527180783815e-07,
+      "loss": 2.5754,
+      "step": 53350
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.730878634639697e-07,
+      "loss": 2.2623,
+      "step": 53351
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.730404551201012e-07,
+      "loss": 3.3383,
+      "step": 53352
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7299304677623263e-07,
+      "loss": 2.3874,
+      "step": 53353
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.7294563843236407e-07,
+      "loss": 3.8252,
+      "step": 53354
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.728982300884956e-07,
+      "loss": 2.6965,
+      "step": 53355
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7285082174462705e-07,
+      "loss": 1.9112,
+      "step": 53356
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7280341340075854e-07,
+      "loss": 2.0152,
+      "step": 53357
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7275600505689e-07,
+      "loss": 1.3597,
+      "step": 53358
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7270859671302153e-07,
+      "loss": 1.6625,
+      "step": 53359
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7266118836915297e-07,
+      "loss": 1.9503,
+      "step": 53360
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7261378002528446e-07,
+      "loss": 2.1491,
+      "step": 53361
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7256637168141595e-07,
+      "loss": 2.5791,
+      "step": 53362
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7251896333754744e-07,
+      "loss": 2.1475,
+      "step": 53363
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.724715549936789e-07,
+      "loss": 2.3639,
+      "step": 53364
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.724241466498103e-07,
+      "loss": 2.1659,
+      "step": 53365
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7237673830594187e-07,
+      "loss": 3.4699,
+      "step": 53366
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7232932996207336e-07,
+      "loss": 0.6969,
+      "step": 53367
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.722819216182048e-07,
+      "loss": 3.4521,
+      "step": 53368
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7223451327433624e-07,
+      "loss": 1.1642,
+      "step": 53369
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.721871049304678e-07,
+      "loss": 0.8406,
+      "step": 53370
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.721396965865993e-07,
+      "loss": 1.3389,
+      "step": 53371
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.720922882427307e-07,
+      "loss": 2.4081,
+      "step": 53372
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7204487989886226e-07,
+      "loss": 3.0965,
+      "step": 53373
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.719974715549937e-07,
+      "loss": 1.1551,
+      "step": 53374
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7195006321112514e-07,
+      "loss": 1.7835,
+      "step": 53375
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7190265486725663e-07,
+      "loss": 3.2098,
+      "step": 53376
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.718552465233882e-07,
+      "loss": 2.4968,
+      "step": 53377
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.718078381795196e-07,
+      "loss": 1.3204,
+      "step": 53378
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7176042983565105e-07,
+      "loss": 2.5152,
+      "step": 53379
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7171302149178254e-07,
+      "loss": 3.2428,
+      "step": 53380
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.716656131479141e-07,
+      "loss": 2.5636,
+      "step": 53381
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7161820480404553e-07,
+      "loss": 1.9864,
+      "step": 53382
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7157079646017697e-07,
+      "loss": 1.3959,
+      "step": 53383
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.715233881163085e-07,
+      "loss": 1.0845,
+      "step": 53384
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7147597977243995e-07,
+      "loss": 2.573,
+      "step": 53385
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7142857142857145e-07,
+      "loss": 2.1586,
+      "step": 53386
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.713811630847029e-07,
+      "loss": 1.7015,
+      "step": 53387
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7133375474083443e-07,
+      "loss": 1.7195,
+      "step": 53388
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7128634639696587e-07,
+      "loss": 1.5559,
+      "step": 53389
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7123893805309736e-07,
+      "loss": 1.8321,
+      "step": 53390
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.711915297092288e-07,
+      "loss": 3.1284,
+      "step": 53391
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7114412136536035e-07,
+      "loss": 1.7906,
+      "step": 53392
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.710967130214918e-07,
+      "loss": 1.718,
+      "step": 53393
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.710493046776232e-07,
+      "loss": 1.2435,
+      "step": 53394
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7100189633375477e-07,
+      "loss": 1.7888,
+      "step": 53395
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7095448798988626e-07,
+      "loss": 2.3776,
+      "step": 53396
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.709070796460177e-07,
+      "loss": 4.4417,
+      "step": 53397
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7085967130214914e-07,
+      "loss": 2.0219,
+      "step": 53398
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.708122629582807e-07,
+      "loss": 1.1369,
+      "step": 53399
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.707648546144122e-07,
+      "loss": 1.3056,
+      "step": 53400
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.707174462705436e-07,
+      "loss": 1.9302,
+      "step": 53401
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7067003792667506e-07,
+      "loss": 1.0681,
+      "step": 53402
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.706226295828066e-07,
+      "loss": 1.4459,
+      "step": 53403
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7057522123893804e-07,
+      "loss": 1.1807,
+      "step": 53404
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7052781289506953e-07,
+      "loss": 1.2869,
+      "step": 53405
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.704804045512011e-07,
+      "loss": 2.4011,
+      "step": 53406
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.704329962073325e-07,
+      "loss": 1.3488,
+      "step": 53407
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7038558786346396e-07,
+      "loss": 1.6768,
+      "step": 53408
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7033817951959545e-07,
+      "loss": 1.5751,
+      "step": 53409
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.70290771175727e-07,
+      "loss": 0.817,
+      "step": 53410
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7024336283185843e-07,
+      "loss": 1.3508,
+      "step": 53411
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7019595448798987e-07,
+      "loss": 1.365,
+      "step": 53412
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7014854614412136e-07,
+      "loss": 2.2557,
+      "step": 53413
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7010113780025286e-07,
+      "loss": 3.5485,
+      "step": 53414
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.7005372945638435e-07,
+      "loss": 1.184,
+      "step": 53415
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.700063211125158e-07,
+      "loss": 2.251,
+      "step": 53416
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.6995891276864733e-07,
+      "loss": 2.6558,
+      "step": 53417
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.6991150442477877e-07,
+      "loss": 1.4749,
+      "step": 53418
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6986409608091026e-07,
+      "loss": 1.9103,
+      "step": 53419
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.698166877370417e-07,
+      "loss": 1.2449,
+      "step": 53420
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6976927939317325e-07,
+      "loss": 2.3747,
+      "step": 53421
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.697218710493047e-07,
+      "loss": 1.5202,
+      "step": 53422
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.696744627054362e-07,
+      "loss": 1.4987,
+      "step": 53423
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.696270543615676e-07,
+      "loss": 1.6105,
+      "step": 53424
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6957964601769916e-07,
+      "loss": 2.0431,
+      "step": 53425
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.695322376738306e-07,
+      "loss": 1.7928,
+      "step": 53426
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6948482932996204e-07,
+      "loss": 2.1733,
+      "step": 53427
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.694374209860936e-07,
+      "loss": 1.8542,
+      "step": 53428
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.693900126422251e-07,
+      "loss": 2.7756,
+      "step": 53429
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.693426042983565e-07,
+      "loss": 3.6041,
+      "step": 53430
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6929519595448796e-07,
+      "loss": 2.3705,
+      "step": 53431
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.692477876106195e-07,
+      "loss": 1.7168,
+      "step": 53432
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6920037926675094e-07,
+      "loss": 1.3517,
+      "step": 53433
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6915297092288243e-07,
+      "loss": 2.6106,
+      "step": 53434
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6910556257901387e-07,
+      "loss": 4.2795,
+      "step": 53435
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.690581542351454e-07,
+      "loss": 1.3497,
+      "step": 53436
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6901074589127686e-07,
+      "loss": 2.0044,
+      "step": 53437
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6896333754740835e-07,
+      "loss": 0.8022,
+      "step": 53438
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.689159292035399e-07,
+      "loss": 2.3753,
+      "step": 53439
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6886852085967133e-07,
+      "loss": 0.8202,
+      "step": 53440
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.688211125158028e-07,
+      "loss": 1.0799,
+      "step": 53441
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6877370417193427e-07,
+      "loss": 1.535,
+      "step": 53442
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6872629582806576e-07,
+      "loss": 1.7151,
+      "step": 53443
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6867888748419725e-07,
+      "loss": 0.8317,
+      "step": 53444
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.686314791403287e-07,
+      "loss": 2.3834,
+      "step": 53445
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6858407079646013e-07,
+      "loss": 1.03,
+      "step": 53446
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.685366624525917e-07,
+      "loss": 2.7475,
+      "step": 53447
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6848925410872317e-07,
+      "loss": 0.9951,
+      "step": 53448
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.684418457648546e-07,
+      "loss": 2.1689,
+      "step": 53449
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6839443742098615e-07,
+      "loss": 4.0718,
+      "step": 53450
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.683470290771176e-07,
+      "loss": 2.7378,
+      "step": 53451
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.682996207332491e-07,
+      "loss": 2.7453,
+      "step": 53452
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.682522123893805e-07,
+      "loss": 1.0113,
+      "step": 53453
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6820480404551207e-07,
+      "loss": 1.2516,
+      "step": 53454
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.681573957016435e-07,
+      "loss": 3.059,
+      "step": 53455
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6810998735777494e-07,
+      "loss": 3.0777,
+      "step": 53456
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6806257901390644e-07,
+      "loss": 3.282,
+      "step": 53457
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.68015170670038e-07,
+      "loss": 1.1582,
+      "step": 53458
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.679677623261694e-07,
+      "loss": 1.0289,
+      "step": 53459
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6792035398230086e-07,
+      "loss": 2.3097,
+      "step": 53460
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.678729456384324e-07,
+      "loss": 0.4079,
+      "step": 53461
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6782553729456384e-07,
+      "loss": 1.3169,
+      "step": 53462
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6777812895069534e-07,
+      "loss": 1.8025,
+      "step": 53463
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.677307206068268e-07,
+      "loss": 1.8035,
+      "step": 53464
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.676833122629583e-07,
+      "loss": 2.7683,
+      "step": 53465
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6763590391908976e-07,
+      "loss": 1.6665,
+      "step": 53466
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6758849557522125e-07,
+      "loss": 0.9829,
+      "step": 53467
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.675410872313527e-07,
+      "loss": 1.5211,
+      "step": 53468
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6749367888748424e-07,
+      "loss": 2.6122,
+      "step": 53469
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.674462705436157e-07,
+      "loss": 0.8845,
+      "step": 53470
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6739886219974717e-07,
+      "loss": 1.7571,
+      "step": 53471
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6735145385587866e-07,
+      "loss": 2.899,
+      "step": 53472
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6730404551201015e-07,
+      "loss": 2.2835,
+      "step": 53473
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.672566371681416e-07,
+      "loss": 1.8973,
+      "step": 53474
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6720922882427303e-07,
+      "loss": 1.3533,
+      "step": 53475
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.671618204804046e-07,
+      "loss": 1.9642,
+      "step": 53476
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6711441213653607e-07,
+      "loss": 0.7295,
+      "step": 53477
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.670670037926675e-07,
+      "loss": 0.4816,
+      "step": 53478
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.6701959544879895e-07,
+      "loss": 2.3323,
+      "step": 53479
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.669721871049305e-07,
+      "loss": 2.5935,
+      "step": 53480
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.66924778761062e-07,
+      "loss": 1.8858,
+      "step": 53481
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.668773704171934e-07,
+      "loss": 2.6627,
+      "step": 53482
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6682996207332497e-07,
+      "loss": 0.6734,
+      "step": 53483
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.667825537294564e-07,
+      "loss": 2.0459,
+      "step": 53484
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6673514538558785e-07,
+      "loss": 0.5659,
+      "step": 53485
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6668773704171934e-07,
+      "loss": 1.0487,
+      "step": 53486
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.666403286978509e-07,
+      "loss": 1.3916,
+      "step": 53487
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.665929203539823e-07,
+      "loss": 0.9731,
+      "step": 53488
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6654551201011376e-07,
+      "loss": 1.6898,
+      "step": 53489
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6649810366624525e-07,
+      "loss": 1.5886,
+      "step": 53490
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.664506953223768e-07,
+      "loss": 2.9309,
+      "step": 53491
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6640328697850824e-07,
+      "loss": 1.7406,
+      "step": 53492
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.663558786346397e-07,
+      "loss": 1.8527,
+      "step": 53493
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.663084702907712e-07,
+      "loss": 1.231,
+      "step": 53494
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6626106194690266e-07,
+      "loss": 1.8292,
+      "step": 53495
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6621365360303416e-07,
+      "loss": 2.3028,
+      "step": 53496
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.661662452591656e-07,
+      "loss": 1.0724,
+      "step": 53497
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6611883691529714e-07,
+      "loss": 1.6644,
+      "step": 53498
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.660714285714286e-07,
+      "loss": 2.0493,
+      "step": 53499
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6602402022756007e-07,
+      "loss": 1.0891,
+      "step": 53500
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.659766118836915e-07,
+      "loss": 0.6694,
+      "step": 53501
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6592920353982306e-07,
+      "loss": 1.545,
+      "step": 53502
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.658817951959545e-07,
+      "loss": 2.6427,
+      "step": 53503
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6583438685208593e-07,
+      "loss": 1.1582,
+      "step": 53504
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.657869785082175e-07,
+      "loss": 1.6538,
+      "step": 53505
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6573957016434897e-07,
+      "loss": 1.7374,
+      "step": 53506
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.656921618204804e-07,
+      "loss": 3.0918,
+      "step": 53507
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6564475347661185e-07,
+      "loss": 1.8141,
+      "step": 53508
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.655973451327434e-07,
+      "loss": 2.1046,
+      "step": 53509
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.655499367888749e-07,
+      "loss": 3.1304,
+      "step": 53510
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.655025284450063e-07,
+      "loss": 2.4753,
+      "step": 53511
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6545512010113777e-07,
+      "loss": 2.7255,
+      "step": 53512
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.654077117572693e-07,
+      "loss": 3.6266,
+      "step": 53513
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6536030341340075e-07,
+      "loss": 1.4153,
+      "step": 53514
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6531289506953224e-07,
+      "loss": 2.5838,
+      "step": 53515
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.652654867256638e-07,
+      "loss": 0.8336,
+      "step": 53516
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.652180783817952e-07,
+      "loss": 1.4214,
+      "step": 53517
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6517067003792667e-07,
+      "loss": 0.8378,
+      "step": 53518
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6512326169405816e-07,
+      "loss": 2.9212,
+      "step": 53519
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.650758533501897e-07,
+      "loss": 1.9107,
+      "step": 53520
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6502844500632114e-07,
+      "loss": 2.0265,
+      "step": 53521
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.649810366624526e-07,
+      "loss": 1.2604,
+      "step": 53522
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6493362831858407e-07,
+      "loss": 2.6946,
+      "step": 53523
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6488621997471557e-07,
+      "loss": 2.8278,
+      "step": 53524
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6483881163084706e-07,
+      "loss": 1.9773,
+      "step": 53525
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.647914032869785e-07,
+      "loss": 2.0258,
+      "step": 53526
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6474399494311004e-07,
+      "loss": 0.8556,
+      "step": 53527
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.646965865992415e-07,
+      "loss": 1.8763,
+      "step": 53528
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6464917825537297e-07,
+      "loss": 0.5161,
+      "step": 53529
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.646017699115044e-07,
+      "loss": 1.3735,
+      "step": 53530
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6455436156763596e-07,
+      "loss": 1.2224,
+      "step": 53531
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.645069532237674e-07,
+      "loss": 2.3958,
+      "step": 53532
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6445954487989884e-07,
+      "loss": 1.542,
+      "step": 53533
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6441213653603033e-07,
+      "loss": 4.1523,
+      "step": 53534
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.643647281921619e-07,
+      "loss": 0.8067,
+      "step": 53535
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.643173198482933e-07,
+      "loss": 1.4138,
+      "step": 53536
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6426991150442475e-07,
+      "loss": 0.789,
+      "step": 53537
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.642225031605563e-07,
+      "loss": 1.724,
+      "step": 53538
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.641750948166878e-07,
+      "loss": 2.5973,
+      "step": 53539
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6412768647281923e-07,
+      "loss": 3.8479,
+      "step": 53540
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6408027812895067e-07,
+      "loss": 0.7577,
+      "step": 53541
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.640328697850822e-07,
+      "loss": 0.6008,
+      "step": 53542
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6398546144121365e-07,
+      "loss": 2.1181,
+      "step": 53543
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.6393805309734514e-07,
+      "loss": 2.2591,
+      "step": 53544
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.638906447534766e-07,
+      "loss": 0.9323,
+      "step": 53545
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6384323640960813e-07,
+      "loss": 2.2105,
+      "step": 53546
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6379582806573957e-07,
+      "loss": 4.1193,
+      "step": 53547
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6374841972187106e-07,
+      "loss": 2.0798,
+      "step": 53548
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.637010113780026e-07,
+      "loss": 1.9433,
+      "step": 53549
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6365360303413404e-07,
+      "loss": 1.0698,
+      "step": 53550
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.636061946902655e-07,
+      "loss": 0.7976,
+      "step": 53551
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.63558786346397e-07,
+      "loss": 3.5923,
+      "step": 53552
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6351137800252847e-07,
+      "loss": 2.0151,
+      "step": 53553
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6346396965865996e-07,
+      "loss": 3.5593,
+      "step": 53554
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.634165613147914e-07,
+      "loss": 2.402,
+      "step": 53555
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6336915297092284e-07,
+      "loss": 1.9629,
+      "step": 53556
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.633217446270544e-07,
+      "loss": 1.2554,
+      "step": 53557
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.632743362831859e-07,
+      "loss": 2.5076,
+      "step": 53558
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.632269279393173e-07,
+      "loss": 3.2047,
+      "step": 53559
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6317951959544875e-07,
+      "loss": 2.0466,
+      "step": 53560
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.631321112515803e-07,
+      "loss": 2.8483,
+      "step": 53561
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.630847029077118e-07,
+      "loss": 1.4949,
+      "step": 53562
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6303729456384323e-07,
+      "loss": 0.7166,
+      "step": 53563
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.629898862199748e-07,
+      "loss": 1.9332,
+      "step": 53564
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.629424778761062e-07,
+      "loss": 1.8528,
+      "step": 53565
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6289506953223765e-07,
+      "loss": 1.8749,
+      "step": 53566
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6284766118836915e-07,
+      "loss": 1.138,
+      "step": 53567
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.628002528445007e-07,
+      "loss": 2.0915,
+      "step": 53568
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6275284450063213e-07,
+      "loss": 1.9829,
+      "step": 53569
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6270543615676357e-07,
+      "loss": 0.8901,
+      "step": 53570
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6265802781289506e-07,
+      "loss": 1.491,
+      "step": 53571
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6261061946902655e-07,
+      "loss": 1.7202,
+      "step": 53572
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6256321112515805e-07,
+      "loss": 1.4813,
+      "step": 53573
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.625158027812895e-07,
+      "loss": 0.7763,
+      "step": 53574
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6246839443742103e-07,
+      "loss": 2.5091,
+      "step": 53575
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6242098609355247e-07,
+      "loss": 2.4313,
+      "step": 53576
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6237357774968396e-07,
+      "loss": 0.4893,
+      "step": 53577
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.623261694058154e-07,
+      "loss": 1.404,
+      "step": 53578
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6227876106194695e-07,
+      "loss": 1.859,
+      "step": 53579
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.622313527180784e-07,
+      "loss": 3.2641,
+      "step": 53580
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.621839443742099e-07,
+      "loss": 2.0118,
+      "step": 53581
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.621365360303413e-07,
+      "loss": 2.2996,
+      "step": 53582
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6208912768647286e-07,
+      "loss": 2.0707,
+      "step": 53583
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.620417193426043e-07,
+      "loss": 2.5235,
+      "step": 53584
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6199431099873574e-07,
+      "loss": 2.6365,
+      "step": 53585
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.619469026548673e-07,
+      "loss": 2.8169,
+      "step": 53586
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.618994943109988e-07,
+      "loss": 1.7501,
+      "step": 53587
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.618520859671302e-07,
+      "loss": 2.0478,
+      "step": 53588
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6180467762326166e-07,
+      "loss": 2.7082,
+      "step": 53589
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.617572692793932e-07,
+      "loss": 3.9027,
+      "step": 53590
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.617098609355247e-07,
+      "loss": 1.4479,
+      "step": 53591
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6166245259165613e-07,
+      "loss": 1.639,
+      "step": 53592
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6161504424778757e-07,
+      "loss": 2.8594,
+      "step": 53593
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.615676359039191e-07,
+      "loss": 1.589,
+      "step": 53594
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6152022756005056e-07,
+      "loss": 1.8176,
+      "step": 53595
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6147281921618205e-07,
+      "loss": 1.3754,
+      "step": 53596
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.614254108723136e-07,
+      "loss": 2.0048,
+      "step": 53597
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6137800252844503e-07,
+      "loss": 1.5532,
+      "step": 53598
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6133059418457647e-07,
+      "loss": 3.1693,
+      "step": 53599
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6128318584070797e-07,
+      "loss": 1.4967,
+      "step": 53600
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6123577749683946e-07,
+      "loss": 2.2606,
+      "step": 53601
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6118836915297095e-07,
+      "loss": 0.9944,
+      "step": 53602
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.611409608091024e-07,
+      "loss": 1.3708,
+      "step": 53603
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6109355246523383e-07,
+      "loss": 0.8658,
+      "step": 53604
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6104614412136537e-07,
+      "loss": 1.8261,
+      "step": 53605
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6099873577749687e-07,
+      "loss": 1.5525,
+      "step": 53606
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.609513274336283e-07,
+      "loss": 2.4536,
+      "step": 53607
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.6090391908975985e-07,
+      "loss": 4.0006,
+      "step": 53608
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.608565107458913e-07,
+      "loss": 1.1318,
+      "step": 53609
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.608091024020228e-07,
+      "loss": 2.152,
+      "step": 53610
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.607616940581542e-07,
+      "loss": 1.1388,
+      "step": 53611
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6071428571428577e-07,
+      "loss": 2.2207,
+      "step": 53612
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.606668773704172e-07,
+      "loss": 0.8936,
+      "step": 53613
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6061946902654864e-07,
+      "loss": 2.6579,
+      "step": 53614
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6057206068268014e-07,
+      "loss": 1.1319,
+      "step": 53615
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.605246523388117e-07,
+      "loss": 1.2993,
+      "step": 53616
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.604772439949431e-07,
+      "loss": 1.988,
+      "step": 53617
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6042983565107456e-07,
+      "loss": 2.844,
+      "step": 53618
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.603824273072061e-07,
+      "loss": 2.0322,
+      "step": 53619
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.603350189633376e-07,
+      "loss": 0.5844,
+      "step": 53620
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6028761061946904e-07,
+      "loss": 3.0993,
+      "step": 53621
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.602402022756005e-07,
+      "loss": 1.2522,
+      "step": 53622
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.60192793931732e-07,
+      "loss": 1.8769,
+      "step": 53623
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6014538558786346e-07,
+      "loss": 2.3652,
+      "step": 53624
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6009797724399495e-07,
+      "loss": 2.9679,
+      "step": 53625
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.600505689001264e-07,
+      "loss": 1.7053,
+      "step": 53626
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.6000316055625794e-07,
+      "loss": 0.8528,
+      "step": 53627
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.599557522123894e-07,
+      "loss": 2.5015,
+      "step": 53628
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5990834386852087e-07,
+      "loss": 2.4157,
+      "step": 53629
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.598609355246524e-07,
+      "loss": 1.5416,
+      "step": 53630
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5981352718078385e-07,
+      "loss": 1.8708,
+      "step": 53631
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.597661188369153e-07,
+      "loss": 1.6625,
+      "step": 53632
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5971871049304673e-07,
+      "loss": 0.9649,
+      "step": 53633
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.596713021491783e-07,
+      "loss": 1.6187,
+      "step": 53634
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5962389380530977e-07,
+      "loss": 2.7272,
+      "step": 53635
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.595764854614412e-07,
+      "loss": 2.1255,
+      "step": 53636
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5952907711757265e-07,
+      "loss": 1.7889,
+      "step": 53637
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.594816687737042e-07,
+      "loss": 1.2606,
+      "step": 53638
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.594342604298357e-07,
+      "loss": 2.001,
+      "step": 53639
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.593868520859671e-07,
+      "loss": 2.5657,
+      "step": 53640
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5933944374209867e-07,
+      "loss": 2.4209,
+      "step": 53641
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.592920353982301e-07,
+      "loss": 1.7598,
+      "step": 53642
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5924462705436155e-07,
+      "loss": 1.8056,
+      "step": 53643
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5919721871049304e-07,
+      "loss": 2.877,
+      "step": 53644
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.591498103666246e-07,
+      "loss": 1.1937,
+      "step": 53645
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.59102402022756e-07,
+      "loss": 0.6077,
+      "step": 53646
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5905499367888746e-07,
+      "loss": 2.2355,
+      "step": 53647
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5900758533501895e-07,
+      "loss": 2.3198,
+      "step": 53648
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.589601769911505e-07,
+      "loss": 1.1175,
+      "step": 53649
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5891276864728194e-07,
+      "loss": 1.267,
+      "step": 53650
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.588653603034134e-07,
+      "loss": 1.6235,
+      "step": 53651
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.588179519595449e-07,
+      "loss": 1.7332,
+      "step": 53652
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5877054361567636e-07,
+      "loss": 2.0866,
+      "step": 53653
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5872313527180785e-07,
+      "loss": 0.8441,
+      "step": 53654
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.586757269279393e-07,
+      "loss": 1.6134,
+      "step": 53655
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5862831858407084e-07,
+      "loss": 1.4175,
+      "step": 53656
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.585809102402023e-07,
+      "loss": 2.2863,
+      "step": 53657
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5853350189633377e-07,
+      "loss": 1.8026,
+      "step": 53658
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.584860935524652e-07,
+      "loss": 1.1359,
+      "step": 53659
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5843868520859675e-07,
+      "loss": 1.7403,
+      "step": 53660
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.583912768647282e-07,
+      "loss": 0.5356,
+      "step": 53661
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.583438685208597e-07,
+      "loss": 2.1752,
+      "step": 53662
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.582964601769912e-07,
+      "loss": 2.4779,
+      "step": 53663
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5824905183312267e-07,
+      "loss": 2.365,
+      "step": 53664
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.582016434892541e-07,
+      "loss": 1.4092,
+      "step": 53665
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5815423514538555e-07,
+      "loss": 2.5233,
+      "step": 53666
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.581068268015171e-07,
+      "loss": 1.7905,
+      "step": 53667
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.580594184576486e-07,
+      "loss": 2.8695,
+      "step": 53668
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5801201011378e-07,
+      "loss": 2.9108,
+      "step": 53669
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5796460176991146e-07,
+      "loss": 2.2686,
+      "step": 53670
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.57917193426043e-07,
+      "loss": 2.309,
+      "step": 53671
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.5786978508217445e-07,
+      "loss": 1.2759,
+      "step": 53672
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5782237673830594e-07,
+      "loss": 2.3088,
+      "step": 53673
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.577749683944375e-07,
+      "loss": 2.3652,
+      "step": 53674
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.577275600505689e-07,
+      "loss": 2.2579,
+      "step": 53675
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5768015170670036e-07,
+      "loss": 1.6263,
+      "step": 53676
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5763274336283186e-07,
+      "loss": 1.0206,
+      "step": 53677
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.575853350189634e-07,
+      "loss": 2.4546,
+      "step": 53678
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5753792667509484e-07,
+      "loss": 1.3964,
+      "step": 53679
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.574905183312263e-07,
+      "loss": 1.434,
+      "step": 53680
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5744310998735777e-07,
+      "loss": 2.0046,
+      "step": 53681
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5739570164348926e-07,
+      "loss": 1.8219,
+      "step": 53682
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5734829329962076e-07,
+      "loss": 1.315,
+      "step": 53683
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.573008849557522e-07,
+      "loss": 1.0098,
+      "step": 53684
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5725347661188374e-07,
+      "loss": 0.6075,
+      "step": 53685
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.572060682680152e-07,
+      "loss": 0.6628,
+      "step": 53686
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5715865992414667e-07,
+      "loss": 0.9537,
+      "step": 53687
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.571112515802781e-07,
+      "loss": 4.2255,
+      "step": 53688
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5706384323640966e-07,
+      "loss": 2.1277,
+      "step": 53689
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.570164348925411e-07,
+      "loss": 1.7466,
+      "step": 53690
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.569690265486726e-07,
+      "loss": 2.8805,
+      "step": 53691
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5692161820480403e-07,
+      "loss": 1.8654,
+      "step": 53692
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5687420986093557e-07,
+      "loss": 0.2979,
+      "step": 53693
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.56826801517067e-07,
+      "loss": 0.8232,
+      "step": 53694
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5677939317319845e-07,
+      "loss": 2.01,
+      "step": 53695
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5673198482933e-07,
+      "loss": 1.923,
+      "step": 53696
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.566845764854615e-07,
+      "loss": 1.4386,
+      "step": 53697
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5663716814159293e-07,
+      "loss": 1.4617,
+      "step": 53698
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5658975979772437e-07,
+      "loss": 1.7184,
+      "step": 53699
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.565423514538559e-07,
+      "loss": 2.8402,
+      "step": 53700
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.564949431099874e-07,
+      "loss": 1.1645,
+      "step": 53701
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5644753476611884e-07,
+      "loss": 0.8436,
+      "step": 53702
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.564001264222503e-07,
+      "loss": 3.0407,
+      "step": 53703
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5635271807838183e-07,
+      "loss": 1.7126,
+      "step": 53704
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5630530973451327e-07,
+      "loss": 2.002,
+      "step": 53705
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5625790139064476e-07,
+      "loss": 1.9098,
+      "step": 53706
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.562104930467763e-07,
+      "loss": 2.9351,
+      "step": 53707
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5616308470290774e-07,
+      "loss": 1.6621,
+      "step": 53708
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.561156763590392e-07,
+      "loss": 0.8677,
+      "step": 53709
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.560682680151707e-07,
+      "loss": 3.7849,
+      "step": 53710
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5602085967130217e-07,
+      "loss": 2.2325,
+      "step": 53711
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5597345132743366e-07,
+      "loss": 1.516,
+      "step": 53712
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.559260429835651e-07,
+      "loss": 1.6646,
+      "step": 53713
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5587863463969654e-07,
+      "loss": 2.5658,
+      "step": 53714
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.558312262958281e-07,
+      "loss": 1.7096,
+      "step": 53715
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.557838179519596e-07,
+      "loss": 1.7185,
+      "step": 53716
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.55736409608091e-07,
+      "loss": 1.5381,
+      "step": 53717
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5568900126422256e-07,
+      "loss": 1.9183,
+      "step": 53718
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.55641592920354e-07,
+      "loss": 2.1286,
+      "step": 53719
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.555941845764855e-07,
+      "loss": 1.9238,
+      "step": 53720
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5554677623261693e-07,
+      "loss": 1.2577,
+      "step": 53721
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.554993678887485e-07,
+      "loss": 0.7919,
+      "step": 53722
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.554519595448799e-07,
+      "loss": 4.2759,
+      "step": 53723
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5540455120101135e-07,
+      "loss": 2.6308,
+      "step": 53724
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5535714285714285e-07,
+      "loss": 1.1906,
+      "step": 53725
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.553097345132744e-07,
+      "loss": 3.9908,
+      "step": 53726
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5526232616940583e-07,
+      "loss": 2.2689,
+      "step": 53727
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5521491782553727e-07,
+      "loss": 2.0426,
+      "step": 53728
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.551675094816688e-07,
+      "loss": 3.1976,
+      "step": 53729
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.551201011378003e-07,
+      "loss": 1.2394,
+      "step": 53730
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5507269279393175e-07,
+      "loss": 2.3895,
+      "step": 53731
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.550252844500632e-07,
+      "loss": 3.2072,
+      "step": 53732
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5497787610619473e-07,
+      "loss": 1.9184,
+      "step": 53733
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5493046776232617e-07,
+      "loss": 1.8459,
+      "step": 53734
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.5488305941845766e-07,
+      "loss": 2.0224,
+      "step": 53735
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.548356510745891e-07,
+      "loss": 1.4518,
+      "step": 53736
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5478824273072065e-07,
+      "loss": 0.5061,
+      "step": 53737
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.547408343868521e-07,
+      "loss": 1.6658,
+      "step": 53738
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.546934260429836e-07,
+      "loss": 1.9012,
+      "step": 53739
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.546460176991151e-07,
+      "loss": 1.8138,
+      "step": 53740
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5459860935524656e-07,
+      "loss": 1.0081,
+      "step": 53741
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.54551201011378e-07,
+      "loss": 0.5721,
+      "step": 53742
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5450379266750944e-07,
+      "loss": 1.7722,
+      "step": 53743
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.54456384323641e-07,
+      "loss": 2.7328,
+      "step": 53744
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.544089759797725e-07,
+      "loss": 1.7489,
+      "step": 53745
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.543615676359039e-07,
+      "loss": 1.9484,
+      "step": 53746
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5431415929203536e-07,
+      "loss": 2.3849,
+      "step": 53747
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.542667509481669e-07,
+      "loss": 2.3534,
+      "step": 53748
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.542193426042984e-07,
+      "loss": 1.8432,
+      "step": 53749
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5417193426042983e-07,
+      "loss": 2.667,
+      "step": 53750
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.541245259165614e-07,
+      "loss": 1.3715,
+      "step": 53751
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.540771175726928e-07,
+      "loss": 0.5772,
+      "step": 53752
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5402970922882426e-07,
+      "loss": 1.4759,
+      "step": 53753
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5398230088495575e-07,
+      "loss": 1.4466,
+      "step": 53754
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.539348925410873e-07,
+      "loss": 2.508,
+      "step": 53755
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5388748419721873e-07,
+      "loss": 1.2151,
+      "step": 53756
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5384007585335017e-07,
+      "loss": 1.6957,
+      "step": 53757
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5379266750948166e-07,
+      "loss": 2.6614,
+      "step": 53758
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.537452591656132e-07,
+      "loss": 2.9713,
+      "step": 53759
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5369785082174465e-07,
+      "loss": 2.8395,
+      "step": 53760
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.536504424778761e-07,
+      "loss": 2.7296,
+      "step": 53761
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5360303413400763e-07,
+      "loss": 1.7211,
+      "step": 53762
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5355562579013907e-07,
+      "loss": 1.8733,
+      "step": 53763
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5350821744627056e-07,
+      "loss": 1.8528,
+      "step": 53764
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.53460809102402e-07,
+      "loss": 1.6557,
+      "step": 53765
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5341340075853355e-07,
+      "loss": 2.1639,
+      "step": 53766
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.53365992414665e-07,
+      "loss": 1.8606,
+      "step": 53767
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.533185840707965e-07,
+      "loss": 1.0324,
+      "step": 53768
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.532711757269279e-07,
+      "loss": 1.526,
+      "step": 53769
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5322376738305946e-07,
+      "loss": 2.3581,
+      "step": 53770
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.531763590391909e-07,
+      "loss": 0.5603,
+      "step": 53771
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5312895069532234e-07,
+      "loss": 2.0066,
+      "step": 53772
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5308154235145384e-07,
+      "loss": 1.9309,
+      "step": 53773
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.530341340075854e-07,
+      "loss": 1.4617,
+      "step": 53774
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.529867256637168e-07,
+      "loss": 1.7906,
+      "step": 53775
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5293931731984826e-07,
+      "loss": 1.9957,
+      "step": 53776
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.528919089759798e-07,
+      "loss": 1.9654,
+      "step": 53777
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.528445006321113e-07,
+      "loss": 2.2938,
+      "step": 53778
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5279709228824274e-07,
+      "loss": 1.6498,
+      "step": 53779
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.527496839443742e-07,
+      "loss": 1.5163,
+      "step": 53780
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.527022756005057e-07,
+      "loss": 1.4673,
+      "step": 53781
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5265486725663716e-07,
+      "loss": 2.208,
+      "step": 53782
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5260745891276865e-07,
+      "loss": 2.0568,
+      "step": 53783
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.525600505689001e-07,
+      "loss": 2.4563,
+      "step": 53784
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5251264222503164e-07,
+      "loss": 1.5353,
+      "step": 53785
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.524652338811631e-07,
+      "loss": 1.5851,
+      "step": 53786
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5241782553729457e-07,
+      "loss": 0.4849,
+      "step": 53787
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.523704171934261e-07,
+      "loss": 2.2876,
+      "step": 53788
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5232300884955755e-07,
+      "loss": 1.5075,
+      "step": 53789
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.52275600505689e-07,
+      "loss": 1.8984,
+      "step": 53790
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.522281921618205e-07,
+      "loss": 2.7257,
+      "step": 53791
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.52180783817952e-07,
+      "loss": 0.3723,
+      "step": 53792
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5213337547408347e-07,
+      "loss": 2.1952,
+      "step": 53793
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.520859671302149e-07,
+      "loss": 0.3814,
+      "step": 53794
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.5203855878634635e-07,
+      "loss": 1.338,
+      "step": 53795
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.519911504424779e-07,
+      "loss": 1.3405,
+      "step": 53796
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.519437420986094e-07,
+      "loss": 3.6571,
+      "step": 53797
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.518963337547408e-07,
+      "loss": 1.1279,
+      "step": 53798
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5184892541087237e-07,
+      "loss": 1.6136,
+      "step": 53799
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.518015170670038e-07,
+      "loss": 1.4962,
+      "step": 53800
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.517541087231353e-07,
+      "loss": 1.571,
+      "step": 53801
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5170670037926674e-07,
+      "loss": 2.7342,
+      "step": 53802
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.516592920353983e-07,
+      "loss": 2.3537,
+      "step": 53803
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.516118836915297e-07,
+      "loss": 2.3418,
+      "step": 53804
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5156447534766116e-07,
+      "loss": 2.8769,
+      "step": 53805
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5151706700379265e-07,
+      "loss": 0.6799,
+      "step": 53806
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.514696586599242e-07,
+      "loss": 1.2386,
+      "step": 53807
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5142225031605564e-07,
+      "loss": 2.3866,
+      "step": 53808
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.513748419721871e-07,
+      "loss": 1.3946,
+      "step": 53809
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.513274336283186e-07,
+      "loss": 1.3653,
+      "step": 53810
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5128002528445006e-07,
+      "loss": 1.8396,
+      "step": 53811
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5123261694058155e-07,
+      "loss": 2.874,
+      "step": 53812
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.51185208596713e-07,
+      "loss": 2.2582,
+      "step": 53813
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5113780025284454e-07,
+      "loss": 1.3441,
+      "step": 53814
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.51090391908976e-07,
+      "loss": 1.9063,
+      "step": 53815
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5104298356510747e-07,
+      "loss": 1.42,
+      "step": 53816
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.509955752212389e-07,
+      "loss": 0.841,
+      "step": 53817
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5094816687737045e-07,
+      "loss": 2.2182,
+      "step": 53818
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.509007585335019e-07,
+      "loss": 1.7271,
+      "step": 53819
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.508533501896334e-07,
+      "loss": 2.7071,
+      "step": 53820
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.508059418457649e-07,
+      "loss": 1.9268,
+      "step": 53821
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5075853350189637e-07,
+      "loss": 2.4317,
+      "step": 53822
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.507111251580278e-07,
+      "loss": 1.0727,
+      "step": 53823
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5066371681415925e-07,
+      "loss": 1.3048,
+      "step": 53824
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.506163084702908e-07,
+      "loss": 1.0737,
+      "step": 53825
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.505689001264223e-07,
+      "loss": 3.2778,
+      "step": 53826
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.505214917825537e-07,
+      "loss": 2.4651,
+      "step": 53827
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5047408343868516e-07,
+      "loss": 2.5902,
+      "step": 53828
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.504266750948167e-07,
+      "loss": 1.9824,
+      "step": 53829
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.503792667509482e-07,
+      "loss": 1.311,
+      "step": 53830
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5033185840707964e-07,
+      "loss": 1.2989,
+      "step": 53831
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.502844500632112e-07,
+      "loss": 2.1927,
+      "step": 53832
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.502370417193426e-07,
+      "loss": 2.4631,
+      "step": 53833
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5018963337547406e-07,
+      "loss": 2.4534,
+      "step": 53834
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5014222503160556e-07,
+      "loss": 1.0658,
+      "step": 53835
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.500948166877371e-07,
+      "loss": 1.2882,
+      "step": 53836
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5004740834386854e-07,
+      "loss": 1.2676,
+      "step": 53837
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.5e-07,
+      "loss": 1.2545,
+      "step": 53838
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4995259165613147e-07,
+      "loss": 1.5843,
+      "step": 53839
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.49905183312263e-07,
+      "loss": 1.7495,
+      "step": 53840
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4985777496839446e-07,
+      "loss": 0.9333,
+      "step": 53841
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.498103666245259e-07,
+      "loss": 1.0712,
+      "step": 53842
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4976295828065744e-07,
+      "loss": 3.2636,
+      "step": 53843
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.497155499367889e-07,
+      "loss": 2.0909,
+      "step": 53844
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4966814159292037e-07,
+      "loss": 0.837,
+      "step": 53845
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.496207332490518e-07,
+      "loss": 2.6279,
+      "step": 53846
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4957332490518336e-07,
+      "loss": 0.8185,
+      "step": 53847
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.495259165613148e-07,
+      "loss": 1.2242,
+      "step": 53848
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.494785082174463e-07,
+      "loss": 1.0215,
+      "step": 53849
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4943109987357773e-07,
+      "loss": 0.9223,
+      "step": 53850
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4938369152970927e-07,
+      "loss": 1.9245,
+      "step": 53851
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.493362831858407e-07,
+      "loss": 1.6981,
+      "step": 53852
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4928887484197215e-07,
+      "loss": 2.3183,
+      "step": 53853
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.492414664981037e-07,
+      "loss": 0.8935,
+      "step": 53854
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.491940581542352e-07,
+      "loss": 1.6802,
+      "step": 53855
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4914664981036663e-07,
+      "loss": 0.6296,
+      "step": 53856
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4909924146649807e-07,
+      "loss": 0.9457,
+      "step": 53857
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.490518331226296e-07,
+      "loss": 0.6281,
+      "step": 53858
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.490044247787611e-07,
+      "loss": 0.7481,
+      "step": 53859
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4895701643489254e-07,
+      "loss": 2.8123,
+      "step": 53860
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.48909608091024e-07,
+      "loss": 1.855,
+      "step": 53861
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.4886219974715553e-07,
+      "loss": 0.9707,
+      "step": 53862
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4881479140328697e-07,
+      "loss": 1.9847,
+      "step": 53863
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4876738305941846e-07,
+      "loss": 1.5244,
+      "step": 53864
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4871997471555e-07,
+      "loss": 2.2709,
+      "step": 53865
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4867256637168144e-07,
+      "loss": 3.2613,
+      "step": 53866
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.486251580278129e-07,
+      "loss": 2.82,
+      "step": 53867
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.485777496839444e-07,
+      "loss": 1.566,
+      "step": 53868
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.485303413400759e-07,
+      "loss": 1.1671,
+      "step": 53869
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4848293299620736e-07,
+      "loss": 2.8878,
+      "step": 53870
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.484355246523388e-07,
+      "loss": 1.9375,
+      "step": 53871
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.483881163084703e-07,
+      "loss": 3.2204,
+      "step": 53872
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.483407079646018e-07,
+      "loss": 2.9577,
+      "step": 53873
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.482932996207333e-07,
+      "loss": 1.2407,
+      "step": 53874
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.482458912768647e-07,
+      "loss": 0.7287,
+      "step": 53875
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4819848293299626e-07,
+      "loss": 0.5261,
+      "step": 53876
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.481510745891277e-07,
+      "loss": 2.0845,
+      "step": 53877
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.481036662452592e-07,
+      "loss": 3.065,
+      "step": 53878
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4805625790139063e-07,
+      "loss": 1.3542,
+      "step": 53879
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.480088495575222e-07,
+      "loss": 1.2287,
+      "step": 53880
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.479614412136536e-07,
+      "loss": 2.0587,
+      "step": 53881
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4791403286978505e-07,
+      "loss": 1.5357,
+      "step": 53882
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4786662452591655e-07,
+      "loss": 0.9381,
+      "step": 53883
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.478192161820481e-07,
+      "loss": 1.7014,
+      "step": 53884
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4777180783817953e-07,
+      "loss": 0.5663,
+      "step": 53885
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4772439949431097e-07,
+      "loss": 0.8647,
+      "step": 53886
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.476769911504425e-07,
+      "loss": 1.7683,
+      "step": 53887
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.47629582806574e-07,
+      "loss": 0.813,
+      "step": 53888
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4758217446270545e-07,
+      "loss": 3.0344,
+      "step": 53889
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.475347661188369e-07,
+      "loss": 0.1316,
+      "step": 53890
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4748735777496843e-07,
+      "loss": 1.0533,
+      "step": 53891
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4743994943109987e-07,
+      "loss": 2.4421,
+      "step": 53892
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4739254108723136e-07,
+      "loss": 1.516,
+      "step": 53893
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.473451327433628e-07,
+      "loss": 1.0945,
+      "step": 53894
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4729772439949435e-07,
+      "loss": 2.6079,
+      "step": 53895
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.472503160556258e-07,
+      "loss": 2.1646,
+      "step": 53896
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.472029077117573e-07,
+      "loss": 2.986,
+      "step": 53897
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.471554993678888e-07,
+      "loss": 2.5512,
+      "step": 53898
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4710809102402026e-07,
+      "loss": 1.7748,
+      "step": 53899
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.470606826801517e-07,
+      "loss": 2.8908,
+      "step": 53900
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.470132743362832e-07,
+      "loss": 0.893,
+      "step": 53901
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.469658659924147e-07,
+      "loss": 1.5777,
+      "step": 53902
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.469184576485462e-07,
+      "loss": 3.1805,
+      "step": 53903
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.468710493046776e-07,
+      "loss": 1.7683,
+      "step": 53904
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4682364096080906e-07,
+      "loss": 3.2411,
+      "step": 53905
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.467762326169406e-07,
+      "loss": 2.6374,
+      "step": 53906
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.467288242730721e-07,
+      "loss": 1.1365,
+      "step": 53907
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4668141592920353e-07,
+      "loss": 1.9151,
+      "step": 53908
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.466340075853351e-07,
+      "loss": 1.9183,
+      "step": 53909
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.465865992414665e-07,
+      "loss": 3.4125,
+      "step": 53910
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4653919089759796e-07,
+      "loss": 1.2875,
+      "step": 53911
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4649178255372945e-07,
+      "loss": 1.7644,
+      "step": 53912
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.46444374209861e-07,
+      "loss": 2.0803,
+      "step": 53913
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4639696586599243e-07,
+      "loss": 1.9502,
+      "step": 53914
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4634955752212387e-07,
+      "loss": 2.0111,
+      "step": 53915
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4630214917825536e-07,
+      "loss": 1.6088,
+      "step": 53916
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.462547408343869e-07,
+      "loss": 1.2777,
+      "step": 53917
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4620733249051835e-07,
+      "loss": 1.9711,
+      "step": 53918
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.461599241466498e-07,
+      "loss": 2.442,
+      "step": 53919
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4611251580278133e-07,
+      "loss": 2.3082,
+      "step": 53920
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4606510745891277e-07,
+      "loss": 1.9294,
+      "step": 53921
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4601769911504426e-07,
+      "loss": 1.48,
+      "step": 53922
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.459702907711757e-07,
+      "loss": 2.2028,
+      "step": 53923
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.4592288242730725e-07,
+      "loss": 2.2704,
+      "step": 53924
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.458754740834387e-07,
+      "loss": 2.0775,
+      "step": 53925
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.458280657395702e-07,
+      "loss": 1.7886,
+      "step": 53926
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.457806573957016e-07,
+      "loss": 2.1914,
+      "step": 53927
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4573324905183316e-07,
+      "loss": 2.7774,
+      "step": 53928
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.456858407079646e-07,
+      "loss": 2.9321,
+      "step": 53929
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.456384323640961e-07,
+      "loss": 1.6946,
+      "step": 53930
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.455910240202276e-07,
+      "loss": 1.9007,
+      "step": 53931
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.455436156763591e-07,
+      "loss": 1.491,
+      "step": 53932
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.454962073324905e-07,
+      "loss": 1.1611,
+      "step": 53933
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4544879898862196e-07,
+      "loss": 1.8872,
+      "step": 53934
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.454013906447535e-07,
+      "loss": 3.1299,
+      "step": 53935
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.45353982300885e-07,
+      "loss": 2.3224,
+      "step": 53936
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4530657395701643e-07,
+      "loss": 0.9384,
+      "step": 53937
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.452591656131479e-07,
+      "loss": 2.3788,
+      "step": 53938
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.452117572692794e-07,
+      "loss": 2.9463,
+      "step": 53939
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.451643489254109e-07,
+      "loss": 1.3743,
+      "step": 53940
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4511694058154235e-07,
+      "loss": 2.842,
+      "step": 53941
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.450695322376739e-07,
+      "loss": 1.2867,
+      "step": 53942
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4502212389380533e-07,
+      "loss": 1.9584,
+      "step": 53943
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.449747155499368e-07,
+      "loss": 2.3423,
+      "step": 53944
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4492730720606827e-07,
+      "loss": 0.5989,
+      "step": 53945
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.448798988621998e-07,
+      "loss": 1.037,
+      "step": 53946
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4483249051833125e-07,
+      "loss": 1.3096,
+      "step": 53947
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.447850821744627e-07,
+      "loss": 0.9296,
+      "step": 53948
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.447376738305942e-07,
+      "loss": 2.7961,
+      "step": 53949
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.446902654867257e-07,
+      "loss": 1.4337,
+      "step": 53950
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4464285714285717e-07,
+      "loss": 3.9998,
+      "step": 53951
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.445954487989886e-07,
+      "loss": 1.5165,
+      "step": 53952
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4454804045512015e-07,
+      "loss": 2.5793,
+      "step": 53953
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.445006321112516e-07,
+      "loss": 0.7034,
+      "step": 53954
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.444532237673831e-07,
+      "loss": 3.2477,
+      "step": 53955
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.444058154235145e-07,
+      "loss": 2.0858,
+      "step": 53956
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4435840707964607e-07,
+      "loss": 1.8648,
+      "step": 53957
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.443109987357775e-07,
+      "loss": 2.1728,
+      "step": 53958
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.44263590391909e-07,
+      "loss": 1.5244,
+      "step": 53959
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4421618204804044e-07,
+      "loss": 2.1499,
+      "step": 53960
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.44168773704172e-07,
+      "loss": 1.8583,
+      "step": 53961
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.441213653603034e-07,
+      "loss": 1.8661,
+      "step": 53962
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4407395701643486e-07,
+      "loss": 0.699,
+      "step": 53963
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.440265486725664e-07,
+      "loss": 1.6576,
+      "step": 53964
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.439791403286979e-07,
+      "loss": 0.7755,
+      "step": 53965
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4393173198482934e-07,
+      "loss": 2.1263,
+      "step": 53966
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.438843236409608e-07,
+      "loss": 1.1568,
+      "step": 53967
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.438369152970923e-07,
+      "loss": 3.3032,
+      "step": 53968
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.437895069532238e-07,
+      "loss": 1.0676,
+      "step": 53969
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4374209860935525e-07,
+      "loss": 1.0039,
+      "step": 53970
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.436946902654867e-07,
+      "loss": 2.6269,
+      "step": 53971
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4364728192161824e-07,
+      "loss": 0.8536,
+      "step": 53972
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.435998735777497e-07,
+      "loss": 1.2067,
+      "step": 53973
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4355246523388117e-07,
+      "loss": 3.7335,
+      "step": 53974
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.435050568900127e-07,
+      "loss": 2.4245,
+      "step": 53975
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4345764854614415e-07,
+      "loss": 2.2582,
+      "step": 53976
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.434102402022756e-07,
+      "loss": 2.3378,
+      "step": 53977
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.433628318584071e-07,
+      "loss": 1.2303,
+      "step": 53978
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4331542351453863e-07,
+      "loss": 1.6796,
+      "step": 53979
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4326801517067007e-07,
+      "loss": 3.7303,
+      "step": 53980
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.432206068268015e-07,
+      "loss": 2.3835,
+      "step": 53981
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4317319848293295e-07,
+      "loss": 0.9757,
+      "step": 53982
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.431257901390645e-07,
+      "loss": 2.2034,
+      "step": 53983
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.43078381795196e-07,
+      "loss": 0.742,
+      "step": 53984
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.430309734513274e-07,
+      "loss": 1.2895,
+      "step": 53985
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.4298356510745897e-07,
+      "loss": 1.5618,
+      "step": 53986
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.429361567635904e-07,
+      "loss": 1.6006,
+      "step": 53987
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.428887484197219e-07,
+      "loss": 2.311,
+      "step": 53988
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4284134007585334e-07,
+      "loss": 2.5414,
+      "step": 53989
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.427939317319849e-07,
+      "loss": 2.0535,
+      "step": 53990
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.427465233881163e-07,
+      "loss": 1.8854,
+      "step": 53991
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4269911504424776e-07,
+      "loss": 1.6216,
+      "step": 53992
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4265170670037926e-07,
+      "loss": 1.5448,
+      "step": 53993
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.426042983565108e-07,
+      "loss": 2.6093,
+      "step": 53994
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4255689001264224e-07,
+      "loss": 2.1244,
+      "step": 53995
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.425094816687737e-07,
+      "loss": 1.0348,
+      "step": 53996
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4246207332490517e-07,
+      "loss": 2.5912,
+      "step": 53997
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.424146649810367e-07,
+      "loss": 1.6811,
+      "step": 53998
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4236725663716816e-07,
+      "loss": 2.701,
+      "step": 53999
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.423198482932996e-07,
+      "loss": 1.2172,
+      "step": 54000
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4227243994943114e-07,
+      "loss": 1.011,
+      "step": 54001
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.422250316055626e-07,
+      "loss": 2.9831,
+      "step": 54002
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4217762326169407e-07,
+      "loss": 0.7739,
+      "step": 54003
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.421302149178255e-07,
+      "loss": 0.3615,
+      "step": 54004
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4208280657395706e-07,
+      "loss": 1.1926,
+      "step": 54005
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.420353982300885e-07,
+      "loss": 1.7534,
+      "step": 54006
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4198798988622e-07,
+      "loss": 2.7241,
+      "step": 54007
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.419405815423514e-07,
+      "loss": 1.3081,
+      "step": 54008
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4189317319848297e-07,
+      "loss": 1.2901,
+      "step": 54009
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.418457648546144e-07,
+      "loss": 1.7385,
+      "step": 54010
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.417983565107459e-07,
+      "loss": 2.9012,
+      "step": 54011
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.417509481668774e-07,
+      "loss": 1.9623,
+      "step": 54012
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.417035398230089e-07,
+      "loss": 2.1601,
+      "step": 54013
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.416561314791403e-07,
+      "loss": 1.2378,
+      "step": 54014
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4160872313527177e-07,
+      "loss": 1.7941,
+      "step": 54015
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.415613147914033e-07,
+      "loss": 1.1227,
+      "step": 54016
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.415139064475348e-07,
+      "loss": 0.6892,
+      "step": 54017
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4146649810366624e-07,
+      "loss": 1.2826,
+      "step": 54018
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.414190897597977e-07,
+      "loss": 4.4739,
+      "step": 54019
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.413716814159292e-07,
+      "loss": 0.759,
+      "step": 54020
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4132427307206067e-07,
+      "loss": 1.7973,
+      "step": 54021
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4127686472819216e-07,
+      "loss": 2.1962,
+      "step": 54022
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.412294563843237e-07,
+      "loss": 1.0807,
+      "step": 54023
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4118204804045514e-07,
+      "loss": 2.6132,
+      "step": 54024
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.411346396965866e-07,
+      "loss": 1.3526,
+      "step": 54025
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.410872313527181e-07,
+      "loss": 2.6457,
+      "step": 54026
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.410398230088496e-07,
+      "loss": 1.7636,
+      "step": 54027
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4099241466498106e-07,
+      "loss": 2.4883,
+      "step": 54028
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.409450063211125e-07,
+      "loss": 0.564,
+      "step": 54029
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.40897597977244e-07,
+      "loss": 1.6382,
+      "step": 54030
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.408501896333755e-07,
+      "loss": 3.3865,
+      "step": 54031
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.40802781289507e-07,
+      "loss": 2.5821,
+      "step": 54032
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.407553729456384e-07,
+      "loss": 1.2036,
+      "step": 54033
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4070796460176996e-07,
+      "loss": 2.1847,
+      "step": 54034
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.406605562579014e-07,
+      "loss": 1.6132,
+      "step": 54035
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.406131479140329e-07,
+      "loss": 2.3625,
+      "step": 54036
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4056573957016433e-07,
+      "loss": 2.4197,
+      "step": 54037
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.405183312262959e-07,
+      "loss": 2.0158,
+      "step": 54038
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.404709228824273e-07,
+      "loss": 0.7894,
+      "step": 54039
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.404235145385588e-07,
+      "loss": 1.2916,
+      "step": 54040
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4037610619469024e-07,
+      "loss": 2.3294,
+      "step": 54041
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.403286978508218e-07,
+      "loss": 2.7301,
+      "step": 54042
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4028128950695323e-07,
+      "loss": 2.2665,
+      "step": 54043
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4023388116308467e-07,
+      "loss": 0.8038,
+      "step": 54044
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.401864728192162e-07,
+      "loss": 1.6919,
+      "step": 54045
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.401390644753477e-07,
+      "loss": 1.5172,
+      "step": 54046
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.4009165613147914e-07,
+      "loss": 2.4968,
+      "step": 54047
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.400442477876106e-07,
+      "loss": 2.012,
+      "step": 54048
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.3999683944374213e-07,
+      "loss": 1.342,
+      "step": 54049
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.399494310998736e-07,
+      "loss": 2.3493,
+      "step": 54050
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.3990202275600506e-07,
+      "loss": 1.5864,
+      "step": 54051
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.398546144121365e-07,
+      "loss": 2.0795,
+      "step": 54052
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3980720606826804e-07,
+      "loss": 1.7981,
+      "step": 54053
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.397597977243995e-07,
+      "loss": 1.53,
+      "step": 54054
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.39712389380531e-07,
+      "loss": 2.546,
+      "step": 54055
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.396649810366625e-07,
+      "loss": 2.173,
+      "step": 54056
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3961757269279396e-07,
+      "loss": 2.2036,
+      "step": 54057
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.395701643489254e-07,
+      "loss": 2.9748,
+      "step": 54058
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.395227560050569e-07,
+      "loss": 2.1149,
+      "step": 54059
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.394753476611884e-07,
+      "loss": 1.9871,
+      "step": 54060
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.394279393173199e-07,
+      "loss": 1.7553,
+      "step": 54061
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.393805309734513e-07,
+      "loss": 1.5149,
+      "step": 54062
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3933312262958275e-07,
+      "loss": 2.4501,
+      "step": 54063
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.392857142857143e-07,
+      "loss": 0.9289,
+      "step": 54064
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.392383059418458e-07,
+      "loss": 0.8053,
+      "step": 54065
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3919089759797723e-07,
+      "loss": 1.7911,
+      "step": 54066
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.391434892541088e-07,
+      "loss": 1.4293,
+      "step": 54067
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.390960809102402e-07,
+      "loss": 1.1718,
+      "step": 54068
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.390486725663717e-07,
+      "loss": 1.4114,
+      "step": 54069
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3900126422250315e-07,
+      "loss": 0.751,
+      "step": 54070
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.389538558786347e-07,
+      "loss": 1.5391,
+      "step": 54071
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3890644753476613e-07,
+      "loss": 1.4361,
+      "step": 54072
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3885903919089757e-07,
+      "loss": 1.3026,
+      "step": 54073
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3881163084702906e-07,
+      "loss": 0.8816,
+      "step": 54074
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.387642225031606e-07,
+      "loss": 2.1487,
+      "step": 54075
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3871681415929205e-07,
+      "loss": 1.6625,
+      "step": 54076
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.386694058154235e-07,
+      "loss": 1.6699,
+      "step": 54077
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3862199747155503e-07,
+      "loss": 2.8731,
+      "step": 54078
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.385745891276865e-07,
+      "loss": 1.8214,
+      "step": 54079
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3852718078381796e-07,
+      "loss": 2.4854,
+      "step": 54080
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.384797724399494e-07,
+      "loss": 0.9591,
+      "step": 54081
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3843236409608095e-07,
+      "loss": 1.772,
+      "step": 54082
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.383849557522124e-07,
+      "loss": 2.2477,
+      "step": 54083
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.383375474083439e-07,
+      "loss": 3.0471,
+      "step": 54084
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.382901390644753e-07,
+      "loss": 0.8532,
+      "step": 54085
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3824273072060686e-07,
+      "loss": 3.1704,
+      "step": 54086
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.381953223767383e-07,
+      "loss": 2.7128,
+      "step": 54087
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.381479140328698e-07,
+      "loss": 1.6038,
+      "step": 54088
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.381005056890013e-07,
+      "loss": 2.0864,
+      "step": 54089
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.380530973451328e-07,
+      "loss": 2.0911,
+      "step": 54090
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.380056890012642e-07,
+      "loss": 1.2122,
+      "step": 54091
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3795828065739566e-07,
+      "loss": 3.5276,
+      "step": 54092
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.379108723135272e-07,
+      "loss": 1.6929,
+      "step": 54093
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.378634639696587e-07,
+      "loss": 1.3859,
+      "step": 54094
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3781605562579013e-07,
+      "loss": 3.0155,
+      "step": 54095
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3776864728192157e-07,
+      "loss": 0.9183,
+      "step": 54096
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.377212389380531e-07,
+      "loss": 2.6623,
+      "step": 54097
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.376738305941846e-07,
+      "loss": 2.8246,
+      "step": 54098
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3762642225031605e-07,
+      "loss": 1.6849,
+      "step": 54099
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.375790139064476e-07,
+      "loss": 1.3281,
+      "step": 54100
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3753160556257903e-07,
+      "loss": 2.105,
+      "step": 54101
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3748419721871047e-07,
+      "loss": 2.3599,
+      "step": 54102
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3743678887484197e-07,
+      "loss": 1.8073,
+      "step": 54103
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.373893805309735e-07,
+      "loss": 2.0865,
+      "step": 54104
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3734197218710495e-07,
+      "loss": 1.1494,
+      "step": 54105
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.372945638432364e-07,
+      "loss": 2.015,
+      "step": 54106
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.372471554993679e-07,
+      "loss": 1.4411,
+      "step": 54107
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.371997471554994e-07,
+      "loss": 2.0224,
+      "step": 54108
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3715233881163087e-07,
+      "loss": 0.5571,
+      "step": 54109
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.371049304677623e-07,
+      "loss": 0.6074,
+      "step": 54110
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3705752212389385e-07,
+      "loss": 1.9125,
+      "step": 54111
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.370101137800253e-07,
+      "loss": 0.9946,
+      "step": 54112
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.369627054361568e-07,
+      "loss": 2.4843,
+      "step": 54113
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.369152970922882e-07,
+      "loss": 1.351,
+      "step": 54114
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.3686788874841977e-07,
+      "loss": 2.0625,
+      "step": 54115
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.368204804045512e-07,
+      "loss": 1.6087,
+      "step": 54116
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.367730720606827e-07,
+      "loss": 2.7737,
+      "step": 54117
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3672566371681414e-07,
+      "loss": 1.1955,
+      "step": 54118
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.366782553729457e-07,
+      "loss": 1.6018,
+      "step": 54119
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.366308470290771e-07,
+      "loss": 0.8544,
+      "step": 54120
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3658343868520856e-07,
+      "loss": 1.4821,
+      "step": 54121
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.365360303413401e-07,
+      "loss": 1.6184,
+      "step": 54122
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.364886219974716e-07,
+      "loss": 2.0806,
+      "step": 54123
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3644121365360304e-07,
+      "loss": 2.1487,
+      "step": 54124
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.363938053097345e-07,
+      "loss": 1.5288,
+      "step": 54125
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.36346396965866e-07,
+      "loss": 2.9935,
+      "step": 54126
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.362989886219975e-07,
+      "loss": 2.2202,
+      "step": 54127
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3625158027812895e-07,
+      "loss": 1.643,
+      "step": 54128
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.362041719342604e-07,
+      "loss": 2.6882,
+      "step": 54129
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3615676359039194e-07,
+      "loss": 2.296,
+      "step": 54130
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.361093552465234e-07,
+      "loss": 2.4992,
+      "step": 54131
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3606194690265487e-07,
+      "loss": 2.0899,
+      "step": 54132
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.360145385587864e-07,
+      "loss": 1.9772,
+      "step": 54133
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3596713021491785e-07,
+      "loss": 1.5092,
+      "step": 54134
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.359197218710493e-07,
+      "loss": 1.9248,
+      "step": 54135
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.358723135271808e-07,
+      "loss": 2.4291,
+      "step": 54136
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3582490518331233e-07,
+      "loss": 2.6195,
+      "step": 54137
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3577749683944377e-07,
+      "loss": 2.6411,
+      "step": 54138
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.357300884955752e-07,
+      "loss": 2.6819,
+      "step": 54139
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.356826801517067e-07,
+      "loss": 2.8637,
+      "step": 54140
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.356352718078382e-07,
+      "loss": 3.4888,
+      "step": 54141
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.355878634639697e-07,
+      "loss": 1.2265,
+      "step": 54142
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.355404551201011e-07,
+      "loss": 1.0874,
+      "step": 54143
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3549304677623267e-07,
+      "loss": 2.1641,
+      "step": 54144
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.354456384323641e-07,
+      "loss": 0.3916,
+      "step": 54145
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.353982300884956e-07,
+      "loss": 2.5718,
+      "step": 54146
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3535082174462704e-07,
+      "loss": 1.3196,
+      "step": 54147
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.353034134007586e-07,
+      "loss": 2.7952,
+      "step": 54148
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3525600505689e-07,
+      "loss": 2.6703,
+      "step": 54149
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.352085967130215e-07,
+      "loss": 1.6885,
+      "step": 54150
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3516118836915295e-07,
+      "loss": 2.6667,
+      "step": 54151
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.351137800252845e-07,
+      "loss": 1.1754,
+      "step": 54152
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3506637168141594e-07,
+      "loss": 1.5881,
+      "step": 54153
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.350189633375474e-07,
+      "loss": 2.6874,
+      "step": 54154
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.349715549936789e-07,
+      "loss": 1.348,
+      "step": 54155
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.349241466498104e-07,
+      "loss": 1.3514,
+      "step": 54156
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3487673830594185e-07,
+      "loss": 0.8571,
+      "step": 54157
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.348293299620733e-07,
+      "loss": 2.5778,
+      "step": 54158
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3478192161820484e-07,
+      "loss": 3.199,
+      "step": 54159
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.347345132743363e-07,
+      "loss": 2.6604,
+      "step": 54160
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3468710493046777e-07,
+      "loss": 2.3917,
+      "step": 54161
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.346396965865992e-07,
+      "loss": 0.7731,
+      "step": 54162
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3459228824273075e-07,
+      "loss": 1.6571,
+      "step": 54163
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.345448798988622e-07,
+      "loss": 2.9526,
+      "step": 54164
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.344974715549937e-07,
+      "loss": 1.7356,
+      "step": 54165
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3445006321112523e-07,
+      "loss": 3.1201,
+      "step": 54166
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3440265486725667e-07,
+      "loss": 0.9843,
+      "step": 54167
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.343552465233881e-07,
+      "loss": 2.735,
+      "step": 54168
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.343078381795196e-07,
+      "loss": 1.6628,
+      "step": 54169
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.342604298356511e-07,
+      "loss": 2.2509,
+      "step": 54170
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.342130214917826e-07,
+      "loss": 1.9017,
+      "step": 54171
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.34165613147914e-07,
+      "loss": 2.1415,
+      "step": 54172
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3411820480404546e-07,
+      "loss": 2.5876,
+      "step": 54173
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.34070796460177e-07,
+      "loss": 2.3331,
+      "step": 54174
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.340233881163085e-07,
+      "loss": 2.1159,
+      "step": 54175
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.3397597977243994e-07,
+      "loss": 1.4813,
+      "step": 54176
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.339285714285715e-07,
+      "loss": 4.5901,
+      "step": 54177
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.338811630847029e-07,
+      "loss": 1.3736,
+      "step": 54178
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.338337547408344e-07,
+      "loss": 2.8874,
+      "step": 54179
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3378634639696586e-07,
+      "loss": 1.2087,
+      "step": 54180
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.337389380530974e-07,
+      "loss": 1.4261,
+      "step": 54181
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3369152970922884e-07,
+      "loss": 1.4742,
+      "step": 54182
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.336441213653603e-07,
+      "loss": 1.7093,
+      "step": 54183
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3359671302149177e-07,
+      "loss": 2.8323,
+      "step": 54184
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.335493046776233e-07,
+      "loss": 1.3036,
+      "step": 54185
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3350189633375476e-07,
+      "loss": 2.128,
+      "step": 54186
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.334544879898862e-07,
+      "loss": 2.5626,
+      "step": 54187
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3340707964601774e-07,
+      "loss": 0.8292,
+      "step": 54188
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3335967130214923e-07,
+      "loss": 1.2083,
+      "step": 54189
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3331226295828067e-07,
+      "loss": 0.7841,
+      "step": 54190
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.332648546144121e-07,
+      "loss": 2.5412,
+      "step": 54191
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3321744627054366e-07,
+      "loss": 1.1627,
+      "step": 54192
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.331700379266751e-07,
+      "loss": 2.2227,
+      "step": 54193
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.331226295828066e-07,
+      "loss": 1.4094,
+      "step": 54194
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3307522123893803e-07,
+      "loss": 1.3897,
+      "step": 54195
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3302781289506957e-07,
+      "loss": 3.6613,
+      "step": 54196
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.32980404551201e-07,
+      "loss": 1.3245,
+      "step": 54197
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.329329962073325e-07,
+      "loss": 1.4279,
+      "step": 54198
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.32885587863464e-07,
+      "loss": 1.9695,
+      "step": 54199
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.328381795195955e-07,
+      "loss": 1.3995,
+      "step": 54200
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3279077117572693e-07,
+      "loss": 2.2043,
+      "step": 54201
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3274336283185837e-07,
+      "loss": 3.304,
+      "step": 54202
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.326959544879899e-07,
+      "loss": 1.0674,
+      "step": 54203
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.326485461441214e-07,
+      "loss": 0.8988,
+      "step": 54204
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3260113780025284e-07,
+      "loss": 2.0463,
+      "step": 54205
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.325537294563843e-07,
+      "loss": 2.5935,
+      "step": 54206
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3250632111251583e-07,
+      "loss": 2.3219,
+      "step": 54207
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.324589127686473e-07,
+      "loss": 1.6934,
+      "step": 54208
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3241150442477876e-07,
+      "loss": 2.697,
+      "step": 54209
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.323640960809103e-07,
+      "loss": 3.1102,
+      "step": 54210
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3231668773704174e-07,
+      "loss": 1.1028,
+      "step": 54211
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.322692793931732e-07,
+      "loss": 3.1169,
+      "step": 54212
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.322218710493047e-07,
+      "loss": 1.6055,
+      "step": 54213
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.321744627054362e-07,
+      "loss": 0.88,
+      "step": 54214
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3212705436156766e-07,
+      "loss": 0.2056,
+      "step": 54215
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.320796460176991e-07,
+      "loss": 0.4915,
+      "step": 54216
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.320322376738306e-07,
+      "loss": 1.9336,
+      "step": 54217
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3198482932996214e-07,
+      "loss": 1.6498,
+      "step": 54218
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.319374209860936e-07,
+      "loss": 1.8628,
+      "step": 54219
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.31890012642225e-07,
+      "loss": 1.1667,
+      "step": 54220
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3184260429835645e-07,
+      "loss": 1.2525,
+      "step": 54221
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.31795195954488e-07,
+      "loss": 2.6025,
+      "step": 54222
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.317477876106195e-07,
+      "loss": 1.6943,
+      "step": 54223
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3170037926675093e-07,
+      "loss": 3.3432,
+      "step": 54224
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.316529709228825e-07,
+      "loss": 1.43,
+      "step": 54225
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.316055625790139e-07,
+      "loss": 0.31,
+      "step": 54226
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.315581542351454e-07,
+      "loss": 0.1712,
+      "step": 54227
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3151074589127685e-07,
+      "loss": 1.7836,
+      "step": 54228
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.314633375474084e-07,
+      "loss": 3.0359,
+      "step": 54229
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3141592920353983e-07,
+      "loss": 3.1529,
+      "step": 54230
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3136852085967127e-07,
+      "loss": 1.1735,
+      "step": 54231
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3132111251580276e-07,
+      "loss": 1.8188,
+      "step": 54232
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.312737041719343e-07,
+      "loss": 1.5744,
+      "step": 54233
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3122629582806575e-07,
+      "loss": 1.1582,
+      "step": 54234
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.311788874841972e-07,
+      "loss": 3.9072,
+      "step": 54235
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3113147914032873e-07,
+      "loss": 3.4612,
+      "step": 54236
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.310840707964602e-07,
+      "loss": 1.2455,
+      "step": 54237
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3103666245259166e-07,
+      "loss": 1.5844,
+      "step": 54238
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.309892541087231e-07,
+      "loss": 1.5526,
+      "step": 54239
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.3094184576485465e-07,
+      "loss": 1.9401,
+      "step": 54240
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.308944374209861e-07,
+      "loss": 2.6381,
+      "step": 54241
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.308470290771176e-07,
+      "loss": 0.3126,
+      "step": 54242
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.30799620733249e-07,
+      "loss": 0.8283,
+      "step": 54243
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.3075221238938056e-07,
+      "loss": 4.0249,
+      "step": 54244
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.30704804045512e-07,
+      "loss": 2.5024,
+      "step": 54245
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.306573957016435e-07,
+      "loss": 1.6657,
+      "step": 54246
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.3060998735777504e-07,
+      "loss": 2.9564,
+      "step": 54247
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.305625790139065e-07,
+      "loss": 2.5033,
+      "step": 54248
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.305151706700379e-07,
+      "loss": 2.2227,
+      "step": 54249
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.304677623261694e-07,
+      "loss": 2.4114,
+      "step": 54250
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.304203539823009e-07,
+      "loss": 2.6843,
+      "step": 54251
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.303729456384324e-07,
+      "loss": 1.6467,
+      "step": 54252
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.3032553729456383e-07,
+      "loss": 0.6492,
+      "step": 54253
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.3027812895069527e-07,
+      "loss": 1.6724,
+      "step": 54254
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.302307206068268e-07,
+      "loss": 0.2191,
+      "step": 54255
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.301833122629583e-07,
+      "loss": 2.2845,
+      "step": 54256
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.3013590391908975e-07,
+      "loss": 1.4383,
+      "step": 54257
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.300884955752213e-07,
+      "loss": 0.9707,
+      "step": 54258
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.3004108723135273e-07,
+      "loss": 2.185,
+      "step": 54259
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2999367888748417e-07,
+      "loss": 2.5062,
+      "step": 54260
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2994627054361566e-07,
+      "loss": 0.733,
+      "step": 54261
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.298988621997472e-07,
+      "loss": 1.0912,
+      "step": 54262
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2985145385587865e-07,
+      "loss": 1.2064,
+      "step": 54263
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.298040455120101e-07,
+      "loss": 1.2132,
+      "step": 54264
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.297566371681416e-07,
+      "loss": 0.8834,
+      "step": 54265
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.297092288242731e-07,
+      "loss": 1.2468,
+      "step": 54266
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2966182048040456e-07,
+      "loss": 1.509,
+      "step": 54267
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.29614412136536e-07,
+      "loss": 2.197,
+      "step": 54268
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2956700379266755e-07,
+      "loss": 2.399,
+      "step": 54269
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.29519595448799e-07,
+      "loss": 2.3929,
+      "step": 54270
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.294721871049305e-07,
+      "loss": 2.2731,
+      "step": 54271
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.294247787610619e-07,
+      "loss": 1.6141,
+      "step": 54272
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2937737041719346e-07,
+      "loss": 0.3777,
+      "step": 54273
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.293299620733249e-07,
+      "loss": 2.0168,
+      "step": 54274
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.292825537294564e-07,
+      "loss": 1.5286,
+      "step": 54275
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2923514538558784e-07,
+      "loss": 2.5872,
+      "step": 54276
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.291877370417194e-07,
+      "loss": 1.6832,
+      "step": 54277
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.291403286978508e-07,
+      "loss": 2.0734,
+      "step": 54278
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.290929203539823e-07,
+      "loss": 1.9932,
+      "step": 54279
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.290455120101138e-07,
+      "loss": 1.4991,
+      "step": 54280
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.289981036662453e-07,
+      "loss": 1.9646,
+      "step": 54281
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2895069532237674e-07,
+      "loss": 1.8517,
+      "step": 54282
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.289032869785082e-07,
+      "loss": 3.062,
+      "step": 54283
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.288558786346397e-07,
+      "loss": 1.7874,
+      "step": 54284
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.288084702907712e-07,
+      "loss": 1.2855,
+      "step": 54285
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2876106194690265e-07,
+      "loss": 2.4289,
+      "step": 54286
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.287136536030341e-07,
+      "loss": 1.7731,
+      "step": 54287
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2866624525916564e-07,
+      "loss": 1.3472,
+      "step": 54288
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2861883691529713e-07,
+      "loss": 3.0987,
+      "step": 54289
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2857142857142857e-07,
+      "loss": 3.2208,
+      "step": 54290
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.285240202275601e-07,
+      "loss": 1.7908,
+      "step": 54291
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2847661188369155e-07,
+      "loss": 2.9551,
+      "step": 54292
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.28429203539823e-07,
+      "loss": 2.3714,
+      "step": 54293
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.283817951959545e-07,
+      "loss": 2.1841,
+      "step": 54294
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2833438685208603e-07,
+      "loss": 2.5732,
+      "step": 54295
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2828697850821747e-07,
+      "loss": 0.9536,
+      "step": 54296
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.282395701643489e-07,
+      "loss": 2.6772,
+      "step": 54297
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.281921618204804e-07,
+      "loss": 3.9826,
+      "step": 54298
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.281447534766119e-07,
+      "loss": 2.1238,
+      "step": 54299
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.280973451327434e-07,
+      "loss": 0.8718,
+      "step": 54300
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.280499367888748e-07,
+      "loss": 1.2591,
+      "step": 54301
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2800252844500637e-07,
+      "loss": 2.2192,
+      "step": 54302
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.279551201011378e-07,
+      "loss": 0.7066,
+      "step": 54303
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.279077117572693e-07,
+      "loss": 2.9262,
+      "step": 54304
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.2786030341340074e-07,
+      "loss": 2.7334,
+      "step": 54305
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.278128950695323e-07,
+      "loss": 2.1065,
+      "step": 54306
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.277654867256637e-07,
+      "loss": 0.1447,
+      "step": 54307
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.277180783817952e-07,
+      "loss": 2.2997,
+      "step": 54308
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2767067003792665e-07,
+      "loss": 1.6543,
+      "step": 54309
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.276232616940582e-07,
+      "loss": 1.9039,
+      "step": 54310
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2757585335018964e-07,
+      "loss": 3.2919,
+      "step": 54311
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.275284450063211e-07,
+      "loss": 1.6191,
+      "step": 54312
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.274810366624526e-07,
+      "loss": 2.0869,
+      "step": 54313
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.274336283185841e-07,
+      "loss": 0.9234,
+      "step": 54314
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2738621997471555e-07,
+      "loss": 2.3315,
+      "step": 54315
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.27338811630847e-07,
+      "loss": 3.3733,
+      "step": 54316
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2729140328697854e-07,
+      "loss": 2.6108,
+      "step": 54317
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2724399494311003e-07,
+      "loss": 2.0813,
+      "step": 54318
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2719658659924147e-07,
+      "loss": 1.5555,
+      "step": 54319
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.271491782553729e-07,
+      "loss": 1.8635,
+      "step": 54320
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2710176991150445e-07,
+      "loss": 2.2614,
+      "step": 54321
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.270543615676359e-07,
+      "loss": 1.216,
+      "step": 54322
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.270069532237674e-07,
+      "loss": 1.778,
+      "step": 54323
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2695954487989893e-07,
+      "loss": 2.4125,
+      "step": 54324
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2691213653603037e-07,
+      "loss": 1.5148,
+      "step": 54325
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.268647281921618e-07,
+      "loss": 0.9615,
+      "step": 54326
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.268173198482933e-07,
+      "loss": 1.1892,
+      "step": 54327
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2676991150442485e-07,
+      "loss": 1.1383,
+      "step": 54328
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.267225031605563e-07,
+      "loss": 0.8962,
+      "step": 54329
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.266750948166877e-07,
+      "loss": 2.5532,
+      "step": 54330
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2662768647281916e-07,
+      "loss": 1.0635,
+      "step": 54331
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.265802781289507e-07,
+      "loss": 0.7753,
+      "step": 54332
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.265328697850822e-07,
+      "loss": 2.9797,
+      "step": 54333
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2648546144121364e-07,
+      "loss": 2.3234,
+      "step": 54334
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.264380530973452e-07,
+      "loss": 1.7865,
+      "step": 54335
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.263906447534766e-07,
+      "loss": 2.4442,
+      "step": 54336
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.263432364096081e-07,
+      "loss": 1.4754,
+      "step": 54337
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2629582806573956e-07,
+      "loss": 2.3569,
+      "step": 54338
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.262484197218711e-07,
+      "loss": 2.4371,
+      "step": 54339
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2620101137800254e-07,
+      "loss": 1.0252,
+      "step": 54340
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.26153603034134e-07,
+      "loss": 1.6481,
+      "step": 54341
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2610619469026547e-07,
+      "loss": 2.3422,
+      "step": 54342
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.26058786346397e-07,
+      "loss": 0.5581,
+      "step": 54343
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2601137800252846e-07,
+      "loss": 1.3965,
+      "step": 54344
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.259639696586599e-07,
+      "loss": 1.5235,
+      "step": 54345
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2591656131479144e-07,
+      "loss": 1.4729,
+      "step": 54346
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2586915297092293e-07,
+      "loss": 1.0772,
+      "step": 54347
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2582174462705437e-07,
+      "loss": 2.176,
+      "step": 54348
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.257743362831858e-07,
+      "loss": 2.7199,
+      "step": 54349
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2572692793931736e-07,
+      "loss": 1.2775,
+      "step": 54350
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.256795195954488e-07,
+      "loss": 3.4299,
+      "step": 54351
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.256321112515803e-07,
+      "loss": 1.952,
+      "step": 54352
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2558470290771173e-07,
+      "loss": 1.0602,
+      "step": 54353
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2553729456384327e-07,
+      "loss": 0.3605,
+      "step": 54354
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.254898862199747e-07,
+      "loss": 2.5017,
+      "step": 54355
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.254424778761062e-07,
+      "loss": 2.6662,
+      "step": 54356
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2539506953223775e-07,
+      "loss": 2.3342,
+      "step": 54357
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.253476611883692e-07,
+      "loss": 1.959,
+      "step": 54358
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2530025284450063e-07,
+      "loss": 2.4429,
+      "step": 54359
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2525284450063207e-07,
+      "loss": 2.1049,
+      "step": 54360
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.252054361567636e-07,
+      "loss": 1.7141,
+      "step": 54361
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.251580278128951e-07,
+      "loss": 2.274,
+      "step": 54362
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2511061946902654e-07,
+      "loss": 2.0568,
+      "step": 54363
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.25063211125158e-07,
+      "loss": 2.0198,
+      "step": 54364
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2501580278128953e-07,
+      "loss": 1.931,
+      "step": 54365
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.24968394437421e-07,
+      "loss": 1.574,
+      "step": 54366
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.2492098609355246e-07,
+      "loss": 3.2904,
+      "step": 54367
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.24873577749684e-07,
+      "loss": 2.5844,
+      "step": 54368
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2482616940581544e-07,
+      "loss": 4.6505,
+      "step": 54369
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.247787610619469e-07,
+      "loss": 1.2701,
+      "step": 54370
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.247313527180784e-07,
+      "loss": 1.6065,
+      "step": 54371
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.246839443742099e-07,
+      "loss": 2.025,
+      "step": 54372
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2463653603034136e-07,
+      "loss": 2.9104,
+      "step": 54373
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.245891276864728e-07,
+      "loss": 1.815,
+      "step": 54374
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.245417193426043e-07,
+      "loss": 0.9876,
+      "step": 54375
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2449431099873584e-07,
+      "loss": 3.6757,
+      "step": 54376
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.244469026548673e-07,
+      "loss": 2.5766,
+      "step": 54377
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.243994943109987e-07,
+      "loss": 2.9297,
+      "step": 54378
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2435208596713026e-07,
+      "loss": 0.7049,
+      "step": 54379
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.243046776232617e-07,
+      "loss": 1.7682,
+      "step": 54380
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.242572692793932e-07,
+      "loss": 2.4329,
+      "step": 54381
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2420986093552463e-07,
+      "loss": 0.2908,
+      "step": 54382
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.241624525916562e-07,
+      "loss": 2.155,
+      "step": 54383
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.241150442477876e-07,
+      "loss": 2.2315,
+      "step": 54384
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.240676359039191e-07,
+      "loss": 1.0561,
+      "step": 54385
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2402022756005055e-07,
+      "loss": 1.2428,
+      "step": 54386
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.239728192161821e-07,
+      "loss": 1.6514,
+      "step": 54387
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2392541087231353e-07,
+      "loss": 1.0717,
+      "step": 54388
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.23878002528445e-07,
+      "loss": 0.4573,
+      "step": 54389
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.238305941845765e-07,
+      "loss": 2.1131,
+      "step": 54390
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.23783185840708e-07,
+      "loss": 3.9858,
+      "step": 54391
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2373577749683945e-07,
+      "loss": 1.7983,
+      "step": 54392
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.236883691529709e-07,
+      "loss": 2.0205,
+      "step": 54393
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2364096080910243e-07,
+      "loss": 1.2783,
+      "step": 54394
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.235935524652339e-07,
+      "loss": 1.6602,
+      "step": 54395
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2354614412136536e-07,
+      "loss": 2.6215,
+      "step": 54396
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.234987357774968e-07,
+      "loss": 4.5769,
+      "step": 54397
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2345132743362835e-07,
+      "loss": 1.8233,
+      "step": 54398
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.234039190897598e-07,
+      "loss": 2.5066,
+      "step": 54399
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.233565107458913e-07,
+      "loss": 1.875,
+      "step": 54400
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.233091024020228e-07,
+      "loss": 1.5597,
+      "step": 54401
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2326169405815426e-07,
+      "loss": 3.3688,
+      "step": 54402
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.232142857142857e-07,
+      "loss": 3.3209,
+      "step": 54403
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.231668773704172e-07,
+      "loss": 1.5708,
+      "step": 54404
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2311946902654874e-07,
+      "loss": 2.9498,
+      "step": 54405
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.230720606826802e-07,
+      "loss": 2.3605,
+      "step": 54406
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.230246523388116e-07,
+      "loss": 2.7212,
+      "step": 54407
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.229772439949431e-07,
+      "loss": 0.925,
+      "step": 54408
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.229298356510746e-07,
+      "loss": 3.0684,
+      "step": 54409
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.228824273072061e-07,
+      "loss": 3.338,
+      "step": 54410
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2283501896333753e-07,
+      "loss": 1.4551,
+      "step": 54411
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.227876106194691e-07,
+      "loss": 0.2016,
+      "step": 54412
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.227402022756005e-07,
+      "loss": 1.565,
+      "step": 54413
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.22692793931732e-07,
+      "loss": 2.5199,
+      "step": 54414
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2264538558786345e-07,
+      "loss": 0.6034,
+      "step": 54415
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.22597977243995e-07,
+      "loss": 3.4977,
+      "step": 54416
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2255056890012643e-07,
+      "loss": 1.8297,
+      "step": 54417
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.225031605562579e-07,
+      "loss": 3.3919,
+      "step": 54418
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2245575221238936e-07,
+      "loss": 2.4814,
+      "step": 54419
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.224083438685209e-07,
+      "loss": 1.5839,
+      "step": 54420
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2236093552465235e-07,
+      "loss": 1.5403,
+      "step": 54421
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.223135271807838e-07,
+      "loss": 3.132,
+      "step": 54422
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2226611883691533e-07,
+      "loss": 1.208,
+      "step": 54423
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.222187104930468e-07,
+      "loss": 2.3287,
+      "step": 54424
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2217130214917826e-07,
+      "loss": 1.2457,
+      "step": 54425
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.221238938053097e-07,
+      "loss": 1.9012,
+      "step": 54426
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2207648546144125e-07,
+      "loss": 0.959,
+      "step": 54427
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2202907711757274e-07,
+      "loss": 2.6621,
+      "step": 54428
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.219816687737042e-07,
+      "loss": 2.2576,
+      "step": 54429
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.219342604298356e-07,
+      "loss": 2.4594,
+      "step": 54430
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.2188685208596716e-07,
+      "loss": 1.4376,
+      "step": 54431
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.218394437420986e-07,
+      "loss": 1.6597,
+      "step": 54432
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.217920353982301e-07,
+      "loss": 2.2667,
+      "step": 54433
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2174462705436153e-07,
+      "loss": 1.7635,
+      "step": 54434
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.216972187104931e-07,
+      "loss": 2.9462,
+      "step": 54435
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.216498103666245e-07,
+      "loss": 2.6371,
+      "step": 54436
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.21602402022756e-07,
+      "loss": 2.1441,
+      "step": 54437
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.215549936788875e-07,
+      "loss": 1.8198,
+      "step": 54438
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.21507585335019e-07,
+      "loss": 2.2783,
+      "step": 54439
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2146017699115043e-07,
+      "loss": 3.7454,
+      "step": 54440
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.214127686472819e-07,
+      "loss": 1.4264,
+      "step": 54441
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.213653603034134e-07,
+      "loss": 1.1195,
+      "step": 54442
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.213179519595449e-07,
+      "loss": 2.288,
+      "step": 54443
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2127054361567635e-07,
+      "loss": 1.0376,
+      "step": 54444
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.212231352718078e-07,
+      "loss": 1.7675,
+      "step": 54445
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2117572692793934e-07,
+      "loss": 0.5163,
+      "step": 54446
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2112831858407083e-07,
+      "loss": 2.5324,
+      "step": 54447
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2108091024020227e-07,
+      "loss": 2.7778,
+      "step": 54448
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.210335018963338e-07,
+      "loss": 2.2197,
+      "step": 54449
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2098609355246525e-07,
+      "loss": 0.8837,
+      "step": 54450
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.209386852085967e-07,
+      "loss": 2.4929,
+      "step": 54451
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.208912768647282e-07,
+      "loss": 0.8848,
+      "step": 54452
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2084386852085973e-07,
+      "loss": 1.6117,
+      "step": 54453
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2079646017699117e-07,
+      "loss": 2.9719,
+      "step": 54454
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.207490518331226e-07,
+      "loss": 2.0399,
+      "step": 54455
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.207016434892541e-07,
+      "loss": 1.7099,
+      "step": 54456
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2065423514538564e-07,
+      "loss": 2.2429,
+      "step": 54457
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.206068268015171e-07,
+      "loss": 1.5333,
+      "step": 54458
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.205594184576485e-07,
+      "loss": 0.8442,
+      "step": 54459
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2051201011378007e-07,
+      "loss": 1.4226,
+      "step": 54460
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.204646017699115e-07,
+      "loss": 1.8091,
+      "step": 54461
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.20417193426043e-07,
+      "loss": 1.2779,
+      "step": 54462
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2036978508217444e-07,
+      "loss": 1.27,
+      "step": 54463
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.20322376738306e-07,
+      "loss": 2.421,
+      "step": 54464
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.202749683944374e-07,
+      "loss": 1.5942,
+      "step": 54465
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.202275600505689e-07,
+      "loss": 1.6413,
+      "step": 54466
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2018015170670035e-07,
+      "loss": 0.6274,
+      "step": 54467
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.201327433628319e-07,
+      "loss": 1.0739,
+      "step": 54468
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.2008533501896334e-07,
+      "loss": 0.5486,
+      "step": 54469
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.200379266750948e-07,
+      "loss": 2.4837,
+      "step": 54470
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.199905183312263e-07,
+      "loss": 3.2517,
+      "step": 54471
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.199431099873578e-07,
+      "loss": 1.1387,
+      "step": 54472
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1989570164348925e-07,
+      "loss": 1.6305,
+      "step": 54473
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.198482932996207e-07,
+      "loss": 1.0678,
+      "step": 54474
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1980088495575224e-07,
+      "loss": 1.5712,
+      "step": 54475
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1975347661188373e-07,
+      "loss": 1.6236,
+      "step": 54476
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1970606826801517e-07,
+      "loss": 1.1738,
+      "step": 54477
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.196586599241466e-07,
+      "loss": 2.5263,
+      "step": 54478
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1961125158027815e-07,
+      "loss": 1.2992,
+      "step": 54479
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.195638432364096e-07,
+      "loss": 2.5618,
+      "step": 54480
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.195164348925411e-07,
+      "loss": 2.9993,
+      "step": 54481
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1946902654867263e-07,
+      "loss": 3.0343,
+      "step": 54482
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1942161820480407e-07,
+      "loss": 1.4342,
+      "step": 54483
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.193742098609355e-07,
+      "loss": 2.3345,
+      "step": 54484
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.19326801517067e-07,
+      "loss": 2.0761,
+      "step": 54485
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1927939317319855e-07,
+      "loss": 0.9549,
+      "step": 54486
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1923198482933e-07,
+      "loss": 3.6917,
+      "step": 54487
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.191845764854614e-07,
+      "loss": 1.4982,
+      "step": 54488
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.191371681415929e-07,
+      "loss": 2.483,
+      "step": 54489
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.190897597977244e-07,
+      "loss": 2.864,
+      "step": 54490
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.190423514538559e-07,
+      "loss": 0.6231,
+      "step": 54491
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.1899494310998734e-07,
+      "loss": 1.7279,
+      "step": 54492
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.189475347661189e-07,
+      "loss": 2.8071,
+      "step": 54493
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.189001264222503e-07,
+      "loss": 0.9909,
+      "step": 54494
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.188527180783818e-07,
+      "loss": 3.4613,
+      "step": 54495
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1880530973451326e-07,
+      "loss": 1.4708,
+      "step": 54496
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.187579013906448e-07,
+      "loss": 3.0278,
+      "step": 54497
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1871049304677624e-07,
+      "loss": 1.818,
+      "step": 54498
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1866308470290773e-07,
+      "loss": 2.3449,
+      "step": 54499
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1861567635903917e-07,
+      "loss": 1.9076,
+      "step": 54500
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.185682680151707e-07,
+      "loss": 2.2686,
+      "step": 54501
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1852085967130216e-07,
+      "loss": 1.8846,
+      "step": 54502
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.184734513274336e-07,
+      "loss": 3.1018,
+      "step": 54503
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1842604298356514e-07,
+      "loss": 0.4978,
+      "step": 54504
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1837863463969663e-07,
+      "loss": 2.5493,
+      "step": 54505
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1833122629582807e-07,
+      "loss": 2.755,
+      "step": 54506
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.182838179519595e-07,
+      "loss": 2.3132,
+      "step": 54507
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1823640960809106e-07,
+      "loss": 1.6002,
+      "step": 54508
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.181890012642225e-07,
+      "loss": 1.5778,
+      "step": 54509
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.18141592920354e-07,
+      "loss": 2.4414,
+      "step": 54510
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1809418457648543e-07,
+      "loss": 0.8765,
+      "step": 54511
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1804677623261697e-07,
+      "loss": 1.4837,
+      "step": 54512
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.179993678887484e-07,
+      "loss": 2.8713,
+      "step": 54513
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.179519595448799e-07,
+      "loss": 1.2968,
+      "step": 54514
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1790455120101145e-07,
+      "loss": 3.9418,
+      "step": 54515
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.178571428571429e-07,
+      "loss": 1.2689,
+      "step": 54516
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1780973451327433e-07,
+      "loss": 1.2965,
+      "step": 54517
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.177623261694058e-07,
+      "loss": 3.3872,
+      "step": 54518
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.177149178255373e-07,
+      "loss": 1.9795,
+      "step": 54519
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.176675094816688e-07,
+      "loss": 1.8824,
+      "step": 54520
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1762010113780024e-07,
+      "loss": 1.1152,
+      "step": 54521
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.175726927939317e-07,
+      "loss": 0.8234,
+      "step": 54522
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1752528445006323e-07,
+      "loss": 1.9285,
+      "step": 54523
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.174778761061947e-07,
+      "loss": 1.9559,
+      "step": 54524
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1743046776232616e-07,
+      "loss": 2.612,
+      "step": 54525
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.173830594184577e-07,
+      "loss": 0.4899,
+      "step": 54526
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1733565107458914e-07,
+      "loss": 2.0808,
+      "step": 54527
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1728824273072063e-07,
+      "loss": 3.0771,
+      "step": 54528
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.172408343868521e-07,
+      "loss": 3.8925,
+      "step": 54529
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.171934260429836e-07,
+      "loss": 1.0928,
+      "step": 54530
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1714601769911506e-07,
+      "loss": 1.3184,
+      "step": 54531
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.170986093552465e-07,
+      "loss": 1.7234,
+      "step": 54532
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.17051201011378e-07,
+      "loss": 1.8747,
+      "step": 54533
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1700379266750953e-07,
+      "loss": 0.7356,
+      "step": 54534
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.16956384323641e-07,
+      "loss": 0.6543,
+      "step": 54535
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.169089759797724e-07,
+      "loss": 1.5312,
+      "step": 54536
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1686156763590396e-07,
+      "loss": 2.5257,
+      "step": 54537
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.168141592920354e-07,
+      "loss": 0.8297,
+      "step": 54538
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.167667509481669e-07,
+      "loss": 2.9595,
+      "step": 54539
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1671934260429833e-07,
+      "loss": 1.4963,
+      "step": 54540
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.166719342604299e-07,
+      "loss": 2.169,
+      "step": 54541
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.166245259165613e-07,
+      "loss": 1.1749,
+      "step": 54542
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.165771175726928e-07,
+      "loss": 1.9372,
+      "step": 54543
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1652970922882424e-07,
+      "loss": 1.8306,
+      "step": 54544
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.164823008849558e-07,
+      "loss": 1.0125,
+      "step": 54545
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1643489254108723e-07,
+      "loss": 2.1462,
+      "step": 54546
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.163874841972187e-07,
+      "loss": 1.2099,
+      "step": 54547
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.163400758533502e-07,
+      "loss": 2.4166,
+      "step": 54548
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.162926675094817e-07,
+      "loss": 1.5026,
+      "step": 54549
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1624525916561314e-07,
+      "loss": 1.2553,
+      "step": 54550
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.161978508217446e-07,
+      "loss": 1.7839,
+      "step": 54551
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1615044247787613e-07,
+      "loss": 1.2728,
+      "step": 54552
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.161030341340076e-07,
+      "loss": 2.268,
+      "step": 54553
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1605562579013906e-07,
+      "loss": 1.0793,
+      "step": 54554
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.160082174462705e-07,
+      "loss": 1.6701,
+      "step": 54555
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1596080910240205e-07,
+      "loss": 1.8022,
+      "step": 54556
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.1591340075853354e-07,
+      "loss": 2.8061,
+      "step": 54557
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.15865992414665e-07,
+      "loss": 1.6958,
+      "step": 54558
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.158185840707965e-07,
+      "loss": 1.9715,
+      "step": 54559
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1577117572692796e-07,
+      "loss": 0.5695,
+      "step": 54560
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.157237673830594e-07,
+      "loss": 1.4491,
+      "step": 54561
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.156763590391909e-07,
+      "loss": 1.4423,
+      "step": 54562
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1562895069532244e-07,
+      "loss": 2.3173,
+      "step": 54563
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.155815423514539e-07,
+      "loss": 1.4648,
+      "step": 54564
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.155341340075853e-07,
+      "loss": 1.859,
+      "step": 54565
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.154867256637168e-07,
+      "loss": 2.5661,
+      "step": 54566
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1543931731984835e-07,
+      "loss": 0.5179,
+      "step": 54567
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.153919089759798e-07,
+      "loss": 1.9546,
+      "step": 54568
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1534450063211123e-07,
+      "loss": 1.2746,
+      "step": 54569
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.152970922882428e-07,
+      "loss": 1.9179,
+      "step": 54570
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.152496839443742e-07,
+      "loss": 0.6235,
+      "step": 54571
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.152022756005057e-07,
+      "loss": 1.3119,
+      "step": 54572
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1515486725663715e-07,
+      "loss": 1.1342,
+      "step": 54573
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.151074589127687e-07,
+      "loss": 2.6202,
+      "step": 54574
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1506005056890013e-07,
+      "loss": 2.1346,
+      "step": 54575
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.150126422250316e-07,
+      "loss": 2.8129,
+      "step": 54576
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1496523388116306e-07,
+      "loss": 1.9281,
+      "step": 54577
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.149178255372946e-07,
+      "loss": 1.2082,
+      "step": 54578
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1487041719342605e-07,
+      "loss": 2.6434,
+      "step": 54579
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.148230088495575e-07,
+      "loss": 0.9336,
+      "step": 54580
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1477560050568903e-07,
+      "loss": 1.6267,
+      "step": 54581
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.147281921618205e-07,
+      "loss": 0.7037,
+      "step": 54582
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1468078381795196e-07,
+      "loss": 1.5779,
+      "step": 54583
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.146333754740834e-07,
+      "loss": 1.2685,
+      "step": 54584
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1458596713021495e-07,
+      "loss": 2.9122,
+      "step": 54585
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1453855878634644e-07,
+      "loss": 2.8214,
+      "step": 54586
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.144911504424779e-07,
+      "loss": 3.2862,
+      "step": 54587
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.144437420986093e-07,
+      "loss": 2.809,
+      "step": 54588
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1439633375474086e-07,
+      "loss": 2.0557,
+      "step": 54589
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.143489254108723e-07,
+      "loss": 1.4916,
+      "step": 54590
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.143015170670038e-07,
+      "loss": 1.8424,
+      "step": 54591
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1425410872313534e-07,
+      "loss": 1.9689,
+      "step": 54592
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.142067003792668e-07,
+      "loss": 1.1782,
+      "step": 54593
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.141592920353982e-07,
+      "loss": 1.1401,
+      "step": 54594
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.141118836915297e-07,
+      "loss": 1.6027,
+      "step": 54595
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1406447534766126e-07,
+      "loss": 2.8677,
+      "step": 54596
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.140170670037927e-07,
+      "loss": 1.8008,
+      "step": 54597
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1396965865992413e-07,
+      "loss": 1.6947,
+      "step": 54598
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.139222503160556e-07,
+      "loss": 2.2861,
+      "step": 54599
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.138748419721871e-07,
+      "loss": 1.3593,
+      "step": 54600
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.138274336283186e-07,
+      "loss": 1.1964,
+      "step": 54601
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1378002528445005e-07,
+      "loss": 1.567,
+      "step": 54602
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.137326169405816e-07,
+      "loss": 0.4896,
+      "step": 54603
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1368520859671303e-07,
+      "loss": 1.5868,
+      "step": 54604
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.136378002528445e-07,
+      "loss": 1.3151,
+      "step": 54605
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1359039190897597e-07,
+      "loss": 2.0948,
+      "step": 54606
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.135429835651075e-07,
+      "loss": 1.0749,
+      "step": 54607
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1349557522123895e-07,
+      "loss": 1.3874,
+      "step": 54608
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.134481668773704e-07,
+      "loss": 1.8308,
+      "step": 54609
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.134007585335019e-07,
+      "loss": 2.5303,
+      "step": 54610
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1335335018963343e-07,
+      "loss": 0.4469,
+      "step": 54611
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1330594184576487e-07,
+      "loss": 2.7575,
+      "step": 54612
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.132585335018963e-07,
+      "loss": 1.1379,
+      "step": 54613
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1321112515802785e-07,
+      "loss": 2.1018,
+      "step": 54614
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1316371681415934e-07,
+      "loss": 3.0604,
+      "step": 54615
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.131163084702908e-07,
+      "loss": 2.1166,
+      "step": 54616
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.130689001264222e-07,
+      "loss": 2.2939,
+      "step": 54617
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1302149178255377e-07,
+      "loss": 2.0031,
+      "step": 54618
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.129740834386852e-07,
+      "loss": 2.7133,
+      "step": 54619
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.129266750948167e-07,
+      "loss": 3.4284,
+      "step": 54620
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.1287926675094814e-07,
+      "loss": 1.4449,
+      "step": 54621
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.128318584070797e-07,
+      "loss": 1.2518,
+      "step": 54622
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.127844500632111e-07,
+      "loss": 0.4145,
+      "step": 54623
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.127370417193426e-07,
+      "loss": 2.6401,
+      "step": 54624
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1268963337547416e-07,
+      "loss": 1.2129,
+      "step": 54625
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.126422250316056e-07,
+      "loss": 0.273,
+      "step": 54626
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1259481668773704e-07,
+      "loss": 0.7913,
+      "step": 54627
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1254740834386853e-07,
+      "loss": 2.3475,
+      "step": 54628
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.125e-07,
+      "loss": 3.029,
+      "step": 54629
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.124525916561315e-07,
+      "loss": 1.4187,
+      "step": 54630
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1240518331226295e-07,
+      "loss": 1.2168,
+      "step": 54631
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.123577749683944e-07,
+      "loss": 1.6497,
+      "step": 54632
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1231036662452594e-07,
+      "loss": 0.881,
+      "step": 54633
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1226295828065743e-07,
+      "loss": 1.4266,
+      "step": 54634
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1221554993678887e-07,
+      "loss": 1.1361,
+      "step": 54635
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.121681415929204e-07,
+      "loss": 1.7559,
+      "step": 54636
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1212073324905185e-07,
+      "loss": 1.8129,
+      "step": 54637
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1207332490518334e-07,
+      "loss": 0.8816,
+      "step": 54638
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.120259165613148e-07,
+      "loss": 2.7617,
+      "step": 54639
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1197850821744633e-07,
+      "loss": 1.4048,
+      "step": 54640
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1193109987357777e-07,
+      "loss": 1.3502,
+      "step": 54641
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.118836915297092e-07,
+      "loss": 2.0935,
+      "step": 54642
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.118362831858407e-07,
+      "loss": 1.1774,
+      "step": 54643
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1178887484197224e-07,
+      "loss": 2.7729,
+      "step": 54644
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.117414664981037e-07,
+      "loss": 1.5813,
+      "step": 54645
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.116940581542351e-07,
+      "loss": 1.3232,
+      "step": 54646
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1164664981036667e-07,
+      "loss": 2.209,
+      "step": 54647
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.115992414664981e-07,
+      "loss": 2.3397,
+      "step": 54648
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.115518331226296e-07,
+      "loss": 2.1798,
+      "step": 54649
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1150442477876104e-07,
+      "loss": 2.3689,
+      "step": 54650
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.114570164348926e-07,
+      "loss": 2.1554,
+      "step": 54651
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.11409608091024e-07,
+      "loss": 1.9203,
+      "step": 54652
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.113621997471555e-07,
+      "loss": 1.0725,
+      "step": 54653
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1131479140328695e-07,
+      "loss": 1.7531,
+      "step": 54654
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.112673830594185e-07,
+      "loss": 1.1752,
+      "step": 54655
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1121997471554994e-07,
+      "loss": 2.671,
+      "step": 54656
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1117256637168143e-07,
+      "loss": 2.6295,
+      "step": 54657
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1112515802781287e-07,
+      "loss": 1.4209,
+      "step": 54658
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.110777496839444e-07,
+      "loss": 1.8175,
+      "step": 54659
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1103034134007585e-07,
+      "loss": 1.5969,
+      "step": 54660
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.109829329962073e-07,
+      "loss": 4.6181,
+      "step": 54661
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1093552465233884e-07,
+      "loss": 0.9776,
+      "step": 54662
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1088811630847033e-07,
+      "loss": 1.19,
+      "step": 54663
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1084070796460177e-07,
+      "loss": 1.98,
+      "step": 54664
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.107932996207332e-07,
+      "loss": 2.2701,
+      "step": 54665
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1074589127686476e-07,
+      "loss": 1.3727,
+      "step": 54666
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1069848293299625e-07,
+      "loss": 2.2549,
+      "step": 54667
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.106510745891277e-07,
+      "loss": 1.2963,
+      "step": 54668
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.106036662452591e-07,
+      "loss": 1.8801,
+      "step": 54669
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1055625790139067e-07,
+      "loss": 1.2874,
+      "step": 54670
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.105088495575221e-07,
+      "loss": 2.7116,
+      "step": 54671
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.104614412136536e-07,
+      "loss": 1.3189,
+      "step": 54672
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1041403286978515e-07,
+      "loss": 1.6935,
+      "step": 54673
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.103666245259166e-07,
+      "loss": 2.0583,
+      "step": 54674
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.10319216182048e-07,
+      "loss": 1.2558,
+      "step": 54675
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.102718078381795e-07,
+      "loss": 3.8944,
+      "step": 54676
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.10224399494311e-07,
+      "loss": 1.3534,
+      "step": 54677
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.101769911504425e-07,
+      "loss": 1.791,
+      "step": 54678
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.1012958280657394e-07,
+      "loss": 0.8261,
+      "step": 54679
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.100821744627054e-07,
+      "loss": 3.3501,
+      "step": 54680
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.100347661188369e-07,
+      "loss": 2.2964,
+      "step": 54681
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.099873577749684e-07,
+      "loss": 2.3889,
+      "step": 54682
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.0993994943109986e-07,
+      "loss": 1.9683,
+      "step": 54683
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.098925410872314e-07,
+      "loss": 1.7564,
+      "step": 54684
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.0984513274336284e-07,
+      "loss": 2.0737,
+      "step": 54685
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0979772439949433e-07,
+      "loss": 2.8917,
+      "step": 54686
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0975031605562577e-07,
+      "loss": 1.6366,
+      "step": 54687
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.097029077117573e-07,
+      "loss": 1.4568,
+      "step": 54688
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0965549936788876e-07,
+      "loss": 0.388,
+      "step": 54689
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.096080910240202e-07,
+      "loss": 1.8663,
+      "step": 54690
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.095606826801517e-07,
+      "loss": 1.8779,
+      "step": 54691
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0951327433628323e-07,
+      "loss": 2.5616,
+      "step": 54692
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0946586599241467e-07,
+      "loss": 1.7971,
+      "step": 54693
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.094184576485461e-07,
+      "loss": 2.4396,
+      "step": 54694
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0937104930467766e-07,
+      "loss": 0.4706,
+      "step": 54695
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0932364096080915e-07,
+      "loss": 0.5011,
+      "step": 54696
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.092762326169406e-07,
+      "loss": 2.3025,
+      "step": 54697
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0922882427307203e-07,
+      "loss": 1.4365,
+      "step": 54698
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0918141592920357e-07,
+      "loss": 1.7914,
+      "step": 54699
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.09134007585335e-07,
+      "loss": 1.2202,
+      "step": 54700
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.090865992414665e-07,
+      "loss": 2.2874,
+      "step": 54701
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0903919089759794e-07,
+      "loss": 1.9097,
+      "step": 54702
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.089917825537295e-07,
+      "loss": 1.4839,
+      "step": 54703
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0894437420986093e-07,
+      "loss": 1.4086,
+      "step": 54704
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.088969658659924e-07,
+      "loss": 0.4862,
+      "step": 54705
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0884955752212397e-07,
+      "loss": 1.9336,
+      "step": 54706
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.088021491782554e-07,
+      "loss": 2.0081,
+      "step": 54707
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0875474083438684e-07,
+      "loss": 1.7309,
+      "step": 54708
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.087073324905183e-07,
+      "loss": 1.9754,
+      "step": 54709
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0865992414664983e-07,
+      "loss": 1.9499,
+      "step": 54710
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.086125158027813e-07,
+      "loss": 2.7191,
+      "step": 54711
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0856510745891276e-07,
+      "loss": 2.2025,
+      "step": 54712
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.085176991150442e-07,
+      "loss": 1.774,
+      "step": 54713
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0847029077117574e-07,
+      "loss": 2.0261,
+      "step": 54714
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0842288242730724e-07,
+      "loss": 2.6578,
+      "step": 54715
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.083754740834387e-07,
+      "loss": 0.4784,
+      "step": 54716
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.083280657395702e-07,
+      "loss": 2.247,
+      "step": 54717
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0828065739570166e-07,
+      "loss": 2.6862,
+      "step": 54718
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.082332490518331e-07,
+      "loss": 1.7502,
+      "step": 54719
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.081858407079646e-07,
+      "loss": 1.7144,
+      "step": 54720
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0813843236409614e-07,
+      "loss": 0.9523,
+      "step": 54721
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.080910240202276e-07,
+      "loss": 1.7065,
+      "step": 54722
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.08043615676359e-07,
+      "loss": 1.6354,
+      "step": 54723
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.079962073324905e-07,
+      "loss": 0.238,
+      "step": 54724
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0794879898862205e-07,
+      "loss": 2.5783,
+      "step": 54725
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.079013906447535e-07,
+      "loss": 1.6146,
+      "step": 54726
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0785398230088493e-07,
+      "loss": 1.2762,
+      "step": 54727
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.078065739570165e-07,
+      "loss": 0.6828,
+      "step": 54728
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.077591656131479e-07,
+      "loss": 2.164,
+      "step": 54729
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.077117572692794e-07,
+      "loss": 2.3059,
+      "step": 54730
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0766434892541085e-07,
+      "loss": 1.2989,
+      "step": 54731
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.076169405815424e-07,
+      "loss": 1.7973,
+      "step": 54732
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0756953223767383e-07,
+      "loss": 0.9321,
+      "step": 54733
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.075221238938053e-07,
+      "loss": 1.4684,
+      "step": 54734
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0747471554993676e-07,
+      "loss": 2.7838,
+      "step": 54735
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.074273072060683e-07,
+      "loss": 0.9967,
+      "step": 54736
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0737989886219975e-07,
+      "loss": 2.5207,
+      "step": 54737
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0733249051833124e-07,
+      "loss": 1.4491,
+      "step": 54738
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0728508217446273e-07,
+      "loss": 0.6113,
+      "step": 54739
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.072376738305942e-07,
+      "loss": 1.1772,
+      "step": 54740
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0719026548672566e-07,
+      "loss": 3.4447,
+      "step": 54741
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.071428571428571e-07,
+      "loss": 0.8095,
+      "step": 54742
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0709544879898865e-07,
+      "loss": 1.0733,
+      "step": 54743
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0704804045512014e-07,
+      "loss": 4.0623,
+      "step": 54744
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.070006321112516e-07,
+      "loss": 0.6269,
+      "step": 54745
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.06953223767383e-07,
+      "loss": 1.6756,
+      "step": 54746
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0690581542351456e-07,
+      "loss": 2.0524,
+      "step": 54747
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.06858407079646e-07,
+      "loss": 1.1555,
+      "step": 54748
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.068109987357775e-07,
+      "loss": 2.5022,
+      "step": 54749
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0676359039190904e-07,
+      "loss": 1.4907,
+      "step": 54750
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.067161820480405e-07,
+      "loss": 1.5031,
+      "step": 54751
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.066687737041719e-07,
+      "loss": 1.1908,
+      "step": 54752
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.066213653603034e-07,
+      "loss": 1.2119,
+      "step": 54753
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0657395701643495e-07,
+      "loss": 1.5227,
+      "step": 54754
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.065265486725664e-07,
+      "loss": 2.1455,
+      "step": 54755
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0647914032869783e-07,
+      "loss": 0.8768,
+      "step": 54756
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.064317319848293e-07,
+      "loss": 2.3955,
+      "step": 54757
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.063843236409608e-07,
+      "loss": 2.8781,
+      "step": 54758
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.063369152970923e-07,
+      "loss": 2.5631,
+      "step": 54759
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0628950695322375e-07,
+      "loss": 1.7706,
+      "step": 54760
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.062420986093553e-07,
+      "loss": 2.2988,
+      "step": 54761
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0619469026548673e-07,
+      "loss": 1.9148,
+      "step": 54762
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.061472819216182e-07,
+      "loss": 2.3882,
+      "step": 54763
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0609987357774966e-07,
+      "loss": 1.0185,
+      "step": 54764
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.060524652338812e-07,
+      "loss": 2.0097,
+      "step": 54765
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0600505689001265e-07,
+      "loss": 1.0252,
+      "step": 54766
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0595764854614414e-07,
+      "loss": 1.4953,
+      "step": 54767
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.059102402022756e-07,
+      "loss": 0.512,
+      "step": 54768
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.058628318584071e-07,
+      "loss": 3.0667,
+      "step": 54769
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0581542351453857e-07,
+      "loss": 1.0916,
+      "step": 54770
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0576801517067e-07,
+      "loss": 2.6155,
+      "step": 54771
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0572060682680155e-07,
+      "loss": 1.7269,
+      "step": 54772
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0567319848293304e-07,
+      "loss": 1.8536,
+      "step": 54773
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.056257901390645e-07,
+      "loss": 2.2862,
+      "step": 54774
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.055783817951959e-07,
+      "loss": 2.8278,
+      "step": 54775
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0553097345132747e-07,
+      "loss": 2.0125,
+      "step": 54776
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0548356510745896e-07,
+      "loss": 1.1519,
+      "step": 54777
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.054361567635904e-07,
+      "loss": 1.3864,
+      "step": 54778
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0538874841972184e-07,
+      "loss": 2.7813,
+      "step": 54779
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.053413400758534e-07,
+      "loss": 0.9803,
+      "step": 54780
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.052939317319848e-07,
+      "loss": 2.0735,
+      "step": 54781
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.052465233881163e-07,
+      "loss": 2.6303,
+      "step": 54782
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0519911504424786e-07,
+      "loss": 1.8756,
+      "step": 54783
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.051517067003793e-07,
+      "loss": 1.6112,
+      "step": 54784
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0510429835651074e-07,
+      "loss": 1.1261,
+      "step": 54785
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0505689001264223e-07,
+      "loss": 1.3892,
+      "step": 54786
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.050094816687737e-07,
+      "loss": 2.4578,
+      "step": 54787
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.049620733249052e-07,
+      "loss": 1.736,
+      "step": 54788
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0491466498103665e-07,
+      "loss": 1.8286,
+      "step": 54789
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.048672566371681e-07,
+      "loss": 1.083,
+      "step": 54790
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0481984829329964e-07,
+      "loss": 1.6714,
+      "step": 54791
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0477243994943113e-07,
+      "loss": 0.8675,
+      "step": 54792
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0472503160556257e-07,
+      "loss": 1.7785,
+      "step": 54793
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.046776232616941e-07,
+      "loss": 2.3446,
+      "step": 54794
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0463021491782555e-07,
+      "loss": 1.7541,
+      "step": 54795
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0458280657395704e-07,
+      "loss": 1.0005,
+      "step": 54796
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.045353982300885e-07,
+      "loss": 1.3739,
+      "step": 54797
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0448798988622003e-07,
+      "loss": 0.8068,
+      "step": 54798
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0444058154235147e-07,
+      "loss": 1.4724,
+      "step": 54799
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.043931731984829e-07,
+      "loss": 2.1756,
+      "step": 54800
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.043457648546144e-07,
+      "loss": 3.1409,
+      "step": 54801
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0429835651074594e-07,
+      "loss": 2.3607,
+      "step": 54802
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.042509481668774e-07,
+      "loss": 1.6777,
+      "step": 54803
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.042035398230088e-07,
+      "loss": 1.2602,
+      "step": 54804
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0415613147914037e-07,
+      "loss": 1.1149,
+      "step": 54805
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0410872313527186e-07,
+      "loss": 2.1856,
+      "step": 54806
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.040613147914033e-07,
+      "loss": 0.6364,
+      "step": 54807
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0401390644753474e-07,
+      "loss": 1.6521,
+      "step": 54808
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.039664981036663e-07,
+      "loss": 3.7899,
+      "step": 54809
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.039190897597977e-07,
+      "loss": 1.9939,
+      "step": 54810
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.038716814159292e-07,
+      "loss": 0.6162,
+      "step": 54811
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.0382427307206065e-07,
+      "loss": 0.6111,
+      "step": 54812
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.037768647281922e-07,
+      "loss": 2.0718,
+      "step": 54813
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0372945638432364e-07,
+      "loss": 2.4548,
+      "step": 54814
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0368204804045513e-07,
+      "loss": 1.5441,
+      "step": 54815
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.036346396965866e-07,
+      "loss": 0.9497,
+      "step": 54816
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.035872313527181e-07,
+      "loss": 3.2189,
+      "step": 54817
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0353982300884955e-07,
+      "loss": 1.3301,
+      "step": 54818
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.03492414664981e-07,
+      "loss": 0.6994,
+      "step": 54819
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0344500632111254e-07,
+      "loss": 2.5016,
+      "step": 54820
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0339759797724403e-07,
+      "loss": 3.2705,
+      "step": 54821
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0335018963337547e-07,
+      "loss": 2.7442,
+      "step": 54822
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.033027812895069e-07,
+      "loss": 1.9271,
+      "step": 54823
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0325537294563845e-07,
+      "loss": 1.3877,
+      "step": 54824
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0320796460176995e-07,
+      "loss": 2.0345,
+      "step": 54825
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.031605562579014e-07,
+      "loss": 2.833,
+      "step": 54826
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0311314791403293e-07,
+      "loss": 1.7357,
+      "step": 54827
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0306573957016437e-07,
+      "loss": 1.2448,
+      "step": 54828
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.030183312262958e-07,
+      "loss": 2.2522,
+      "step": 54829
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.029709228824273e-07,
+      "loss": 1.2328,
+      "step": 54830
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0292351453855885e-07,
+      "loss": 2.7243,
+      "step": 54831
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.028761061946903e-07,
+      "loss": 0.3757,
+      "step": 54832
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.028286978508217e-07,
+      "loss": 2.3314,
+      "step": 54833
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.027812895069532e-07,
+      "loss": 2.3023,
+      "step": 54834
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0273388116308476e-07,
+      "loss": 1.4923,
+      "step": 54835
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.026864728192162e-07,
+      "loss": 1.8054,
+      "step": 54836
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0263906447534764e-07,
+      "loss": 0.7765,
+      "step": 54837
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.025916561314792e-07,
+      "loss": 0.9805,
+      "step": 54838
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.025442477876106e-07,
+      "loss": 2.9038,
+      "step": 54839
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.024968394437421e-07,
+      "loss": 0.8505,
+      "step": 54840
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0244943109987356e-07,
+      "loss": 2.8423,
+      "step": 54841
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.024020227560051e-07,
+      "loss": 0.4542,
+      "step": 54842
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0235461441213654e-07,
+      "loss": 2.5739,
+      "step": 54843
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0230720606826803e-07,
+      "loss": 1.5371,
+      "step": 54844
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0225979772439947e-07,
+      "loss": 1.2972,
+      "step": 54845
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.02212389380531e-07,
+      "loss": 2.3134,
+      "step": 54846
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0216498103666246e-07,
+      "loss": 0.444,
+      "step": 54847
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.021175726927939e-07,
+      "loss": 2.2179,
+      "step": 54848
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0207016434892544e-07,
+      "loss": 1.6574,
+      "step": 54849
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0202275600505693e-07,
+      "loss": 1.3268,
+      "step": 54850
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0197534766118837e-07,
+      "loss": 1.3477,
+      "step": 54851
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.019279393173198e-07,
+      "loss": 1.7723,
+      "step": 54852
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0188053097345136e-07,
+      "loss": 0.9383,
+      "step": 54853
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0183312262958285e-07,
+      "loss": 1.6806,
+      "step": 54854
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.017857142857143e-07,
+      "loss": 2.3274,
+      "step": 54855
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0173830594184573e-07,
+      "loss": 1.5826,
+      "step": 54856
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0169089759797727e-07,
+      "loss": 0.5919,
+      "step": 54857
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.016434892541087e-07,
+      "loss": 0.7441,
+      "step": 54858
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.015960809102402e-07,
+      "loss": 1.5395,
+      "step": 54859
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0154867256637175e-07,
+      "loss": 1.4846,
+      "step": 54860
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.015012642225032e-07,
+      "loss": 0.6713,
+      "step": 54861
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0145385587863463e-07,
+      "loss": 1.1637,
+      "step": 54862
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.014064475347661e-07,
+      "loss": 2.4421,
+      "step": 54863
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0135903919089766e-07,
+      "loss": 1.9555,
+      "step": 54864
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.013116308470291e-07,
+      "loss": 1.1244,
+      "step": 54865
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0126422250316054e-07,
+      "loss": 2.272,
+      "step": 54866
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0121681415929204e-07,
+      "loss": 3.4383,
+      "step": 54867
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0116940581542353e-07,
+      "loss": 1.9748,
+      "step": 54868
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.01121997471555e-07,
+      "loss": 2.56,
+      "step": 54869
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0107458912768646e-07,
+      "loss": 2.6212,
+      "step": 54870
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.010271807838179e-07,
+      "loss": 0.6812,
+      "step": 54871
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0097977243994944e-07,
+      "loss": 1.8339,
+      "step": 54872
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0093236409608094e-07,
+      "loss": 1.241,
+      "step": 54873
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.008849557522124e-07,
+      "loss": 1.0348,
+      "step": 54874
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.008375474083439e-07,
+      "loss": 2.5647,
+      "step": 54875
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0079013906447536e-07,
+      "loss": 1.1522,
+      "step": 54876
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0074273072060685e-07,
+      "loss": 0.3114,
+      "step": 54877
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.006953223767383e-07,
+      "loss": 2.205,
+      "step": 54878
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0064791403286984e-07,
+      "loss": 1.088,
+      "step": 54879
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.006005056890013e-07,
+      "loss": 1.4008,
+      "step": 54880
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.005530973451327e-07,
+      "loss": 2.4421,
+      "step": 54881
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.005056890012642e-07,
+      "loss": 1.1981,
+      "step": 54882
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0045828065739575e-07,
+      "loss": 2.0973,
+      "step": 54883
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.004108723135272e-07,
+      "loss": 1.8653,
+      "step": 54884
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0036346396965863e-07,
+      "loss": 1.8105,
+      "step": 54885
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.003160556257902e-07,
+      "loss": 1.6485,
+      "step": 54886
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.002686472819216e-07,
+      "loss": 2.8953,
+      "step": 54887
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.002212389380531e-07,
+      "loss": 1.048,
+      "step": 54888
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0017383059418455e-07,
+      "loss": 2.0132,
+      "step": 54889
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.001264222503161e-07,
+      "loss": 2.0392,
+      "step": 54890
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.0007901390644753e-07,
+      "loss": 2.2597,
+      "step": 54891
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.00031605562579e-07,
+      "loss": 2.7579,
+      "step": 54892
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9998419721871046e-07,
+      "loss": 0.6055,
+      "step": 54893
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.99936788874842e-07,
+      "loss": 3.5995,
+      "step": 54894
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9988938053097345e-07,
+      "loss": 1.7199,
+      "step": 54895
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9984197218710494e-07,
+      "loss": 3.224,
+      "step": 54896
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9979456384323643e-07,
+      "loss": 2.4582,
+      "step": 54897
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.997471554993679e-07,
+      "loss": 2.6884,
+      "step": 54898
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9969974715549936e-07,
+      "loss": 1.592,
+      "step": 54899
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.996523388116308e-07,
+      "loss": 1.9452,
+      "step": 54900
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9960493046776235e-07,
+      "loss": 1.6661,
+      "step": 54901
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9955752212389384e-07,
+      "loss": 2.4549,
+      "step": 54902
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.995101137800253e-07,
+      "loss": 0.6343,
+      "step": 54903
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.994627054361567e-07,
+      "loss": 2.9063,
+      "step": 54904
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9941529709228826e-07,
+      "loss": 2.1272,
+      "step": 54905
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9936788874841975e-07,
+      "loss": 1.2014,
+      "step": 54906
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.993204804045512e-07,
+      "loss": 1.8913,
+      "step": 54907
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9927307206068274e-07,
+      "loss": 2.9405,
+      "step": 54908
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.992256637168142e-07,
+      "loss": 2.4207,
+      "step": 54909
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.991782553729456e-07,
+      "loss": 1.0737,
+      "step": 54910
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.991308470290771e-07,
+      "loss": 1.7058,
+      "step": 54911
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9908343868520865e-07,
+      "loss": 0.9491,
+      "step": 54912
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.990360303413401e-07,
+      "loss": 1.9988,
+      "step": 54913
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9898862199747153e-07,
+      "loss": 0.6356,
+      "step": 54914
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.98941213653603e-07,
+      "loss": 4.2494,
+      "step": 54915
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9889380530973457e-07,
+      "loss": 2.4546,
+      "step": 54916
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.98846396965866e-07,
+      "loss": 2.457,
+      "step": 54917
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9879898862199745e-07,
+      "loss": 0.6632,
+      "step": 54918
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.98751580278129e-07,
+      "loss": 1.6983,
+      "step": 54919
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9870417193426043e-07,
+      "loss": 0.8306,
+      "step": 54920
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.986567635903919e-07,
+      "loss": 2.6849,
+      "step": 54921
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9860935524652336e-07,
+      "loss": 2.8522,
+      "step": 54922
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.985619469026549e-07,
+      "loss": 1.9356,
+      "step": 54923
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9851453855878635e-07,
+      "loss": 1.5616,
+      "step": 54924
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9846713021491784e-07,
+      "loss": 2.5241,
+      "step": 54925
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.984197218710493e-07,
+      "loss": 1.0029,
+      "step": 54926
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.983723135271808e-07,
+      "loss": 1.8953,
+      "step": 54927
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9832490518331226e-07,
+      "loss": 2.6937,
+      "step": 54928
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.982774968394437e-07,
+      "loss": 1.8951,
+      "step": 54929
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9823008849557525e-07,
+      "loss": 1.2825,
+      "step": 54930
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9818268015170674e-07,
+      "loss": 1.82,
+      "step": 54931
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.981352718078382e-07,
+      "loss": 1.1565,
+      "step": 54932
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.980878634639696e-07,
+      "loss": 0.8604,
+      "step": 54933
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9804045512010116e-07,
+      "loss": 1.7827,
+      "step": 54934
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9799304677623266e-07,
+      "loss": 0.9532,
+      "step": 54935
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.979456384323641e-07,
+      "loss": 1.0067,
+      "step": 54936
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9789823008849554e-07,
+      "loss": 1.1603,
+      "step": 54937
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.978508217446271e-07,
+      "loss": 2.4554,
+      "step": 54938
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.978034134007585e-07,
+      "loss": 1.7021,
+      "step": 54939
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9775600505689e-07,
+      "loss": 1.4996,
+      "step": 54940
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9770859671302156e-07,
+      "loss": 2.1014,
+      "step": 54941
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.97661188369153e-07,
+      "loss": 2.1545,
+      "step": 54942
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9761378002528444e-07,
+      "loss": 3.0955,
+      "step": 54943
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9756637168141593e-07,
+      "loss": 1.1234,
+      "step": 54944
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9751896333754747e-07,
+      "loss": 2.0608,
+      "step": 54945
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.974715549936789e-07,
+      "loss": 2.6508,
+      "step": 54946
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9742414664981035e-07,
+      "loss": 1.2398,
+      "step": 54947
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9737673830594184e-07,
+      "loss": 1.1292,
+      "step": 54948
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9732932996207334e-07,
+      "loss": 1.2801,
+      "step": 54949
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9728192161820483e-07,
+      "loss": 2.7144,
+      "step": 54950
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9723451327433627e-07,
+      "loss": 2.511,
+      "step": 54951
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.971871049304678e-07,
+      "loss": 1.0042,
+      "step": 54952
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9713969658659925e-07,
+      "loss": 3.9992,
+      "step": 54953
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9709228824273074e-07,
+      "loss": 0.7616,
+      "step": 54954
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.970448798988622e-07,
+      "loss": 1.8046,
+      "step": 54955
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9699747155499373e-07,
+      "loss": 0.8581,
+      "step": 54956
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9695006321112517e-07,
+      "loss": 1.9263,
+      "step": 54957
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.969026548672566e-07,
+      "loss": 2.268,
+      "step": 54958
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.968552465233881e-07,
+      "loss": 2.0416,
+      "step": 54959
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9680783817951964e-07,
+      "loss": 2.4775,
+      "step": 54960
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.967604298356511e-07,
+      "loss": 1.8508,
+      "step": 54961
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.967130214917825e-07,
+      "loss": 1.8901,
+      "step": 54962
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9666561314791407e-07,
+      "loss": 2.2071,
+      "step": 54963
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9661820480404556e-07,
+      "loss": 1.6088,
+      "step": 54964
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.96570796460177e-07,
+      "loss": 3.4824,
+      "step": 54965
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9652338811630844e-07,
+      "loss": 3.0466,
+      "step": 54966
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9647597977244e-07,
+      "loss": 2.2846,
+      "step": 54967
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.964285714285714e-07,
+      "loss": 3.9986,
+      "step": 54968
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.963811630847029e-07,
+      "loss": 1.8535,
+      "step": 54969
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9633375474083435e-07,
+      "loss": 2.6659,
+      "step": 54970
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.962863463969659e-07,
+      "loss": 3.4218,
+      "step": 54971
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9623893805309734e-07,
+      "loss": 1.9157,
+      "step": 54972
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9619152970922883e-07,
+      "loss": 1.4081,
+      "step": 54973
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.961441213653604e-07,
+      "loss": 2.8387,
+      "step": 54974
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.960967130214918e-07,
+      "loss": 1.1578,
+      "step": 54975
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9604930467762325e-07,
+      "loss": 1.4792,
+      "step": 54976
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9600189633375475e-07,
+      "loss": 1.1774,
+      "step": 54977
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9595448798988624e-07,
+      "loss": 2.3063,
+      "step": 54978
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9590707964601773e-07,
+      "loss": 1.0664,
+      "step": 54979
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9585967130214917e-07,
+      "loss": 2.3477,
+      "step": 54980
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.958122629582806e-07,
+      "loss": 1.8356,
+      "step": 54981
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9576485461441215e-07,
+      "loss": 2.3711,
+      "step": 54982
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9571744627054365e-07,
+      "loss": 1.0595,
+      "step": 54983
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.956700379266751e-07,
+      "loss": 2.2824,
+      "step": 54984
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9562262958280663e-07,
+      "loss": 1.5194,
+      "step": 54985
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9557522123893807e-07,
+      "loss": 0.6771,
+      "step": 54986
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.955278128950695e-07,
+      "loss": 0.5161,
+      "step": 54987
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.95480404551201e-07,
+      "loss": 1.8172,
+      "step": 54988
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9543299620733255e-07,
+      "loss": 1.3512,
+      "step": 54989
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.95385587863464e-07,
+      "loss": 0.5718,
+      "step": 54990
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.953381795195954e-07,
+      "loss": 1.204,
+      "step": 54991
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.952907711757269e-07,
+      "loss": 2.8398,
+      "step": 54992
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9524336283185846e-07,
+      "loss": 2.4411,
+      "step": 54993
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.951959544879899e-07,
+      "loss": 1.3458,
+      "step": 54994
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9514854614412134e-07,
+      "loss": 1.3465,
+      "step": 54995
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.951011378002529e-07,
+      "loss": 1.9895,
+      "step": 54996
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.950537294563843e-07,
+      "loss": 2.2836,
+      "step": 54997
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.950063211125158e-07,
+      "loss": 2.426,
+      "step": 54998
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9495891276864726e-07,
+      "loss": 1.7225,
+      "step": 54999
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.949115044247788e-07,
+      "loss": 2.7987,
+      "step": 55000
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9486409608091024e-07,
+      "loss": 3.361,
+      "step": 55001
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.9481668773704173e-07,
+      "loss": 1.9038,
+      "step": 55002
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9476927939317317e-07,
+      "loss": 2.1713,
+      "step": 55003
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.947218710493047e-07,
+      "loss": 0.9282,
+      "step": 55004
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9467446270543616e-07,
+      "loss": 2.5021,
+      "step": 55005
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9462705436156765e-07,
+      "loss": 1.3965,
+      "step": 55006
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9457964601769914e-07,
+      "loss": 0.5302,
+      "step": 55007
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9453223767383063e-07,
+      "loss": 1.7371,
+      "step": 55008
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9448482932996207e-07,
+      "loss": 2.022,
+      "step": 55009
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.944374209860935e-07,
+      "loss": 2.5531,
+      "step": 55010
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9439001264222506e-07,
+      "loss": 3.5571,
+      "step": 55011
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9434260429835655e-07,
+      "loss": 2.2324,
+      "step": 55012
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.94295195954488e-07,
+      "loss": 0.6356,
+      "step": 55013
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9424778761061943e-07,
+      "loss": 1.712,
+      "step": 55014
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9420037926675097e-07,
+      "loss": 0.9901,
+      "step": 55015
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9415297092288246e-07,
+      "loss": 2.6219,
+      "step": 55016
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.941055625790139e-07,
+      "loss": 2.1654,
+      "step": 55017
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9405815423514545e-07,
+      "loss": 1.1443,
+      "step": 55018
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.940107458912769e-07,
+      "loss": 0.7693,
+      "step": 55019
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9396333754740833e-07,
+      "loss": 2.9524,
+      "step": 55020
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.939159292035398e-07,
+      "loss": 1.1347,
+      "step": 55021
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9386852085967136e-07,
+      "loss": 1.4939,
+      "step": 55022
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.938211125158028e-07,
+      "loss": 1.3122,
+      "step": 55023
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9377370417193424e-07,
+      "loss": 1.7177,
+      "step": 55024
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9372629582806573e-07,
+      "loss": 1.3035,
+      "step": 55025
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9367888748419723e-07,
+      "loss": 2.2382,
+      "step": 55026
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.936314791403287e-07,
+      "loss": 3.5043,
+      "step": 55027
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9358407079646016e-07,
+      "loss": 2.1621,
+      "step": 55028
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.935366624525917e-07,
+      "loss": 2.1997,
+      "step": 55029
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9348925410872314e-07,
+      "loss": 1.3842,
+      "step": 55030
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9344184576485464e-07,
+      "loss": 3.2854,
+      "step": 55031
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.933944374209861e-07,
+      "loss": 3.9626,
+      "step": 55032
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.933470290771176e-07,
+      "loss": 0.157,
+      "step": 55033
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9329962073324906e-07,
+      "loss": 1.4829,
+      "step": 55034
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9325221238938055e-07,
+      "loss": 1.567,
+      "step": 55035
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.93204804045512e-07,
+      "loss": 0.8217,
+      "step": 55036
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9315739570164354e-07,
+      "loss": 2.4131,
+      "step": 55037
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.93109987357775e-07,
+      "loss": 0.6981,
+      "step": 55038
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.930625790139064e-07,
+      "loss": 1.1152,
+      "step": 55039
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9301517067003796e-07,
+      "loss": 2.4507,
+      "step": 55040
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9296776232616945e-07,
+      "loss": 1.6321,
+      "step": 55041
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.929203539823009e-07,
+      "loss": 0.9981,
+      "step": 55042
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9287294563843233e-07,
+      "loss": 1.1386,
+      "step": 55043
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.928255372945639e-07,
+      "loss": 2.1738,
+      "step": 55044
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9277812895069537e-07,
+      "loss": 2.2498,
+      "step": 55045
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.927307206068268e-07,
+      "loss": 0.8344,
+      "step": 55046
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9268331226295825e-07,
+      "loss": 1.9801,
+      "step": 55047
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.926359039190898e-07,
+      "loss": 0.7413,
+      "step": 55048
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9258849557522123e-07,
+      "loss": 2.2471,
+      "step": 55049
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.925410872313527e-07,
+      "loss": 2.9987,
+      "step": 55050
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9249367888748427e-07,
+      "loss": 2.3987,
+      "step": 55051
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.924462705436157e-07,
+      "loss": 0.8572,
+      "step": 55052
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9239886219974715e-07,
+      "loss": 2.0649,
+      "step": 55053
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9235145385587864e-07,
+      "loss": 1.1097,
+      "step": 55054
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.923040455120102e-07,
+      "loss": 1.5711,
+      "step": 55055
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.922566371681416e-07,
+      "loss": 2.5543,
+      "step": 55056
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9220922882427306e-07,
+      "loss": 1.2812,
+      "step": 55057
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.921618204804045e-07,
+      "loss": 1.1747,
+      "step": 55058
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9211441213653605e-07,
+      "loss": 2.9523,
+      "step": 55059
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9206700379266754e-07,
+      "loss": 3.5786,
+      "step": 55060
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.92019595448799e-07,
+      "loss": 0.9349,
+      "step": 55061
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.919721871049305e-07,
+      "loss": 1.9194,
+      "step": 55062
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9192477876106196e-07,
+      "loss": 1.2213,
+      "step": 55063
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.9187737041719345e-07,
+      "loss": 0.949,
+      "step": 55064
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.918299620733249e-07,
+      "loss": 1.8569,
+      "step": 55065
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9178255372945644e-07,
+      "loss": 1.7516,
+      "step": 55066
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.917351453855879e-07,
+      "loss": 3.818,
+      "step": 55067
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.916877370417193e-07,
+      "loss": 1.5812,
+      "step": 55068
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.916403286978508e-07,
+      "loss": 1.3681,
+      "step": 55069
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9159292035398235e-07,
+      "loss": 1.6262,
+      "step": 55070
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.915455120101138e-07,
+      "loss": 1.7811,
+      "step": 55071
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9149810366624523e-07,
+      "loss": 3.7988,
+      "step": 55072
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.914506953223768e-07,
+      "loss": 2.2072,
+      "step": 55073
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9140328697850827e-07,
+      "loss": 3.3408,
+      "step": 55074
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.913558786346397e-07,
+      "loss": 1.1754,
+      "step": 55075
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9130847029077115e-07,
+      "loss": 2.1637,
+      "step": 55076
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.912610619469027e-07,
+      "loss": 1.0918,
+      "step": 55077
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9121365360303413e-07,
+      "loss": 0.8798,
+      "step": 55078
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.911662452591656e-07,
+      "loss": 1.2147,
+      "step": 55079
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9111883691529706e-07,
+      "loss": 1.972,
+      "step": 55080
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.910714285714286e-07,
+      "loss": 3.1771,
+      "step": 55081
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9102402022756005e-07,
+      "loss": 2.7806,
+      "step": 55082
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9097661188369154e-07,
+      "loss": 1.7494,
+      "step": 55083
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.909292035398231e-07,
+      "loss": 3.4009,
+      "step": 55084
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.908817951959545e-07,
+      "loss": 1.4673,
+      "step": 55085
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9083438685208596e-07,
+      "loss": 0.5521,
+      "step": 55086
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9078697850821746e-07,
+      "loss": 1.6539,
+      "step": 55087
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9073957016434895e-07,
+      "loss": 2.2062,
+      "step": 55088
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9069216182048044e-07,
+      "loss": 0.6576,
+      "step": 55089
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.906447534766119e-07,
+      "loss": 1.5028,
+      "step": 55090
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.905973451327433e-07,
+      "loss": 0.6213,
+      "step": 55091
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9054993678887486e-07,
+      "loss": 2.612,
+      "step": 55092
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9050252844500636e-07,
+      "loss": 1.5932,
+      "step": 55093
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.904551201011378e-07,
+      "loss": 2.0318,
+      "step": 55094
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9040771175726923e-07,
+      "loss": 1.7519,
+      "step": 55095
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.903603034134008e-07,
+      "loss": 3.2099,
+      "step": 55096
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.903128950695322e-07,
+      "loss": 2.0311,
+      "step": 55097
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.902654867256637e-07,
+      "loss": 1.8856,
+      "step": 55098
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9021807838179526e-07,
+      "loss": 2.027,
+      "step": 55099
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.901706700379267e-07,
+      "loss": 2.2192,
+      "step": 55100
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9012326169405813e-07,
+      "loss": 1.6194,
+      "step": 55101
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9007585335018963e-07,
+      "loss": 2.5559,
+      "step": 55102
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.9002844500632117e-07,
+      "loss": 0.4711,
+      "step": 55103
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.899810366624526e-07,
+      "loss": 2.5042,
+      "step": 55104
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8993362831858405e-07,
+      "loss": 1.589,
+      "step": 55105
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8988621997471554e-07,
+      "loss": 2.6735,
+      "step": 55106
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8983881163084703e-07,
+      "loss": 2.2021,
+      "step": 55107
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8979140328697853e-07,
+      "loss": 2.1558,
+      "step": 55108
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8974399494310997e-07,
+      "loss": 3.5741,
+      "step": 55109
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.896965865992415e-07,
+      "loss": 2.997,
+      "step": 55110
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8964917825537295e-07,
+      "loss": 2.1375,
+      "step": 55111
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8960176991150444e-07,
+      "loss": 2.1629,
+      "step": 55112
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.895543615676359e-07,
+      "loss": 3.0809,
+      "step": 55113
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8950695322376743e-07,
+      "loss": 2.6066,
+      "step": 55114
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8945954487989887e-07,
+      "loss": 1.8632,
+      "step": 55115
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8941213653603036e-07,
+      "loss": 2.3002,
+      "step": 55116
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.893647281921618e-07,
+      "loss": 1.2361,
+      "step": 55117
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8931731984829334e-07,
+      "loss": 3.5443,
+      "step": 55118
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.892699115044248e-07,
+      "loss": 1.6491,
+      "step": 55119
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.892225031605562e-07,
+      "loss": 1.4363,
+      "step": 55120
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8917509481668777e-07,
+      "loss": 1.0321,
+      "step": 55121
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8912768647281926e-07,
+      "loss": 1.909,
+      "step": 55122
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.890802781289507e-07,
+      "loss": 2.1479,
+      "step": 55123
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8903286978508214e-07,
+      "loss": 1.8125,
+      "step": 55124
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.889854614412137e-07,
+      "loss": 2.2314,
+      "step": 55125
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.889380530973451e-07,
+      "loss": 1.5351,
+      "step": 55126
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.888906447534766e-07,
+      "loss": 2.9616,
+      "step": 55127
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 3.8884323640960805e-07,
+      "loss": 1.782,
+      "step": 55128
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.887958280657396e-07,
+      "loss": 1.1448,
+      "step": 55129
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8874841972187104e-07,
+      "loss": 3.2181,
+      "step": 55130
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8870101137800253e-07,
+      "loss": 2.2798,
+      "step": 55131
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.886536030341341e-07,
+      "loss": 1.6703,
+      "step": 55132
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.886061946902655e-07,
+      "loss": 2.5276,
+      "step": 55133
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8855878634639695e-07,
+      "loss": 0.9153,
+      "step": 55134
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8851137800252844e-07,
+      "loss": 2.061,
+      "step": 55135
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8846396965865994e-07,
+      "loss": 3.2036,
+      "step": 55136
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8841656131479143e-07,
+      "loss": 1.6756,
+      "step": 55137
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8836915297092287e-07,
+      "loss": 2.2131,
+      "step": 55138
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.883217446270543e-07,
+      "loss": 1.8534,
+      "step": 55139
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8827433628318585e-07,
+      "loss": 1.9148,
+      "step": 55140
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8822692793931735e-07,
+      "loss": 2.3199,
+      "step": 55141
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.881795195954488e-07,
+      "loss": 1.4855,
+      "step": 55142
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8813211125158033e-07,
+      "loss": 1.5535,
+      "step": 55143
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8808470290771177e-07,
+      "loss": 1.9656,
+      "step": 55144
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8803729456384326e-07,
+      "loss": 3.0151,
+      "step": 55145
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.879898862199747e-07,
+      "loss": 1.4525,
+      "step": 55146
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8794247787610625e-07,
+      "loss": 4.3234,
+      "step": 55147
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.878950695322377e-07,
+      "loss": 1.6015,
+      "step": 55148
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.878476611883691e-07,
+      "loss": 1.4159,
+      "step": 55149
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.878002528445006e-07,
+      "loss": 1.1218,
+      "step": 55150
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8775284450063216e-07,
+      "loss": 2.1992,
+      "step": 55151
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.877054361567636e-07,
+      "loss": 2.6057,
+      "step": 55152
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8765802781289504e-07,
+      "loss": 0.333,
+      "step": 55153
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.876106194690266e-07,
+      "loss": 2.2386,
+      "step": 55154
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.875632111251581e-07,
+      "loss": 0.9196,
+      "step": 55155
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.875158027812895e-07,
+      "loss": 3.0376,
+      "step": 55156
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8746839443742096e-07,
+      "loss": 2.1596,
+      "step": 55157
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.874209860935525e-07,
+      "loss": 1.7697,
+      "step": 55158
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8737357774968394e-07,
+      "loss": 1.626,
+      "step": 55159
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8732616940581543e-07,
+      "loss": 0.6416,
+      "step": 55160
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8727876106194687e-07,
+      "loss": 3.2965,
+      "step": 55161
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.872313527180784e-07,
+      "loss": 0.7513,
+      "step": 55162
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8718394437420986e-07,
+      "loss": 2.0744,
+      "step": 55163
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8713653603034135e-07,
+      "loss": 1.8394,
+      "step": 55164
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8708912768647284e-07,
+      "loss": 1.5057,
+      "step": 55165
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8704171934260433e-07,
+      "loss": 1.6923,
+      "step": 55166
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8699431099873577e-07,
+      "loss": 1.7118,
+      "step": 55167
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.869469026548672e-07,
+      "loss": 2.3197,
+      "step": 55168
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8689949431099876e-07,
+      "loss": 1.8684,
+      "step": 55169
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8685208596713025e-07,
+      "loss": 1.8763,
+      "step": 55170
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.868046776232617e-07,
+      "loss": 0.5325,
+      "step": 55171
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.867572692793931e-07,
+      "loss": 1.5542,
+      "step": 55172
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8670986093552467e-07,
+      "loss": 1.5694,
+      "step": 55173
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8666245259165616e-07,
+      "loss": 1.1213,
+      "step": 55174
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.866150442477876e-07,
+      "loss": 2.3647,
+      "step": 55175
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8656763590391915e-07,
+      "loss": 1.555,
+      "step": 55176
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.865202275600506e-07,
+      "loss": 2.3028,
+      "step": 55177
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.86472819216182e-07,
+      "loss": 2.0128,
+      "step": 55178
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.864254108723135e-07,
+      "loss": 2.5215,
+      "step": 55179
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8637800252844506e-07,
+      "loss": 1.7738,
+      "step": 55180
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.863305941845765e-07,
+      "loss": 2.109,
+      "step": 55181
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8628318584070794e-07,
+      "loss": 2.4904,
+      "step": 55182
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8623577749683943e-07,
+      "loss": 1.8203,
+      "step": 55183
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.86188369152971e-07,
+      "loss": 1.8818,
+      "step": 55184
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.861409608091024e-07,
+      "loss": 2.2405,
+      "step": 55185
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8609355246523386e-07,
+      "loss": 2.7385,
+      "step": 55186
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.860461441213654e-07,
+      "loss": 3.0482,
+      "step": 55187
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8599873577749684e-07,
+      "loss": 3.098,
+      "step": 55188
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8595132743362833e-07,
+      "loss": 1.9504,
+      "step": 55189
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8590391908975977e-07,
+      "loss": 0.4098,
+      "step": 55190
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.858565107458913e-07,
+      "loss": 3.0592,
+      "step": 55191
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.8580910240202276e-07,
+      "loss": 2.266,
+      "step": 55192
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8576169405815425e-07,
+      "loss": 2.1607,
+      "step": 55193
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.857142857142857e-07,
+      "loss": 1.5858,
+      "step": 55194
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8566687737041723e-07,
+      "loss": 2.3634,
+      "step": 55195
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.856194690265487e-07,
+      "loss": 1.54,
+      "step": 55196
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.855720606826801e-07,
+      "loss": 2.0488,
+      "step": 55197
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8552465233881166e-07,
+      "loss": 1.3347,
+      "step": 55198
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8547724399494315e-07,
+      "loss": 1.5988,
+      "step": 55199
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.854298356510746e-07,
+      "loss": 1.0544,
+      "step": 55200
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8538242730720603e-07,
+      "loss": 2.793,
+      "step": 55201
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.853350189633376e-07,
+      "loss": 1.7517,
+      "step": 55202
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8528761061946907e-07,
+      "loss": 1.6738,
+      "step": 55203
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.852402022756005e-07,
+      "loss": 1.0234,
+      "step": 55204
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8519279393173194e-07,
+      "loss": 2.8005,
+      "step": 55205
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.851453855878635e-07,
+      "loss": 2.4831,
+      "step": 55206
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8509797724399493e-07,
+      "loss": 1.852,
+      "step": 55207
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.850505689001264e-07,
+      "loss": 2.1063,
+      "step": 55208
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8500316055625797e-07,
+      "loss": 1.2925,
+      "step": 55209
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.849557522123894e-07,
+      "loss": 1.4705,
+      "step": 55210
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8490834386852084e-07,
+      "loss": 2.1723,
+      "step": 55211
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8486093552465234e-07,
+      "loss": 1.5635,
+      "step": 55212
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.848135271807839e-07,
+      "loss": 2.0058,
+      "step": 55213
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.847661188369153e-07,
+      "loss": 0.8293,
+      "step": 55214
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8471871049304676e-07,
+      "loss": 2.1265,
+      "step": 55215
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8467130214917825e-07,
+      "loss": 1.9091,
+      "step": 55216
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8462389380530974e-07,
+      "loss": 1.1095,
+      "step": 55217
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8457648546144124e-07,
+      "loss": 1.8913,
+      "step": 55218
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.845290771175727e-07,
+      "loss": 1.7442,
+      "step": 55219
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.844816687737042e-07,
+      "loss": 2.5276,
+      "step": 55220
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8443426042983566e-07,
+      "loss": 2.5788,
+      "step": 55221
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8438685208596715e-07,
+      "loss": 0.722,
+      "step": 55222
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.843394437420986e-07,
+      "loss": 1.2435,
+      "step": 55223
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8429203539823014e-07,
+      "loss": 2.6472,
+      "step": 55224
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.842446270543616e-07,
+      "loss": 3.1358,
+      "step": 55225
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8419721871049307e-07,
+      "loss": 1.7725,
+      "step": 55226
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.841498103666245e-07,
+      "loss": 3.7805,
+      "step": 55227
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8410240202275605e-07,
+      "loss": 1.1999,
+      "step": 55228
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.840549936788875e-07,
+      "loss": 2.0114,
+      "step": 55229
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8400758533501893e-07,
+      "loss": 1.6822,
+      "step": 55230
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.839601769911505e-07,
+      "loss": 1.6964,
+      "step": 55231
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8391276864728197e-07,
+      "loss": 1.0251,
+      "step": 55232
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.838653603034134e-07,
+      "loss": 1.3261,
+      "step": 55233
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8381795195954485e-07,
+      "loss": 4.0935,
+      "step": 55234
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.837705436156764e-07,
+      "loss": 1.4315,
+      "step": 55235
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8372313527180783e-07,
+      "loss": 2.1212,
+      "step": 55236
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.836757269279393e-07,
+      "loss": 0.776,
+      "step": 55237
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8362831858407076e-07,
+      "loss": 2.588,
+      "step": 55238
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.835809102402023e-07,
+      "loss": 1.2112,
+      "step": 55239
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8353350189633375e-07,
+      "loss": 1.5948,
+      "step": 55240
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8348609355246524e-07,
+      "loss": 2.6699,
+      "step": 55241
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.834386852085968e-07,
+      "loss": 1.9622,
+      "step": 55242
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.833912768647282e-07,
+      "loss": 1.6082,
+      "step": 55243
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8334386852085966e-07,
+      "loss": 3.2379,
+      "step": 55244
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8329646017699115e-07,
+      "loss": 1.5916,
+      "step": 55245
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8324905183312265e-07,
+      "loss": 1.4423,
+      "step": 55246
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8320164348925414e-07,
+      "loss": 2.4578,
+      "step": 55247
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.831542351453856e-07,
+      "loss": 2.5607,
+      "step": 55248
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.83106826801517e-07,
+      "loss": 2.4884,
+      "step": 55249
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8305941845764856e-07,
+      "loss": 1.4401,
+      "step": 55250
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8301201011378006e-07,
+      "loss": 2.4515,
+      "step": 55251
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.829646017699115e-07,
+      "loss": 1.0756,
+      "step": 55252
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8291719342604304e-07,
+      "loss": 2.4259,
+      "step": 55253
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.828697850821745e-07,
+      "loss": 1.9428,
+      "step": 55254
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.8282237673830597e-07,
+      "loss": 1.2935,
+      "step": 55255
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.827749683944374e-07,
+      "loss": 0.3443,
+      "step": 55256
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8272756005056896e-07,
+      "loss": 1.4259,
+      "step": 55257
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.826801517067004e-07,
+      "loss": 3.4974,
+      "step": 55258
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8263274336283183e-07,
+      "loss": 1.36,
+      "step": 55259
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.825853350189633e-07,
+      "loss": 0.7735,
+      "step": 55260
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8253792667509487e-07,
+      "loss": 1.1735,
+      "step": 55261
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.824905183312263e-07,
+      "loss": 1.7923,
+      "step": 55262
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8244310998735775e-07,
+      "loss": 3.4135,
+      "step": 55263
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.823957016434893e-07,
+      "loss": 2.4636,
+      "step": 55264
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.823482932996208e-07,
+      "loss": 2.6133,
+      "step": 55265
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.823008849557522e-07,
+      "loss": 2.7114,
+      "step": 55266
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8225347661188367e-07,
+      "loss": 1.8003,
+      "step": 55267
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.822060682680152e-07,
+      "loss": 2.9505,
+      "step": 55268
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8215865992414665e-07,
+      "loss": 2.1776,
+      "step": 55269
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8211125158027814e-07,
+      "loss": 1.3362,
+      "step": 55270
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.820638432364096e-07,
+      "loss": 0.5837,
+      "step": 55271
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.820164348925411e-07,
+      "loss": 2.3916,
+      "step": 55272
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8196902654867257e-07,
+      "loss": 0.8162,
+      "step": 55273
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8192161820480406e-07,
+      "loss": 2.0927,
+      "step": 55274
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8187420986093555e-07,
+      "loss": 0.3539,
+      "step": 55275
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8182680151706704e-07,
+      "loss": 2.1642,
+      "step": 55276
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.817793931731985e-07,
+      "loss": 2.263,
+      "step": 55277
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.817319848293299e-07,
+      "loss": 1.4418,
+      "step": 55278
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8168457648546147e-07,
+      "loss": 2.6486,
+      "step": 55279
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8163716814159296e-07,
+      "loss": 2.0444,
+      "step": 55280
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.815897597977244e-07,
+      "loss": 0.8306,
+      "step": 55281
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8154235145385584e-07,
+      "loss": 2.4921,
+      "step": 55282
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.814949431099874e-07,
+      "loss": 1.1224,
+      "step": 55283
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8144753476611887e-07,
+      "loss": 2.734,
+      "step": 55284
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.814001264222503e-07,
+      "loss": 0.2405,
+      "step": 55285
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8135271807838186e-07,
+      "loss": 1.4386,
+      "step": 55286
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.813053097345133e-07,
+      "loss": 0.9207,
+      "step": 55287
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8125790139064474e-07,
+      "loss": 0.8782,
+      "step": 55288
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8121049304677623e-07,
+      "loss": 2.5709,
+      "step": 55289
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.811630847029078e-07,
+      "loss": 2.0548,
+      "step": 55290
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.811156763590392e-07,
+      "loss": 1.8244,
+      "step": 55291
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8106826801517065e-07,
+      "loss": 1.9644,
+      "step": 55292
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8102085967130214e-07,
+      "loss": 3.3,
+      "step": 55293
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.809734513274337e-07,
+      "loss": 1.9004,
+      "step": 55294
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8092604298356513e-07,
+      "loss": 3.0475,
+      "step": 55295
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8087863463969657e-07,
+      "loss": 1.3248,
+      "step": 55296
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.808312262958281e-07,
+      "loss": 2.2301,
+      "step": 55297
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8078381795195955e-07,
+      "loss": 1.5548,
+      "step": 55298
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8073640960809104e-07,
+      "loss": 1.9148,
+      "step": 55299
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.806890012642225e-07,
+      "loss": 1.0077,
+      "step": 55300
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8064159292035403e-07,
+      "loss": 2.5296,
+      "step": 55301
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8059418457648547e-07,
+      "loss": 1.3152,
+      "step": 55302
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8054677623261696e-07,
+      "loss": 2.0847,
+      "step": 55303
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.804993678887484e-07,
+      "loss": 1.9667,
+      "step": 55304
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8045195954487994e-07,
+      "loss": 2.3405,
+      "step": 55305
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.804045512010114e-07,
+      "loss": 1.4516,
+      "step": 55306
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.803571428571428e-07,
+      "loss": 1.2056,
+      "step": 55307
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8030973451327437e-07,
+      "loss": 2.2887,
+      "step": 55308
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8026232616940586e-07,
+      "loss": 1.323,
+      "step": 55309
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.802149178255373e-07,
+      "loss": 1.9129,
+      "step": 55310
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.8016750948166874e-07,
+      "loss": 2.1986,
+      "step": 55311
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.801201011378003e-07,
+      "loss": 0.2276,
+      "step": 55312
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.800726927939318e-07,
+      "loss": 2.6997,
+      "step": 55313
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.800252844500632e-07,
+      "loss": 2.6133,
+      "step": 55314
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.7997787610619465e-07,
+      "loss": 0.9105,
+      "step": 55315
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.799304677623262e-07,
+      "loss": 2.1163,
+      "step": 55316
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.7988305941845764e-07,
+      "loss": 2.1459,
+      "step": 55317
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.7983565107458913e-07,
+      "loss": 1.4564,
+      "step": 55318
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7978824273072057e-07,
+      "loss": 1.4302,
+      "step": 55319
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.797408343868521e-07,
+      "loss": 1.7533,
+      "step": 55320
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7969342604298355e-07,
+      "loss": 2.1493,
+      "step": 55321
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7964601769911505e-07,
+      "loss": 2.6183,
+      "step": 55322
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.795986093552466e-07,
+      "loss": 0.8457,
+      "step": 55323
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7955120101137803e-07,
+      "loss": 0.5524,
+      "step": 55324
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7950379266750947e-07,
+      "loss": 2.1865,
+      "step": 55325
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7945638432364096e-07,
+      "loss": 1.3648,
+      "step": 55326
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7940897597977245e-07,
+      "loss": 0.9549,
+      "step": 55327
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7936156763590395e-07,
+      "loss": 0.4236,
+      "step": 55328
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.793141592920354e-07,
+      "loss": 3.2093,
+      "step": 55329
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.792667509481668e-07,
+      "loss": 3.7381,
+      "step": 55330
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7921934260429837e-07,
+      "loss": 1.0341,
+      "step": 55331
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7917193426042986e-07,
+      "loss": 1.0436,
+      "step": 55332
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.791245259165613e-07,
+      "loss": 3.0786,
+      "step": 55333
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7907711757269285e-07,
+      "loss": 1.6445,
+      "step": 55334
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.790297092288243e-07,
+      "loss": 2.1492,
+      "step": 55335
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.789823008849557e-07,
+      "loss": 1.8467,
+      "step": 55336
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.789348925410872e-07,
+      "loss": 1.3056,
+      "step": 55337
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7888748419721876e-07,
+      "loss": 2.434,
+      "step": 55338
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.788400758533502e-07,
+      "loss": 0.6636,
+      "step": 55339
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7879266750948164e-07,
+      "loss": 1.5638,
+      "step": 55340
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7874525916561313e-07,
+      "loss": 2.2039,
+      "step": 55341
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.786978508217447e-07,
+      "loss": 1.3187,
+      "step": 55342
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.786504424778761e-07,
+      "loss": 2.2263,
+      "step": 55343
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7860303413400756e-07,
+      "loss": 1.2979,
+      "step": 55344
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.785556257901391e-07,
+      "loss": 2.5214,
+      "step": 55345
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7850821744627054e-07,
+      "loss": 1.4697,
+      "step": 55346
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7846080910240203e-07,
+      "loss": 1.4429,
+      "step": 55347
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7841340075853347e-07,
+      "loss": 0.5463,
+      "step": 55348
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.78365992414665e-07,
+      "loss": 2.7754,
+      "step": 55349
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7831858407079646e-07,
+      "loss": 2.2782,
+      "step": 55350
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7827117572692795e-07,
+      "loss": 2.3561,
+      "step": 55351
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.782237673830594e-07,
+      "loss": 1.6895,
+      "step": 55352
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7817635903919093e-07,
+      "loss": 1.9227,
+      "step": 55353
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7812895069532237e-07,
+      "loss": 2.0451,
+      "step": 55354
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7808154235145387e-07,
+      "loss": 0.9426,
+      "step": 55355
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7803413400758536e-07,
+      "loss": 2.7635,
+      "step": 55356
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7798672566371685e-07,
+      "loss": 3.3078,
+      "step": 55357
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.779393173198483e-07,
+      "loss": 1.5602,
+      "step": 55358
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7789190897597973e-07,
+      "loss": 1.8794,
+      "step": 55359
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7784450063211127e-07,
+      "loss": 3.4621,
+      "step": 55360
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7779709228824277e-07,
+      "loss": 1.4153,
+      "step": 55361
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.777496839443742e-07,
+      "loss": 0.9128,
+      "step": 55362
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7770227560050564e-07,
+      "loss": 2.3884,
+      "step": 55363
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.776548672566372e-07,
+      "loss": 1.6941,
+      "step": 55364
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.776074589127687e-07,
+      "loss": 2.9033,
+      "step": 55365
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.775600505689001e-07,
+      "loss": 1.928,
+      "step": 55366
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7751264222503167e-07,
+      "loss": 0.6196,
+      "step": 55367
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.774652338811631e-07,
+      "loss": 2.5345,
+      "step": 55368
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7741782553729454e-07,
+      "loss": 0.1879,
+      "step": 55369
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7737041719342604e-07,
+      "loss": 1.2158,
+      "step": 55370
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.773230088495576e-07,
+      "loss": 0.8566,
+      "step": 55371
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.77275600505689e-07,
+      "loss": 2.6209,
+      "step": 55372
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7722819216182046e-07,
+      "loss": 1.8104,
+      "step": 55373
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7718078381795195e-07,
+      "loss": 1.0788,
+      "step": 55374
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7713337547408344e-07,
+      "loss": 1.8761,
+      "step": 55375
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7708596713021494e-07,
+      "loss": 2.4644,
+      "step": 55376
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.770385587863464e-07,
+      "loss": 2.951,
+      "step": 55377
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.769911504424779e-07,
+      "loss": 2.0262,
+      "step": 55378
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7694374209860936e-07,
+      "loss": 1.1274,
+      "step": 55379
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7689633375474085e-07,
+      "loss": 1.3515,
+      "step": 55380
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.768489254108723e-07,
+      "loss": 1.7141,
+      "step": 55381
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.7680151706700384e-07,
+      "loss": 0.6564,
+      "step": 55382
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.767541087231353e-07,
+      "loss": 2.5895,
+      "step": 55383
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7670670037926677e-07,
+      "loss": 1.7709,
+      "step": 55384
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.766592920353982e-07,
+      "loss": 0.8045,
+      "step": 55385
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7661188369152975e-07,
+      "loss": 0.8082,
+      "step": 55386
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.765644753476612e-07,
+      "loss": 2.9308,
+      "step": 55387
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7651706700379263e-07,
+      "loss": 3.2613,
+      "step": 55388
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.764696586599242e-07,
+      "loss": 1.8115,
+      "step": 55389
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7642225031605567e-07,
+      "loss": 1.0056,
+      "step": 55390
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.763748419721871e-07,
+      "loss": 2.2465,
+      "step": 55391
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7632743362831855e-07,
+      "loss": 1.1275,
+      "step": 55392
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.762800252844501e-07,
+      "loss": 2.2083,
+      "step": 55393
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.762326169405816e-07,
+      "loss": 1.6389,
+      "step": 55394
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.76185208596713e-07,
+      "loss": 1.6419,
+      "step": 55395
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7613780025284446e-07,
+      "loss": 3.9654,
+      "step": 55396
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.76090391908976e-07,
+      "loss": 1.4034,
+      "step": 55397
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7604298356510745e-07,
+      "loss": 2.1454,
+      "step": 55398
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7599557522123894e-07,
+      "loss": 1.3518,
+      "step": 55399
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.759481668773705e-07,
+      "loss": 1.1541,
+      "step": 55400
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.759007585335019e-07,
+      "loss": 1.21,
+      "step": 55401
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7585335018963336e-07,
+      "loss": 3.1691,
+      "step": 55402
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7580594184576485e-07,
+      "loss": 1.8113,
+      "step": 55403
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.757585335018964e-07,
+      "loss": 2.1439,
+      "step": 55404
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7571112515802784e-07,
+      "loss": 2.5827,
+      "step": 55405
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.756637168141593e-07,
+      "loss": 2.762,
+      "step": 55406
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.756163084702907e-07,
+      "loss": 1.5275,
+      "step": 55407
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7556890012642226e-07,
+      "loss": 1.0543,
+      "step": 55408
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7552149178255375e-07,
+      "loss": 2.5494,
+      "step": 55409
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.754740834386852e-07,
+      "loss": 1.3313,
+      "step": 55410
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7542667509481674e-07,
+      "loss": 1.9545,
+      "step": 55411
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.753792667509482e-07,
+      "loss": 0.6324,
+      "step": 55412
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7533185840707967e-07,
+      "loss": 1.8247,
+      "step": 55413
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.752844500632111e-07,
+      "loss": 2.5342,
+      "step": 55414
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7523704171934265e-07,
+      "loss": 1.9505,
+      "step": 55415
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.751896333754741e-07,
+      "loss": 1.3928,
+      "step": 55416
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7514222503160553e-07,
+      "loss": 1.4666,
+      "step": 55417
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.75094816687737e-07,
+      "loss": 2.2204,
+      "step": 55418
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7504740834386857e-07,
+      "loss": 1.2516,
+      "step": 55419
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.75e-07,
+      "loss": 0.851,
+      "step": 55420
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.749525916561315e-07,
+      "loss": 1.8553,
+      "step": 55421
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7490518331226294e-07,
+      "loss": 1.2205,
+      "step": 55422
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.748577749683945e-07,
+      "loss": 0.9771,
+      "step": 55423
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.748103666245259e-07,
+      "loss": 2.3402,
+      "step": 55424
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.747629582806574e-07,
+      "loss": 3.1006,
+      "step": 55425
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7471554993678886e-07,
+      "loss": 2.2969,
+      "step": 55426
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7466814159292035e-07,
+      "loss": 3.1131,
+      "step": 55427
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7462073324905184e-07,
+      "loss": 2.33,
+      "step": 55428
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7457332490518333e-07,
+      "loss": 1.6207,
+      "step": 55429
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7452591656131477e-07,
+      "loss": 0.7581,
+      "step": 55430
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7447850821744626e-07,
+      "loss": 2.473,
+      "step": 55431
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7443109987357776e-07,
+      "loss": 0.5098,
+      "step": 55432
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7438369152970925e-07,
+      "loss": 1.2883,
+      "step": 55433
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7433628318584074e-07,
+      "loss": 2.4982,
+      "step": 55434
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.742888748419722e-07,
+      "loss": 1.6634,
+      "step": 55435
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7424146649810367e-07,
+      "loss": 3.7742,
+      "step": 55436
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7419405815423516e-07,
+      "loss": 1.8741,
+      "step": 55437
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7414664981036666e-07,
+      "loss": 1.8363,
+      "step": 55438
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.740992414664981e-07,
+      "loss": 1.6808,
+      "step": 55439
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.740518331226296e-07,
+      "loss": 1.9229,
+      "step": 55440
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7400442477876103e-07,
+      "loss": 2.1035,
+      "step": 55441
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7395701643489257e-07,
+      "loss": 0.9891,
+      "step": 55442
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.7390960809102406e-07,
+      "loss": 3.4317,
+      "step": 55443
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.738621997471555e-07,
+      "loss": 2.0284,
+      "step": 55444
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.73814791403287e-07,
+      "loss": 1.8843,
+      "step": 55445
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7376738305941844e-07,
+      "loss": 2.247,
+      "step": 55446
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7371997471555e-07,
+      "loss": 1.0287,
+      "step": 55447
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.736725663716814e-07,
+      "loss": 2.3265,
+      "step": 55448
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.736251580278129e-07,
+      "loss": 2.5003,
+      "step": 55449
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7357774968394435e-07,
+      "loss": 1.6925,
+      "step": 55450
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7353034134007584e-07,
+      "loss": 1.6704,
+      "step": 55451
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7348293299620734e-07,
+      "loss": 1.0461,
+      "step": 55452
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7343552465233883e-07,
+      "loss": 1.7563,
+      "step": 55453
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.733881163084703e-07,
+      "loss": 2.7388,
+      "step": 55454
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7334070796460176e-07,
+      "loss": 1.3999,
+      "step": 55455
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7329329962073325e-07,
+      "loss": 1.9192,
+      "step": 55456
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7324589127686474e-07,
+      "loss": 3.2811,
+      "step": 55457
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7319848293299624e-07,
+      "loss": 3.2598,
+      "step": 55458
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.731510745891277e-07,
+      "loss": 2.5965,
+      "step": 55459
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7310366624525917e-07,
+      "loss": 0.6107,
+      "step": 55460
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7305625790139066e-07,
+      "loss": 0.9032,
+      "step": 55461
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7300884955752215e-07,
+      "loss": 2.3628,
+      "step": 55462
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.729614412136536e-07,
+      "loss": 2.6653,
+      "step": 55463
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.729140328697851e-07,
+      "loss": 1.8714,
+      "step": 55464
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.728666245259166e-07,
+      "loss": 2.2515,
+      "step": 55465
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7281921618204807e-07,
+      "loss": 1.5583,
+      "step": 55466
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7277180783817956e-07,
+      "loss": 0.8405,
+      "step": 55467
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.72724399494311e-07,
+      "loss": 1.6735,
+      "step": 55468
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.726769911504425e-07,
+      "loss": 1.4266,
+      "step": 55469
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.72629582806574e-07,
+      "loss": 2.137,
+      "step": 55470
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.725821744627055e-07,
+      "loss": 0.9994,
+      "step": 55471
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.725347661188369e-07,
+      "loss": 1.9498,
+      "step": 55472
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.724873577749684e-07,
+      "loss": 1.9958,
+      "step": 55473
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7243994943109985e-07,
+      "loss": 1.0747,
+      "step": 55474
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7239254108723134e-07,
+      "loss": 2.443,
+      "step": 55475
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7234513274336283e-07,
+      "loss": 2.3602,
+      "step": 55476
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.722977243994943e-07,
+      "loss": 2.0099,
+      "step": 55477
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.722503160556258e-07,
+      "loss": 1.9186,
+      "step": 55478
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7220290771175725e-07,
+      "loss": 1.853,
+      "step": 55479
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7215549936788875e-07,
+      "loss": 0.7291,
+      "step": 55480
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7210809102402024e-07,
+      "loss": 1.8071,
+      "step": 55481
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7206068268015173e-07,
+      "loss": 1.0736,
+      "step": 55482
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7201327433628317e-07,
+      "loss": 1.246,
+      "step": 55483
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7196586599241466e-07,
+      "loss": 3.648,
+      "step": 55484
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7191845764854615e-07,
+      "loss": 1.0034,
+      "step": 55485
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7187104930467765e-07,
+      "loss": 1.1033,
+      "step": 55486
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.718236409608091e-07,
+      "loss": 2.2496,
+      "step": 55487
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.717762326169406e-07,
+      "loss": 2.049,
+      "step": 55488
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7172882427307207e-07,
+      "loss": 1.3937,
+      "step": 55489
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7168141592920356e-07,
+      "loss": 2.8951,
+      "step": 55490
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7163400758533505e-07,
+      "loss": 2.7234,
+      "step": 55491
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.715865992414665e-07,
+      "loss": 2.0066,
+      "step": 55492
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.71539190897598e-07,
+      "loss": 1.9424,
+      "step": 55493
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.714917825537295e-07,
+      "loss": 0.4669,
+      "step": 55494
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7144437420986097e-07,
+      "loss": 1.0464,
+      "step": 55495
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.713969658659924e-07,
+      "loss": 1.475,
+      "step": 55496
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.713495575221239e-07,
+      "loss": 0.8889,
+      "step": 55497
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7130214917825534e-07,
+      "loss": 1.9634,
+      "step": 55498
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.712547408343869e-07,
+      "loss": 1.4323,
+      "step": 55499
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.712073324905184e-07,
+      "loss": 1.3763,
+      "step": 55500
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.711599241466498e-07,
+      "loss": 1.2721,
+      "step": 55501
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.711125158027813e-07,
+      "loss": 2.5682,
+      "step": 55502
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7106510745891275e-07,
+      "loss": 1.9849,
+      "step": 55503
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.710176991150443e-07,
+      "loss": 1.3062,
+      "step": 55504
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7097029077117573e-07,
+      "loss": 1.2549,
+      "step": 55505
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.709228824273072e-07,
+      "loss": 2.2104,
+      "step": 55506
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7087547408343866e-07,
+      "loss": 1.612,
+      "step": 55507
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.7082806573957016e-07,
+      "loss": 1.7287,
+      "step": 55508
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7078065739570165e-07,
+      "loss": 2.1913,
+      "step": 55509
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7073324905183314e-07,
+      "loss": 0.7016,
+      "step": 55510
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7068584070796463e-07,
+      "loss": 2.9219,
+      "step": 55511
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7063843236409607e-07,
+      "loss": 2.4212,
+      "step": 55512
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7059102402022756e-07,
+      "loss": 1.2684,
+      "step": 55513
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7054361567635906e-07,
+      "loss": 1.148,
+      "step": 55514
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7049620733249055e-07,
+      "loss": 1.8546,
+      "step": 55515
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.70448798988622e-07,
+      "loss": 2.2266,
+      "step": 55516
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.704013906447535e-07,
+      "loss": 2.2571,
+      "step": 55517
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7035398230088497e-07,
+      "loss": 1.9892,
+      "step": 55518
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7030657395701646e-07,
+      "loss": 1.0006,
+      "step": 55519
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.702591656131479e-07,
+      "loss": 1.015,
+      "step": 55520
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.702117572692794e-07,
+      "loss": 1.2824,
+      "step": 55521
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.701643489254109e-07,
+      "loss": 1.1197,
+      "step": 55522
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.701169405815424e-07,
+      "loss": 3.818,
+      "step": 55523
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.7006953223767387e-07,
+      "loss": 2.114,
+      "step": 55524
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.700221238938053e-07,
+      "loss": 2.3,
+      "step": 55525
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.699747155499368e-07,
+      "loss": 1.4205,
+      "step": 55526
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6992730720606824e-07,
+      "loss": 2.2067,
+      "step": 55527
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.698798988621998e-07,
+      "loss": 2.0533,
+      "step": 55528
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6983249051833123e-07,
+      "loss": 1.6473,
+      "step": 55529
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.697850821744627e-07,
+      "loss": 1.2946,
+      "step": 55530
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6973767383059416e-07,
+      "loss": 1.4992,
+      "step": 55531
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6969026548672565e-07,
+      "loss": 2.7378,
+      "step": 55532
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.696428571428572e-07,
+      "loss": 1.9567,
+      "step": 55533
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6959544879898864e-07,
+      "loss": 0.9483,
+      "step": 55534
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6954804045512013e-07,
+      "loss": 1.8514,
+      "step": 55535
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6950063211125157e-07,
+      "loss": 1.6122,
+      "step": 55536
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6945322376738306e-07,
+      "loss": 2.3228,
+      "step": 55537
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6940581542351455e-07,
+      "loss": 1.2961,
+      "step": 55538
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6935840707964604e-07,
+      "loss": 1.0406,
+      "step": 55539
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.693109987357775e-07,
+      "loss": 2.3968,
+      "step": 55540
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.69263590391909e-07,
+      "loss": 1.5288,
+      "step": 55541
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6921618204804047e-07,
+      "loss": 4.5226,
+      "step": 55542
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6916877370417196e-07,
+      "loss": 0.4563,
+      "step": 55543
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6912136536030345e-07,
+      "loss": 3.2698,
+      "step": 55544
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.690739570164349e-07,
+      "loss": 2.1309,
+      "step": 55545
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.690265486725664e-07,
+      "loss": 2.3359,
+      "step": 55546
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.689791403286979e-07,
+      "loss": 2.0733,
+      "step": 55547
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6893173198482937e-07,
+      "loss": 1.9264,
+      "step": 55548
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.688843236409608e-07,
+      "loss": 1.2564,
+      "step": 55549
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.688369152970923e-07,
+      "loss": 1.9627,
+      "step": 55550
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6878950695322374e-07,
+      "loss": 1.2667,
+      "step": 55551
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.687420986093553e-07,
+      "loss": 1.4581,
+      "step": 55552
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.686946902654867e-07,
+      "loss": 3.4923,
+      "step": 55553
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.686472819216182e-07,
+      "loss": 0.8554,
+      "step": 55554
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.685998735777497e-07,
+      "loss": 1.7069,
+      "step": 55555
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6855246523388115e-07,
+      "loss": 1.6248,
+      "step": 55556
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.685050568900127e-07,
+      "loss": 1.1616,
+      "step": 55557
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6845764854614413e-07,
+      "loss": 1.1369,
+      "step": 55558
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.684102402022756e-07,
+      "loss": 1.8614,
+      "step": 55559
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6836283185840706e-07,
+      "loss": 1.5523,
+      "step": 55560
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6831542351453855e-07,
+      "loss": 2.2302,
+      "step": 55561
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6826801517067005e-07,
+      "loss": 1.3802,
+      "step": 55562
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6822060682680154e-07,
+      "loss": 0.7012,
+      "step": 55563
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.68173198482933e-07,
+      "loss": 1.0722,
+      "step": 55564
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6812579013906447e-07,
+      "loss": 2.6848,
+      "step": 55565
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6807838179519596e-07,
+      "loss": 2.2178,
+      "step": 55566
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6803097345132745e-07,
+      "loss": 2.9344,
+      "step": 55567
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6798356510745895e-07,
+      "loss": 0.9977,
+      "step": 55568
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.679361567635904e-07,
+      "loss": 1.3127,
+      "step": 55569
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.678887484197219e-07,
+      "loss": 1.3956,
+      "step": 55570
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6784134007585337e-07,
+      "loss": 1.9129,
+      "step": 55571
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.6779393173198486e-07,
+      "loss": 2.6548,
+      "step": 55572
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.677465233881163e-07,
+      "loss": 1.6906,
+      "step": 55573
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.676991150442478e-07,
+      "loss": 1.0039,
+      "step": 55574
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6765170670037923e-07,
+      "loss": 0.8529,
+      "step": 55575
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.676042983565108e-07,
+      "loss": 2.616,
+      "step": 55576
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6755689001264227e-07,
+      "loss": 2.8665,
+      "step": 55577
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.675094816687737e-07,
+      "loss": 1.7099,
+      "step": 55578
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.674620733249052e-07,
+      "loss": 1.3856,
+      "step": 55579
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6741466498103664e-07,
+      "loss": 2.8353,
+      "step": 55580
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.673672566371682e-07,
+      "loss": 1.0815,
+      "step": 55581
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.673198482932996e-07,
+      "loss": 2.7161,
+      "step": 55582
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.672724399494311e-07,
+      "loss": 1.918,
+      "step": 55583
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6722503160556256e-07,
+      "loss": 1.7852,
+      "step": 55584
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6717762326169405e-07,
+      "loss": 1.2152,
+      "step": 55585
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6713021491782554e-07,
+      "loss": 1.9446,
+      "step": 55586
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6708280657395703e-07,
+      "loss": 2.2485,
+      "step": 55587
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6703539823008847e-07,
+      "loss": 2.7045,
+      "step": 55588
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6698798988621996e-07,
+      "loss": 2.0968,
+      "step": 55589
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6694058154235146e-07,
+      "loss": 2.0806,
+      "step": 55590
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6689317319848295e-07,
+      "loss": 1.1132,
+      "step": 55591
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6684576485461444e-07,
+      "loss": 0.6443,
+      "step": 55592
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.667983565107459e-07,
+      "loss": 0.8531,
+      "step": 55593
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6675094816687737e-07,
+      "loss": 1.6913,
+      "step": 55594
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6670353982300886e-07,
+      "loss": 3.3624,
+      "step": 55595
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6665613147914036e-07,
+      "loss": 0.5588,
+      "step": 55596
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.666087231352718e-07,
+      "loss": 2.0394,
+      "step": 55597
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.665613147914033e-07,
+      "loss": 1.4321,
+      "step": 55598
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.665139064475348e-07,
+      "loss": 2.4097,
+      "step": 55599
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6646649810366627e-07,
+      "loss": 0.8821,
+      "step": 55600
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6641908975979776e-07,
+      "loss": 1.2334,
+      "step": 55601
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.663716814159292e-07,
+      "loss": 3.1597,
+      "step": 55602
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.663242730720607e-07,
+      "loss": 1.1786,
+      "step": 55603
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.662768647281922e-07,
+      "loss": 1.6502,
+      "step": 55604
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.662294563843237e-07,
+      "loss": 1.0651,
+      "step": 55605
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.661820480404551e-07,
+      "loss": 3.0233,
+      "step": 55606
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.661346396965866e-07,
+      "loss": 1.7368,
+      "step": 55607
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6608723135271805e-07,
+      "loss": 1.9372,
+      "step": 55608
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.660398230088496e-07,
+      "loss": 1.4428,
+      "step": 55609
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6599241466498103e-07,
+      "loss": 1.1936,
+      "step": 55610
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6594500632111253e-07,
+      "loss": 3.4473,
+      "step": 55611
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.65897597977244e-07,
+      "loss": 1.8973,
+      "step": 55612
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6585018963337546e-07,
+      "loss": 0.84,
+      "step": 55613
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6580278128950695e-07,
+      "loss": 2.2597,
+      "step": 55614
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6575537294563844e-07,
+      "loss": 2.5893,
+      "step": 55615
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6570796460176994e-07,
+      "loss": 2.1244,
+      "step": 55616
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.656605562579014e-07,
+      "loss": 2.4847,
+      "step": 55617
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6561314791403287e-07,
+      "loss": 1.4296,
+      "step": 55618
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6556573957016436e-07,
+      "loss": 2.5806,
+      "step": 55619
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6551833122629585e-07,
+      "loss": 1.2665,
+      "step": 55620
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.654709228824273e-07,
+      "loss": 2.299,
+      "step": 55621
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.654235145385588e-07,
+      "loss": 1.968,
+      "step": 55622
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.653761061946903e-07,
+      "loss": 2.6755,
+      "step": 55623
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6532869785082177e-07,
+      "loss": 1.656,
+      "step": 55624
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6528128950695326e-07,
+      "loss": 3.0465,
+      "step": 55625
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.652338811630847e-07,
+      "loss": 3.1869,
+      "step": 55626
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.651864728192162e-07,
+      "loss": 1.4011,
+      "step": 55627
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.651390644753477e-07,
+      "loss": 0.8829,
+      "step": 55628
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.650916561314792e-07,
+      "loss": 2.2718,
+      "step": 55629
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.650442477876106e-07,
+      "loss": 0.9875,
+      "step": 55630
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.649968394437421e-07,
+      "loss": 2.8655,
+      "step": 55631
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.6494943109987355e-07,
+      "loss": 1.9377,
+      "step": 55632
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.649020227560051e-07,
+      "loss": 1.3446,
+      "step": 55633
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.648546144121366e-07,
+      "loss": 2.4688,
+      "step": 55634
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.64807206068268e-07,
+      "loss": 0.9852,
+      "step": 55635
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.647597977243995e-07,
+      "loss": 0.8874,
+      "step": 55636
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6471238938053095e-07,
+      "loss": 2.1727,
+      "step": 55637
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.646649810366625e-07,
+      "loss": 1.181,
+      "step": 55638
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6461757269279394e-07,
+      "loss": 2.258,
+      "step": 55639
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6457016434892543e-07,
+      "loss": 2.5662,
+      "step": 55640
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6452275600505687e-07,
+      "loss": 0.5861,
+      "step": 55641
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6447534766118836e-07,
+      "loss": 3.8676,
+      "step": 55642
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6442793931731985e-07,
+      "loss": 0.9565,
+      "step": 55643
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6438053097345135e-07,
+      "loss": 1.9297,
+      "step": 55644
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6433312262958284e-07,
+      "loss": 1.5567,
+      "step": 55645
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.642857142857143e-07,
+      "loss": 1.1542,
+      "step": 55646
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6423830594184577e-07,
+      "loss": 0.6386,
+      "step": 55647
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6419089759797726e-07,
+      "loss": 2.5307,
+      "step": 55648
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6414348925410875e-07,
+      "loss": 3.2345,
+      "step": 55649
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.640960809102402e-07,
+      "loss": 2.8555,
+      "step": 55650
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.640486725663717e-07,
+      "loss": 1.5242,
+      "step": 55651
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.640012642225032e-07,
+      "loss": 1.6167,
+      "step": 55652
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6395385587863467e-07,
+      "loss": 1.522,
+      "step": 55653
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.639064475347661e-07,
+      "loss": 1.9403,
+      "step": 55654
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.638590391908976e-07,
+      "loss": 1.9513,
+      "step": 55655
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.638116308470291e-07,
+      "loss": 2.6562,
+      "step": 55656
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.637642225031606e-07,
+      "loss": 1.749,
+      "step": 55657
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.637168141592921e-07,
+      "loss": 2.058,
+      "step": 55658
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.636694058154235e-07,
+      "loss": 3.0702,
+      "step": 55659
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.63621997471555e-07,
+      "loss": 2.6284,
+      "step": 55660
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6357458912768645e-07,
+      "loss": 0.7302,
+      "step": 55661
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.63527180783818e-07,
+      "loss": 1.8042,
+      "step": 55662
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6347977243994943e-07,
+      "loss": 1.366,
+      "step": 55663
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.634323640960809e-07,
+      "loss": 2.6088,
+      "step": 55664
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6338495575221236e-07,
+      "loss": 0.7791,
+      "step": 55665
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6333754740834386e-07,
+      "loss": 1.9925,
+      "step": 55666
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.632901390644754e-07,
+      "loss": 3.0456,
+      "step": 55667
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6324273072060684e-07,
+      "loss": 2.0138,
+      "step": 55668
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6319532237673833e-07,
+      "loss": 0.7548,
+      "step": 55669
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6314791403286977e-07,
+      "loss": 2.2695,
+      "step": 55670
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6310050568900126e-07,
+      "loss": 2.7166,
+      "step": 55671
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6305309734513276e-07,
+      "loss": 2.3524,
+      "step": 55672
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6300568900126425e-07,
+      "loss": 1.16,
+      "step": 55673
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.629582806573957e-07,
+      "loss": 2.1832,
+      "step": 55674
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.629108723135272e-07,
+      "loss": 3.8287,
+      "step": 55675
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6286346396965867e-07,
+      "loss": 1.1419,
+      "step": 55676
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6281605562579016e-07,
+      "loss": 1.783,
+      "step": 55677
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6276864728192166e-07,
+      "loss": 1.9133,
+      "step": 55678
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.627212389380531e-07,
+      "loss": 1.9575,
+      "step": 55679
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.626738305941846e-07,
+      "loss": 2.1661,
+      "step": 55680
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.626264222503161e-07,
+      "loss": 1.8841,
+      "step": 55681
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6257901390644757e-07,
+      "loss": 2.4232,
+      "step": 55682
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.62531605562579e-07,
+      "loss": 2.5106,
+      "step": 55683
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.624841972187105e-07,
+      "loss": 2.0873,
+      "step": 55684
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6243678887484194e-07,
+      "loss": 1.4778,
+      "step": 55685
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.623893805309735e-07,
+      "loss": 1.4627,
+      "step": 55686
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6234197218710493e-07,
+      "loss": 1.5042,
+      "step": 55687
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.622945638432364e-07,
+      "loss": 1.2394,
+      "step": 55688
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.622471554993679e-07,
+      "loss": 1.9511,
+      "step": 55689
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6219974715549935e-07,
+      "loss": 1.9746,
+      "step": 55690
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.621523388116309e-07,
+      "loss": 1.2172,
+      "step": 55691
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6210493046776233e-07,
+      "loss": 3.298,
+      "step": 55692
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6205752212389383e-07,
+      "loss": 2.4722,
+      "step": 55693
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6201011378002527e-07,
+      "loss": 2.0265,
+      "step": 55694
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6196270543615676e-07,
+      "loss": 1.3698,
+      "step": 55695
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6191529709228825e-07,
+      "loss": 2.944,
+      "step": 55696
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.6186788874841974e-07,
+      "loss": 2.6299,
+      "step": 55697
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.618204804045512e-07,
+      "loss": 2.3888,
+      "step": 55698
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.617730720606827e-07,
+      "loss": 0.7318,
+      "step": 55699
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6172566371681417e-07,
+      "loss": 2.5967,
+      "step": 55700
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6167825537294566e-07,
+      "loss": 1.9375,
+      "step": 55701
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6163084702907715e-07,
+      "loss": 2.2791,
+      "step": 55702
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.615834386852086e-07,
+      "loss": 1.1379,
+      "step": 55703
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.615360303413401e-07,
+      "loss": 2.2344,
+      "step": 55704
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.614886219974716e-07,
+      "loss": 0.8775,
+      "step": 55705
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6144121365360307e-07,
+      "loss": 2.1264,
+      "step": 55706
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.613938053097345e-07,
+      "loss": 2.1047,
+      "step": 55707
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.61346396965866e-07,
+      "loss": 2.1493,
+      "step": 55708
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.612989886219975e-07,
+      "loss": 2.3825,
+      "step": 55709
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.61251580278129e-07,
+      "loss": 2.0248,
+      "step": 55710
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.612041719342604e-07,
+      "loss": 2.5048,
+      "step": 55711
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.611567635903919e-07,
+      "loss": 2.1505,
+      "step": 55712
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.611093552465234e-07,
+      "loss": 2.2817,
+      "step": 55713
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.610619469026549e-07,
+      "loss": 3.1702,
+      "step": 55714
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.610145385587864e-07,
+      "loss": 1.5403,
+      "step": 55715
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6096713021491783e-07,
+      "loss": 2.2674,
+      "step": 55716
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.609197218710493e-07,
+      "loss": 2.6786,
+      "step": 55717
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6087231352718076e-07,
+      "loss": 1.9901,
+      "step": 55718
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6082490518331225e-07,
+      "loss": 1.223,
+      "step": 55719
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6077749683944374e-07,
+      "loss": 3.2344,
+      "step": 55720
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6073008849557524e-07,
+      "loss": 2.2585,
+      "step": 55721
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.606826801517067e-07,
+      "loss": 3.0593,
+      "step": 55722
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6063527180783817e-07,
+      "loss": 0.7933,
+      "step": 55723
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6058786346396966e-07,
+      "loss": 2.7055,
+      "step": 55724
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6054045512010115e-07,
+      "loss": 2.4463,
+      "step": 55725
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6049304677623265e-07,
+      "loss": 1.4055,
+      "step": 55726
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.604456384323641e-07,
+      "loss": 0.4624,
+      "step": 55727
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.603982300884956e-07,
+      "loss": 1.8578,
+      "step": 55728
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6035082174462707e-07,
+      "loss": 1.5414,
+      "step": 55729
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6030341340075856e-07,
+      "loss": 2.1056,
+      "step": 55730
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6025600505689e-07,
+      "loss": 1.5341,
+      "step": 55731
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.602085967130215e-07,
+      "loss": 1.9343,
+      "step": 55732
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.60161188369153e-07,
+      "loss": 1.813,
+      "step": 55733
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.601137800252845e-07,
+      "loss": 2.3505,
+      "step": 55734
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.6006637168141597e-07,
+      "loss": 1.7631,
+      "step": 55735
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.600189633375474e-07,
+      "loss": 1.8913,
+      "step": 55736
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.599715549936789e-07,
+      "loss": 1.5573,
+      "step": 55737
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.599241466498104e-07,
+      "loss": 4.4555,
+      "step": 55738
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.598767383059419e-07,
+      "loss": 0.8575,
+      "step": 55739
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.598293299620733e-07,
+      "loss": 0.8301,
+      "step": 55740
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.597819216182048e-07,
+      "loss": 2.2031,
+      "step": 55741
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5973451327433626e-07,
+      "loss": 2.1914,
+      "step": 55742
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.596871049304678e-07,
+      "loss": 1.5715,
+      "step": 55743
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5963969658659924e-07,
+      "loss": 1.2281,
+      "step": 55744
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5959228824273073e-07,
+      "loss": 2.3339,
+      "step": 55745
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.595448798988622e-07,
+      "loss": 2.9336,
+      "step": 55746
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5949747155499366e-07,
+      "loss": 1.7944,
+      "step": 55747
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.594500632111252e-07,
+      "loss": 1.4328,
+      "step": 55748
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5940265486725665e-07,
+      "loss": 3.4309,
+      "step": 55749
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5935524652338814e-07,
+      "loss": 1.017,
+      "step": 55750
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.593078381795196e-07,
+      "loss": 2.592,
+      "step": 55751
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5926042983565107e-07,
+      "loss": 1.6792,
+      "step": 55752
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5921302149178256e-07,
+      "loss": 1.9913,
+      "step": 55753
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5916561314791406e-07,
+      "loss": 2.9256,
+      "step": 55754
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.591182048040455e-07,
+      "loss": 2.0576,
+      "step": 55755
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.59070796460177e-07,
+      "loss": 1.6977,
+      "step": 55756
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.590233881163085e-07,
+      "loss": 2.4649,
+      "step": 55757
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5897597977243997e-07,
+      "loss": 2.3672,
+      "step": 55758
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.5892857142857146e-07,
+      "loss": 1.8734,
+      "step": 55759
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.588811630847029e-07,
+      "loss": 1.9135,
+      "step": 55760
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.588337547408344e-07,
+      "loss": 1.6014,
+      "step": 55761
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.587863463969659e-07,
+      "loss": 2.0511,
+      "step": 55762
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.587389380530974e-07,
+      "loss": 0.7511,
+      "step": 55763
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.586915297092288e-07,
+      "loss": 1.4288,
+      "step": 55764
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.586441213653603e-07,
+      "loss": 1.7773,
+      "step": 55765
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5859671302149175e-07,
+      "loss": 2.963,
+      "step": 55766
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.585493046776233e-07,
+      "loss": 0.4217,
+      "step": 55767
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.585018963337548e-07,
+      "loss": 1.2391,
+      "step": 55768
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.584544879898862e-07,
+      "loss": 1.4911,
+      "step": 55769
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.584070796460177e-07,
+      "loss": 0.541,
+      "step": 55770
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5835967130214916e-07,
+      "loss": 2.5383,
+      "step": 55771
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.583122629582807e-07,
+      "loss": 0.5858,
+      "step": 55772
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5826485461441214e-07,
+      "loss": 1.1526,
+      "step": 55773
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5821744627054363e-07,
+      "loss": 1.7605,
+      "step": 55774
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5817003792667507e-07,
+      "loss": 1.4575,
+      "step": 55775
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5812262958280657e-07,
+      "loss": 2.347,
+      "step": 55776
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5807522123893806e-07,
+      "loss": 2.417,
+      "step": 55777
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5802781289506955e-07,
+      "loss": 0.7103,
+      "step": 55778
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5798040455120104e-07,
+      "loss": 2.6625,
+      "step": 55779
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.579329962073325e-07,
+      "loss": 3.3499,
+      "step": 55780
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.57885587863464e-07,
+      "loss": 2.0771,
+      "step": 55781
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5783817951959547e-07,
+      "loss": 2.481,
+      "step": 55782
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5779077117572696e-07,
+      "loss": 1.4004,
+      "step": 55783
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.577433628318584e-07,
+      "loss": 2.0183,
+      "step": 55784
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.576959544879899e-07,
+      "loss": 2.2199,
+      "step": 55785
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.576485461441214e-07,
+      "loss": 1.5722,
+      "step": 55786
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.576011378002529e-07,
+      "loss": 1.1419,
+      "step": 55787
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.575537294563843e-07,
+      "loss": 1.0922,
+      "step": 55788
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.575063211125158e-07,
+      "loss": 2.3596,
+      "step": 55789
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.574589127686473e-07,
+      "loss": 2.3755,
+      "step": 55790
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.574115044247788e-07,
+      "loss": 1.4146,
+      "step": 55791
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.573640960809103e-07,
+      "loss": 2.6368,
+      "step": 55792
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.573166877370417e-07,
+      "loss": 2.3573,
+      "step": 55793
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.572692793931732e-07,
+      "loss": 0.2735,
+      "step": 55794
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5722187104930465e-07,
+      "loss": 2.1426,
+      "step": 55795
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.571744627054362e-07,
+      "loss": 1.9393,
+      "step": 55796
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5712705436156764e-07,
+      "loss": 0.9723,
+      "step": 55797
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5707964601769913e-07,
+      "loss": 1.987,
+      "step": 55798
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5703223767383057e-07,
+      "loss": 0.8164,
+      "step": 55799
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5698482932996206e-07,
+      "loss": 2.1386,
+      "step": 55800
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.569374209860936e-07,
+      "loss": 2.301,
+      "step": 55801
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5689001264222504e-07,
+      "loss": 1.4301,
+      "step": 55802
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5684260429835654e-07,
+      "loss": 2.0419,
+      "step": 55803
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.56795195954488e-07,
+      "loss": 2.2849,
+      "step": 55804
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5674778761061947e-07,
+      "loss": 1.2466,
+      "step": 55805
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5670037926675096e-07,
+      "loss": 2.1297,
+      "step": 55806
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5665297092288245e-07,
+      "loss": 4.1504,
+      "step": 55807
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.566055625790139e-07,
+      "loss": 1.2791,
+      "step": 55808
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.565581542351454e-07,
+      "loss": 0.9126,
+      "step": 55809
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.565107458912769e-07,
+      "loss": 2.0661,
+      "step": 55810
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5646333754740837e-07,
+      "loss": 1.9819,
+      "step": 55811
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.564159292035398e-07,
+      "loss": 3.3766,
+      "step": 55812
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.563685208596713e-07,
+      "loss": 2.3665,
+      "step": 55813
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.563211125158028e-07,
+      "loss": 2.8795,
+      "step": 55814
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.562737041719343e-07,
+      "loss": 1.0028,
+      "step": 55815
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.562262958280658e-07,
+      "loss": 0.8347,
+      "step": 55816
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.561788874841972e-07,
+      "loss": 1.9168,
+      "step": 55817
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.561314791403287e-07,
+      "loss": 2.2531,
+      "step": 55818
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5608407079646015e-07,
+      "loss": 1.5841,
+      "step": 55819
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.560366624525917e-07,
+      "loss": 1.3263,
+      "step": 55820
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5598925410872313e-07,
+      "loss": 1.8728,
+      "step": 55821
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.559418457648546e-07,
+      "loss": 0.8301,
+      "step": 55822
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5589443742098606e-07,
+      "loss": 2.503,
+      "step": 55823
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5584702907711755e-07,
+      "loss": 1.7303,
+      "step": 55824
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.557996207332491e-07,
+      "loss": 2.5236,
+      "step": 55825
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5575221238938054e-07,
+      "loss": 1.9445,
+      "step": 55826
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5570480404551203e-07,
+      "loss": 1.6951,
+      "step": 55827
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5565739570164347e-07,
+      "loss": 2.3761,
+      "step": 55828
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5560998735777496e-07,
+      "loss": 2.4655,
+      "step": 55829
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5556257901390646e-07,
+      "loss": 2.7116,
+      "step": 55830
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5551517067003795e-07,
+      "loss": 2.2996,
+      "step": 55831
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.554677623261694e-07,
+      "loss": 2.2342,
+      "step": 55832
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.554203539823009e-07,
+      "loss": 2.0295,
+      "step": 55833
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5537294563843237e-07,
+      "loss": 0.7455,
+      "step": 55834
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5532553729456386e-07,
+      "loss": 2.1399,
+      "step": 55835
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5527812895069536e-07,
+      "loss": 3.3608,
+      "step": 55836
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.552307206068268e-07,
+      "loss": 2.7278,
+      "step": 55837
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.551833122629583e-07,
+      "loss": 0.7392,
+      "step": 55838
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.551359039190898e-07,
+      "loss": 1.426,
+      "step": 55839
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5508849557522127e-07,
+      "loss": 2.3301,
+      "step": 55840
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.550410872313527e-07,
+      "loss": 2.6726,
+      "step": 55841
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.549936788874842e-07,
+      "loss": 1.8682,
+      "step": 55842
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.549462705436157e-07,
+      "loss": 1.5625,
+      "step": 55843
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.548988621997472e-07,
+      "loss": 0.9443,
+      "step": 55844
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.548514538558786e-07,
+      "loss": 3.8815,
+      "step": 55845
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.548040455120101e-07,
+      "loss": 2.2931,
+      "step": 55846
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.547566371681416e-07,
+      "loss": 3.0044,
+      "step": 55847
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.547092288242731e-07,
+      "loss": 1.3583,
+      "step": 55848
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.546618204804046e-07,
+      "loss": 2.135,
+      "step": 55849
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5461441213653603e-07,
+      "loss": 2.172,
+      "step": 55850
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.545670037926675e-07,
+      "loss": 1.7016,
+      "step": 55851
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5451959544879897e-07,
+      "loss": 2.5088,
+      "step": 55852
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.544721871049305e-07,
+      "loss": 1.4283,
+      "step": 55853
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5442477876106195e-07,
+      "loss": 1.8361,
+      "step": 55854
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5437737041719344e-07,
+      "loss": 3.585,
+      "step": 55855
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.543299620733249e-07,
+      "loss": 2.8356,
+      "step": 55856
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5428255372945637e-07,
+      "loss": 1.4601,
+      "step": 55857
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5423514538558787e-07,
+      "loss": 1.0345,
+      "step": 55858
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5418773704171936e-07,
+      "loss": 4.1673,
+      "step": 55859
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5414032869785085e-07,
+      "loss": 1.1031,
+      "step": 55860
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.540929203539823e-07,
+      "loss": 2.2852,
+      "step": 55861
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.540455120101138e-07,
+      "loss": 2.423,
+      "step": 55862
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5399810366624527e-07,
+      "loss": 0.5705,
+      "step": 55863
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5395069532237677e-07,
+      "loss": 2.4278,
+      "step": 55864
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.539032869785082e-07,
+      "loss": 1.8645,
+      "step": 55865
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.538558786346397e-07,
+      "loss": 2.3594,
+      "step": 55866
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.538084702907712e-07,
+      "loss": 1.7697,
+      "step": 55867
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.537610619469027e-07,
+      "loss": 1.4797,
+      "step": 55868
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5371365360303417e-07,
+      "loss": 3.0833,
+      "step": 55869
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.536662452591656e-07,
+      "loss": 0.7881,
+      "step": 55870
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.536188369152971e-07,
+      "loss": 1.5163,
+      "step": 55871
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.535714285714286e-07,
+      "loss": 2.844,
+      "step": 55872
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.535240202275601e-07,
+      "loss": 1.4422,
+      "step": 55873
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5347661188369153e-07,
+      "loss": 0.9317,
+      "step": 55874
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.53429203539823e-07,
+      "loss": 1.0668,
+      "step": 55875
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5338179519595446e-07,
+      "loss": 1.6576,
+      "step": 55876
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.53334386852086e-07,
+      "loss": 3.0073,
+      "step": 55877
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5328697850821744e-07,
+      "loss": 1.8011,
+      "step": 55878
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5323957016434894e-07,
+      "loss": 2.4193,
+      "step": 55879
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5319216182048043e-07,
+      "loss": 1.8434,
+      "step": 55880
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5314475347661187e-07,
+      "loss": 1.6588,
+      "step": 55881
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.530973451327434e-07,
+      "loss": 1.2305,
+      "step": 55882
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5304993678887485e-07,
+      "loss": 3.6705,
+      "step": 55883
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5300252844500634e-07,
+      "loss": 2.5612,
+      "step": 55884
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.529551201011378e-07,
+      "loss": 2.2435,
+      "step": 55885
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.529077117572693e-07,
+      "loss": 1.6571,
+      "step": 55886
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5286030341340077e-07,
+      "loss": 0.9972,
+      "step": 55887
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.5281289506953226e-07,
+      "loss": 0.8505,
+      "step": 55888
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.527654867256637e-07,
+      "loss": 0.8416,
+      "step": 55889
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.527180783817952e-07,
+      "loss": 3.057,
+      "step": 55890
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.526706700379267e-07,
+      "loss": 1.6648,
+      "step": 55891
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.526232616940582e-07,
+      "loss": 2.5878,
+      "step": 55892
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5257585335018967e-07,
+      "loss": 2.3585,
+      "step": 55893
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.525284450063211e-07,
+      "loss": 1.6967,
+      "step": 55894
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.524810366624526e-07,
+      "loss": 1.4078,
+      "step": 55895
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.524336283185841e-07,
+      "loss": 1.0795,
+      "step": 55896
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.523862199747156e-07,
+      "loss": 1.8483,
+      "step": 55897
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.52338811630847e-07,
+      "loss": 1.2754,
+      "step": 55898
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.522914032869785e-07,
+      "loss": 1.6766,
+      "step": 55899
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5224399494310995e-07,
+      "loss": 1.9635,
+      "step": 55900
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.521965865992415e-07,
+      "loss": 1.9892,
+      "step": 55901
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.52149178255373e-07,
+      "loss": 2.399,
+      "step": 55902
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5210176991150443e-07,
+      "loss": 1.2669,
+      "step": 55903
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.520543615676359e-07,
+      "loss": 2.4266,
+      "step": 55904
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5200695322376736e-07,
+      "loss": 0.7701,
+      "step": 55905
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.519595448798989e-07,
+      "loss": 2.5513,
+      "step": 55906
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5191213653603035e-07,
+      "loss": 1.8107,
+      "step": 55907
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5186472819216184e-07,
+      "loss": 1.4877,
+      "step": 55908
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.518173198482933e-07,
+      "loss": 0.8474,
+      "step": 55909
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5176991150442477e-07,
+      "loss": 1.0395,
+      "step": 55910
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5172250316055626e-07,
+      "loss": 3.1721,
+      "step": 55911
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5167509481668775e-07,
+      "loss": 1.9101,
+      "step": 55912
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5162768647281925e-07,
+      "loss": 0.8939,
+      "step": 55913
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.515802781289507e-07,
+      "loss": 2.7842,
+      "step": 55914
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.515328697850822e-07,
+      "loss": 1.6936,
+      "step": 55915
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5148546144121367e-07,
+      "loss": 1.659,
+      "step": 55916
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5143805309734516e-07,
+      "loss": 1.3751,
+      "step": 55917
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.513906447534766e-07,
+      "loss": 2.3925,
+      "step": 55918
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.513432364096081e-07,
+      "loss": 2.0434,
+      "step": 55919
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.512958280657396e-07,
+      "loss": 1.1913,
+      "step": 55920
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.512484197218711e-07,
+      "loss": 1.39,
+      "step": 55921
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.512010113780025e-07,
+      "loss": 1.8383,
+      "step": 55922
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.51153603034134e-07,
+      "loss": 1.1501,
+      "step": 55923
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5110619469026545e-07,
+      "loss": 1.6538,
+      "step": 55924
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.51058786346397e-07,
+      "loss": 1.3289,
+      "step": 55925
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.510113780025285e-07,
+      "loss": 1.1105,
+      "step": 55926
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.509639696586599e-07,
+      "loss": 0.8331,
+      "step": 55927
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.509165613147914e-07,
+      "loss": 2.101,
+      "step": 55928
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5086915297092286e-07,
+      "loss": 0.7204,
+      "step": 55929
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.508217446270544e-07,
+      "loss": 1.5728,
+      "step": 55930
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5077433628318584e-07,
+      "loss": 1.8093,
+      "step": 55931
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5072692793931733e-07,
+      "loss": 3.0293,
+      "step": 55932
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5067951959544877e-07,
+      "loss": 1.3907,
+      "step": 55933
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5063211125158026e-07,
+      "loss": 0.712,
+      "step": 55934
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5058470290771176e-07,
+      "loss": 1.0822,
+      "step": 55935
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5053729456384325e-07,
+      "loss": 1.3229,
+      "step": 55936
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5048988621997474e-07,
+      "loss": 1.1717,
+      "step": 55937
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.504424778761062e-07,
+      "loss": 1.9006,
+      "step": 55938
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5039506953223767e-07,
+      "loss": 1.0841,
+      "step": 55939
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5034766118836917e-07,
+      "loss": 1.3191,
+      "step": 55940
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5030025284450066e-07,
+      "loss": 0.9607,
+      "step": 55941
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.502528445006321e-07,
+      "loss": 1.4262,
+      "step": 55942
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.502054361567636e-07,
+      "loss": 2.2782,
+      "step": 55943
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.501580278128951e-07,
+      "loss": 0.8293,
+      "step": 55944
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.5011061946902657e-07,
+      "loss": 2.4705,
+      "step": 55945
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.50063211125158e-07,
+      "loss": 1.7705,
+      "step": 55946
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.500158027812895e-07,
+      "loss": 2.3769,
+      "step": 55947
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.49968394437421e-07,
+      "loss": 1.8854,
+      "step": 55948
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.499209860935525e-07,
+      "loss": 1.9366,
+      "step": 55949
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.49873577749684e-07,
+      "loss": 3.3814,
+      "step": 55950
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.498261694058154e-07,
+      "loss": 2.0252,
+      "step": 55951
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.497787610619469e-07,
+      "loss": 1.9688,
+      "step": 55952
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.497313527180784e-07,
+      "loss": 1.9636,
+      "step": 55953
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.496839443742099e-07,
+      "loss": 3.1431,
+      "step": 55954
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4963653603034134e-07,
+      "loss": 2.0768,
+      "step": 55955
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4958912768647283e-07,
+      "loss": 1.5837,
+      "step": 55956
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4954171934260427e-07,
+      "loss": 1.0201,
+      "step": 55957
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4949431099873576e-07,
+      "loss": 2.1324,
+      "step": 55958
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.494469026548673e-07,
+      "loss": 0.9427,
+      "step": 55959
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4939949431099874e-07,
+      "loss": 0.4768,
+      "step": 55960
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4935208596713024e-07,
+      "loss": 2.4701,
+      "step": 55961
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.493046776232617e-07,
+      "loss": 2.7028,
+      "step": 55962
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4925726927939317e-07,
+      "loss": 1.7904,
+      "step": 55963
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4920986093552466e-07,
+      "loss": 2.7594,
+      "step": 55964
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4916245259165615e-07,
+      "loss": 2.6095,
+      "step": 55965
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.491150442477876e-07,
+      "loss": 1.2554,
+      "step": 55966
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.490676359039191e-07,
+      "loss": 1.7162,
+      "step": 55967
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.490202275600506e-07,
+      "loss": 0.8159,
+      "step": 55968
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4897281921618207e-07,
+      "loss": 2.413,
+      "step": 55969
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4892541087231356e-07,
+      "loss": 2.099,
+      "step": 55970
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.48878002528445e-07,
+      "loss": 1.2222,
+      "step": 55971
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.488305941845765e-07,
+      "loss": 1.4364,
+      "step": 55972
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.48783185840708e-07,
+      "loss": 0.664,
+      "step": 55973
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.487357774968395e-07,
+      "loss": 0.7028,
+      "step": 55974
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.486883691529709e-07,
+      "loss": 2.6338,
+      "step": 55975
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.486409608091024e-07,
+      "loss": 2.687,
+      "step": 55976
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.485935524652339e-07,
+      "loss": 1.3854,
+      "step": 55977
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.485461441213654e-07,
+      "loss": 2.7512,
+      "step": 55978
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4849873577749683e-07,
+      "loss": 1.4064,
+      "step": 55979
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.484513274336283e-07,
+      "loss": 0.8747,
+      "step": 55980
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.484039190897598e-07,
+      "loss": 1.2532,
+      "step": 55981
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.483565107458913e-07,
+      "loss": 1.0186,
+      "step": 55982
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.483091024020228e-07,
+      "loss": 4.0286,
+      "step": 55983
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4826169405815424e-07,
+      "loss": 0.3154,
+      "step": 55984
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4821428571428573e-07,
+      "loss": 1.768,
+      "step": 55985
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4816687737041717e-07,
+      "loss": 1.6795,
+      "step": 55986
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.481194690265487e-07,
+      "loss": 2.2512,
+      "step": 55987
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4807206068268015e-07,
+      "loss": 0.6288,
+      "step": 55988
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4802465233881165e-07,
+      "loss": 0.4156,
+      "step": 55989
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.479772439949431e-07,
+      "loss": 1.2226,
+      "step": 55990
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.479298356510746e-07,
+      "loss": 1.4534,
+      "step": 55991
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.478824273072061e-07,
+      "loss": 1.5075,
+      "step": 55992
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4783501896333756e-07,
+      "loss": 1.6913,
+      "step": 55993
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4778761061946905e-07,
+      "loss": 1.9955,
+      "step": 55994
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.477402022756005e-07,
+      "loss": 2.0238,
+      "step": 55995
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.47692793931732e-07,
+      "loss": 2.0636,
+      "step": 55996
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.476453855878635e-07,
+      "loss": 1.5988,
+      "step": 55997
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4759797724399497e-07,
+      "loss": 1.6212,
+      "step": 55998
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.475505689001264e-07,
+      "loss": 1.5385,
+      "step": 55999
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.475031605562579e-07,
+      "loss": 2.6481,
+      "step": 56000
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.474557522123894e-07,
+      "loss": 1.4014,
+      "step": 56001
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.474083438685209e-07,
+      "loss": 1.6769,
+      "step": 56002
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.473609355246524e-07,
+      "loss": 2.6535,
+      "step": 56003
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.473135271807838e-07,
+      "loss": 1.1986,
+      "step": 56004
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.472661188369153e-07,
+      "loss": 1.9291,
+      "step": 56005
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.472187104930468e-07,
+      "loss": 2.6373,
+      "step": 56006
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.471713021491783e-07,
+      "loss": 2.6469,
+      "step": 56007
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4712389380530973e-07,
+      "loss": 1.5134,
+      "step": 56008
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.470764854614412e-07,
+      "loss": 1.4475,
+      "step": 56009
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4702907711757266e-07,
+      "loss": 2.2865,
+      "step": 56010
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.469816687737042e-07,
+      "loss": 3.5625,
+      "step": 56011
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4693426042983565e-07,
+      "loss": 1.5913,
+      "step": 56012
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4688685208596714e-07,
+      "loss": 2.7559,
+      "step": 56013
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4683944374209863e-07,
+      "loss": 1.8105,
+      "step": 56014
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.4679203539823007e-07,
+      "loss": 0.9129,
+      "step": 56015
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.467446270543616e-07,
+      "loss": 2.6018,
+      "step": 56016
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4669721871049306e-07,
+      "loss": 1.9983,
+      "step": 56017
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4664981036662455e-07,
+      "loss": 1.213,
+      "step": 56018
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.46602402022756e-07,
+      "loss": 0.7619,
+      "step": 56019
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.465549936788875e-07,
+      "loss": 1.1668,
+      "step": 56020
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4650758533501897e-07,
+      "loss": 1.458,
+      "step": 56021
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4646017699115046e-07,
+      "loss": 4.9724,
+      "step": 56022
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.464127686472819e-07,
+      "loss": 1.2545,
+      "step": 56023
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.463653603034134e-07,
+      "loss": 1.8125,
+      "step": 56024
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.463179519595449e-07,
+      "loss": 3.1103,
+      "step": 56025
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.462705436156764e-07,
+      "loss": 1.4859,
+      "step": 56026
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4622313527180787e-07,
+      "loss": 1.1608,
+      "step": 56027
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.461757269279393e-07,
+      "loss": 2.8141,
+      "step": 56028
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.461283185840708e-07,
+      "loss": 1.3197,
+      "step": 56029
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.460809102402023e-07,
+      "loss": 2.7114,
+      "step": 56030
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.460335018963338e-07,
+      "loss": 1.3036,
+      "step": 56031
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4598609355246523e-07,
+      "loss": 1.9563,
+      "step": 56032
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.459386852085967e-07,
+      "loss": 1.4803,
+      "step": 56033
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4589127686472816e-07,
+      "loss": 1.7368,
+      "step": 56034
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.458438685208597e-07,
+      "loss": 0.7492,
+      "step": 56035
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4579646017699114e-07,
+      "loss": 1.9936,
+      "step": 56036
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4574905183312264e-07,
+      "loss": 2.1258,
+      "step": 56037
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4570164348925413e-07,
+      "loss": 0.7311,
+      "step": 56038
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4565423514538557e-07,
+      "loss": 2.6797,
+      "step": 56039
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.456068268015171e-07,
+      "loss": 3.188,
+      "step": 56040
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4555941845764855e-07,
+      "loss": 1.1299,
+      "step": 56041
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4551201011378004e-07,
+      "loss": 1.8812,
+      "step": 56042
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.454646017699115e-07,
+      "loss": 2.9771,
+      "step": 56043
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.45417193426043e-07,
+      "loss": 3.5586,
+      "step": 56044
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4536978508217447e-07,
+      "loss": 1.935,
+      "step": 56045
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4532237673830596e-07,
+      "loss": 1.0568,
+      "step": 56046
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.452749683944374e-07,
+      "loss": 2.3983,
+      "step": 56047
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.452275600505689e-07,
+      "loss": 1.1013,
+      "step": 56048
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.451801517067004e-07,
+      "loss": 1.0414,
+      "step": 56049
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.451327433628319e-07,
+      "loss": 1.6025,
+      "step": 56050
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4508533501896337e-07,
+      "loss": 4.4324,
+      "step": 56051
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.450379266750948e-07,
+      "loss": 0.8823,
+      "step": 56052
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.449905183312263e-07,
+      "loss": 0.9362,
+      "step": 56053
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.449431099873578e-07,
+      "loss": 3.0155,
+      "step": 56054
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.448957016434893e-07,
+      "loss": 2.3559,
+      "step": 56055
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.448482932996207e-07,
+      "loss": 1.5332,
+      "step": 56056
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.448008849557522e-07,
+      "loss": 2.9458,
+      "step": 56057
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.447534766118837e-07,
+      "loss": 2.4058,
+      "step": 56058
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.447060682680152e-07,
+      "loss": 2.1727,
+      "step": 56059
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.446586599241467e-07,
+      "loss": 2.1205,
+      "step": 56060
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4461125158027813e-07,
+      "loss": 1.735,
+      "step": 56061
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.445638432364096e-07,
+      "loss": 2.1469,
+      "step": 56062
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4451643489254106e-07,
+      "loss": 0.9855,
+      "step": 56063
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.444690265486726e-07,
+      "loss": 1.937,
+      "step": 56064
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4442161820480405e-07,
+      "loss": 3.2009,
+      "step": 56065
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4437420986093554e-07,
+      "loss": 2.2173,
+      "step": 56066
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.44326801517067e-07,
+      "loss": 1.2896,
+      "step": 56067
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4427939317319847e-07,
+      "loss": 2.7152,
+      "step": 56068
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4423198482932996e-07,
+      "loss": 2.1994,
+      "step": 56069
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4418457648546145e-07,
+      "loss": 2.1465,
+      "step": 56070
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4413716814159295e-07,
+      "loss": 1.604,
+      "step": 56071
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.440897597977244e-07,
+      "loss": 2.1102,
+      "step": 56072
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.440423514538559e-07,
+      "loss": 1.5511,
+      "step": 56073
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4399494310998737e-07,
+      "loss": 1.9509,
+      "step": 56074
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.4394753476611886e-07,
+      "loss": 2.3415,
+      "step": 56075
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.439001264222503e-07,
+      "loss": 0.4299,
+      "step": 56076
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.438527180783818e-07,
+      "loss": 2.0475,
+      "step": 56077
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.438053097345133e-07,
+      "loss": 2.3293,
+      "step": 56078
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.437579013906448e-07,
+      "loss": 0.988,
+      "step": 56079
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.437104930467762e-07,
+      "loss": 1.6253,
+      "step": 56080
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.436630847029077e-07,
+      "loss": 2.9965,
+      "step": 56081
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.436156763590392e-07,
+      "loss": 2.8129,
+      "step": 56082
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.435682680151707e-07,
+      "loss": 1.9446,
+      "step": 56083
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.435208596713022e-07,
+      "loss": 2.4473,
+      "step": 56084
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.434734513274336e-07,
+      "loss": 1.6336,
+      "step": 56085
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.434260429835651e-07,
+      "loss": 2.2451,
+      "step": 56086
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.433786346396966e-07,
+      "loss": 0.7657,
+      "step": 56087
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.433312262958281e-07,
+      "loss": 1.8518,
+      "step": 56088
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4328381795195954e-07,
+      "loss": 2.1793,
+      "step": 56089
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4323640960809103e-07,
+      "loss": 2.0967,
+      "step": 56090
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4318900126422247e-07,
+      "loss": 4.176,
+      "step": 56091
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.43141592920354e-07,
+      "loss": 0.8303,
+      "step": 56092
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.430941845764855e-07,
+      "loss": 2.2353,
+      "step": 56093
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4304677623261695e-07,
+      "loss": 1.5944,
+      "step": 56094
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4299936788874844e-07,
+      "loss": 1.5227,
+      "step": 56095
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.429519595448799e-07,
+      "loss": 1.8457,
+      "step": 56096
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.429045512010114e-07,
+      "loss": 2.022,
+      "step": 56097
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4285714285714286e-07,
+      "loss": 1.2729,
+      "step": 56098
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4280973451327436e-07,
+      "loss": 1.6901,
+      "step": 56099
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.427623261694058e-07,
+      "loss": 0.3676,
+      "step": 56100
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.427149178255373e-07,
+      "loss": 2.4174,
+      "step": 56101
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.426675094816688e-07,
+      "loss": 1.2841,
+      "step": 56102
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4262010113780027e-07,
+      "loss": 1.4767,
+      "step": 56103
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4257269279393176e-07,
+      "loss": 1.1619,
+      "step": 56104
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.425252844500632e-07,
+      "loss": 2.0427,
+      "step": 56105
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.424778761061947e-07,
+      "loss": 2.4272,
+      "step": 56106
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.424304677623262e-07,
+      "loss": 1.9336,
+      "step": 56107
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.423830594184577e-07,
+      "loss": 1.1081,
+      "step": 56108
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.423356510745891e-07,
+      "loss": 3.7663,
+      "step": 56109
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.422882427307206e-07,
+      "loss": 1.2074,
+      "step": 56110
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.422408343868521e-07,
+      "loss": 1.301,
+      "step": 56111
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.421934260429836e-07,
+      "loss": 1.6077,
+      "step": 56112
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4214601769911504e-07,
+      "loss": 0.5668,
+      "step": 56113
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4209860935524653e-07,
+      "loss": 2.6481,
+      "step": 56114
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.42051201011378e-07,
+      "loss": 0.62,
+      "step": 56115
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.420037926675095e-07,
+      "loss": 1.2935,
+      "step": 56116
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.41956384323641e-07,
+      "loss": 3.7716,
+      "step": 56117
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4190897597977244e-07,
+      "loss": 2.04,
+      "step": 56118
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4186156763590394e-07,
+      "loss": 1.3669,
+      "step": 56119
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.418141592920354e-07,
+      "loss": 1.0495,
+      "step": 56120
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.417667509481669e-07,
+      "loss": 1.8085,
+      "step": 56121
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4171934260429836e-07,
+      "loss": 1.5711,
+      "step": 56122
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4167193426042985e-07,
+      "loss": 1.9069,
+      "step": 56123
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.416245259165613e-07,
+      "loss": 1.0214,
+      "step": 56124
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.415771175726928e-07,
+      "loss": 1.284,
+      "step": 56125
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4152970922882433e-07,
+      "loss": 1.4665,
+      "step": 56126
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4148230088495577e-07,
+      "loss": 1.7126,
+      "step": 56127
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4143489254108726e-07,
+      "loss": 1.9493,
+      "step": 56128
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.413874841972187e-07,
+      "loss": 2.1828,
+      "step": 56129
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.413400758533502e-07,
+      "loss": 1.5683,
+      "step": 56130
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.412926675094817e-07,
+      "loss": 1.9883,
+      "step": 56131
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.412452591656132e-07,
+      "loss": 1.6938,
+      "step": 56132
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.411978508217446e-07,
+      "loss": 1.4992,
+      "step": 56133
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.411504424778761e-07,
+      "loss": 2.8045,
+      "step": 56134
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.411030341340076e-07,
+      "loss": 1.6932,
+      "step": 56135
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.410556257901391e-07,
+      "loss": 2.2632,
+      "step": 56136
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4100821744627053e-07,
+      "loss": 1.3047,
+      "step": 56137
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.40960809102402e-07,
+      "loss": 2.5206,
+      "step": 56138
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.409134007585335e-07,
+      "loss": 2.6332,
+      "step": 56139
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.40865992414665e-07,
+      "loss": 1.6663,
+      "step": 56140
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.408185840707965e-07,
+      "loss": 1.2892,
+      "step": 56141
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.4077117572692794e-07,
+      "loss": 2.7519,
+      "step": 56142
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4072376738305943e-07,
+      "loss": 1.986,
+      "step": 56143
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4067635903919087e-07,
+      "loss": 1.9928,
+      "step": 56144
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.406289506953224e-07,
+      "loss": 2.1001,
+      "step": 56145
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4058154235145385e-07,
+      "loss": 2.2268,
+      "step": 56146
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4053413400758535e-07,
+      "loss": 2.3701,
+      "step": 56147
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.404867256637168e-07,
+      "loss": 0.8802,
+      "step": 56148
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.404393173198483e-07,
+      "loss": 2.4636,
+      "step": 56149
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.403919089759798e-07,
+      "loss": 1.7382,
+      "step": 56150
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4034450063211126e-07,
+      "loss": 1.2598,
+      "step": 56151
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4029709228824275e-07,
+      "loss": 0.9302,
+      "step": 56152
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.402496839443742e-07,
+      "loss": 2.1377,
+      "step": 56153
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.402022756005057e-07,
+      "loss": 2.297,
+      "step": 56154
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.401548672566372e-07,
+      "loss": 2.8856,
+      "step": 56155
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.4010745891276867e-07,
+      "loss": 2.6872,
+      "step": 56156
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.400600505689001e-07,
+      "loss": 1.7618,
+      "step": 56157
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.400126422250316e-07,
+      "loss": 1.8316,
+      "step": 56158
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.399652338811631e-07,
+      "loss": 2.976,
+      "step": 56159
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.399178255372946e-07,
+      "loss": 1.1257,
+      "step": 56160
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.398704171934261e-07,
+      "loss": 2.1768,
+      "step": 56161
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.398230088495575e-07,
+      "loss": 1.705,
+      "step": 56162
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.39775600505689e-07,
+      "loss": 2.4901,
+      "step": 56163
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.397281921618205e-07,
+      "loss": 2.4141,
+      "step": 56164
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.39680783817952e-07,
+      "loss": 2.6315,
+      "step": 56165
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3963337547408343e-07,
+      "loss": 1.1401,
+      "step": 56166
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.395859671302149e-07,
+      "loss": 2.833,
+      "step": 56167
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3953855878634636e-07,
+      "loss": 3.06,
+      "step": 56168
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.394911504424779e-07,
+      "loss": 0.762,
+      "step": 56169
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3944374209860935e-07,
+      "loss": 2.627,
+      "step": 56170
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3939633375474084e-07,
+      "loss": 2.0833,
+      "step": 56171
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3934892541087233e-07,
+      "loss": 0.3864,
+      "step": 56172
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3930151706700377e-07,
+      "loss": 2.6013,
+      "step": 56173
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.392541087231353e-07,
+      "loss": 1.4358,
+      "step": 56174
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3920670037926676e-07,
+      "loss": 1.7035,
+      "step": 56175
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3915929203539825e-07,
+      "loss": 1.3295,
+      "step": 56176
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.391118836915297e-07,
+      "loss": 2.3926,
+      "step": 56177
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.390644753476612e-07,
+      "loss": 2.1731,
+      "step": 56178
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3901706700379267e-07,
+      "loss": 2.9555,
+      "step": 56179
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3896965865992416e-07,
+      "loss": 0.6314,
+      "step": 56180
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.389222503160556e-07,
+      "loss": 2.6841,
+      "step": 56181
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.388748419721871e-07,
+      "loss": 2.0774,
+      "step": 56182
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.388274336283186e-07,
+      "loss": 0.7237,
+      "step": 56183
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.387800252844501e-07,
+      "loss": 1.0565,
+      "step": 56184
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3873261694058157e-07,
+      "loss": 1.9532,
+      "step": 56185
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.38685208596713e-07,
+      "loss": 1.3411,
+      "step": 56186
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.386378002528445e-07,
+      "loss": 1.0044,
+      "step": 56187
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.38590391908976e-07,
+      "loss": 0.9224,
+      "step": 56188
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.385429835651075e-07,
+      "loss": 2.4046,
+      "step": 56189
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3849557522123893e-07,
+      "loss": 1.5746,
+      "step": 56190
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.384481668773704e-07,
+      "loss": 1.717,
+      "step": 56191
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.384007585335019e-07,
+      "loss": 1.2785,
+      "step": 56192
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.383533501896334e-07,
+      "loss": 2.0482,
+      "step": 56193
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.383059418457649e-07,
+      "loss": 1.9268,
+      "step": 56194
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3825853350189633e-07,
+      "loss": 1.8414,
+      "step": 56195
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3821112515802783e-07,
+      "loss": 2.909,
+      "step": 56196
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.381637168141593e-07,
+      "loss": 2.0292,
+      "step": 56197
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.381163084702908e-07,
+      "loss": 2.4504,
+      "step": 56198
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3806890012642225e-07,
+      "loss": 1.9769,
+      "step": 56199
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3802149178255374e-07,
+      "loss": 0.403,
+      "step": 56200
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.379740834386852e-07,
+      "loss": 1.4646,
+      "step": 56201
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.379266750948167e-07,
+      "loss": 2.4756,
+      "step": 56202
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3787926675094817e-07,
+      "loss": 1.9823,
+      "step": 56203
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3783185840707966e-07,
+      "loss": 1.4169,
+      "step": 56204
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.3778445006321115e-07,
+      "loss": 1.0895,
+      "step": 56205
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.377370417193426e-07,
+      "loss": 1.4467,
+      "step": 56206
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.376896333754741e-07,
+      "loss": 1.6891,
+      "step": 56207
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.376422250316056e-07,
+      "loss": 0.8045,
+      "step": 56208
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3759481668773707e-07,
+      "loss": 0.8842,
+      "step": 56209
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.375474083438685e-07,
+      "loss": 2.946,
+      "step": 56210
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.375e-07,
+      "loss": 1.393,
+      "step": 56211
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.374525916561315e-07,
+      "loss": 3.4172,
+      "step": 56212
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.37405183312263e-07,
+      "loss": 0.7655,
+      "step": 56213
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.373577749683944e-07,
+      "loss": 1.3206,
+      "step": 56214
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.373103666245259e-07,
+      "loss": 1.5673,
+      "step": 56215
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.372629582806574e-07,
+      "loss": 2.4024,
+      "step": 56216
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.372155499367889e-07,
+      "loss": 1.4517,
+      "step": 56217
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.371681415929204e-07,
+      "loss": 1.9067,
+      "step": 56218
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3712073324905183e-07,
+      "loss": 3.1523,
+      "step": 56219
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.370733249051833e-07,
+      "loss": 1.9049,
+      "step": 56220
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.370259165613148e-07,
+      "loss": 1.6513,
+      "step": 56221
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.369785082174463e-07,
+      "loss": 2.3968,
+      "step": 56222
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3693109987357775e-07,
+      "loss": 1.9254,
+      "step": 56223
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3688369152970924e-07,
+      "loss": 0.8504,
+      "step": 56224
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.368362831858407e-07,
+      "loss": 1.5403,
+      "step": 56225
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.367888748419722e-07,
+      "loss": 2.298,
+      "step": 56226
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.367414664981037e-07,
+      "loss": 2.4544,
+      "step": 56227
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3669405815423515e-07,
+      "loss": 2.5248,
+      "step": 56228
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3664664981036665e-07,
+      "loss": 1.2297,
+      "step": 56229
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.365992414664981e-07,
+      "loss": 1.9369,
+      "step": 56230
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3655183312262963e-07,
+      "loss": 1.5505,
+      "step": 56231
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3650442477876107e-07,
+      "loss": 1.0589,
+      "step": 56232
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3645701643489256e-07,
+      "loss": 2.3123,
+      "step": 56233
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.36409608091024e-07,
+      "loss": 3.114,
+      "step": 56234
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.363621997471555e-07,
+      "loss": 2.0176,
+      "step": 56235
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.36314791403287e-07,
+      "loss": 2.0208,
+      "step": 56236
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.362673830594185e-07,
+      "loss": 2.122,
+      "step": 56237
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3621997471554997e-07,
+      "loss": 1.5801,
+      "step": 56238
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.361725663716814e-07,
+      "loss": 1.5831,
+      "step": 56239
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.361251580278129e-07,
+      "loss": 1.8287,
+      "step": 56240
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.360777496839444e-07,
+      "loss": 1.4716,
+      "step": 56241
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.360303413400759e-07,
+      "loss": 3.1719,
+      "step": 56242
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.359829329962073e-07,
+      "loss": 0.9058,
+      "step": 56243
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.359355246523388e-07,
+      "loss": 0.913,
+      "step": 56244
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.358881163084703e-07,
+      "loss": 1.9363,
+      "step": 56245
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.358407079646018e-07,
+      "loss": 0.65,
+      "step": 56246
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3579329962073324e-07,
+      "loss": 2.0218,
+      "step": 56247
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3574589127686473e-07,
+      "loss": 0.7827,
+      "step": 56248
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3569848293299617e-07,
+      "loss": 2.0623,
+      "step": 56249
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.356510745891277e-07,
+      "loss": 1.5247,
+      "step": 56250
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.356036662452592e-07,
+      "loss": 1.7611,
+      "step": 56251
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3555625790139065e-07,
+      "loss": 1.8084,
+      "step": 56252
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3550884955752214e-07,
+      "loss": 2.5555,
+      "step": 56253
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.354614412136536e-07,
+      "loss": 1.3762,
+      "step": 56254
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.354140328697851e-07,
+      "loss": 1.5605,
+      "step": 56255
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3536662452591656e-07,
+      "loss": 1.7613,
+      "step": 56256
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3531921618204806e-07,
+      "loss": 2.4249,
+      "step": 56257
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.352718078381795e-07,
+      "loss": 2.0642,
+      "step": 56258
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.35224399494311e-07,
+      "loss": 2.2673,
+      "step": 56259
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.351769911504425e-07,
+      "loss": 2.2794,
+      "step": 56260
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3512958280657397e-07,
+      "loss": 4.0468,
+      "step": 56261
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.3508217446270546e-07,
+      "loss": 2.5465,
+      "step": 56262
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.350347661188369e-07,
+      "loss": 1.1374,
+      "step": 56263
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.349873577749684e-07,
+      "loss": 1.2745,
+      "step": 56264
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.349399494310999e-07,
+      "loss": 1.9317,
+      "step": 56265
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.348925410872314e-07,
+      "loss": 0.4851,
+      "step": 56266
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.348451327433628e-07,
+      "loss": 2.0395,
+      "step": 56267
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.347977243994943e-07,
+      "loss": 1.1592,
+      "step": 56268
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.347503160556258e-07,
+      "loss": 1.0781,
+      "step": 56269
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.347029077117573e-07,
+      "loss": 3.3782,
+      "step": 56270
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3465549936788873e-07,
+      "loss": 2.1303,
+      "step": 56271
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3460809102402023e-07,
+      "loss": 1.0107,
+      "step": 56272
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.345606826801517e-07,
+      "loss": 1.9935,
+      "step": 56273
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.345132743362832e-07,
+      "loss": 2.4765,
+      "step": 56274
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.344658659924147e-07,
+      "loss": 1.5667,
+      "step": 56275
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3441845764854614e-07,
+      "loss": 1.409,
+      "step": 56276
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3437104930467763e-07,
+      "loss": 1.6945,
+      "step": 56277
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.343236409608091e-07,
+      "loss": 3.4481,
+      "step": 56278
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.342762326169406e-07,
+      "loss": 1.7682,
+      "step": 56279
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3422882427307206e-07,
+      "loss": 1.3556,
+      "step": 56280
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3418141592920355e-07,
+      "loss": 1.9389,
+      "step": 56281
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.34134007585335e-07,
+      "loss": 1.2393,
+      "step": 56282
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.340865992414665e-07,
+      "loss": 0.6175,
+      "step": 56283
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3403919089759803e-07,
+      "loss": 0.9821,
+      "step": 56284
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3399178255372947e-07,
+      "loss": 1.529,
+      "step": 56285
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3394437420986096e-07,
+      "loss": 1.0305,
+      "step": 56286
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.338969658659924e-07,
+      "loss": 1.0836,
+      "step": 56287
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.338495575221239e-07,
+      "loss": 2.7832,
+      "step": 56288
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.338021491782554e-07,
+      "loss": 2.4335,
+      "step": 56289
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.337547408343869e-07,
+      "loss": 1.9848,
+      "step": 56290
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.337073324905183e-07,
+      "loss": 2.1473,
+      "step": 56291
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.336599241466498e-07,
+      "loss": 1.52,
+      "step": 56292
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.336125158027813e-07,
+      "loss": 2.9,
+      "step": 56293
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.335651074589128e-07,
+      "loss": 2.8472,
+      "step": 56294
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.335176991150443e-07,
+      "loss": 1.6288,
+      "step": 56295
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.334702907711757e-07,
+      "loss": 2.3309,
+      "step": 56296
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.334228824273072e-07,
+      "loss": 3.2971,
+      "step": 56297
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.333754740834387e-07,
+      "loss": 3.2027,
+      "step": 56298
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.333280657395702e-07,
+      "loss": 2.0981,
+      "step": 56299
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3328065739570164e-07,
+      "loss": 1.2985,
+      "step": 56300
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3323324905183313e-07,
+      "loss": 1.3704,
+      "step": 56301
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.331858407079646e-07,
+      "loss": 3.5346,
+      "step": 56302
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.331384323640961e-07,
+      "loss": 0.648,
+      "step": 56303
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3309102402022755e-07,
+      "loss": 2.0699,
+      "step": 56304
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3304361567635904e-07,
+      "loss": 2.1,
+      "step": 56305
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3299620733249054e-07,
+      "loss": 2.2402,
+      "step": 56306
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.32948798988622e-07,
+      "loss": 2.2816,
+      "step": 56307
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.329013906447535e-07,
+      "loss": 1.2338,
+      "step": 56308
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3285398230088496e-07,
+      "loss": 1.1303,
+      "step": 56309
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3280657395701645e-07,
+      "loss": 2.5279,
+      "step": 56310
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.327591656131479e-07,
+      "loss": 1.3143,
+      "step": 56311
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.327117572692794e-07,
+      "loss": 1.9633,
+      "step": 56312
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.326643489254109e-07,
+      "loss": 1.9281,
+      "step": 56313
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3261694058154237e-07,
+      "loss": 2.5118,
+      "step": 56314
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.325695322376738e-07,
+      "loss": 2.5129,
+      "step": 56315
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.325221238938053e-07,
+      "loss": 2.0355,
+      "step": 56316
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.324747155499368e-07,
+      "loss": 2.188,
+      "step": 56317
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.324273072060683e-07,
+      "loss": 2.4763,
+      "step": 56318
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.323798988621998e-07,
+      "loss": 1.2371,
+      "step": 56319
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.323324905183312e-07,
+      "loss": 2.721,
+      "step": 56320
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.322850821744627e-07,
+      "loss": 2.6415,
+      "step": 56321
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.322376738305942e-07,
+      "loss": 2.6908,
+      "step": 56322
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.321902654867257e-07,
+      "loss": 3.0662,
+      "step": 56323
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3214285714285713e-07,
+      "loss": 2.1953,
+      "step": 56324
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.320954487989886e-07,
+      "loss": 3.2771,
+      "step": 56325
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.320480404551201e-07,
+      "loss": 3.3061,
+      "step": 56326
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.320006321112516e-07,
+      "loss": 1.4224,
+      "step": 56327
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.319532237673831e-07,
+      "loss": 1.8883,
+      "step": 56328
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3190581542351454e-07,
+      "loss": 1.3206,
+      "step": 56329
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.3185840707964603e-07,
+      "loss": 2.6459,
+      "step": 56330
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.318109987357775e-07,
+      "loss": 1.6901,
+      "step": 56331
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.31763590391909e-07,
+      "loss": 1.845,
+      "step": 56332
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3171618204804046e-07,
+      "loss": 0.5151,
+      "step": 56333
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3166877370417195e-07,
+      "loss": 1.8888,
+      "step": 56334
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.316213653603034e-07,
+      "loss": 1.3996,
+      "step": 56335
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3157395701643493e-07,
+      "loss": 2.2283,
+      "step": 56336
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3152654867256637e-07,
+      "loss": 2.3277,
+      "step": 56337
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3147914032869786e-07,
+      "loss": 1.6872,
+      "step": 56338
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3143173198482936e-07,
+      "loss": 2.3502,
+      "step": 56339
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.313843236409608e-07,
+      "loss": 1.8412,
+      "step": 56340
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.313369152970923e-07,
+      "loss": 2.3582,
+      "step": 56341
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.312895069532238e-07,
+      "loss": 1.4898,
+      "step": 56342
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3124209860935527e-07,
+      "loss": 1.7976,
+      "step": 56343
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.311946902654867e-07,
+      "loss": 1.9203,
+      "step": 56344
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.311472819216182e-07,
+      "loss": 1.5404,
+      "step": 56345
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.310998735777497e-07,
+      "loss": 0.9237,
+      "step": 56346
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.310524652338812e-07,
+      "loss": 1.5904,
+      "step": 56347
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.310050568900126e-07,
+      "loss": 1.9324,
+      "step": 56348
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.309576485461441e-07,
+      "loss": 2.3057,
+      "step": 56349
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.309102402022756e-07,
+      "loss": 4.0915,
+      "step": 56350
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.308628318584071e-07,
+      "loss": 2.3502,
+      "step": 56351
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.308154235145386e-07,
+      "loss": 1.8082,
+      "step": 56352
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3076801517067003e-07,
+      "loss": 2.8344,
+      "step": 56353
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.307206068268015e-07,
+      "loss": 2.3907,
+      "step": 56354
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.30673198482933e-07,
+      "loss": 1.8284,
+      "step": 56355
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.306257901390645e-07,
+      "loss": 1.1808,
+      "step": 56356
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3057838179519595e-07,
+      "loss": 2.4555,
+      "step": 56357
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3053097345132744e-07,
+      "loss": 1.6218,
+      "step": 56358
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.304835651074589e-07,
+      "loss": 2.8623,
+      "step": 56359
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.304361567635904e-07,
+      "loss": 2.0232,
+      "step": 56360
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3038874841972187e-07,
+      "loss": 1.5479,
+      "step": 56361
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3034134007585336e-07,
+      "loss": 1.0108,
+      "step": 56362
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3029393173198485e-07,
+      "loss": 1.2458,
+      "step": 56363
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.302465233881163e-07,
+      "loss": 1.177,
+      "step": 56364
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3019911504424783e-07,
+      "loss": 1.3166,
+      "step": 56365
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.301517067003793e-07,
+      "loss": 2.0386,
+      "step": 56366
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.3010429835651077e-07,
+      "loss": 2.9005,
+      "step": 56367
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.300568900126422e-07,
+      "loss": 2.4876,
+      "step": 56368
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.300094816687737e-07,
+      "loss": 0.8833,
+      "step": 56369
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.299620733249052e-07,
+      "loss": 2.5527,
+      "step": 56370
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.299146649810367e-07,
+      "loss": 1.2706,
+      "step": 56371
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.298672566371681e-07,
+      "loss": 0.7762,
+      "step": 56372
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.298198482932996e-07,
+      "loss": 2.7761,
+      "step": 56373
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.297724399494311e-07,
+      "loss": 0.6851,
+      "step": 56374
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.297250316055626e-07,
+      "loss": 0.6784,
+      "step": 56375
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.296776232616941e-07,
+      "loss": 2.0636,
+      "step": 56376
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2963021491782553e-07,
+      "loss": 2.0542,
+      "step": 56377
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.29582806573957e-07,
+      "loss": 1.5004,
+      "step": 56378
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.295353982300885e-07,
+      "loss": 1.0285,
+      "step": 56379
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2948798988622e-07,
+      "loss": 1.2272,
+      "step": 56380
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2944058154235144e-07,
+      "loss": 2.3181,
+      "step": 56381
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2939317319848294e-07,
+      "loss": 1.599,
+      "step": 56382
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.293457648546144e-07,
+      "loss": 1.3375,
+      "step": 56383
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.292983565107459e-07,
+      "loss": 3.3945,
+      "step": 56384
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.292509481668774e-07,
+      "loss": 0.755,
+      "step": 56385
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2920353982300885e-07,
+      "loss": 2.7014,
+      "step": 56386
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2915613147914034e-07,
+      "loss": 2.465,
+      "step": 56387
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.291087231352718e-07,
+      "loss": 2.5543,
+      "step": 56388
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2906131479140333e-07,
+      "loss": 1.7983,
+      "step": 56389
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2901390644753477e-07,
+      "loss": 2.823,
+      "step": 56390
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.2896649810366626e-07,
+      "loss": 1.5675,
+      "step": 56391
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.289190897597977e-07,
+      "loss": 0.6575,
+      "step": 56392
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.288716814159292e-07,
+      "loss": 1.9992,
+      "step": 56393
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.288242730720607e-07,
+      "loss": 1.9735,
+      "step": 56394
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.287768647281922e-07,
+      "loss": 0.9056,
+      "step": 56395
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2872945638432367e-07,
+      "loss": 2.4163,
+      "step": 56396
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.286820480404551e-07,
+      "loss": 1.161,
+      "step": 56397
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.286346396965866e-07,
+      "loss": 1.9327,
+      "step": 56398
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.285872313527181e-07,
+      "loss": 2.1333,
+      "step": 56399
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.285398230088496e-07,
+      "loss": 1.1341,
+      "step": 56400
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.28492414664981e-07,
+      "loss": 2.9213,
+      "step": 56401
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.284450063211125e-07,
+      "loss": 1.5246,
+      "step": 56402
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.28397597977244e-07,
+      "loss": 1.1103,
+      "step": 56403
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.283501896333755e-07,
+      "loss": 1.1011,
+      "step": 56404
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2830278128950694e-07,
+      "loss": 2.3162,
+      "step": 56405
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2825537294563843e-07,
+      "loss": 2.0449,
+      "step": 56406
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.282079646017699e-07,
+      "loss": 0.6299,
+      "step": 56407
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.281605562579014e-07,
+      "loss": 1.2736,
+      "step": 56408
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.281131479140329e-07,
+      "loss": 1.4949,
+      "step": 56409
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2806573957016435e-07,
+      "loss": 2.8561,
+      "step": 56410
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2801833122629584e-07,
+      "loss": 1.384,
+      "step": 56411
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.279709228824273e-07,
+      "loss": 1.997,
+      "step": 56412
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.279235145385588e-07,
+      "loss": 2.3401,
+      "step": 56413
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2787610619469026e-07,
+      "loss": 1.7419,
+      "step": 56414
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2782869785082176e-07,
+      "loss": 2.1897,
+      "step": 56415
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.277812895069532e-07,
+      "loss": 1.83,
+      "step": 56416
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.277338811630847e-07,
+      "loss": 2.9796,
+      "step": 56417
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2768647281921623e-07,
+      "loss": 2.1991,
+      "step": 56418
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2763906447534767e-07,
+      "loss": 1.4008,
+      "step": 56419
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2759165613147916e-07,
+      "loss": 2.531,
+      "step": 56420
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.275442477876106e-07,
+      "loss": 2.3725,
+      "step": 56421
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.274968394437421e-07,
+      "loss": 1.8583,
+      "step": 56422
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.274494310998736e-07,
+      "loss": 0.6159,
+      "step": 56423
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.274020227560051e-07,
+      "loss": 0.7779,
+      "step": 56424
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.273546144121365e-07,
+      "loss": 2.085,
+      "step": 56425
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.27307206068268e-07,
+      "loss": 1.3862,
+      "step": 56426
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.272597977243995e-07,
+      "loss": 4.0634,
+      "step": 56427
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.27212389380531e-07,
+      "loss": 1.4175,
+      "step": 56428
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.271649810366625e-07,
+      "loss": 2.9001,
+      "step": 56429
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.271175726927939e-07,
+      "loss": 2.3175,
+      "step": 56430
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.270701643489254e-07,
+      "loss": 0.8559,
+      "step": 56431
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.270227560050569e-07,
+      "loss": 1.3118,
+      "step": 56432
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.269753476611884e-07,
+      "loss": 0.694,
+      "step": 56433
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2692793931731984e-07,
+      "loss": 2.0841,
+      "step": 56434
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2688053097345133e-07,
+      "loss": 1.6811,
+      "step": 56435
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.268331226295828e-07,
+      "loss": 0.927,
+      "step": 56436
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.267857142857143e-07,
+      "loss": 2.4179,
+      "step": 56437
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2673830594184576e-07,
+      "loss": 1.1077,
+      "step": 56438
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2669089759797725e-07,
+      "loss": 3.046,
+      "step": 56439
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2664348925410874e-07,
+      "loss": 1.3645,
+      "step": 56440
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2659608091024023e-07,
+      "loss": 1.0265,
+      "step": 56441
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.265486725663717e-07,
+      "loss": 0.7387,
+      "step": 56442
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2650126422250317e-07,
+      "loss": 3.8306,
+      "step": 56443
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2645385587863466e-07,
+      "loss": 2.3031,
+      "step": 56444
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.264064475347661e-07,
+      "loss": 2.7373,
+      "step": 56445
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.263590391908976e-07,
+      "loss": 2.4894,
+      "step": 56446
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.263116308470291e-07,
+      "loss": 1.8042,
+      "step": 56447
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2626422250316057e-07,
+      "loss": 1.6923,
+      "step": 56448
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.26216814159292e-07,
+      "loss": 1.6878,
+      "step": 56449
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.261694058154235e-07,
+      "loss": 3.1515,
+      "step": 56450
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.26121997471555e-07,
+      "loss": 1.5124,
+      "step": 56451
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.260745891276865e-07,
+      "loss": 0.6435,
+      "step": 56452
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.26027180783818e-07,
+      "loss": 2.3882,
+      "step": 56453
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.259797724399494e-07,
+      "loss": 0.9962,
+      "step": 56454
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.259323640960809e-07,
+      "loss": 1.9891,
+      "step": 56455
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.258849557522124e-07,
+      "loss": 0.4048,
+      "step": 56456
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.258375474083439e-07,
+      "loss": 2.8281,
+      "step": 56457
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.2579013906447534e-07,
+      "loss": 1.4331,
+      "step": 56458
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2574273072060683e-07,
+      "loss": 1.6436,
+      "step": 56459
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.256953223767383e-07,
+      "loss": 1.6053,
+      "step": 56460
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.256479140328698e-07,
+      "loss": 2.5081,
+      "step": 56461
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.256005056890013e-07,
+      "loss": 2.4076,
+      "step": 56462
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2555309734513274e-07,
+      "loss": 0.8445,
+      "step": 56463
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2550568900126424e-07,
+      "loss": 2.283,
+      "step": 56464
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2545828065739573e-07,
+      "loss": 1.6078,
+      "step": 56465
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.254108723135272e-07,
+      "loss": 2.492,
+      "step": 56466
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2536346396965866e-07,
+      "loss": 2.0189,
+      "step": 56467
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2531605562579015e-07,
+      "loss": 3.1329,
+      "step": 56468
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.252686472819216e-07,
+      "loss": 1.5779,
+      "step": 56469
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2522123893805314e-07,
+      "loss": 1.9031,
+      "step": 56470
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.251738305941846e-07,
+      "loss": 1.7659,
+      "step": 56471
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2512642225031607e-07,
+      "loss": 1.8873,
+      "step": 56472
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.250790139064475e-07,
+      "loss": 1.887,
+      "step": 56473
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.25031605562579e-07,
+      "loss": 0.4626,
+      "step": 56474
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2498419721871054e-07,
+      "loss": 2.9578,
+      "step": 56475
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.24936788874842e-07,
+      "loss": 1.8513,
+      "step": 56476
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.248893805309735e-07,
+      "loss": 1.0264,
+      "step": 56477
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.248419721871049e-07,
+      "loss": 2.6912,
+      "step": 56478
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.247945638432364e-07,
+      "loss": 3.1775,
+      "step": 56479
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.247471554993679e-07,
+      "loss": 1.3468,
+      "step": 56480
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.246997471554994e-07,
+      "loss": 1.915,
+      "step": 56481
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2465233881163083e-07,
+      "loss": 1.7638,
+      "step": 56482
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.246049304677623e-07,
+      "loss": 3.7617,
+      "step": 56483
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.245575221238938e-07,
+      "loss": 2.7396,
+      "step": 56484
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.245101137800253e-07,
+      "loss": 1.9429,
+      "step": 56485
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.244627054361568e-07,
+      "loss": 1.0208,
+      "step": 56486
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2441529709228824e-07,
+      "loss": 2.564,
+      "step": 56487
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2436788874841973e-07,
+      "loss": 1.0028,
+      "step": 56488
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.243204804045512e-07,
+      "loss": 1.2527,
+      "step": 56489
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.242730720606827e-07,
+      "loss": 1.7387,
+      "step": 56490
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2422566371681415e-07,
+      "loss": 3.8147,
+      "step": 56491
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2417825537294565e-07,
+      "loss": 1.7241,
+      "step": 56492
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.241308470290771e-07,
+      "loss": 0.8576,
+      "step": 56493
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2408343868520863e-07,
+      "loss": 1.1046,
+      "step": 56494
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2403603034134007e-07,
+      "loss": 1.6796,
+      "step": 56495
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2398862199747156e-07,
+      "loss": 3.8323,
+      "step": 56496
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2394121365360305e-07,
+      "loss": 0.9564,
+      "step": 56497
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.238938053097345e-07,
+      "loss": 0.9287,
+      "step": 56498
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2384639696586604e-07,
+      "loss": 1.9993,
+      "step": 56499
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.237989886219975e-07,
+      "loss": 1.9511,
+      "step": 56500
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2375158027812897e-07,
+      "loss": 1.9881,
+      "step": 56501
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.237041719342604e-07,
+      "loss": 2.0931,
+      "step": 56502
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.236567635903919e-07,
+      "loss": 2.7518,
+      "step": 56503
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.236093552465234e-07,
+      "loss": 2.7435,
+      "step": 56504
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.235619469026549e-07,
+      "loss": 1.1243,
+      "step": 56505
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.235145385587863e-07,
+      "loss": 2.7929,
+      "step": 56506
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.234671302149178e-07,
+      "loss": 1.9968,
+      "step": 56507
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.234197218710493e-07,
+      "loss": 1.6417,
+      "step": 56508
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.233723135271808e-07,
+      "loss": 1.4617,
+      "step": 56509
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.233249051833123e-07,
+      "loss": 1.1659,
+      "step": 56510
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2327749683944373e-07,
+      "loss": 3.4371,
+      "step": 56511
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.232300884955752e-07,
+      "loss": 1.227,
+      "step": 56512
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.231826801517067e-07,
+      "loss": 1.8494,
+      "step": 56513
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.231352718078382e-07,
+      "loss": 0.7813,
+      "step": 56514
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2308786346396965e-07,
+      "loss": 2.8877,
+      "step": 56515
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2304045512010114e-07,
+      "loss": 2.5437,
+      "step": 56516
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.229930467762326e-07,
+      "loss": 1.8284,
+      "step": 56517
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.229456384323641e-07,
+      "loss": 1.7648,
+      "step": 56518
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.228982300884956e-07,
+      "loss": 1.6991,
+      "step": 56519
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2285082174462706e-07,
+      "loss": 0.9124,
+      "step": 56520
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2280341340075855e-07,
+      "loss": 2.0076,
+      "step": 56521
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.2275600505689e-07,
+      "loss": 2.3749,
+      "step": 56522
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2270859671302153e-07,
+      "loss": 1.9482,
+      "step": 56523
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2266118836915297e-07,
+      "loss": 1.8898,
+      "step": 56524
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2261378002528447e-07,
+      "loss": 3.5097,
+      "step": 56525
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.225663716814159e-07,
+      "loss": 1.4785,
+      "step": 56526
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.225189633375474e-07,
+      "loss": 0.6874,
+      "step": 56527
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.224715549936789e-07,
+      "loss": 2.1334,
+      "step": 56528
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.224241466498104e-07,
+      "loss": 1.2825,
+      "step": 56529
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2237673830594187e-07,
+      "loss": 1.7535,
+      "step": 56530
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.223293299620733e-07,
+      "loss": 2.0238,
+      "step": 56531
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.222819216182048e-07,
+      "loss": 1.2179,
+      "step": 56532
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.222345132743363e-07,
+      "loss": 2.1287,
+      "step": 56533
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.221871049304678e-07,
+      "loss": 1.2218,
+      "step": 56534
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2213969658659923e-07,
+      "loss": 2.7023,
+      "step": 56535
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.220922882427307e-07,
+      "loss": 2.8896,
+      "step": 56536
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.220448798988622e-07,
+      "loss": 0.8863,
+      "step": 56537
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.219974715549937e-07,
+      "loss": 1.9833,
+      "step": 56538
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2195006321112514e-07,
+      "loss": 2.1344,
+      "step": 56539
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2190265486725664e-07,
+      "loss": 2.2275,
+      "step": 56540
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2185524652338813e-07,
+      "loss": 3.6929,
+      "step": 56541
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.218078381795196e-07,
+      "loss": 2.0754,
+      "step": 56542
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.217604298356511e-07,
+      "loss": 1.8307,
+      "step": 56543
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2171302149178255e-07,
+      "loss": 1.8594,
+      "step": 56544
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2166561314791404e-07,
+      "loss": 1.496,
+      "step": 56545
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2161820480404554e-07,
+      "loss": 3.4425,
+      "step": 56546
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2157079646017703e-07,
+      "loss": 2.3104,
+      "step": 56547
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2152338811630847e-07,
+      "loss": 2.0822,
+      "step": 56548
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2147597977243996e-07,
+      "loss": 2.7928,
+      "step": 56549
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.214285714285714e-07,
+      "loss": 1.4599,
+      "step": 56550
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.213811630847029e-07,
+      "loss": 3.1748,
+      "step": 56551
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2133375474083444e-07,
+      "loss": 2.1728,
+      "step": 56552
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.212863463969659e-07,
+      "loss": 1.5835,
+      "step": 56553
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2123893805309737e-07,
+      "loss": 0.8479,
+      "step": 56554
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.211915297092288e-07,
+      "loss": 0.9955,
+      "step": 56555
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.211441213653603e-07,
+      "loss": 1.8701,
+      "step": 56556
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.210967130214918e-07,
+      "loss": 2.1284,
+      "step": 56557
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.210493046776233e-07,
+      "loss": 1.2132,
+      "step": 56558
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.210018963337547e-07,
+      "loss": 3.5887,
+      "step": 56559
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.209544879898862e-07,
+      "loss": 3.7079,
+      "step": 56560
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.209070796460177e-07,
+      "loss": 0.3974,
+      "step": 56561
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.208596713021492e-07,
+      "loss": 2.1196,
+      "step": 56562
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.208122629582807e-07,
+      "loss": 2.7964,
+      "step": 56563
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2076485461441213e-07,
+      "loss": 4.9421,
+      "step": 56564
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.207174462705436e-07,
+      "loss": 3.1219,
+      "step": 56565
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.206700379266751e-07,
+      "loss": 0.8942,
+      "step": 56566
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.206226295828066e-07,
+      "loss": 1.2953,
+      "step": 56567
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2057522123893805e-07,
+      "loss": 1.1443,
+      "step": 56568
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2052781289506954e-07,
+      "loss": 1.4518,
+      "step": 56569
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2048040455120103e-07,
+      "loss": 2.0011,
+      "step": 56570
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.204329962073325e-07,
+      "loss": 1.2758,
+      "step": 56571
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2038558786346396e-07,
+      "loss": 1.707,
+      "step": 56572
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2033817951959545e-07,
+      "loss": 1.3654,
+      "step": 56573
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.202907711757269e-07,
+      "loss": 2.5594,
+      "step": 56574
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2024336283185844e-07,
+      "loss": 2.1281,
+      "step": 56575
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2019595448798993e-07,
+      "loss": 1.8159,
+      "step": 56576
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2014854614412137e-07,
+      "loss": 2.419,
+      "step": 56577
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2010113780025286e-07,
+      "loss": 0.5296,
+      "step": 56578
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.200537294563843e-07,
+      "loss": 2.732,
+      "step": 56579
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.2000632111251585e-07,
+      "loss": 2.2586,
+      "step": 56580
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.199589127686473e-07,
+      "loss": 1.0714,
+      "step": 56581
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.199115044247788e-07,
+      "loss": 1.5473,
+      "step": 56582
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.198640960809102e-07,
+      "loss": 0.8219,
+      "step": 56583
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.198166877370417e-07,
+      "loss": 3.3588,
+      "step": 56584
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.197692793931732e-07,
+      "loss": 1.9479,
+      "step": 56585
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.197218710493047e-07,
+      "loss": 1.7916,
+      "step": 56586
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.196744627054362e-07,
+      "loss": 2.7946,
+      "step": 56587
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.196270543615676e-07,
+      "loss": 0.8983,
+      "step": 56588
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.195796460176991e-07,
+      "loss": 2.1575,
+      "step": 56589
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.195322376738306e-07,
+      "loss": 1.4852,
+      "step": 56590
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.194848293299621e-07,
+      "loss": 1.9034,
+      "step": 56591
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1943742098609354e-07,
+      "loss": 1.6601,
+      "step": 56592
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1939001264222503e-07,
+      "loss": 0.8132,
+      "step": 56593
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.193426042983565e-07,
+      "loss": 2.729,
+      "step": 56594
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.19295195954488e-07,
+      "loss": 1.2105,
+      "step": 56595
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1924778761061946e-07,
+      "loss": 0.9235,
+      "step": 56596
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1920037926675095e-07,
+      "loss": 1.8588,
+      "step": 56597
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1915297092288244e-07,
+      "loss": 1.492,
+      "step": 56598
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1910556257901393e-07,
+      "loss": 1.3089,
+      "step": 56599
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.190581542351454e-07,
+      "loss": 1.4039,
+      "step": 56600
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1901074589127686e-07,
+      "loss": 0.3606,
+      "step": 56601
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1896333754740836e-07,
+      "loss": 3.5759,
+      "step": 56602
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.189159292035398e-07,
+      "loss": 2.8662,
+      "step": 56603
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1886852085967134e-07,
+      "loss": 2.4706,
+      "step": 56604
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.188211125158028e-07,
+      "loss": 3.4104,
+      "step": 56605
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1877370417193427e-07,
+      "loss": 3.8793,
+      "step": 56606
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.187262958280657e-07,
+      "loss": 2.311,
+      "step": 56607
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.186788874841972e-07,
+      "loss": 2.8317,
+      "step": 56608
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1863147914032875e-07,
+      "loss": 1.7878,
+      "step": 56609
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.185840707964602e-07,
+      "loss": 1.0183,
+      "step": 56610
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.185366624525917e-07,
+      "loss": 1.7432,
+      "step": 56611
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.184892541087231e-07,
+      "loss": 1.3384,
+      "step": 56612
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.184418457648546e-07,
+      "loss": 1.5127,
+      "step": 56613
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.183944374209861e-07,
+      "loss": 1.1516,
+      "step": 56614
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.183470290771176e-07,
+      "loss": 2.5388,
+      "step": 56615
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1829962073324904e-07,
+      "loss": 1.716,
+      "step": 56616
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1825221238938053e-07,
+      "loss": 2.9778,
+      "step": 56617
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.18204804045512e-07,
+      "loss": 1.3219,
+      "step": 56618
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.181573957016435e-07,
+      "loss": 2.9543,
+      "step": 56619
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.18109987357775e-07,
+      "loss": 2.2586,
+      "step": 56620
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1806257901390644e-07,
+      "loss": 0.8556,
+      "step": 56621
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1801517067003794e-07,
+      "loss": 2.9434,
+      "step": 56622
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1796776232616943e-07,
+      "loss": 3.2529,
+      "step": 56623
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.179203539823009e-07,
+      "loss": 1.035,
+      "step": 56624
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1787294563843236e-07,
+      "loss": 1.2618,
+      "step": 56625
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1782553729456385e-07,
+      "loss": 1.3162,
+      "step": 56626
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.177781289506953e-07,
+      "loss": 0.9947,
+      "step": 56627
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1773072060682684e-07,
+      "loss": 2.7015,
+      "step": 56628
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.176833122629583e-07,
+      "loss": 2.0563,
+      "step": 56629
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1763590391908977e-07,
+      "loss": 1.4932,
+      "step": 56630
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1758849557522126e-07,
+      "loss": 2.0084,
+      "step": 56631
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.175410872313527e-07,
+      "loss": 1.6061,
+      "step": 56632
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1749367888748424e-07,
+      "loss": 1.7941,
+      "step": 56633
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.174462705436157e-07,
+      "loss": 1.6708,
+      "step": 56634
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.173988621997472e-07,
+      "loss": 0.6142,
+      "step": 56635
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.173514538558786e-07,
+      "loss": 4.2991,
+      "step": 56636
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.173040455120101e-07,
+      "loss": 0.8369,
+      "step": 56637
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.172566371681416e-07,
+      "loss": 0.9731,
+      "step": 56638
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.172092288242731e-07,
+      "loss": 2.6947,
+      "step": 56639
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1716182048040453e-07,
+      "loss": 2.0762,
+      "step": 56640
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.17114412136536e-07,
+      "loss": 2.4372,
+      "step": 56641
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.170670037926675e-07,
+      "loss": 2.54,
+      "step": 56642
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.17019595448799e-07,
+      "loss": 1.4669,
+      "step": 56643
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.169721871049305e-07,
+      "loss": 1.5133,
+      "step": 56644
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1692477876106194e-07,
+      "loss": 0.1613,
+      "step": 56645
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.1687737041719343e-07,
+      "loss": 1.3914,
+      "step": 56646
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.168299620733249e-07,
+      "loss": 2.3399,
+      "step": 56647
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.167825537294564e-07,
+      "loss": 2.8444,
+      "step": 56648
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1673514538558785e-07,
+      "loss": 3.3985,
+      "step": 56649
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1668773704171935e-07,
+      "loss": 0.861,
+      "step": 56650
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.166403286978508e-07,
+      "loss": 2.3013,
+      "step": 56651
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1659292035398233e-07,
+      "loss": 2.2231,
+      "step": 56652
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.165455120101138e-07,
+      "loss": 2.3198,
+      "step": 56653
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1649810366624526e-07,
+      "loss": 1.7848,
+      "step": 56654
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1645069532237675e-07,
+      "loss": 3.0128,
+      "step": 56655
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.164032869785082e-07,
+      "loss": 2.2699,
+      "step": 56656
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1635587863463974e-07,
+      "loss": 0.7813,
+      "step": 56657
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.163084702907712e-07,
+      "loss": 0.9885,
+      "step": 56658
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1626106194690267e-07,
+      "loss": 2.5431,
+      "step": 56659
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.162136536030341e-07,
+      "loss": 2.5736,
+      "step": 56660
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.161662452591656e-07,
+      "loss": 2.841,
+      "step": 56661
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.161188369152971e-07,
+      "loss": 1.7817,
+      "step": 56662
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.160714285714286e-07,
+      "loss": 0.9397,
+      "step": 56663
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.160240202275601e-07,
+      "loss": 0.9342,
+      "step": 56664
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.159766118836915e-07,
+      "loss": 1.8934,
+      "step": 56665
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.15929203539823e-07,
+      "loss": 2.8994,
+      "step": 56666
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.158817951959545e-07,
+      "loss": 2.4377,
+      "step": 56667
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.15834386852086e-07,
+      "loss": 2.7752,
+      "step": 56668
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1578697850821743e-07,
+      "loss": 1.6155,
+      "step": 56669
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.157395701643489e-07,
+      "loss": 2.3246,
+      "step": 56670
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.156921618204804e-07,
+      "loss": 1.2533,
+      "step": 56671
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.156447534766119e-07,
+      "loss": 1.6142,
+      "step": 56672
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1559734513274335e-07,
+      "loss": 1.8538,
+      "step": 56673
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1554993678887484e-07,
+      "loss": 0.9705,
+      "step": 56674
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1550252844500633e-07,
+      "loss": 1.5059,
+      "step": 56675
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.154551201011378e-07,
+      "loss": 2.4691,
+      "step": 56676
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.154077117572693e-07,
+      "loss": 1.5017,
+      "step": 56677
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1536030341340076e-07,
+      "loss": 2.1635,
+      "step": 56678
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1531289506953225e-07,
+      "loss": 2.5114,
+      "step": 56679
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1526548672566374e-07,
+      "loss": 2.3737,
+      "step": 56680
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1521807838179523e-07,
+      "loss": 1.6175,
+      "step": 56681
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1517067003792667e-07,
+      "loss": 1.054,
+      "step": 56682
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1512326169405816e-07,
+      "loss": 1.4581,
+      "step": 56683
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.150758533501896e-07,
+      "loss": 1.7438,
+      "step": 56684
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1502844500632115e-07,
+      "loss": 1.5925,
+      "step": 56685
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.149810366624526e-07,
+      "loss": 2.3304,
+      "step": 56686
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.149336283185841e-07,
+      "loss": 1.0091,
+      "step": 56687
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1488621997471557e-07,
+      "loss": 2.7662,
+      "step": 56688
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.14838811630847e-07,
+      "loss": 0.7585,
+      "step": 56689
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.147914032869785e-07,
+      "loss": 1.5536,
+      "step": 56690
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1474399494311e-07,
+      "loss": 2.4938,
+      "step": 56691
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.146965865992415e-07,
+      "loss": 2.0335,
+      "step": 56692
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1464917825537293e-07,
+      "loss": 1.9518,
+      "step": 56693
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.146017699115044e-07,
+      "loss": 1.3784,
+      "step": 56694
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.145543615676359e-07,
+      "loss": 1.2736,
+      "step": 56695
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.145069532237674e-07,
+      "loss": 2.998,
+      "step": 56696
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1445954487989884e-07,
+      "loss": 5.5419,
+      "step": 56697
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1441213653603034e-07,
+      "loss": 1.0654,
+      "step": 56698
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1436472819216183e-07,
+      "loss": 2.1138,
+      "step": 56699
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.143173198482933e-07,
+      "loss": 3.614,
+      "step": 56700
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.142699115044248e-07,
+      "loss": 1.5485,
+      "step": 56701
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1422250316055625e-07,
+      "loss": 1.3164,
+      "step": 56702
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1417509481668774e-07,
+      "loss": 3.1011,
+      "step": 56703
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1412768647281924e-07,
+      "loss": 0.9454,
+      "step": 56704
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1408027812895073e-07,
+      "loss": 1.9057,
+      "step": 56705
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1403286978508217e-07,
+      "loss": 1.6904,
+      "step": 56706
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1398546144121366e-07,
+      "loss": 1.6693,
+      "step": 56707
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.139380530973451e-07,
+      "loss": 1.1635,
+      "step": 56708
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1389064475347664e-07,
+      "loss": 1.5443,
+      "step": 56709
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1384323640960814e-07,
+      "loss": 1.8681,
+      "step": 56710
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.137958280657396e-07,
+      "loss": 2.45,
+      "step": 56711
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.1374841972187107e-07,
+      "loss": 1.1948,
+      "step": 56712
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.137010113780025e-07,
+      "loss": 2.1588,
+      "step": 56713
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1365360303413405e-07,
+      "loss": 0.7481,
+      "step": 56714
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.136061946902655e-07,
+      "loss": 1.4071,
+      "step": 56715
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.13558786346397e-07,
+      "loss": 1.3085,
+      "step": 56716
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.135113780025284e-07,
+      "loss": 2.0701,
+      "step": 56717
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.134639696586599e-07,
+      "loss": 1.3019,
+      "step": 56718
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.134165613147914e-07,
+      "loss": 1.8824,
+      "step": 56719
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.133691529709229e-07,
+      "loss": 1.524,
+      "step": 56720
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.133217446270544e-07,
+      "loss": 1.2296,
+      "step": 56721
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1327433628318583e-07,
+      "loss": 1.6046,
+      "step": 56722
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.132269279393173e-07,
+      "loss": 2.4522,
+      "step": 56723
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.131795195954488e-07,
+      "loss": 0.2844,
+      "step": 56724
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.131321112515803e-07,
+      "loss": 0.2874,
+      "step": 56725
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1308470290771175e-07,
+      "loss": 2.7544,
+      "step": 56726
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1303729456384324e-07,
+      "loss": 2.4775,
+      "step": 56727
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1298988621997473e-07,
+      "loss": 1.6722,
+      "step": 56728
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.129424778761062e-07,
+      "loss": 1.2589,
+      "step": 56729
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1289506953223766e-07,
+      "loss": 1.1322,
+      "step": 56730
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1284766118836915e-07,
+      "loss": 0.7747,
+      "step": 56731
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1280025284450065e-07,
+      "loss": 3.5525,
+      "step": 56732
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1275284450063214e-07,
+      "loss": 3.1523,
+      "step": 56733
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1270543615676363e-07,
+      "loss": 3.0387,
+      "step": 56734
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1265802781289507e-07,
+      "loss": 1.0978,
+      "step": 56735
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1261061946902656e-07,
+      "loss": 3.3542,
+      "step": 56736
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.12563211125158e-07,
+      "loss": 1.2267,
+      "step": 56737
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1251580278128955e-07,
+      "loss": 3.5303,
+      "step": 56738
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.12468394437421e-07,
+      "loss": 2.0464,
+      "step": 56739
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.124209860935525e-07,
+      "loss": 0.7248,
+      "step": 56740
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.123735777496839e-07,
+      "loss": 2.1734,
+      "step": 56741
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.123261694058154e-07,
+      "loss": 3.514,
+      "step": 56742
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1227876106194695e-07,
+      "loss": 0.8324,
+      "step": 56743
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.122313527180784e-07,
+      "loss": 1.5289,
+      "step": 56744
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.121839443742099e-07,
+      "loss": 0.9614,
+      "step": 56745
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.121365360303413e-07,
+      "loss": 1.1397,
+      "step": 56746
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.120891276864728e-07,
+      "loss": 2.6191,
+      "step": 56747
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.120417193426043e-07,
+      "loss": 2.4628,
+      "step": 56748
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.119943109987358e-07,
+      "loss": 2.0336,
+      "step": 56749
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1194690265486724e-07,
+      "loss": 0.2926,
+      "step": 56750
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1189949431099873e-07,
+      "loss": 1.7498,
+      "step": 56751
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.118520859671302e-07,
+      "loss": 1.8644,
+      "step": 56752
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.118046776232617e-07,
+      "loss": 1.8108,
+      "step": 56753
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.117572692793932e-07,
+      "loss": 0.4812,
+      "step": 56754
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1170986093552465e-07,
+      "loss": 1.284,
+      "step": 56755
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1166245259165614e-07,
+      "loss": 2.1867,
+      "step": 56756
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1161504424778763e-07,
+      "loss": 1.0838,
+      "step": 56757
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.115676359039191e-07,
+      "loss": 2.5916,
+      "step": 56758
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1152022756005056e-07,
+      "loss": 1.7945,
+      "step": 56759
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1147281921618206e-07,
+      "loss": 3.4508,
+      "step": 56760
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.114254108723135e-07,
+      "loss": 1.3136,
+      "step": 56761
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1137800252844504e-07,
+      "loss": 0.9809,
+      "step": 56762
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.113305941845765e-07,
+      "loss": 1.7854,
+      "step": 56763
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1128318584070797e-07,
+      "loss": 1.615,
+      "step": 56764
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1123577749683946e-07,
+      "loss": 0.8601,
+      "step": 56765
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.111883691529709e-07,
+      "loss": 1.2514,
+      "step": 56766
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1114096080910245e-07,
+      "loss": 1.2396,
+      "step": 56767
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.110935524652339e-07,
+      "loss": 0.8846,
+      "step": 56768
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.110461441213654e-07,
+      "loss": 2.6464,
+      "step": 56769
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.109987357774968e-07,
+      "loss": 2.8076,
+      "step": 56770
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.109513274336283e-07,
+      "loss": 2.5684,
+      "step": 56771
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.109039190897598e-07,
+      "loss": 3.428,
+      "step": 56772
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.108565107458913e-07,
+      "loss": 2.0452,
+      "step": 56773
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1080910240202273e-07,
+      "loss": 1.6924,
+      "step": 56774
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.1076169405815423e-07,
+      "loss": 0.8089,
+      "step": 56775
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.107142857142857e-07,
+      "loss": 2.1289,
+      "step": 56776
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.106668773704172e-07,
+      "loss": 1.1357,
+      "step": 56777
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.106194690265487e-07,
+      "loss": 3.5316,
+      "step": 56778
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1057206068268014e-07,
+      "loss": 2.1214,
+      "step": 56779
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1052465233881163e-07,
+      "loss": 1.9904,
+      "step": 56780
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1047724399494313e-07,
+      "loss": 1.3595,
+      "step": 56781
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.104298356510746e-07,
+      "loss": 1.7816,
+      "step": 56782
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1038242730720606e-07,
+      "loss": 0.2894,
+      "step": 56783
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1033501896333755e-07,
+      "loss": 1.3397,
+      "step": 56784
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1028761061946904e-07,
+      "loss": 1.9786,
+      "step": 56785
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1024020227560054e-07,
+      "loss": 1.3377,
+      "step": 56786
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1019279393173203e-07,
+      "loss": 2.7618,
+      "step": 56787
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1014538558786347e-07,
+      "loss": 1.6967,
+      "step": 56788
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1009797724399496e-07,
+      "loss": 1.2748,
+      "step": 56789
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.100505689001264e-07,
+      "loss": 1.8925,
+      "step": 56790
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.1000316055625794e-07,
+      "loss": 1.7132,
+      "step": 56791
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.099557522123894e-07,
+      "loss": 1.9076,
+      "step": 56792
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.099083438685209e-07,
+      "loss": 1.7972,
+      "step": 56793
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.098609355246523e-07,
+      "loss": 0.8639,
+      "step": 56794
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.098135271807838e-07,
+      "loss": 2.1956,
+      "step": 56795
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.097661188369153e-07,
+      "loss": 1.9996,
+      "step": 56796
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.097187104930468e-07,
+      "loss": 1.7813,
+      "step": 56797
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0967130214917823e-07,
+      "loss": 1.8828,
+      "step": 56798
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.096238938053097e-07,
+      "loss": 2.3535,
+      "step": 56799
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.095764854614412e-07,
+      "loss": 0.9101,
+      "step": 56800
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.095290771175727e-07,
+      "loss": 0.8308,
+      "step": 56801
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.094816687737042e-07,
+      "loss": 1.1007,
+      "step": 56802
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0943426042983564e-07,
+      "loss": 1.539,
+      "step": 56803
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0938685208596713e-07,
+      "loss": 2.824,
+      "step": 56804
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.093394437420986e-07,
+      "loss": 2.2402,
+      "step": 56805
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.092920353982301e-07,
+      "loss": 3.5072,
+      "step": 56806
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0924462705436155e-07,
+      "loss": 1.5735,
+      "step": 56807
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0919721871049305e-07,
+      "loss": 1.8597,
+      "step": 56808
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0914981036662454e-07,
+      "loss": 1.1118,
+      "step": 56809
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0910240202275603e-07,
+      "loss": 3.2913,
+      "step": 56810
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.090549936788875e-07,
+      "loss": 1.4096,
+      "step": 56811
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0900758533501896e-07,
+      "loss": 2.7123,
+      "step": 56812
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0896017699115045e-07,
+      "loss": 4.7824,
+      "step": 56813
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0891276864728195e-07,
+      "loss": 1.5827,
+      "step": 56814
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0886536030341344e-07,
+      "loss": 2.4605,
+      "step": 56815
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.088179519595449e-07,
+      "loss": 2.6057,
+      "step": 56816
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0877054361567637e-07,
+      "loss": 1.5054,
+      "step": 56817
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.087231352718078e-07,
+      "loss": 1.9966,
+      "step": 56818
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0867572692793935e-07,
+      "loss": 1.134,
+      "step": 56819
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.086283185840708e-07,
+      "loss": 2.6747,
+      "step": 56820
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.085809102402023e-07,
+      "loss": 0.8283,
+      "step": 56821
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.085335018963338e-07,
+      "loss": 1.0511,
+      "step": 56822
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.084860935524652e-07,
+      "loss": 2.2285,
+      "step": 56823
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0843868520859676e-07,
+      "loss": 2.4527,
+      "step": 56824
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.083912768647282e-07,
+      "loss": 1.6539,
+      "step": 56825
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.083438685208597e-07,
+      "loss": 1.2064,
+      "step": 56826
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0829646017699113e-07,
+      "loss": 1.2019,
+      "step": 56827
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.082490518331226e-07,
+      "loss": 0.841,
+      "step": 56828
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.082016434892541e-07,
+      "loss": 1.5241,
+      "step": 56829
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.081542351453856e-07,
+      "loss": 4.2327,
+      "step": 56830
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0810682680151705e-07,
+      "loss": 2.5506,
+      "step": 56831
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0805941845764854e-07,
+      "loss": 2.3074,
+      "step": 56832
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0801201011378003e-07,
+      "loss": 1.2097,
+      "step": 56833
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.079646017699115e-07,
+      "loss": 2.2139,
+      "step": 56834
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.07917193426043e-07,
+      "loss": 2.3164,
+      "step": 56835
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0786978508217446e-07,
+      "loss": 1.9713,
+      "step": 56836
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0782237673830595e-07,
+      "loss": 1.5782,
+      "step": 56837
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.0777496839443744e-07,
+      "loss": 1.3818,
+      "step": 56838
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0772756005056893e-07,
+      "loss": 2.8571,
+      "step": 56839
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0768015170670037e-07,
+      "loss": 1.5185,
+      "step": 56840
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0763274336283186e-07,
+      "loss": 1.7904,
+      "step": 56841
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.075853350189633e-07,
+      "loss": 2.9405,
+      "step": 56842
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0753792667509485e-07,
+      "loss": 1.274,
+      "step": 56843
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0749051833122634e-07,
+      "loss": 2.0592,
+      "step": 56844
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.074431099873578e-07,
+      "loss": 1.2534,
+      "step": 56845
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0739570164348927e-07,
+      "loss": 1.5359,
+      "step": 56846
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.073482932996207e-07,
+      "loss": 1.5436,
+      "step": 56847
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0730088495575226e-07,
+      "loss": 2.0827,
+      "step": 56848
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.072534766118837e-07,
+      "loss": 1.9955,
+      "step": 56849
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.072060682680152e-07,
+      "loss": 2.1837,
+      "step": 56850
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0715865992414663e-07,
+      "loss": 1.3774,
+      "step": 56851
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.071112515802781e-07,
+      "loss": 2.893,
+      "step": 56852
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.070638432364096e-07,
+      "loss": 2.6254,
+      "step": 56853
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.070164348925411e-07,
+      "loss": 2.4865,
+      "step": 56854
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.069690265486726e-07,
+      "loss": 1.7586,
+      "step": 56855
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0692161820480403e-07,
+      "loss": 1.7945,
+      "step": 56856
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0687420986093553e-07,
+      "loss": 1.5919,
+      "step": 56857
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.06826801517067e-07,
+      "loss": 1.9537,
+      "step": 56858
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.067793931731985e-07,
+      "loss": 3.0335,
+      "step": 56859
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0673198482932995e-07,
+      "loss": 1.821,
+      "step": 56860
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0668457648546144e-07,
+      "loss": 1.5495,
+      "step": 56861
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0663716814159293e-07,
+      "loss": 1.4001,
+      "step": 56862
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0658975979772443e-07,
+      "loss": 1.0009,
+      "step": 56863
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0654235145385587e-07,
+      "loss": 4.2283,
+      "step": 56864
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0649494310998736e-07,
+      "loss": 0.8449,
+      "step": 56865
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0644753476611885e-07,
+      "loss": 1.435,
+      "step": 56866
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0640012642225034e-07,
+      "loss": 0.7833,
+      "step": 56867
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0635271807838183e-07,
+      "loss": 1.119,
+      "step": 56868
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.063053097345133e-07,
+      "loss": 1.8521,
+      "step": 56869
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0625790139064477e-07,
+      "loss": 1.0475,
+      "step": 56870
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.062104930467762e-07,
+      "loss": 0.9572,
+      "step": 56871
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0616308470290775e-07,
+      "loss": 2.0429,
+      "step": 56872
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.061156763590392e-07,
+      "loss": 2.3845,
+      "step": 56873
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.060682680151707e-07,
+      "loss": 1.6448,
+      "step": 56874
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.060208596713021e-07,
+      "loss": 2.4187,
+      "step": 56875
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.059734513274336e-07,
+      "loss": 1.8989,
+      "step": 56876
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0592604298356516e-07,
+      "loss": 2.1517,
+      "step": 56877
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.058786346396966e-07,
+      "loss": 1.7025,
+      "step": 56878
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.058312262958281e-07,
+      "loss": 2.0099,
+      "step": 56879
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0578381795195953e-07,
+      "loss": 1.5348,
+      "step": 56880
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.05736409608091e-07,
+      "loss": 1.8405,
+      "step": 56881
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.056890012642225e-07,
+      "loss": 1.2869,
+      "step": 56882
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.05641592920354e-07,
+      "loss": 0.9106,
+      "step": 56883
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0559418457648544e-07,
+      "loss": 2.0592,
+      "step": 56884
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0554677623261694e-07,
+      "loss": 1.5352,
+      "step": 56885
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0549936788874843e-07,
+      "loss": 1.7817,
+      "step": 56886
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.054519595448799e-07,
+      "loss": 1.8696,
+      "step": 56887
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.054045512010114e-07,
+      "loss": 2.2798,
+      "step": 56888
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0535714285714285e-07,
+      "loss": 1.3852,
+      "step": 56889
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0530973451327434e-07,
+      "loss": 0.8683,
+      "step": 56890
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0526232616940584e-07,
+      "loss": 1.1605,
+      "step": 56891
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0521491782553733e-07,
+      "loss": 1.7835,
+      "step": 56892
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0516750948166877e-07,
+      "loss": 3.7935,
+      "step": 56893
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0512010113780026e-07,
+      "loss": 1.891,
+      "step": 56894
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.050726927939317e-07,
+      "loss": 2.2529,
+      "step": 56895
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0502528445006325e-07,
+      "loss": 1.8,
+      "step": 56896
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.049778761061947e-07,
+      "loss": 1.862,
+      "step": 56897
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.049304677623262e-07,
+      "loss": 3.6793,
+      "step": 56898
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0488305941845767e-07,
+      "loss": 0.8866,
+      "step": 56899
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.048356510745891e-07,
+      "loss": 2.4163,
+      "step": 56900
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.0478824273072065e-07,
+      "loss": 1.031,
+      "step": 56901
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.047408343868521e-07,
+      "loss": 2.6599,
+      "step": 56902
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.046934260429836e-07,
+      "loss": 2.002,
+      "step": 56903
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.04646017699115e-07,
+      "loss": 1.8142,
+      "step": 56904
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.045986093552465e-07,
+      "loss": 1.6863,
+      "step": 56905
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.04551201011378e-07,
+      "loss": 1.854,
+      "step": 56906
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.045037926675095e-07,
+      "loss": 2.8329,
+      "step": 56907
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0445638432364094e-07,
+      "loss": 1.6766,
+      "step": 56908
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0440897597977243e-07,
+      "loss": 1.9061,
+      "step": 56909
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.043615676359039e-07,
+      "loss": 2.0226,
+      "step": 56910
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.043141592920354e-07,
+      "loss": 2.5639,
+      "step": 56911
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.042667509481669e-07,
+      "loss": 1.4916,
+      "step": 56912
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0421934260429835e-07,
+      "loss": 3.1192,
+      "step": 56913
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0417193426042984e-07,
+      "loss": 1.7391,
+      "step": 56914
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0412452591656133e-07,
+      "loss": 0.8014,
+      "step": 56915
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.040771175726928e-07,
+      "loss": 1.0931,
+      "step": 56916
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0402970922882426e-07,
+      "loss": 0.8661,
+      "step": 56917
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0398230088495576e-07,
+      "loss": 2.9852,
+      "step": 56918
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0393489254108725e-07,
+      "loss": 3.2077,
+      "step": 56919
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0388748419721874e-07,
+      "loss": 1.4313,
+      "step": 56920
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.038400758533502e-07,
+      "loss": 1.8671,
+      "step": 56921
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0379266750948167e-07,
+      "loss": 2.1186,
+      "step": 56922
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0374525916561316e-07,
+      "loss": 1.7993,
+      "step": 56923
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0369785082174466e-07,
+      "loss": 0.7869,
+      "step": 56924
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0365044247787615e-07,
+      "loss": 1.9605,
+      "step": 56925
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.036030341340076e-07,
+      "loss": 1.1858,
+      "step": 56926
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.035556257901391e-07,
+      "loss": 0.4479,
+      "step": 56927
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.035082174462705e-07,
+      "loss": 1.5605,
+      "step": 56928
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0346080910240206e-07,
+      "loss": 1.9198,
+      "step": 56929
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.034134007585335e-07,
+      "loss": 1.9213,
+      "step": 56930
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.03365992414665e-07,
+      "loss": 1.2933,
+      "step": 56931
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0331858407079643e-07,
+      "loss": 2.8724,
+      "step": 56932
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.032711757269279e-07,
+      "loss": 2.2942,
+      "step": 56933
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.032237673830594e-07,
+      "loss": 4.0922,
+      "step": 56934
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.031763590391909e-07,
+      "loss": 0.8273,
+      "step": 56935
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.031289506953224e-07,
+      "loss": 0.6748,
+      "step": 56936
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0308154235145384e-07,
+      "loss": 2.6686,
+      "step": 56937
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0303413400758533e-07,
+      "loss": 0.9574,
+      "step": 56938
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.029867256637168e-07,
+      "loss": 2.4757,
+      "step": 56939
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.029393173198483e-07,
+      "loss": 1.061,
+      "step": 56940
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0289190897597976e-07,
+      "loss": 1.0614,
+      "step": 56941
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0284450063211125e-07,
+      "loss": 2.062,
+      "step": 56942
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0279709228824274e-07,
+      "loss": 2.4817,
+      "step": 56943
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0274968394437423e-07,
+      "loss": 1.9665,
+      "step": 56944
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.027022756005057e-07,
+      "loss": 1.8082,
+      "step": 56945
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0265486725663717e-07,
+      "loss": 1.9017,
+      "step": 56946
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0260745891276866e-07,
+      "loss": 3.0057,
+      "step": 56947
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0256005056890015e-07,
+      "loss": 2.1337,
+      "step": 56948
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0251264222503164e-07,
+      "loss": 1.8907,
+      "step": 56949
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.024652338811631e-07,
+      "loss": 2.6285,
+      "step": 56950
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.024178255372946e-07,
+      "loss": 1.5438,
+      "step": 56951
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.02370417193426e-07,
+      "loss": 1.1468,
+      "step": 56952
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0232300884955756e-07,
+      "loss": 2.1929,
+      "step": 56953
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.02275600505689e-07,
+      "loss": 2.8386,
+      "step": 56954
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.022281921618205e-07,
+      "loss": 1.4943,
+      "step": 56955
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.02180783817952e-07,
+      "loss": 2.6507,
+      "step": 56956
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.021333754740834e-07,
+      "loss": 1.1017,
+      "step": 56957
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0208596713021497e-07,
+      "loss": 1.2689,
+      "step": 56958
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.020385587863464e-07,
+      "loss": 2.9452,
+      "step": 56959
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.019911504424779e-07,
+      "loss": 1.5436,
+      "step": 56960
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0194374209860934e-07,
+      "loss": 2.7836,
+      "step": 56961
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0189633375474083e-07,
+      "loss": 3.0615,
+      "step": 56962
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.018489254108723e-07,
+      "loss": 1.0344,
+      "step": 56963
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.018015170670038e-07,
+      "loss": 2.2787,
+      "step": 56964
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.0175410872313525e-07,
+      "loss": 1.3306,
+      "step": 56965
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0170670037926674e-07,
+      "loss": 2.3337,
+      "step": 56966
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0165929203539824e-07,
+      "loss": 1.6017,
+      "step": 56967
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0161188369152973e-07,
+      "loss": 1.4403,
+      "step": 56968
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.015644753476612e-07,
+      "loss": 3.7196,
+      "step": 56969
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0151706700379266e-07,
+      "loss": 0.4802,
+      "step": 56970
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0146965865992415e-07,
+      "loss": 2.1724,
+      "step": 56971
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0142225031605564e-07,
+      "loss": 1.1896,
+      "step": 56972
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0137484197218714e-07,
+      "loss": 3.0364,
+      "step": 56973
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.013274336283186e-07,
+      "loss": 2.8655,
+      "step": 56974
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0128002528445007e-07,
+      "loss": 0.8001,
+      "step": 56975
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.012326169405815e-07,
+      "loss": 2.3689,
+      "step": 56976
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0118520859671305e-07,
+      "loss": 1.6268,
+      "step": 56977
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0113780025284454e-07,
+      "loss": 1.8498,
+      "step": 56978
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.01090391908976e-07,
+      "loss": 2.8219,
+      "step": 56979
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.010429835651075e-07,
+      "loss": 1.4264,
+      "step": 56980
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.009955752212389e-07,
+      "loss": 2.6927,
+      "step": 56981
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0094816687737046e-07,
+      "loss": 0.5573,
+      "step": 56982
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.009007585335019e-07,
+      "loss": 1.2813,
+      "step": 56983
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.008533501896334e-07,
+      "loss": 1.3157,
+      "step": 56984
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0080594184576483e-07,
+      "loss": 2.2744,
+      "step": 56985
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.007585335018963e-07,
+      "loss": 1.5898,
+      "step": 56986
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.007111251580278e-07,
+      "loss": 2.0536,
+      "step": 56987
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.006637168141593e-07,
+      "loss": 0.8063,
+      "step": 56988
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.006163084702908e-07,
+      "loss": 3.2491,
+      "step": 56989
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0056890012642224e-07,
+      "loss": 2.2192,
+      "step": 56990
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0052149178255373e-07,
+      "loss": 2.1437,
+      "step": 56991
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.004740834386852e-07,
+      "loss": 1.3519,
+      "step": 56992
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.004266750948167e-07,
+      "loss": 1.6041,
+      "step": 56993
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0037926675094815e-07,
+      "loss": 2.217,
+      "step": 56994
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0033185840707965e-07,
+      "loss": 2.01,
+      "step": 56995
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0028445006321114e-07,
+      "loss": 1.428,
+      "step": 56996
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0023704171934263e-07,
+      "loss": 3.3371,
+      "step": 56997
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.5822380106571936,
+      "eval_loss": 1.7887498140335083,
+      "eval_runtime": 154.1436,
+      "eval_samples_per_second": 18.262,
+      "eval_steps_per_second": 4.567,
+      "step": 56997
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0018963337547407e-07,
+      "loss": 1.798,
+      "step": 56998
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0014222503160556e-07,
+      "loss": 1.337,
+      "step": 56999
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0009481668773706e-07,
+      "loss": 2.3761,
+      "step": 57000
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0004740834386855e-07,
+      "loss": 1.463,
+      "step": 57001
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.0000000000000004e-07,
+      "loss": 1.3675,
+      "step": 57002
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.999525916561315e-07,
+      "loss": 2.0444,
+      "step": 57003
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9990518331226297e-07,
+      "loss": 2.1344,
+      "step": 57004
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.998577749683944e-07,
+      "loss": 2.0671,
+      "step": 57005
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9981036662452596e-07,
+      "loss": 2.1489,
+      "step": 57006
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.997629582806574e-07,
+      "loss": 0.8759,
+      "step": 57007
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.997155499367889e-07,
+      "loss": 1.9895,
+      "step": 57008
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.996681415929203e-07,
+      "loss": 2.1328,
+      "step": 57009
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.996207332490518e-07,
+      "loss": 1.136,
+      "step": 57010
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9957332490518336e-07,
+      "loss": 2.2877,
+      "step": 57011
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.995259165613148e-07,
+      "loss": 2.0432,
+      "step": 57012
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.994785082174463e-07,
+      "loss": 1.335,
+      "step": 57013
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9943109987357773e-07,
+      "loss": 1.5296,
+      "step": 57014
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.993836915297092e-07,
+      "loss": 0.6618,
+      "step": 57015
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.993362831858407e-07,
+      "loss": 1.63,
+      "step": 57016
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.992888748419722e-07,
+      "loss": 1.7825,
+      "step": 57017
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9924146649810365e-07,
+      "loss": 1.5043,
+      "step": 57018
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9919405815423514e-07,
+      "loss": 2.1598,
+      "step": 57019
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9914664981036663e-07,
+      "loss": 2.3409,
+      "step": 57020
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.990992414664981e-07,
+      "loss": 1.3244,
+      "step": 57021
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9905183312262957e-07,
+      "loss": 2.1654,
+      "step": 57022
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9900442477876106e-07,
+      "loss": 1.6903,
+      "step": 57023
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9895701643489255e-07,
+      "loss": 1.0431,
+      "step": 57024
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9890960809102404e-07,
+      "loss": 1.6214,
+      "step": 57025
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9886219974715553e-07,
+      "loss": 2.7409,
+      "step": 57026
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9881479140328697e-07,
+      "loss": 2.7499,
+      "step": 57027
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.9876738305941847e-07,
+      "loss": 2.0755,
+      "step": 57028
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9871997471554996e-07,
+      "loss": 0.8767,
+      "step": 57029
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9867256637168145e-07,
+      "loss": 2.2837,
+      "step": 57030
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.986251580278129e-07,
+      "loss": 1.2347,
+      "step": 57031
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.985777496839444e-07,
+      "loss": 2.2393,
+      "step": 57032
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.985303413400758e-07,
+      "loss": 1.9792,
+      "step": 57033
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.984829329962073e-07,
+      "loss": 2.3567,
+      "step": 57034
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9843552465233886e-07,
+      "loss": 1.3823,
+      "step": 57035
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.983881163084703e-07,
+      "loss": 1.1565,
+      "step": 57036
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.983407079646018e-07,
+      "loss": 2.3075,
+      "step": 57037
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9829329962073323e-07,
+      "loss": 1.2075,
+      "step": 57038
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.982458912768647e-07,
+      "loss": 1.7294,
+      "step": 57039
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.981984829329962e-07,
+      "loss": 2.3014,
+      "step": 57040
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.981510745891277e-07,
+      "loss": 2.6805,
+      "step": 57041
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9810366624525914e-07,
+      "loss": 1.8834,
+      "step": 57042
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9805625790139064e-07,
+      "loss": 1.1901,
+      "step": 57043
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9800884955752213e-07,
+      "loss": 0.5896,
+      "step": 57044
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.979614412136536e-07,
+      "loss": 0.2565,
+      "step": 57045
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.979140328697851e-07,
+      "loss": 2.1036,
+      "step": 57046
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9786662452591655e-07,
+      "loss": 1.3711,
+      "step": 57047
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9781921618204804e-07,
+      "loss": 3.0123,
+      "step": 57048
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9777180783817954e-07,
+      "loss": 1.0428,
+      "step": 57049
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9772439949431103e-07,
+      "loss": 0.6625,
+      "step": 57050
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9767699115044247e-07,
+      "loss": 1.191,
+      "step": 57051
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9762958280657396e-07,
+      "loss": 1.8325,
+      "step": 57052
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9758217446270545e-07,
+      "loss": 2.2696,
+      "step": 57053
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9753476611883694e-07,
+      "loss": 1.9625,
+      "step": 57054
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.974873577749684e-07,
+      "loss": 2.2383,
+      "step": 57055
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.974399494310999e-07,
+      "loss": 1.1193,
+      "step": 57056
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9739254108723137e-07,
+      "loss": 1.9239,
+      "step": 57057
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9734513274336286e-07,
+      "loss": 1.0201,
+      "step": 57058
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9729772439949435e-07,
+      "loss": 2.5497,
+      "step": 57059
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.972503160556258e-07,
+      "loss": 1.612,
+      "step": 57060
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.972029077117573e-07,
+      "loss": 2.927,
+      "step": 57061
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.971554993678887e-07,
+      "loss": 2.2119,
+      "step": 57062
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9710809102402027e-07,
+      "loss": 1.5054,
+      "step": 57063
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.970606826801517e-07,
+      "loss": 0.1556,
+      "step": 57064
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.970132743362832e-07,
+      "loss": 1.8236,
+      "step": 57065
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9696586599241464e-07,
+      "loss": 0.6462,
+      "step": 57066
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9691845764854613e-07,
+      "loss": 1.3289,
+      "step": 57067
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.968710493046777e-07,
+      "loss": 1.4833,
+      "step": 57068
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.968236409608091e-07,
+      "loss": 0.9844,
+      "step": 57069
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.967762326169406e-07,
+      "loss": 2.0863,
+      "step": 57070
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9672882427307205e-07,
+      "loss": 2.3143,
+      "step": 57071
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9668141592920354e-07,
+      "loss": 1.2636,
+      "step": 57072
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9663400758533503e-07,
+      "loss": 1.7724,
+      "step": 57073
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.965865992414665e-07,
+      "loss": 1.3928,
+      "step": 57074
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9653919089759796e-07,
+      "loss": 3.2433,
+      "step": 57075
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9649178255372945e-07,
+      "loss": 1.9414,
+      "step": 57076
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9644437420986095e-07,
+      "loss": 1.4221,
+      "step": 57077
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9639696586599244e-07,
+      "loss": 1.3976,
+      "step": 57078
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9634955752212393e-07,
+      "loss": 1.9365,
+      "step": 57079
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9630214917825537e-07,
+      "loss": 3.1242,
+      "step": 57080
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9625474083438686e-07,
+      "loss": 1.0561,
+      "step": 57081
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9620733249051835e-07,
+      "loss": 1.3907,
+      "step": 57082
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9615992414664985e-07,
+      "loss": 3.2106,
+      "step": 57083
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.961125158027813e-07,
+      "loss": 2.3989,
+      "step": 57084
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.960651074589128e-07,
+      "loss": 0.3032,
+      "step": 57085
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.960176991150442e-07,
+      "loss": 2.0707,
+      "step": 57086
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.9597029077117576e-07,
+      "loss": 1.3389,
+      "step": 57087
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.959228824273072e-07,
+      "loss": 1.6295,
+      "step": 57088
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.958754740834387e-07,
+      "loss": 1.4224,
+      "step": 57089
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.958280657395702e-07,
+      "loss": 1.7161,
+      "step": 57090
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.957806573957016e-07,
+      "loss": 0.812,
+      "step": 57091
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9573324905183317e-07,
+      "loss": 0.9686,
+      "step": 57092
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.956858407079646e-07,
+      "loss": 0.8392,
+      "step": 57093
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.956384323640961e-07,
+      "loss": 1.5924,
+      "step": 57094
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9559102402022754e-07,
+      "loss": 0.6339,
+      "step": 57095
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9554361567635903e-07,
+      "loss": 1.4688,
+      "step": 57096
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.954962073324905e-07,
+      "loss": 2.5984,
+      "step": 57097
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.95448798988622e-07,
+      "loss": 2.8776,
+      "step": 57098
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9540139064475346e-07,
+      "loss": 1.0501,
+      "step": 57099
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9535398230088495e-07,
+      "loss": 2.7828,
+      "step": 57100
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9530657395701644e-07,
+      "loss": 1.6774,
+      "step": 57101
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9525916561314793e-07,
+      "loss": 1.3,
+      "step": 57102
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.952117572692794e-07,
+      "loss": 1.8643,
+      "step": 57103
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9516434892541086e-07,
+      "loss": 1.1496,
+      "step": 57104
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9511694058154236e-07,
+      "loss": 1.0897,
+      "step": 57105
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9506953223767385e-07,
+      "loss": 2.7159,
+      "step": 57106
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9502212389380534e-07,
+      "loss": 1.2555,
+      "step": 57107
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.949747155499368e-07,
+      "loss": 1.307,
+      "step": 57108
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9492730720606827e-07,
+      "loss": 1.5006,
+      "step": 57109
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.948798988621997e-07,
+      "loss": 0.8568,
+      "step": 57110
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9483249051833126e-07,
+      "loss": 2.3796,
+      "step": 57111
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9478508217446275e-07,
+      "loss": 1.0646,
+      "step": 57112
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.947376738305942e-07,
+      "loss": 1.6581,
+      "step": 57113
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.946902654867257e-07,
+      "loss": 2.2931,
+      "step": 57114
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.946428571428571e-07,
+      "loss": 3.4702,
+      "step": 57115
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9459544879898867e-07,
+      "loss": 1.4457,
+      "step": 57116
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.945480404551201e-07,
+      "loss": 2.5355,
+      "step": 57117
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.945006321112516e-07,
+      "loss": 3.3582,
+      "step": 57118
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9445322376738304e-07,
+      "loss": 2.2519,
+      "step": 57119
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9440581542351453e-07,
+      "loss": 2.5224,
+      "step": 57120
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.94358407079646e-07,
+      "loss": 2.6179,
+      "step": 57121
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.943109987357775e-07,
+      "loss": 1.7486,
+      "step": 57122
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9426359039190895e-07,
+      "loss": 1.4549,
+      "step": 57123
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9421618204804044e-07,
+      "loss": 1.7887,
+      "step": 57124
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9416877370417194e-07,
+      "loss": 0.2535,
+      "step": 57125
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9412136536030343e-07,
+      "loss": 1.1153,
+      "step": 57126
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.940739570164349e-07,
+      "loss": 2.5111,
+      "step": 57127
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9402654867256636e-07,
+      "loss": 1.9482,
+      "step": 57128
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9397914032869785e-07,
+      "loss": 2.3989,
+      "step": 57129
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9393173198482934e-07,
+      "loss": 1.3324,
+      "step": 57130
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9388432364096084e-07,
+      "loss": 1.583,
+      "step": 57131
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.938369152970923e-07,
+      "loss": 1.4025,
+      "step": 57132
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9378950695322377e-07,
+      "loss": 2.0464,
+      "step": 57133
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9374209860935526e-07,
+      "loss": 1.0651,
+      "step": 57134
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9369469026548675e-07,
+      "loss": 2.0666,
+      "step": 57135
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9364728192161824e-07,
+      "loss": 1.1908,
+      "step": 57136
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.935998735777497e-07,
+      "loss": 2.1032,
+      "step": 57137
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.935524652338812e-07,
+      "loss": 1.7779,
+      "step": 57138
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.935050568900126e-07,
+      "loss": 0.3255,
+      "step": 57139
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9345764854614416e-07,
+      "loss": 2.3135,
+      "step": 57140
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.934102402022756e-07,
+      "loss": 0.5803,
+      "step": 57141
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.933628318584071e-07,
+      "loss": 2.3222,
+      "step": 57142
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9331542351453853e-07,
+      "loss": 2.5192,
+      "step": 57143
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9326801517067e-07,
+      "loss": 0.7244,
+      "step": 57144
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.932206068268015e-07,
+      "loss": 1.9889,
+      "step": 57145
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.93173198482933e-07,
+      "loss": 3.1043,
+      "step": 57146
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.931257901390645e-07,
+      "loss": 1.5179,
+      "step": 57147
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9307838179519594e-07,
+      "loss": 1.6262,
+      "step": 57148
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9303097345132743e-07,
+      "loss": 1.4938,
+      "step": 57149
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.929835651074589e-07,
+      "loss": 2.2447,
+      "step": 57150
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.929361567635904e-07,
+      "loss": 3.0602,
+      "step": 57151
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9288874841972185e-07,
+      "loss": 1.9239,
+      "step": 57152
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9284134007585335e-07,
+      "loss": 4.2762,
+      "step": 57153
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9279393173198484e-07,
+      "loss": 2.1634,
+      "step": 57154
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.9274652338811633e-07,
+      "loss": 2.0628,
+      "step": 57155
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9269911504424777e-07,
+      "loss": 1.8545,
+      "step": 57156
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9265170670037926e-07,
+      "loss": 2.0381,
+      "step": 57157
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9260429835651075e-07,
+      "loss": 0.3871,
+      "step": 57158
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9255689001264225e-07,
+      "loss": 3.0768,
+      "step": 57159
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9250948166877374e-07,
+      "loss": 2.1106,
+      "step": 57160
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.924620733249052e-07,
+      "loss": 2.1832,
+      "step": 57161
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9241466498103667e-07,
+      "loss": 0.6199,
+      "step": 57162
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9236725663716816e-07,
+      "loss": 1.9187,
+      "step": 57163
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9231984829329965e-07,
+      "loss": 2.4262,
+      "step": 57164
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.922724399494311e-07,
+      "loss": 0.9495,
+      "step": 57165
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.922250316055626e-07,
+      "loss": 3.6911,
+      "step": 57166
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.92177623261694e-07,
+      "loss": 0.9891,
+      "step": 57167
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9213021491782557e-07,
+      "loss": 1.7459,
+      "step": 57168
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9208280657395706e-07,
+      "loss": 0.9923,
+      "step": 57169
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.920353982300885e-07,
+      "loss": 1.6581,
+      "step": 57170
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9198798988622e-07,
+      "loss": 1.8456,
+      "step": 57171
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9194058154235143e-07,
+      "loss": 1.0776,
+      "step": 57172
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.918931731984829e-07,
+      "loss": 0.3166,
+      "step": 57173
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.918457648546144e-07,
+      "loss": 0.7202,
+      "step": 57174
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.917983565107459e-07,
+      "loss": 0.9384,
+      "step": 57175
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9175094816687735e-07,
+      "loss": 2.0841,
+      "step": 57176
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9170353982300884e-07,
+      "loss": 3.3395,
+      "step": 57177
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9165613147914033e-07,
+      "loss": 1.555,
+      "step": 57178
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.916087231352718e-07,
+      "loss": 1.012,
+      "step": 57179
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.915613147914033e-07,
+      "loss": 3.0097,
+      "step": 57180
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9151390644753476e-07,
+      "loss": 1.9989,
+      "step": 57181
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9146649810366625e-07,
+      "loss": 0.8549,
+      "step": 57182
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9141908975979774e-07,
+      "loss": 2.9764,
+      "step": 57183
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9137168141592923e-07,
+      "loss": 2.0899,
+      "step": 57184
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9132427307206067e-07,
+      "loss": 3.4059,
+      "step": 57185
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9127686472819216e-07,
+      "loss": 1.048,
+      "step": 57186
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9122945638432366e-07,
+      "loss": 2.0007,
+      "step": 57187
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9118204804045515e-07,
+      "loss": 1.8794,
+      "step": 57188
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.911346396965866e-07,
+      "loss": 2.1803,
+      "step": 57189
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.910872313527181e-07,
+      "loss": 2.2057,
+      "step": 57190
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9103982300884957e-07,
+      "loss": 1.093,
+      "step": 57191
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9099241466498106e-07,
+      "loss": 2.1399,
+      "step": 57192
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9094500632111256e-07,
+      "loss": 1.1615,
+      "step": 57193
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.90897597977244e-07,
+      "loss": 1.2164,
+      "step": 57194
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.908501896333755e-07,
+      "loss": 1.7801,
+      "step": 57195
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9080278128950693e-07,
+      "loss": 1.7004,
+      "step": 57196
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9075537294563847e-07,
+      "loss": 1.4351,
+      "step": 57197
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.907079646017699e-07,
+      "loss": 2.3203,
+      "step": 57198
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.906605562579014e-07,
+      "loss": 0.4691,
+      "step": 57199
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9061314791403284e-07,
+      "loss": 1.1787,
+      "step": 57200
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9056573957016434e-07,
+      "loss": 1.3658,
+      "step": 57201
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.905183312262959e-07,
+      "loss": 3.2065,
+      "step": 57202
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.904709228824273e-07,
+      "loss": 0.672,
+      "step": 57203
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.904235145385588e-07,
+      "loss": 0.6668,
+      "step": 57204
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9037610619469025e-07,
+      "loss": 1.2244,
+      "step": 57205
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9032869785082174e-07,
+      "loss": 1.0799,
+      "step": 57206
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9028128950695324e-07,
+      "loss": 1.7257,
+      "step": 57207
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9023388116308473e-07,
+      "loss": 3.2111,
+      "step": 57208
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9018647281921617e-07,
+      "loss": 3.8411,
+      "step": 57209
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9013906447534766e-07,
+      "loss": 3.2851,
+      "step": 57210
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9009165613147915e-07,
+      "loss": 1.5443,
+      "step": 57211
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.9004424778761064e-07,
+      "loss": 1.8794,
+      "step": 57212
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8999683944374214e-07,
+      "loss": 1.5046,
+      "step": 57213
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.899494310998736e-07,
+      "loss": 2.5938,
+      "step": 57214
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8990202275600507e-07,
+      "loss": 2.88,
+      "step": 57215
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8985461441213656e-07,
+      "loss": 2.3673,
+      "step": 57216
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8980720606826805e-07,
+      "loss": 1.7007,
+      "step": 57217
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.897597977243995e-07,
+      "loss": 2.1212,
+      "step": 57218
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.89712389380531e-07,
+      "loss": 1.2455,
+      "step": 57219
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.896649810366624e-07,
+      "loss": 2.1055,
+      "step": 57220
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8961757269279397e-07,
+      "loss": 1.2965,
+      "step": 57221
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.895701643489254e-07,
+      "loss": 1.2939,
+      "step": 57222
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.895227560050569e-07,
+      "loss": 3.4893,
+      "step": 57223
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.894753476611884e-07,
+      "loss": 1.3966,
+      "step": 57224
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8942793931731983e-07,
+      "loss": 0.5205,
+      "step": 57225
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.893805309734514e-07,
+      "loss": 1.5789,
+      "step": 57226
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.893331226295828e-07,
+      "loss": 1.5847,
+      "step": 57227
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.892857142857143e-07,
+      "loss": 1.3213,
+      "step": 57228
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8923830594184575e-07,
+      "loss": 2.2373,
+      "step": 57229
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8919089759797724e-07,
+      "loss": 2.3144,
+      "step": 57230
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8914348925410873e-07,
+      "loss": 2.3447,
+      "step": 57231
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.890960809102402e-07,
+      "loss": 1.1016,
+      "step": 57232
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8904867256637166e-07,
+      "loss": 1.6515,
+      "step": 57233
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8900126422250315e-07,
+      "loss": 1.9118,
+      "step": 57234
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8895385587863465e-07,
+      "loss": 1.5531,
+      "step": 57235
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8890644753476614e-07,
+      "loss": 2.3459,
+      "step": 57236
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8885903919089763e-07,
+      "loss": 2.5916,
+      "step": 57237
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8881163084702907e-07,
+      "loss": 1.8687,
+      "step": 57238
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8876422250316056e-07,
+      "loss": 1.2483,
+      "step": 57239
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8871681415929205e-07,
+      "loss": 1.6352,
+      "step": 57240
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8866940581542355e-07,
+      "loss": 1.9972,
+      "step": 57241
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.88621997471555e-07,
+      "loss": 1.2475,
+      "step": 57242
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.885745891276865e-07,
+      "loss": 1.2234,
+      "step": 57243
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.885271807838179e-07,
+      "loss": 1.768,
+      "step": 57244
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8847977243994946e-07,
+      "loss": 1.4768,
+      "step": 57245
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.884323640960809e-07,
+      "loss": 1.9392,
+      "step": 57246
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.883849557522124e-07,
+      "loss": 1.1913,
+      "step": 57247
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.883375474083439e-07,
+      "loss": 1.6991,
+      "step": 57248
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.882901390644753e-07,
+      "loss": 0.9489,
+      "step": 57249
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8824273072060687e-07,
+      "loss": 1.6908,
+      "step": 57250
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.881953223767383e-07,
+      "loss": 1.3609,
+      "step": 57251
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.881479140328698e-07,
+      "loss": 2.448,
+      "step": 57252
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8810050568900124e-07,
+      "loss": 0.8837,
+      "step": 57253
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8805309734513273e-07,
+      "loss": 1.6683,
+      "step": 57254
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.880056890012642e-07,
+      "loss": 3.9896,
+      "step": 57255
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.879582806573957e-07,
+      "loss": 1.7898,
+      "step": 57256
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8791087231352716e-07,
+      "loss": 1.4776,
+      "step": 57257
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8786346396965865e-07,
+      "loss": 2.4773,
+      "step": 57258
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8781605562579014e-07,
+      "loss": 1.6754,
+      "step": 57259
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8776864728192163e-07,
+      "loss": 0.699,
+      "step": 57260
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.877212389380531e-07,
+      "loss": 0.7218,
+      "step": 57261
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8767383059418456e-07,
+      "loss": 3.8892,
+      "step": 57262
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8762642225031606e-07,
+      "loss": 1.5744,
+      "step": 57263
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8757901390644755e-07,
+      "loss": 2.7127,
+      "step": 57264
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8753160556257904e-07,
+      "loss": 1.1831,
+      "step": 57265
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.874841972187105e-07,
+      "loss": 1.4118,
+      "step": 57266
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8743678887484197e-07,
+      "loss": 0.4535,
+      "step": 57267
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8738938053097346e-07,
+      "loss": 1.3049,
+      "step": 57268
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8734197218710496e-07,
+      "loss": 1.3661,
+      "step": 57269
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8729456384323645e-07,
+      "loss": 2.5857,
+      "step": 57270
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.872471554993679e-07,
+      "loss": 1.5478,
+      "step": 57271
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.871997471554994e-07,
+      "loss": 0.8009,
+      "step": 57272
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8715233881163087e-07,
+      "loss": 1.7781,
+      "step": 57273
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8710493046776236e-07,
+      "loss": 0.832,
+      "step": 57274
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.870575221238938e-07,
+      "loss": 1.265,
+      "step": 57275
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.870101137800253e-07,
+      "loss": 1.1632,
+      "step": 57276
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8696270543615674e-07,
+      "loss": 1.2592,
+      "step": 57277
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8691529709228823e-07,
+      "loss": 2.7647,
+      "step": 57278
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.868678887484197e-07,
+      "loss": 1.3074,
+      "step": 57279
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.868204804045512e-07,
+      "loss": 1.7479,
+      "step": 57280
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.867730720606827e-07,
+      "loss": 2.4066,
+      "step": 57281
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8672566371681414e-07,
+      "loss": 2.6181,
+      "step": 57282
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8667825537294564e-07,
+      "loss": 0.9467,
+      "step": 57283
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8663084702907713e-07,
+      "loss": 3.2057,
+      "step": 57284
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.865834386852086e-07,
+      "loss": 1.9029,
+      "step": 57285
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8653603034134006e-07,
+      "loss": 2.4449,
+      "step": 57286
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8648862199747155e-07,
+      "loss": 2.3418,
+      "step": 57287
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8644121365360304e-07,
+      "loss": 3.2245,
+      "step": 57288
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8639380530973454e-07,
+      "loss": 0.8318,
+      "step": 57289
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.86346396965866e-07,
+      "loss": 1.6042,
+      "step": 57290
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8629898862199747e-07,
+      "loss": 1.214,
+      "step": 57291
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8625158027812896e-07,
+      "loss": 3.2719,
+      "step": 57292
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8620417193426045e-07,
+      "loss": 1.2548,
+      "step": 57293
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8615676359039194e-07,
+      "loss": 0.6698,
+      "step": 57294
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.861093552465234e-07,
+      "loss": 2.7827,
+      "step": 57295
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.860619469026549e-07,
+      "loss": 0.4385,
+      "step": 57296
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8601453855878637e-07,
+      "loss": 3.2405,
+      "step": 57297
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8596713021491786e-07,
+      "loss": 2.1274,
+      "step": 57298
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.859197218710493e-07,
+      "loss": 1.2704,
+      "step": 57299
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.858723135271808e-07,
+      "loss": 1.9603,
+      "step": 57300
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8582490518331223e-07,
+      "loss": 2.2134,
+      "step": 57301
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.857774968394438e-07,
+      "loss": 0.9219,
+      "step": 57302
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8573008849557527e-07,
+      "loss": 2.3923,
+      "step": 57303
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.856826801517067e-07,
+      "loss": 2.7919,
+      "step": 57304
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.856352718078382e-07,
+      "loss": 2.746,
+      "step": 57305
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8558786346396964e-07,
+      "loss": 1.182,
+      "step": 57306
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.855404551201012e-07,
+      "loss": 1.3271,
+      "step": 57307
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.854930467762326e-07,
+      "loss": 3.7078,
+      "step": 57308
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.854456384323641e-07,
+      "loss": 2.7321,
+      "step": 57309
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8539823008849555e-07,
+      "loss": 1.2442,
+      "step": 57310
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8535082174462705e-07,
+      "loss": 1.698,
+      "step": 57311
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8530341340075854e-07,
+      "loss": 0.9779,
+      "step": 57312
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8525600505689003e-07,
+      "loss": 3.3166,
+      "step": 57313
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.852085967130215e-07,
+      "loss": 1.9804,
+      "step": 57314
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8516118836915296e-07,
+      "loss": 3.1931,
+      "step": 57315
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8511378002528445e-07,
+      "loss": 1.2859,
+      "step": 57316
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8506637168141595e-07,
+      "loss": 1.5855,
+      "step": 57317
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8501896333754744e-07,
+      "loss": 1.9764,
+      "step": 57318
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.849715549936789e-07,
+      "loss": 1.8326,
+      "step": 57319
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8492414664981037e-07,
+      "loss": 1.1383,
+      "step": 57320
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8487673830594186e-07,
+      "loss": 1.8629,
+      "step": 57321
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8482932996207335e-07,
+      "loss": 2.036,
+      "step": 57322
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.847819216182048e-07,
+      "loss": 2.4822,
+      "step": 57323
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.847345132743363e-07,
+      "loss": 1.0668,
+      "step": 57324
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.846871049304678e-07,
+      "loss": 1.2502,
+      "step": 57325
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8463969658659927e-07,
+      "loss": 2.0779,
+      "step": 57326
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8459228824273076e-07,
+      "loss": 1.6002,
+      "step": 57327
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.845448798988622e-07,
+      "loss": 0.6778,
+      "step": 57328
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.844974715549937e-07,
+      "loss": 1.0134,
+      "step": 57329
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8445006321112513e-07,
+      "loss": 1.0494,
+      "step": 57330
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.844026548672567e-07,
+      "loss": 2.4669,
+      "step": 57331
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.843552465233881e-07,
+      "loss": 2.2895,
+      "step": 57332
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.843078381795196e-07,
+      "loss": 2.0532,
+      "step": 57333
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8426042983565105e-07,
+      "loss": 1.6836,
+      "step": 57334
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8421302149178254e-07,
+      "loss": 1.6659,
+      "step": 57335
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.841656131479141e-07,
+      "loss": 1.6591,
+      "step": 57336
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.841182048040455e-07,
+      "loss": 2.2207,
+      "step": 57337
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.84070796460177e-07,
+      "loss": 2.2004,
+      "step": 57338
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8402338811630846e-07,
+      "loss": 3.698,
+      "step": 57339
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8397597977243995e-07,
+      "loss": 1.8918,
+      "step": 57340
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8392857142857144e-07,
+      "loss": 2.1756,
+      "step": 57341
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8388116308470293e-07,
+      "loss": 1.1446,
+      "step": 57342
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8383375474083437e-07,
+      "loss": 2.3948,
+      "step": 57343
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8378634639696586e-07,
+      "loss": 1.0346,
+      "step": 57344
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8373893805309736e-07,
+      "loss": 1.2284,
+      "step": 57345
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8369152970922885e-07,
+      "loss": 1.2171,
+      "step": 57346
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.836441213653603e-07,
+      "loss": 1.798,
+      "step": 57347
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.835967130214918e-07,
+      "loss": 0.837,
+      "step": 57348
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8354930467762327e-07,
+      "loss": 1.932,
+      "step": 57349
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8350189633375476e-07,
+      "loss": 2.9736,
+      "step": 57350
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8345448798988626e-07,
+      "loss": 1.6875,
+      "step": 57351
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.834070796460177e-07,
+      "loss": 2.4509,
+      "step": 57352
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.833596713021492e-07,
+      "loss": 0.6303,
+      "step": 57353
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8331226295828063e-07,
+      "loss": 1.0229,
+      "step": 57354
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8326485461441217e-07,
+      "loss": 2.3622,
+      "step": 57355
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.832174462705436e-07,
+      "loss": 2.4728,
+      "step": 57356
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.831700379266751e-07,
+      "loss": 0.2229,
+      "step": 57357
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8312262958280654e-07,
+      "loss": 0.9048,
+      "step": 57358
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8307522123893803e-07,
+      "loss": 0.9971,
+      "step": 57359
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.830278128950696e-07,
+      "loss": 2.0784,
+      "step": 57360
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.82980404551201e-07,
+      "loss": 1.3835,
+      "step": 57361
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.829329962073325e-07,
+      "loss": 2.5218,
+      "step": 57362
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8288558786346395e-07,
+      "loss": 0.9216,
+      "step": 57363
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8283817951959544e-07,
+      "loss": 2.9325,
+      "step": 57364
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8279077117572693e-07,
+      "loss": 0.7989,
+      "step": 57365
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8274336283185843e-07,
+      "loss": 1.0023,
+      "step": 57366
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8269595448798987e-07,
+      "loss": 1.2011,
+      "step": 57367
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8264854614412136e-07,
+      "loss": 1.9509,
+      "step": 57368
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8260113780025285e-07,
+      "loss": 2.5116,
+      "step": 57369
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8255372945638434e-07,
+      "loss": 1.5192,
+      "step": 57370
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8250632111251584e-07,
+      "loss": 1.0972,
+      "step": 57371
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.824589127686473e-07,
+      "loss": 0.4949,
+      "step": 57372
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8241150442477877e-07,
+      "loss": 1.4589,
+      "step": 57373
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8236409608091026e-07,
+      "loss": 1.2271,
+      "step": 57374
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8231668773704175e-07,
+      "loss": 1.6301,
+      "step": 57375
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.822692793931732e-07,
+      "loss": 1.7487,
+      "step": 57376
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.822218710493047e-07,
+      "loss": 2.4314,
+      "step": 57377
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.821744627054362e-07,
+      "loss": 2.0423,
+      "step": 57378
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8212705436156767e-07,
+      "loss": 1.4976,
+      "step": 57379
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.820796460176991e-07,
+      "loss": 1.0046,
+      "step": 57380
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.820322376738306e-07,
+      "loss": 0.7025,
+      "step": 57381
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.819848293299621e-07,
+      "loss": 3.0837,
+      "step": 57382
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8193742098609353e-07,
+      "loss": 3.1177,
+      "step": 57383
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.818900126422251e-07,
+      "loss": 2.3121,
+      "step": 57384
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.818426042983565e-07,
+      "loss": 2.1464,
+      "step": 57385
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.81795195954488e-07,
+      "loss": 0.7834,
+      "step": 57386
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8174778761061945e-07,
+      "loss": 3.2351,
+      "step": 57387
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8170037926675094e-07,
+      "loss": 1.2416,
+      "step": 57388
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8165297092288243e-07,
+      "loss": 1.0559,
+      "step": 57389
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.816055625790139e-07,
+      "loss": 1.9146,
+      "step": 57390
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8155815423514536e-07,
+      "loss": 0.9364,
+      "step": 57391
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8151074589127685e-07,
+      "loss": 2.514,
+      "step": 57392
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8146333754740835e-07,
+      "loss": 1.6795,
+      "step": 57393
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8141592920353984e-07,
+      "loss": 1.3605,
+      "step": 57394
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8136852085967133e-07,
+      "loss": 1.3753,
+      "step": 57395
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8132111251580277e-07,
+      "loss": 2.1176,
+      "step": 57396
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8127370417193426e-07,
+      "loss": 3.058,
+      "step": 57397
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8122629582806575e-07,
+      "loss": 1.2506,
+      "step": 57398
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8117888748419725e-07,
+      "loss": 2.0641,
+      "step": 57399
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.811314791403287e-07,
+      "loss": 1.2928,
+      "step": 57400
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.810840707964602e-07,
+      "loss": 2.2087,
+      "step": 57401
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8103666245259167e-07,
+      "loss": 2.0445,
+      "step": 57402
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8098925410872316e-07,
+      "loss": 1.3034,
+      "step": 57403
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8094184576485465e-07,
+      "loss": 3.8649,
+      "step": 57404
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.808944374209861e-07,
+      "loss": 2.1701,
+      "step": 57405
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.808470290771176e-07,
+      "loss": 1.5298,
+      "step": 57406
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.807996207332491e-07,
+      "loss": 2.5773,
+      "step": 57407
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8075221238938057e-07,
+      "loss": 3.2789,
+      "step": 57408
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.80704804045512e-07,
+      "loss": 2.0103,
+      "step": 57409
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.806573957016435e-07,
+      "loss": 0.6995,
+      "step": 57410
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8060998735777494e-07,
+      "loss": 2.4401,
+      "step": 57411
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.805625790139065e-07,
+      "loss": 2.3164,
+      "step": 57412
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.805151706700379e-07,
+      "loss": 0.8946,
+      "step": 57413
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.804677623261694e-07,
+      "loss": 2.105,
+      "step": 57414
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.804203539823009e-07,
+      "loss": 3.1243,
+      "step": 57415
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8037294563843235e-07,
+      "loss": 0.6424,
+      "step": 57416
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8032553729456384e-07,
+      "loss": 1.2743,
+      "step": 57417
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8027812895069533e-07,
+      "loss": 1.1795,
+      "step": 57418
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.802307206068268e-07,
+      "loss": 1.7496,
+      "step": 57419
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8018331226295826e-07,
+      "loss": 1.408,
+      "step": 57420
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8013590391908976e-07,
+      "loss": 3.3199,
+      "step": 57421
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8008849557522125e-07,
+      "loss": 1.2271,
+      "step": 57422
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8004108723135274e-07,
+      "loss": 2.5194,
+      "step": 57423
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.799936788874842e-07,
+      "loss": 0.5025,
+      "step": 57424
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7994627054361567e-07,
+      "loss": 1.0603,
+      "step": 57425
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7989886219974716e-07,
+      "loss": 1.4223,
+      "step": 57426
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7985145385587866e-07,
+      "loss": 1.6711,
+      "step": 57427
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7980404551201015e-07,
+      "loss": 2.484,
+      "step": 57428
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.797566371681416e-07,
+      "loss": 1.5039,
+      "step": 57429
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.797092288242731e-07,
+      "loss": 2.4784,
+      "step": 57430
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7966182048040457e-07,
+      "loss": 2.2886,
+      "step": 57431
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7961441213653606e-07,
+      "loss": 1.2846,
+      "step": 57432
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.795670037926675e-07,
+      "loss": 2.1035,
+      "step": 57433
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.79519595448799e-07,
+      "loss": 1.576,
+      "step": 57434
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7947218710493043e-07,
+      "loss": 0.6236,
+      "step": 57435
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.79424778761062e-07,
+      "loss": 1.3531,
+      "step": 57436
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7937737041719347e-07,
+      "loss": 3.9122,
+      "step": 57437
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.793299620733249e-07,
+      "loss": 0.6337,
+      "step": 57438
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.792825537294564e-07,
+      "loss": 1.729,
+      "step": 57439
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7923514538558784e-07,
+      "loss": 2.0672,
+      "step": 57440
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.791877370417194e-07,
+      "loss": 3.4152,
+      "step": 57441
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7914032869785083e-07,
+      "loss": 1.1354,
+      "step": 57442
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.790929203539823e-07,
+      "loss": 1.77,
+      "step": 57443
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7904551201011376e-07,
+      "loss": 2.0099,
+      "step": 57444
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7899810366624525e-07,
+      "loss": 1.6056,
+      "step": 57445
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7895069532237674e-07,
+      "loss": 0.7777,
+      "step": 57446
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7890328697850823e-07,
+      "loss": 1.3103,
+      "step": 57447
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7885587863463973e-07,
+      "loss": 3.1732,
+      "step": 57448
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7880847029077117e-07,
+      "loss": 2.7879,
+      "step": 57449
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7876106194690266e-07,
+      "loss": 1.0385,
+      "step": 57450
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7871365360303415e-07,
+      "loss": 2.083,
+      "step": 57451
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7866624525916564e-07,
+      "loss": 1.5336,
+      "step": 57452
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.786188369152971e-07,
+      "loss": 1.4198,
+      "step": 57453
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.785714285714286e-07,
+      "loss": 2.618,
+      "step": 57454
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7852402022756007e-07,
+      "loss": 1.7443,
+      "step": 57455
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7847661188369156e-07,
+      "loss": 1.9436,
+      "step": 57456
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.78429203539823e-07,
+      "loss": 1.8369,
+      "step": 57457
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.783817951959545e-07,
+      "loss": 1.1486,
+      "step": 57458
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7833438685208593e-07,
+      "loss": 1.435,
+      "step": 57459
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.782869785082175e-07,
+      "loss": 1.2506,
+      "step": 57460
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7823957016434897e-07,
+      "loss": 1.8916,
+      "step": 57461
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.781921618204804e-07,
+      "loss": 1.4563,
+      "step": 57462
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.781447534766119e-07,
+      "loss": 3.2039,
+      "step": 57463
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7809734513274334e-07,
+      "loss": 1.241,
+      "step": 57464
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.780499367888749e-07,
+      "loss": 0.9484,
+      "step": 57465
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.780025284450063e-07,
+      "loss": 0.9215,
+      "step": 57466
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.779551201011378e-07,
+      "loss": 2.4788,
+      "step": 57467
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7790771175726925e-07,
+      "loss": 1.868,
+      "step": 57468
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7786030341340074e-07,
+      "loss": 0.8713,
+      "step": 57469
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7781289506953224e-07,
+      "loss": 0.8838,
+      "step": 57470
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7776548672566373e-07,
+      "loss": 1.48,
+      "step": 57471
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.777180783817952e-07,
+      "loss": 1.6957,
+      "step": 57472
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7767067003792666e-07,
+      "loss": 0.6526,
+      "step": 57473
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7762326169405815e-07,
+      "loss": 1.7,
+      "step": 57474
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7757585335018964e-07,
+      "loss": 3.4642,
+      "step": 57475
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7752844500632114e-07,
+      "loss": 1.5038,
+      "step": 57476
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.774810366624526e-07,
+      "loss": 1.2231,
+      "step": 57477
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7743362831858407e-07,
+      "loss": 1.5174,
+      "step": 57478
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7738621997471556e-07,
+      "loss": 3.0949,
+      "step": 57479
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7733881163084705e-07,
+      "loss": 2.9018,
+      "step": 57480
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.772914032869785e-07,
+      "loss": 1.7801,
+      "step": 57481
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7724399494311e-07,
+      "loss": 1.5271,
+      "step": 57482
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.771965865992415e-07,
+      "loss": 1.6298,
+      "step": 57483
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7714917825537297e-07,
+      "loss": 1.9938,
+      "step": 57484
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7710176991150446e-07,
+      "loss": 3.1601,
+      "step": 57485
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.770543615676359e-07,
+      "loss": 2.6673,
+      "step": 57486
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.770069532237674e-07,
+      "loss": 2.4815,
+      "step": 57487
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7695954487989883e-07,
+      "loss": 1.0421,
+      "step": 57488
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.769121365360304e-07,
+      "loss": 3.2737,
+      "step": 57489
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.768647281921618e-07,
+      "loss": 1.4372,
+      "step": 57490
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.768173198482933e-07,
+      "loss": 1.5191,
+      "step": 57491
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7676991150442475e-07,
+      "loss": 1.3571,
+      "step": 57492
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7672250316055624e-07,
+      "loss": 2.451,
+      "step": 57493
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.766750948166878e-07,
+      "loss": 1.5408,
+      "step": 57494
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.766276864728192e-07,
+      "loss": 1.3871,
+      "step": 57495
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.765802781289507e-07,
+      "loss": 2.2275,
+      "step": 57496
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7653286978508216e-07,
+      "loss": 1.0062,
+      "step": 57497
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7648546144121365e-07,
+      "loss": 1.4036,
+      "step": 57498
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7643805309734514e-07,
+      "loss": 4.0229,
+      "step": 57499
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7639064475347663e-07,
+      "loss": 3.1418,
+      "step": 57500
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7634323640960807e-07,
+      "loss": 1.2867,
+      "step": 57501
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7629582806573956e-07,
+      "loss": 1.8917,
+      "step": 57502
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7624841972187106e-07,
+      "loss": 2.6816,
+      "step": 57503
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7620101137800255e-07,
+      "loss": 3.143,
+      "step": 57504
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7615360303413404e-07,
+      "loss": 2.499,
+      "step": 57505
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.761061946902655e-07,
+      "loss": 1.4353,
+      "step": 57506
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7605878634639697e-07,
+      "loss": 3.4636,
+      "step": 57507
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7601137800252846e-07,
+      "loss": 0.4993,
+      "step": 57508
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7596396965865996e-07,
+      "loss": 2.4281,
+      "step": 57509
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.759165613147914e-07,
+      "loss": 3.0072,
+      "step": 57510
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.758691529709229e-07,
+      "loss": 0.5861,
+      "step": 57511
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.758217446270544e-07,
+      "loss": 1.7104,
+      "step": 57512
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7577433628318587e-07,
+      "loss": 1.8323,
+      "step": 57513
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.757269279393173e-07,
+      "loss": 2.6396,
+      "step": 57514
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.756795195954488e-07,
+      "loss": 1.6934,
+      "step": 57515
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.756321112515803e-07,
+      "loss": 2.3636,
+      "step": 57516
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.755847029077118e-07,
+      "loss": 2.284,
+      "step": 57517
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.755372945638433e-07,
+      "loss": 1.3165,
+      "step": 57518
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.754898862199747e-07,
+      "loss": 1.0388,
+      "step": 57519
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.754424778761062e-07,
+      "loss": 1.6853,
+      "step": 57520
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7539506953223765e-07,
+      "loss": 2.1775,
+      "step": 57521
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7534766118836914e-07,
+      "loss": 1.0503,
+      "step": 57522
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7530025284450063e-07,
+      "loss": 1.5603,
+      "step": 57523
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.752528445006321e-07,
+      "loss": 1.8781,
+      "step": 57524
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7520543615676357e-07,
+      "loss": 1.4884,
+      "step": 57525
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7515802781289506e-07,
+      "loss": 1.6501,
+      "step": 57526
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7511061946902655e-07,
+      "loss": 1.0023,
+      "step": 57527
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7506321112515804e-07,
+      "loss": 2.7273,
+      "step": 57528
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7501580278128953e-07,
+      "loss": 1.6964,
+      "step": 57529
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7496839443742097e-07,
+      "loss": 2.0168,
+      "step": 57530
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7492098609355247e-07,
+      "loss": 2.1691,
+      "step": 57531
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7487357774968396e-07,
+      "loss": 1.1104,
+      "step": 57532
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7482616940581545e-07,
+      "loss": 1.9091,
+      "step": 57533
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.747787610619469e-07,
+      "loss": 3.1718,
+      "step": 57534
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.747313527180784e-07,
+      "loss": 1.6514,
+      "step": 57535
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.746839443742099e-07,
+      "loss": 3.0117,
+      "step": 57536
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7463653603034137e-07,
+      "loss": 2.1573,
+      "step": 57537
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7458912768647286e-07,
+      "loss": 1.4182,
+      "step": 57538
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.745417193426043e-07,
+      "loss": 1.8052,
+      "step": 57539
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.744943109987358e-07,
+      "loss": 2.6138,
+      "step": 57540
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.744469026548673e-07,
+      "loss": 2.6014,
+      "step": 57541
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.743994943109988e-07,
+      "loss": 1.734,
+      "step": 57542
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.743520859671302e-07,
+      "loss": 2.2023,
+      "step": 57543
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.743046776232617e-07,
+      "loss": 1.773,
+      "step": 57544
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7425726927939314e-07,
+      "loss": 2.3796,
+      "step": 57545
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.742098609355247e-07,
+      "loss": 1.1478,
+      "step": 57546
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7416245259165613e-07,
+      "loss": 1.7641,
+      "step": 57547
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.741150442477876e-07,
+      "loss": 1.9777,
+      "step": 57548
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.740676359039191e-07,
+      "loss": 1.0441,
+      "step": 57549
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7402022756005055e-07,
+      "loss": 0.7799,
+      "step": 57550
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.739728192161821e-07,
+      "loss": 1.8122,
+      "step": 57551
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7392541087231354e-07,
+      "loss": 0.3313,
+      "step": 57552
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7387800252844503e-07,
+      "loss": 0.8828,
+      "step": 57553
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7383059418457647e-07,
+      "loss": 1.1419,
+      "step": 57554
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7378318584070796e-07,
+      "loss": 1.8113,
+      "step": 57555
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7373577749683945e-07,
+      "loss": 2.8982,
+      "step": 57556
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7368836915297094e-07,
+      "loss": 1.7544,
+      "step": 57557
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.736409608091024e-07,
+      "loss": 0.8889,
+      "step": 57558
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.735935524652339e-07,
+      "loss": 2.2809,
+      "step": 57559
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7354614412136537e-07,
+      "loss": 1.5191,
+      "step": 57560
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7349873577749686e-07,
+      "loss": 1.8546,
+      "step": 57561
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7345132743362835e-07,
+      "loss": 4.2583,
+      "step": 57562
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.734039190897598e-07,
+      "loss": 1.6136,
+      "step": 57563
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.733565107458913e-07,
+      "loss": 2.3678,
+      "step": 57564
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.733091024020228e-07,
+      "loss": 1.5917,
+      "step": 57565
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7326169405815427e-07,
+      "loss": 3.3416,
+      "step": 57566
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.732142857142857e-07,
+      "loss": 1.6557,
+      "step": 57567
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.731668773704172e-07,
+      "loss": 2.149,
+      "step": 57568
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7311946902654864e-07,
+      "loss": 1.2693,
+      "step": 57569
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.730720606826802e-07,
+      "loss": 2.2575,
+      "step": 57570
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.730246523388116e-07,
+      "loss": 1.9706,
+      "step": 57571
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.729772439949431e-07,
+      "loss": 1.4234,
+      "step": 57572
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.729298356510746e-07,
+      "loss": 1.4151,
+      "step": 57573
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7288242730720605e-07,
+      "loss": 2.3311,
+      "step": 57574
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.728350189633376e-07,
+      "loss": 1.6694,
+      "step": 57575
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7278761061946903e-07,
+      "loss": 1.8986,
+      "step": 57576
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.727402022756005e-07,
+      "loss": 0.9694,
+      "step": 57577
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7269279393173196e-07,
+      "loss": 1.537,
+      "step": 57578
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7264538558786345e-07,
+      "loss": 1.4769,
+      "step": 57579
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7259797724399495e-07,
+      "loss": 0.8791,
+      "step": 57580
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7255056890012644e-07,
+      "loss": 0.8489,
+      "step": 57581
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.725031605562579e-07,
+      "loss": 1.5917,
+      "step": 57582
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7245575221238937e-07,
+      "loss": 1.9567,
+      "step": 57583
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7240834386852086e-07,
+      "loss": 3.7863,
+      "step": 57584
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7236093552465236e-07,
+      "loss": 1.3116,
+      "step": 57585
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7231352718078385e-07,
+      "loss": 1.7582,
+      "step": 57586
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.722661188369153e-07,
+      "loss": 3.6799,
+      "step": 57587
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.722187104930468e-07,
+      "loss": 1.2611,
+      "step": 57588
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7217130214917827e-07,
+      "loss": 2.1263,
+      "step": 57589
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7212389380530976e-07,
+      "loss": 2.6672,
+      "step": 57590
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.720764854614412e-07,
+      "loss": 1.4696,
+      "step": 57591
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.720290771175727e-07,
+      "loss": 1.0652,
+      "step": 57592
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7198166877370413e-07,
+      "loss": 2.0801,
+      "step": 57593
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.719342604298357e-07,
+      "loss": 1.0925,
+      "step": 57594
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7188685208596717e-07,
+      "loss": 0.5697,
+      "step": 57595
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.718394437420986e-07,
+      "loss": 0.9347,
+      "step": 57596
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.717920353982301e-07,
+      "loss": 1.5095,
+      "step": 57597
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7174462705436154e-07,
+      "loss": 2.3209,
+      "step": 57598
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.716972187104931e-07,
+      "loss": 1.8907,
+      "step": 57599
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.716498103666245e-07,
+      "loss": 1.8135,
+      "step": 57600
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.71602402022756e-07,
+      "loss": 2.6507,
+      "step": 57601
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7155499367888746e-07,
+      "loss": 0.9727,
+      "step": 57602
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7150758533501895e-07,
+      "loss": 3.1665,
+      "step": 57603
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7146017699115044e-07,
+      "loss": 1.4378,
+      "step": 57604
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7141276864728193e-07,
+      "loss": 2.4194,
+      "step": 57605
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.713653603034134e-07,
+      "loss": 1.1915,
+      "step": 57606
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7131795195954487e-07,
+      "loss": 2.5713,
+      "step": 57607
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7127054361567636e-07,
+      "loss": 1.6919,
+      "step": 57608
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7122313527180785e-07,
+      "loss": 1.1967,
+      "step": 57609
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7117572692793934e-07,
+      "loss": 0.9423,
+      "step": 57610
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.711283185840708e-07,
+      "loss": 2.1369,
+      "step": 57611
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7108091024020227e-07,
+      "loss": 1.1544,
+      "step": 57612
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7103350189633377e-07,
+      "loss": 2.1612,
+      "step": 57613
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7098609355246526e-07,
+      "loss": 2.2601,
+      "step": 57614
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.709386852085967e-07,
+      "loss": 1.4524,
+      "step": 57615
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.708912768647282e-07,
+      "loss": 0.5186,
+      "step": 57616
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.708438685208597e-07,
+      "loss": 1.4526,
+      "step": 57617
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7079646017699117e-07,
+      "loss": 0.7374,
+      "step": 57618
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7074905183312267e-07,
+      "loss": 0.8221,
+      "step": 57619
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.707016434892541e-07,
+      "loss": 1.2395,
+      "step": 57620
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.706542351453856e-07,
+      "loss": 1.483,
+      "step": 57621
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.706068268015171e-07,
+      "loss": 2.7818,
+      "step": 57622
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.705594184576486e-07,
+      "loss": 1.996,
+      "step": 57623
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7051201011378e-07,
+      "loss": 0.7826,
+      "step": 57624
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.704646017699115e-07,
+      "loss": 2.9715,
+      "step": 57625
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7041719342604295e-07,
+      "loss": 2.7228,
+      "step": 57626
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7036978508217444e-07,
+      "loss": 2.2695,
+      "step": 57627
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.70322376738306e-07,
+      "loss": 2.0149,
+      "step": 57628
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7027496839443743e-07,
+      "loss": 2.0927,
+      "step": 57629
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.702275600505689e-07,
+      "loss": 1.0982,
+      "step": 57630
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7018015170670036e-07,
+      "loss": 1.8043,
+      "step": 57631
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7013274336283185e-07,
+      "loss": 1.9734,
+      "step": 57632
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7008533501896334e-07,
+      "loss": 2.8266,
+      "step": 57633
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7003792667509484e-07,
+      "loss": 1.7869,
+      "step": 57634
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.699905183312263e-07,
+      "loss": 1.002,
+      "step": 57635
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6994310998735777e-07,
+      "loss": 1.8855,
+      "step": 57636
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6989570164348926e-07,
+      "loss": 2.0749,
+      "step": 57637
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6984829329962075e-07,
+      "loss": 1.5739,
+      "step": 57638
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6980088495575224e-07,
+      "loss": 3.5003,
+      "step": 57639
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.697534766118837e-07,
+      "loss": 1.2435,
+      "step": 57640
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.697060682680152e-07,
+      "loss": 2.5468,
+      "step": 57641
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6965865992414667e-07,
+      "loss": 2.3281,
+      "step": 57642
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6961125158027816e-07,
+      "loss": 2.2077,
+      "step": 57643
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.695638432364096e-07,
+      "loss": 2.6966,
+      "step": 57644
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.695164348925411e-07,
+      "loss": 0.6902,
+      "step": 57645
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.694690265486726e-07,
+      "loss": 3.6114,
+      "step": 57646
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.694216182048041e-07,
+      "loss": 1.2422,
+      "step": 57647
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.693742098609355e-07,
+      "loss": 1.1997,
+      "step": 57648
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.69326801517067e-07,
+      "loss": 1.5008,
+      "step": 57649
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.692793931731985e-07,
+      "loss": 2.0354,
+      "step": 57650
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6923198482933e-07,
+      "loss": 2.6727,
+      "step": 57651
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.691845764854615e-07,
+      "loss": 1.767,
+      "step": 57652
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.691371681415929e-07,
+      "loss": 2.7578,
+      "step": 57653
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.690897597977244e-07,
+      "loss": 2.4695,
+      "step": 57654
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6904235145385585e-07,
+      "loss": 0.4534,
+      "step": 57655
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.689949431099874e-07,
+      "loss": 2.1624,
+      "step": 57656
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6894753476611884e-07,
+      "loss": 0.9418,
+      "step": 57657
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6890012642225033e-07,
+      "loss": 1.1956,
+      "step": 57658
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6885271807838177e-07,
+      "loss": 0.9765,
+      "step": 57659
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6880530973451326e-07,
+      "loss": 1.756,
+      "step": 57660
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.6875790139064475e-07,
+      "loss": 2.0938,
+      "step": 57661
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6871049304677625e-07,
+      "loss": 3.2939,
+      "step": 57662
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6866308470290774e-07,
+      "loss": 2.75,
+      "step": 57663
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.686156763590392e-07,
+      "loss": 2.2211,
+      "step": 57664
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6856826801517067e-07,
+      "loss": 0.3234,
+      "step": 57665
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6852085967130216e-07,
+      "loss": 2.5074,
+      "step": 57666
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6847345132743365e-07,
+      "loss": 1.1697,
+      "step": 57667
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.684260429835651e-07,
+      "loss": 2.7101,
+      "step": 57668
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.683786346396966e-07,
+      "loss": 1.7613,
+      "step": 57669
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.683312262958281e-07,
+      "loss": 2.2933,
+      "step": 57670
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6828381795195957e-07,
+      "loss": 1.0392,
+      "step": 57671
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6823640960809106e-07,
+      "loss": 2.7877,
+      "step": 57672
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.681890012642225e-07,
+      "loss": 2.4488,
+      "step": 57673
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.68141592920354e-07,
+      "loss": 1.9406,
+      "step": 57674
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.680941845764855e-07,
+      "loss": 2.9498,
+      "step": 57675
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.68046776232617e-07,
+      "loss": 2.4521,
+      "step": 57676
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.679993678887484e-07,
+      "loss": 2.7028,
+      "step": 57677
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.679519595448799e-07,
+      "loss": 1.0871,
+      "step": 57678
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6790455120101135e-07,
+      "loss": 1.4913,
+      "step": 57679
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.678571428571429e-07,
+      "loss": 1.6976,
+      "step": 57680
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6780973451327433e-07,
+      "loss": 1.7236,
+      "step": 57681
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.677623261694058e-07,
+      "loss": 1.6498,
+      "step": 57682
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6771491782553726e-07,
+      "loss": 3.5196,
+      "step": 57683
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6766750948166876e-07,
+      "loss": 1.2525,
+      "step": 57684
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.676201011378003e-07,
+      "loss": 2.2436,
+      "step": 57685
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6757269279393174e-07,
+      "loss": 2.0561,
+      "step": 57686
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6752528445006323e-07,
+      "loss": 1.2037,
+      "step": 57687
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6747787610619467e-07,
+      "loss": 3.7705,
+      "step": 57688
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6743046776232616e-07,
+      "loss": 1.5965,
+      "step": 57689
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6738305941845766e-07,
+      "loss": 1.2236,
+      "step": 57690
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6733565107458915e-07,
+      "loss": 1.006,
+      "step": 57691
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.672882427307206e-07,
+      "loss": 2.2868,
+      "step": 57692
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.672408343868521e-07,
+      "loss": 2.7934,
+      "step": 57693
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6719342604298357e-07,
+      "loss": 1.2943,
+      "step": 57694
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6714601769911507e-07,
+      "loss": 1.4521,
+      "step": 57695
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6709860935524656e-07,
+      "loss": 0.8842,
+      "step": 57696
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.67051201011378e-07,
+      "loss": 1.5506,
+      "step": 57697
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.670037926675095e-07,
+      "loss": 1.9769,
+      "step": 57698
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.66956384323641e-07,
+      "loss": 2.0384,
+      "step": 57699
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6690897597977247e-07,
+      "loss": 1.4473,
+      "step": 57700
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.668615676359039e-07,
+      "loss": 2.4224,
+      "step": 57701
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.668141592920354e-07,
+      "loss": 0.9272,
+      "step": 57702
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6676675094816684e-07,
+      "loss": 2.373,
+      "step": 57703
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.667193426042984e-07,
+      "loss": 2.2953,
+      "step": 57704
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6667193426042983e-07,
+      "loss": 1.4789,
+      "step": 57705
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.666245259165613e-07,
+      "loss": 2.0241,
+      "step": 57706
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.665771175726928e-07,
+      "loss": 1.6804,
+      "step": 57707
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6652970922882425e-07,
+      "loss": 3.2905,
+      "step": 57708
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.664823008849558e-07,
+      "loss": 2.7733,
+      "step": 57709
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6643489254108724e-07,
+      "loss": 3.5122,
+      "step": 57710
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6638748419721873e-07,
+      "loss": 1.4092,
+      "step": 57711
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6634007585335017e-07,
+      "loss": 1.2804,
+      "step": 57712
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6629266750948166e-07,
+      "loss": 1.644,
+      "step": 57713
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6624525916561315e-07,
+      "loss": 1.9728,
+      "step": 57714
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6619785082174464e-07,
+      "loss": 1.3614,
+      "step": 57715
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.661504424778761e-07,
+      "loss": 2.3623,
+      "step": 57716
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.661030341340076e-07,
+      "loss": 1.954,
+      "step": 57717
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6605562579013907e-07,
+      "loss": 1.0396,
+      "step": 57718
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6600821744627056e-07,
+      "loss": 1.0681,
+      "step": 57719
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6596080910240205e-07,
+      "loss": 1.6134,
+      "step": 57720
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.659134007585335e-07,
+      "loss": 1.5008,
+      "step": 57721
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.65865992414665e-07,
+      "loss": 1.1538,
+      "step": 57722
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.658185840707965e-07,
+      "loss": 1.4853,
+      "step": 57723
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.6577117572692797e-07,
+      "loss": 0.6856,
+      "step": 57724
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.657237673830594e-07,
+      "loss": 3.338,
+      "step": 57725
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.656763590391909e-07,
+      "loss": 1.9004,
+      "step": 57726
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6562895069532234e-07,
+      "loss": 1.7689,
+      "step": 57727
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.655815423514539e-07,
+      "loss": 2.1814,
+      "step": 57728
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.655341340075854e-07,
+      "loss": 1.0084,
+      "step": 57729
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.654867256637168e-07,
+      "loss": 1.3204,
+      "step": 57730
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.654393173198483e-07,
+      "loss": 1.3584,
+      "step": 57731
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6539190897597975e-07,
+      "loss": 1.845,
+      "step": 57732
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.653445006321113e-07,
+      "loss": 2.0506,
+      "step": 57733
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6529709228824273e-07,
+      "loss": 2.4492,
+      "step": 57734
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.652496839443742e-07,
+      "loss": 1.6595,
+      "step": 57735
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6520227560050566e-07,
+      "loss": 1.0679,
+      "step": 57736
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6515486725663715e-07,
+      "loss": 3.2011,
+      "step": 57737
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6510745891276865e-07,
+      "loss": 1.4335,
+      "step": 57738
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6506005056890014e-07,
+      "loss": 1.8511,
+      "step": 57739
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6501264222503163e-07,
+      "loss": 2.3246,
+      "step": 57740
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6496523388116307e-07,
+      "loss": 1.5544,
+      "step": 57741
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6491782553729456e-07,
+      "loss": 3.0279,
+      "step": 57742
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6487041719342605e-07,
+      "loss": 1.5908,
+      "step": 57743
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6482300884955755e-07,
+      "loss": 2.1223,
+      "step": 57744
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.64775600505689e-07,
+      "loss": 2.319,
+      "step": 57745
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.647281921618205e-07,
+      "loss": 2.0894,
+      "step": 57746
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6468078381795197e-07,
+      "loss": 2.1639,
+      "step": 57747
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6463337547408346e-07,
+      "loss": 2.5488,
+      "step": 57748
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.645859671302149e-07,
+      "loss": 2.4966,
+      "step": 57749
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.645385587863464e-07,
+      "loss": 1.4244,
+      "step": 57750
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.644911504424779e-07,
+      "loss": 0.7039,
+      "step": 57751
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.644437420986094e-07,
+      "loss": 0.9152,
+      "step": 57752
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6439633375474087e-07,
+      "loss": 1.8869,
+      "step": 57753
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.643489254108723e-07,
+      "loss": 1.7343,
+      "step": 57754
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.643015170670038e-07,
+      "loss": 1.5947,
+      "step": 57755
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.642541087231353e-07,
+      "loss": 1.0101,
+      "step": 57756
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.642067003792668e-07,
+      "loss": 2.8017,
+      "step": 57757
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.641592920353982e-07,
+      "loss": 1.9008,
+      "step": 57758
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.641118836915297e-07,
+      "loss": 2.7697,
+      "step": 57759
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6406447534766116e-07,
+      "loss": 0.3912,
+      "step": 57760
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.640170670037927e-07,
+      "loss": 2.2949,
+      "step": 57761
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.639696586599242e-07,
+      "loss": 1.9518,
+      "step": 57762
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6392225031605563e-07,
+      "loss": 2.4594,
+      "step": 57763
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.638748419721871e-07,
+      "loss": 2.5851,
+      "step": 57764
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6382743362831856e-07,
+      "loss": 1.2294,
+      "step": 57765
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6378002528445006e-07,
+      "loss": 1.6887,
+      "step": 57766
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6373261694058155e-07,
+      "loss": 2.489,
+      "step": 57767
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6368520859671304e-07,
+      "loss": 3.2965,
+      "step": 57768
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.636378002528445e-07,
+      "loss": 0.784,
+      "step": 57769
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6359039190897597e-07,
+      "loss": 1.3742,
+      "step": 57770
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6354298356510746e-07,
+      "loss": 1.3954,
+      "step": 57771
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6349557522123896e-07,
+      "loss": 1.8644,
+      "step": 57772
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6344816687737045e-07,
+      "loss": 1.8549,
+      "step": 57773
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.634007585335019e-07,
+      "loss": 2.7629,
+      "step": 57774
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.633533501896334e-07,
+      "loss": 2.0061,
+      "step": 57775
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6330594184576487e-07,
+      "loss": 2.0539,
+      "step": 57776
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6325853350189636e-07,
+      "loss": 1.6358,
+      "step": 57777
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.632111251580278e-07,
+      "loss": 2.8006,
+      "step": 57778
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.631637168141593e-07,
+      "loss": 0.9581,
+      "step": 57779
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.631163084702908e-07,
+      "loss": 2.0151,
+      "step": 57780
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.630689001264223e-07,
+      "loss": 2.4586,
+      "step": 57781
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.630214917825537e-07,
+      "loss": 2.6367,
+      "step": 57782
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.629740834386852e-07,
+      "loss": 0.961,
+      "step": 57783
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6292667509481665e-07,
+      "loss": 2.7969,
+      "step": 57784
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.628792667509482e-07,
+      "loss": 1.5655,
+      "step": 57785
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.628318584070797e-07,
+      "loss": 1.7057,
+      "step": 57786
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.6278445006321113e-07,
+      "loss": 2.5385,
+      "step": 57787
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.627370417193426e-07,
+      "loss": 1.4059,
+      "step": 57788
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6268963337547406e-07,
+      "loss": 0.7725,
+      "step": 57789
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.626422250316056e-07,
+      "loss": 2.0634,
+      "step": 57790
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6259481668773704e-07,
+      "loss": 2.4937,
+      "step": 57791
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6254740834386854e-07,
+      "loss": 2.796,
+      "step": 57792
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.625e-07,
+      "loss": 2.9304,
+      "step": 57793
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6245259165613147e-07,
+      "loss": 1.2468,
+      "step": 57794
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6240518331226296e-07,
+      "loss": 2.0964,
+      "step": 57795
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6235777496839445e-07,
+      "loss": 2.1568,
+      "step": 57796
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6231036662452594e-07,
+      "loss": 2.7851,
+      "step": 57797
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.622629582806574e-07,
+      "loss": 1.6314,
+      "step": 57798
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.622155499367889e-07,
+      "loss": 2.0708,
+      "step": 57799
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6216814159292037e-07,
+      "loss": 1.4325,
+      "step": 57800
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6212073324905186e-07,
+      "loss": 2.019,
+      "step": 57801
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.620733249051833e-07,
+      "loss": 0.7868,
+      "step": 57802
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.620259165613148e-07,
+      "loss": 1.6746,
+      "step": 57803
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.619785082174463e-07,
+      "loss": 1.9263,
+      "step": 57804
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.619310998735778e-07,
+      "loss": 2.3176,
+      "step": 57805
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.618836915297092e-07,
+      "loss": 1.756,
+      "step": 57806
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.618362831858407e-07,
+      "loss": 2.3438,
+      "step": 57807
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.617888748419722e-07,
+      "loss": 1.9741,
+      "step": 57808
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.617414664981037e-07,
+      "loss": 2.5507,
+      "step": 57809
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.616940581542352e-07,
+      "loss": 1.6647,
+      "step": 57810
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.616466498103666e-07,
+      "loss": 2.1907,
+      "step": 57811
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.615992414664981e-07,
+      "loss": 1.5025,
+      "step": 57812
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6155183312262955e-07,
+      "loss": 1.863,
+      "step": 57813
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.615044247787611e-07,
+      "loss": 2.0552,
+      "step": 57814
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6145701643489254e-07,
+      "loss": 1.4777,
+      "step": 57815
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6140960809102403e-07,
+      "loss": 2.9761,
+      "step": 57816
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6136219974715547e-07,
+      "loss": 2.7722,
+      "step": 57817
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6131479140328696e-07,
+      "loss": 2.8318,
+      "step": 57818
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.612673830594185e-07,
+      "loss": 1.4922,
+      "step": 57819
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6121997471554995e-07,
+      "loss": 2.0308,
+      "step": 57820
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6117256637168144e-07,
+      "loss": 2.5278,
+      "step": 57821
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.611251580278129e-07,
+      "loss": 1.299,
+      "step": 57822
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6107774968394437e-07,
+      "loss": 1.7794,
+      "step": 57823
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6103034134007586e-07,
+      "loss": 2.488,
+      "step": 57824
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6098293299620735e-07,
+      "loss": 1.9199,
+      "step": 57825
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.609355246523388e-07,
+      "loss": 1.7276,
+      "step": 57826
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.608881163084703e-07,
+      "loss": 1.4744,
+      "step": 57827
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.608407079646018e-07,
+      "loss": 2.5983,
+      "step": 57828
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6079329962073327e-07,
+      "loss": 1.6048,
+      "step": 57829
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6074589127686476e-07,
+      "loss": 1.9648,
+      "step": 57830
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.606984829329962e-07,
+      "loss": 1.5388,
+      "step": 57831
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.606510745891277e-07,
+      "loss": 1.1619,
+      "step": 57832
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.606036662452592e-07,
+      "loss": 2.4811,
+      "step": 57833
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.605562579013907e-07,
+      "loss": 2.0051,
+      "step": 57834
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.605088495575221e-07,
+      "loss": 0.9687,
+      "step": 57835
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.604614412136536e-07,
+      "loss": 3.492,
+      "step": 57836
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6041403286978505e-07,
+      "loss": 1.4405,
+      "step": 57837
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.603666245259166e-07,
+      "loss": 1.6197,
+      "step": 57838
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6031921618204803e-07,
+      "loss": 1.4953,
+      "step": 57839
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.602718078381795e-07,
+      "loss": 1.4228,
+      "step": 57840
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.60224399494311e-07,
+      "loss": 1.5108,
+      "step": 57841
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6017699115044246e-07,
+      "loss": 1.3285,
+      "step": 57842
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.60129582806574e-07,
+      "loss": 3.582,
+      "step": 57843
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6008217446270544e-07,
+      "loss": 1.2904,
+      "step": 57844
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.6003476611883693e-07,
+      "loss": 1.0701,
+      "step": 57845
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.5998735777496837e-07,
+      "loss": 1.1704,
+      "step": 57846
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.5993994943109986e-07,
+      "loss": 3.7018,
+      "step": 57847
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.5989254108723136e-07,
+      "loss": 0.9952,
+      "step": 57848
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.5984513274336285e-07,
+      "loss": 1.2731,
+      "step": 57849
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.597977243994943e-07,
+      "loss": 0.9877,
+      "step": 57850
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.597503160556258e-07,
+      "loss": 1.8172,
+      "step": 57851
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5970290771175727e-07,
+      "loss": 2.1825,
+      "step": 57852
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5965549936788876e-07,
+      "loss": 2.3051,
+      "step": 57853
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5960809102402026e-07,
+      "loss": 1.7133,
+      "step": 57854
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.595606826801517e-07,
+      "loss": 1.3872,
+      "step": 57855
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.595132743362832e-07,
+      "loss": 3.9112,
+      "step": 57856
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.594658659924147e-07,
+      "loss": 1.7527,
+      "step": 57857
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5941845764854617e-07,
+      "loss": 2.9433,
+      "step": 57858
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.593710493046776e-07,
+      "loss": 1.8127,
+      "step": 57859
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.593236409608091e-07,
+      "loss": 0.92,
+      "step": 57860
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.592762326169406e-07,
+      "loss": 1.1705,
+      "step": 57861
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.592288242730721e-07,
+      "loss": 1.9737,
+      "step": 57862
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.591814159292036e-07,
+      "loss": 1.5694,
+      "step": 57863
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.59134007585335e-07,
+      "loss": 2.1348,
+      "step": 57864
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.590865992414665e-07,
+      "loss": 2.6866,
+      "step": 57865
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5903919089759795e-07,
+      "loss": 1.7041,
+      "step": 57866
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.589917825537295e-07,
+      "loss": 1.8262,
+      "step": 57867
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5894437420986094e-07,
+      "loss": 2.2812,
+      "step": 57868
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5889696586599243e-07,
+      "loss": 3.5808,
+      "step": 57869
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5884955752212387e-07,
+      "loss": 3.1233,
+      "step": 57870
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5880214917825536e-07,
+      "loss": 1.9609,
+      "step": 57871
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5875474083438685e-07,
+      "loss": 1.0798,
+      "step": 57872
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5870733249051834e-07,
+      "loss": 2.2504,
+      "step": 57873
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5865992414664984e-07,
+      "loss": 1.1224,
+      "step": 57874
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.586125158027813e-07,
+      "loss": 2.0828,
+      "step": 57875
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5856510745891277e-07,
+      "loss": 1.7368,
+      "step": 57876
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5851769911504426e-07,
+      "loss": 1.9607,
+      "step": 57877
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5847029077117575e-07,
+      "loss": 1.873,
+      "step": 57878
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.584228824273072e-07,
+      "loss": 2.3618,
+      "step": 57879
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.583754740834387e-07,
+      "loss": 2.8499,
+      "step": 57880
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.583280657395702e-07,
+      "loss": 2.8659,
+      "step": 57881
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5828065739570167e-07,
+      "loss": 1.6081,
+      "step": 57882
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.582332490518331e-07,
+      "loss": 1.6414,
+      "step": 57883
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.581858407079646e-07,
+      "loss": 1.0317,
+      "step": 57884
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.581384323640961e-07,
+      "loss": 1.2684,
+      "step": 57885
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.580910240202276e-07,
+      "loss": 2.4055,
+      "step": 57886
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.580436156763591e-07,
+      "loss": 1.0362,
+      "step": 57887
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.579962073324905e-07,
+      "loss": 2.5776,
+      "step": 57888
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.57948798988622e-07,
+      "loss": 1.9829,
+      "step": 57889
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.579013906447535e-07,
+      "loss": 1.698,
+      "step": 57890
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.57853982300885e-07,
+      "loss": 3.1437,
+      "step": 57891
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5780657395701643e-07,
+      "loss": 1.5777,
+      "step": 57892
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.577591656131479e-07,
+      "loss": 2.1692,
+      "step": 57893
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5771175726927936e-07,
+      "loss": 2.4191,
+      "step": 57894
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.576643489254109e-07,
+      "loss": 1.6347,
+      "step": 57895
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5761694058154235e-07,
+      "loss": 2.1647,
+      "step": 57896
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5756953223767384e-07,
+      "loss": 1.9706,
+      "step": 57897
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5752212389380533e-07,
+      "loss": 0.302,
+      "step": 57898
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5747471554993677e-07,
+      "loss": 3.0793,
+      "step": 57899
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.574273072060683e-07,
+      "loss": 2.7519,
+      "step": 57900
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5737989886219975e-07,
+      "loss": 1.2163,
+      "step": 57901
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5733249051833125e-07,
+      "loss": 1.0584,
+      "step": 57902
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.572850821744627e-07,
+      "loss": 1.6431,
+      "step": 57903
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.572376738305942e-07,
+      "loss": 2.8637,
+      "step": 57904
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5719026548672567e-07,
+      "loss": 1.6895,
+      "step": 57905
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5714285714285716e-07,
+      "loss": 3.1892,
+      "step": 57906
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.570954487989886e-07,
+      "loss": 0.8142,
+      "step": 57907
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.570480404551201e-07,
+      "loss": 1.8897,
+      "step": 57908
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.570006321112516e-07,
+      "loss": 2.5041,
+      "step": 57909
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.569532237673831e-07,
+      "loss": 1.1836,
+      "step": 57910
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.5690581542351457e-07,
+      "loss": 1.0503,
+      "step": 57911
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.56858407079646e-07,
+      "loss": 2.5079,
+      "step": 57912
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.568109987357775e-07,
+      "loss": 1.3615,
+      "step": 57913
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.56763590391909e-07,
+      "loss": 0.956,
+      "step": 57914
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.567161820480405e-07,
+      "loss": 1.9107,
+      "step": 57915
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.566687737041719e-07,
+      "loss": 1.1336,
+      "step": 57916
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.566213653603034e-07,
+      "loss": 0.979,
+      "step": 57917
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5657395701643486e-07,
+      "loss": 1.4354,
+      "step": 57918
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.565265486725664e-07,
+      "loss": 3.4862,
+      "step": 57919
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.564791403286979e-07,
+      "loss": 2.0967,
+      "step": 57920
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5643173198482933e-07,
+      "loss": 0.7638,
+      "step": 57921
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.563843236409608e-07,
+      "loss": 1.5415,
+      "step": 57922
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5633691529709226e-07,
+      "loss": 3.1299,
+      "step": 57923
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.562895069532238e-07,
+      "loss": 1.6697,
+      "step": 57924
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5624209860935525e-07,
+      "loss": 2.6818,
+      "step": 57925
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5619469026548674e-07,
+      "loss": 2.601,
+      "step": 57926
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.561472819216182e-07,
+      "loss": 1.0537,
+      "step": 57927
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5609987357774967e-07,
+      "loss": 1.0617,
+      "step": 57928
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5605246523388116e-07,
+      "loss": 1.73,
+      "step": 57929
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5600505689001266e-07,
+      "loss": 1.5882,
+      "step": 57930
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5595764854614415e-07,
+      "loss": 0.601,
+      "step": 57931
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.559102402022756e-07,
+      "loss": 1.6687,
+      "step": 57932
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.558628318584071e-07,
+      "loss": 2.2904,
+      "step": 57933
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5581542351453857e-07,
+      "loss": 3.7258,
+      "step": 57934
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5576801517067006e-07,
+      "loss": 2.5593,
+      "step": 57935
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.557206068268015e-07,
+      "loss": 1.8697,
+      "step": 57936
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.55673198482933e-07,
+      "loss": 2.2135,
+      "step": 57937
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.556257901390645e-07,
+      "loss": 2.8041,
+      "step": 57938
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.55578381795196e-07,
+      "loss": 1.0611,
+      "step": 57939
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.555309734513274e-07,
+      "loss": 1.6461,
+      "step": 57940
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.554835651074589e-07,
+      "loss": 1.6648,
+      "step": 57941
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.554361567635904e-07,
+      "loss": 1.1992,
+      "step": 57942
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.553887484197219e-07,
+      "loss": 1.3554,
+      "step": 57943
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.553413400758534e-07,
+      "loss": 3.0834,
+      "step": 57944
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5529393173198483e-07,
+      "loss": 0.6638,
+      "step": 57945
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.552465233881163e-07,
+      "loss": 1.6206,
+      "step": 57946
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5519911504424776e-07,
+      "loss": 1.2165,
+      "step": 57947
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.551517067003793e-07,
+      "loss": 1.4813,
+      "step": 57948
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5510429835651074e-07,
+      "loss": 1.693,
+      "step": 57949
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5505689001264223e-07,
+      "loss": 1.8513,
+      "step": 57950
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.550094816687737e-07,
+      "loss": 1.2851,
+      "step": 57951
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5496207332490517e-07,
+      "loss": 1.4331,
+      "step": 57952
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.549146649810367e-07,
+      "loss": 2.1084,
+      "step": 57953
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5486725663716815e-07,
+      "loss": 0.1589,
+      "step": 57954
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5481984829329964e-07,
+      "loss": 0.6047,
+      "step": 57955
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.547724399494311e-07,
+      "loss": 4.1066,
+      "step": 57956
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.547250316055626e-07,
+      "loss": 1.9761,
+      "step": 57957
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5467762326169407e-07,
+      "loss": 1.7867,
+      "step": 57958
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5463021491782556e-07,
+      "loss": 1.7223,
+      "step": 57959
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.54582806573957e-07,
+      "loss": 1.6319,
+      "step": 57960
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.545353982300885e-07,
+      "loss": 0.8452,
+      "step": 57961
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5448798988622e-07,
+      "loss": 0.6764,
+      "step": 57962
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.544405815423515e-07,
+      "loss": 1.3919,
+      "step": 57963
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5439317319848297e-07,
+      "loss": 2.7144,
+      "step": 57964
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.543457648546144e-07,
+      "loss": 0.9692,
+      "step": 57965
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.542983565107459e-07,
+      "loss": 1.6071,
+      "step": 57966
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.542509481668774e-07,
+      "loss": 1.5491,
+      "step": 57967
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.542035398230089e-07,
+      "loss": 0.2977,
+      "step": 57968
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.541561314791403e-07,
+      "loss": 1.9664,
+      "step": 57969
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.541087231352718e-07,
+      "loss": 1.095,
+      "step": 57970
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5406131479140325e-07,
+      "loss": 0.935,
+      "step": 57971
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.540139064475348e-07,
+      "loss": 0.772,
+      "step": 57972
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5396649810366624e-07,
+      "loss": 2.6385,
+      "step": 57973
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5391908975979773e-07,
+      "loss": 2.9345,
+      "step": 57974
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.538716814159292e-07,
+      "loss": 1.4874,
+      "step": 57975
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5382427307206066e-07,
+      "loss": 2.9098,
+      "step": 57976
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.537768647281922e-07,
+      "loss": 2.7962,
+      "step": 57977
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.5372945638432365e-07,
+      "loss": 1.2429,
+      "step": 57978
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5368204804045514e-07,
+      "loss": 1.7824,
+      "step": 57979
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.536346396965866e-07,
+      "loss": 2.0083,
+      "step": 57980
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5358723135271807e-07,
+      "loss": 1.2274,
+      "step": 57981
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5353982300884956e-07,
+      "loss": 1.9179,
+      "step": 57982
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5349241466498105e-07,
+      "loss": 4.2019,
+      "step": 57983
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.534450063211125e-07,
+      "loss": 4.1924,
+      "step": 57984
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.53397597977244e-07,
+      "loss": 1.6025,
+      "step": 57985
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.533501896333755e-07,
+      "loss": 2.2811,
+      "step": 57986
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5330278128950697e-07,
+      "loss": 1.0802,
+      "step": 57987
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5325537294563846e-07,
+      "loss": 0.7761,
+      "step": 57988
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.532079646017699e-07,
+      "loss": 2.2493,
+      "step": 57989
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.531605562579014e-07,
+      "loss": 1.1567,
+      "step": 57990
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.531131479140329e-07,
+      "loss": 1.4622,
+      "step": 57991
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.530657395701644e-07,
+      "loss": 2.0371,
+      "step": 57992
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.530183312262958e-07,
+      "loss": 0.4387,
+      "step": 57993
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.529709228824273e-07,
+      "loss": 1.4817,
+      "step": 57994
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.529235145385588e-07,
+      "loss": 1.675,
+      "step": 57995
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.528761061946903e-07,
+      "loss": 1.9149,
+      "step": 57996
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.528286978508218e-07,
+      "loss": 1.482,
+      "step": 57997
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.527812895069532e-07,
+      "loss": 0.792,
+      "step": 57998
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.527338811630847e-07,
+      "loss": 2.7064,
+      "step": 57999
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.526864728192162e-07,
+      "loss": 2.8591,
+      "step": 58000
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.526390644753477e-07,
+      "loss": 1.8607,
+      "step": 58001
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5259165613147914e-07,
+      "loss": 1.033,
+      "step": 58002
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5254424778761063e-07,
+      "loss": 2.3797,
+      "step": 58003
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5249683944374207e-07,
+      "loss": 1.0317,
+      "step": 58004
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.524494310998736e-07,
+      "loss": 3.1691,
+      "step": 58005
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5240202275600506e-07,
+      "loss": 2.3989,
+      "step": 58006
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5235461441213655e-07,
+      "loss": 2.4419,
+      "step": 58007
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.52307206068268e-07,
+      "loss": 2.4943,
+      "step": 58008
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.522597977243995e-07,
+      "loss": 1.6659,
+      "step": 58009
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5221238938053097e-07,
+      "loss": 0.9897,
+      "step": 58010
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5216498103666246e-07,
+      "loss": 3.0523,
+      "step": 58011
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5211757269279396e-07,
+      "loss": 2.6416,
+      "step": 58012
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.520701643489254e-07,
+      "loss": 2.4548,
+      "step": 58013
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.520227560050569e-07,
+      "loss": 0.8484,
+      "step": 58014
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.519753476611884e-07,
+      "loss": 2.3581,
+      "step": 58015
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5192793931731987e-07,
+      "loss": 4.0927,
+      "step": 58016
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.518805309734513e-07,
+      "loss": 0.892,
+      "step": 58017
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.518331226295828e-07,
+      "loss": 2.4189,
+      "step": 58018
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.517857142857143e-07,
+      "loss": 0.8692,
+      "step": 58019
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.517383059418458e-07,
+      "loss": 1.1851,
+      "step": 58020
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.516908975979773e-07,
+      "loss": 3.5494,
+      "step": 58021
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.516434892541087e-07,
+      "loss": 2.2593,
+      "step": 58022
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.515960809102402e-07,
+      "loss": 1.7647,
+      "step": 58023
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.515486725663717e-07,
+      "loss": 1.6439,
+      "step": 58024
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.515012642225032e-07,
+      "loss": 1.021,
+      "step": 58025
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5145385587863463e-07,
+      "loss": 3.2794,
+      "step": 58026
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5140644753476613e-07,
+      "loss": 4.7204,
+      "step": 58027
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5135903919089757e-07,
+      "loss": 1.5029,
+      "step": 58028
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.513116308470291e-07,
+      "loss": 0.4594,
+      "step": 58029
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5126422250316055e-07,
+      "loss": 1.5997,
+      "step": 58030
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5121681415929204e-07,
+      "loss": 2.0041,
+      "step": 58031
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5116940581542353e-07,
+      "loss": 1.6329,
+      "step": 58032
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.51121997471555e-07,
+      "loss": 1.0645,
+      "step": 58033
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.510745891276865e-07,
+      "loss": 0.9155,
+      "step": 58034
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5102718078381796e-07,
+      "loss": 1.6565,
+      "step": 58035
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5097977243994945e-07,
+      "loss": 1.736,
+      "step": 58036
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.509323640960809e-07,
+      "loss": 0.9291,
+      "step": 58037
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.508849557522124e-07,
+      "loss": 2.9079,
+      "step": 58038
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.508375474083439e-07,
+      "loss": 3.4201,
+      "step": 58039
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.5079013906447537e-07,
+      "loss": 1.6966,
+      "step": 58040
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.507427307206068e-07,
+      "loss": 1.2648,
+      "step": 58041
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.506953223767383e-07,
+      "loss": 0.6305,
+      "step": 58042
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.506479140328698e-07,
+      "loss": 2.3954,
+      "step": 58043
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.506005056890013e-07,
+      "loss": 2.1058,
+      "step": 58044
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.505530973451328e-07,
+      "loss": 0.7616,
+      "step": 58045
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.505056890012642e-07,
+      "loss": 1.2283,
+      "step": 58046
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.504582806573957e-07,
+      "loss": 2.7266,
+      "step": 58047
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.504108723135272e-07,
+      "loss": 2.1839,
+      "step": 58048
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.503634639696587e-07,
+      "loss": 1.6496,
+      "step": 58049
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.5031605562579013e-07,
+      "loss": 2.6625,
+      "step": 58050
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.502686472819216e-07,
+      "loss": 0.5465,
+      "step": 58051
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.5022123893805306e-07,
+      "loss": 1.5457,
+      "step": 58052
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.501738305941846e-07,
+      "loss": 1.1912,
+      "step": 58053
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.501264222503161e-07,
+      "loss": 1.932,
+      "step": 58054
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.5007901390644754e-07,
+      "loss": 1.1582,
+      "step": 58055
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.5003160556257903e-07,
+      "loss": 1.8307,
+      "step": 58056
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4998419721871047e-07,
+      "loss": 1.3933,
+      "step": 58057
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.49936788874842e-07,
+      "loss": 0.4367,
+      "step": 58058
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4988938053097345e-07,
+      "loss": 1.6192,
+      "step": 58059
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4984197218710494e-07,
+      "loss": 2.1404,
+      "step": 58060
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.497945638432364e-07,
+      "loss": 2.2281,
+      "step": 58061
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.497471554993679e-07,
+      "loss": 1.6441,
+      "step": 58062
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4969974715549937e-07,
+      "loss": 1.7447,
+      "step": 58063
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4965233881163086e-07,
+      "loss": 2.4644,
+      "step": 58064
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4960493046776235e-07,
+      "loss": 2.7164,
+      "step": 58065
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.495575221238938e-07,
+      "loss": 2.3917,
+      "step": 58066
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.495101137800253e-07,
+      "loss": 0.9401,
+      "step": 58067
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.494627054361568e-07,
+      "loss": 1.3531,
+      "step": 58068
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4941529709228827e-07,
+      "loss": 1.4948,
+      "step": 58069
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.493678887484197e-07,
+      "loss": 2.2692,
+      "step": 58070
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.493204804045512e-07,
+      "loss": 2.2866,
+      "step": 58071
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.492730720606827e-07,
+      "loss": 1.3954,
+      "step": 58072
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.492256637168142e-07,
+      "loss": 1.5764,
+      "step": 58073
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.491782553729456e-07,
+      "loss": 1.2351,
+      "step": 58074
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.491308470290771e-07,
+      "loss": 1.2727,
+      "step": 58075
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.490834386852086e-07,
+      "loss": 2.8945,
+      "step": 58076
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.490360303413401e-07,
+      "loss": 0.9668,
+      "step": 58077
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.489886219974716e-07,
+      "loss": 3.4012,
+      "step": 58078
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4894121365360303e-07,
+      "loss": 3.3046,
+      "step": 58079
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.488938053097345e-07,
+      "loss": 1.6054,
+      "step": 58080
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4884639696586596e-07,
+      "loss": 3.7604,
+      "step": 58081
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.487989886219975e-07,
+      "loss": 2.8416,
+      "step": 58082
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4875158027812895e-07,
+      "loss": 0.4521,
+      "step": 58083
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4870417193426044e-07,
+      "loss": 0.8102,
+      "step": 58084
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.486567635903919e-07,
+      "loss": 1.5905,
+      "step": 58085
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4860935524652337e-07,
+      "loss": 1.8998,
+      "step": 58086
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.485619469026549e-07,
+      "loss": 0.8991,
+      "step": 58087
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4851453855878636e-07,
+      "loss": 1.8886,
+      "step": 58088
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4846713021491785e-07,
+      "loss": 1.4477,
+      "step": 58089
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.484197218710493e-07,
+      "loss": 2.7822,
+      "step": 58090
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.483723135271808e-07,
+      "loss": 1.9441,
+      "step": 58091
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4832490518331227e-07,
+      "loss": 2.6919,
+      "step": 58092
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4827749683944376e-07,
+      "loss": 0.9156,
+      "step": 58093
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.482300884955752e-07,
+      "loss": 1.3055,
+      "step": 58094
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.481826801517067e-07,
+      "loss": 2.91,
+      "step": 58095
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.481352718078382e-07,
+      "loss": 1.1212,
+      "step": 58096
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.480878634639697e-07,
+      "loss": 1.0654,
+      "step": 58097
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4804045512010117e-07,
+      "loss": 3.1953,
+      "step": 58098
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.479930467762326e-07,
+      "loss": 1.0877,
+      "step": 58099
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.479456384323641e-07,
+      "loss": 1.4502,
+      "step": 58100
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.478982300884956e-07,
+      "loss": 1.5997,
+      "step": 58101
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.478508217446271e-07,
+      "loss": 0.9916,
+      "step": 58102
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.478034134007585e-07,
+      "loss": 2.2615,
+      "step": 58103
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.4775600505689e-07,
+      "loss": 2.989,
+      "step": 58104
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.477085967130215e-07,
+      "loss": 2.37,
+      "step": 58105
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.47661188369153e-07,
+      "loss": 1.0596,
+      "step": 58106
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4761378002528444e-07,
+      "loss": 2.8463,
+      "step": 58107
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4756637168141593e-07,
+      "loss": 1.9624,
+      "step": 58108
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.475189633375474e-07,
+      "loss": 1.3127,
+      "step": 58109
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4747155499367887e-07,
+      "loss": 2.6553,
+      "step": 58110
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.474241466498104e-07,
+      "loss": 1.7925,
+      "step": 58111
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4737673830594185e-07,
+      "loss": 1.0005,
+      "step": 58112
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4732932996207334e-07,
+      "loss": 2.3685,
+      "step": 58113
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.472819216182048e-07,
+      "loss": 1.8317,
+      "step": 58114
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4723451327433627e-07,
+      "loss": 0.6923,
+      "step": 58115
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4718710493046777e-07,
+      "loss": 2.0916,
+      "step": 58116
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4713969658659926e-07,
+      "loss": 1.9252,
+      "step": 58117
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.470922882427307e-07,
+      "loss": 0.4466,
+      "step": 58118
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.470448798988622e-07,
+      "loss": 0.8774,
+      "step": 58119
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.469974715549937e-07,
+      "loss": 1.4576,
+      "step": 58120
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.469500632111252e-07,
+      "loss": 0.1727,
+      "step": 58121
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4690265486725667e-07,
+      "loss": 1.9633,
+      "step": 58122
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.468552465233881e-07,
+      "loss": 1.2205,
+      "step": 58123
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.468078381795196e-07,
+      "loss": 3.0946,
+      "step": 58124
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.467604298356511e-07,
+      "loss": 1.8852,
+      "step": 58125
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.467130214917826e-07,
+      "loss": 0.8772,
+      "step": 58126
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.46665613147914e-07,
+      "loss": 1.9706,
+      "step": 58127
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.466182048040455e-07,
+      "loss": 1.4454,
+      "step": 58128
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.46570796460177e-07,
+      "loss": 0.9975,
+      "step": 58129
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.465233881163085e-07,
+      "loss": 2.717,
+      "step": 58130
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4647597977243994e-07,
+      "loss": 0.8852,
+      "step": 58131
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4642857142857143e-07,
+      "loss": 3.0449,
+      "step": 58132
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.463811630847029e-07,
+      "loss": 0.3391,
+      "step": 58133
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.463337547408344e-07,
+      "loss": 0.7701,
+      "step": 58134
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.462863463969659e-07,
+      "loss": 2.3545,
+      "step": 58135
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4623893805309734e-07,
+      "loss": 1.5214,
+      "step": 58136
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4619152970922884e-07,
+      "loss": 1.0811,
+      "step": 58137
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.461441213653603e-07,
+      "loss": 2.1069,
+      "step": 58138
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.460967130214918e-07,
+      "loss": 0.8435,
+      "step": 58139
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4604930467762326e-07,
+      "loss": 2.3898,
+      "step": 58140
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4600189633375475e-07,
+      "loss": 1.1588,
+      "step": 58141
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.459544879898862e-07,
+      "loss": 0.8838,
+      "step": 58142
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.459070796460177e-07,
+      "loss": 2.1673,
+      "step": 58143
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4585967130214923e-07,
+      "loss": 0.9212,
+      "step": 58144
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4581226295828067e-07,
+      "loss": 1.6637,
+      "step": 58145
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4576485461441216e-07,
+      "loss": 2.4176,
+      "step": 58146
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.457174462705436e-07,
+      "loss": 2.9139,
+      "step": 58147
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.456700379266751e-07,
+      "loss": 0.5321,
+      "step": 58148
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.456226295828066e-07,
+      "loss": 2.4225,
+      "step": 58149
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.455752212389381e-07,
+      "loss": 1.4037,
+      "step": 58150
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.455278128950695e-07,
+      "loss": 0.6104,
+      "step": 58151
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.45480404551201e-07,
+      "loss": 2.2047,
+      "step": 58152
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.454329962073325e-07,
+      "loss": 0.7717,
+      "step": 58153
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.45385587863464e-07,
+      "loss": 1.3774,
+      "step": 58154
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.453381795195955e-07,
+      "loss": 2.4275,
+      "step": 58155
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.452907711757269e-07,
+      "loss": 2.2074,
+      "step": 58156
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.452433628318584e-07,
+      "loss": 2.1381,
+      "step": 58157
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.451959544879899e-07,
+      "loss": 1.5947,
+      "step": 58158
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.451485461441214e-07,
+      "loss": 0.7416,
+      "step": 58159
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4510113780025284e-07,
+      "loss": 1.8274,
+      "step": 58160
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4505372945638433e-07,
+      "loss": 2.2685,
+      "step": 58161
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4500632111251577e-07,
+      "loss": 1.8449,
+      "step": 58162
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.449589127686473e-07,
+      "loss": 1.1975,
+      "step": 58163
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4491150442477875e-07,
+      "loss": 1.0172,
+      "step": 58164
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4486409608091025e-07,
+      "loss": 2.3626,
+      "step": 58165
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.4481668773704174e-07,
+      "loss": 1.6312,
+      "step": 58166
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.447692793931732e-07,
+      "loss": 2.5283,
+      "step": 58167
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.447218710493047e-07,
+      "loss": 2.4244,
+      "step": 58168
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4467446270543616e-07,
+      "loss": 2.0748,
+      "step": 58169
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4462705436156766e-07,
+      "loss": 2.4481,
+      "step": 58170
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.445796460176991e-07,
+      "loss": 2.106,
+      "step": 58171
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.445322376738306e-07,
+      "loss": 1.4115,
+      "step": 58172
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.444848293299621e-07,
+      "loss": 1.8379,
+      "step": 58173
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4443742098609357e-07,
+      "loss": 1.8781,
+      "step": 58174
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.44390012642225e-07,
+      "loss": 1.6137,
+      "step": 58175
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.443426042983565e-07,
+      "loss": 1.8858,
+      "step": 58176
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.44295195954488e-07,
+      "loss": 1.838,
+      "step": 58177
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.442477876106195e-07,
+      "loss": 0.606,
+      "step": 58178
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.44200379266751e-07,
+      "loss": 3.4267,
+      "step": 58179
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.441529709228824e-07,
+      "loss": 1.7985,
+      "step": 58180
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.441055625790139e-07,
+      "loss": 1.4869,
+      "step": 58181
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.440581542351454e-07,
+      "loss": 1.7564,
+      "step": 58182
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.440107458912769e-07,
+      "loss": 2.1511,
+      "step": 58183
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4396333754740833e-07,
+      "loss": 1.2078,
+      "step": 58184
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.439159292035398e-07,
+      "loss": 0.8306,
+      "step": 58185
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4386852085967127e-07,
+      "loss": 2.2511,
+      "step": 58186
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.438211125158028e-07,
+      "loss": 1.851,
+      "step": 58187
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.437737041719343e-07,
+      "loss": 1.5458,
+      "step": 58188
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4372629582806574e-07,
+      "loss": 1.2729,
+      "step": 58189
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4367888748419723e-07,
+      "loss": 2.5229,
+      "step": 58190
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4363147914032867e-07,
+      "loss": 1.9455,
+      "step": 58191
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.435840707964602e-07,
+      "loss": 2.584,
+      "step": 58192
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4353666245259166e-07,
+      "loss": 1.3947,
+      "step": 58193
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4348925410872315e-07,
+      "loss": 1.1923,
+      "step": 58194
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.434418457648546e-07,
+      "loss": 2.2257,
+      "step": 58195
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.433944374209861e-07,
+      "loss": 4.6392,
+      "step": 58196
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4334702907711757e-07,
+      "loss": 1.138,
+      "step": 58197
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4329962073324907e-07,
+      "loss": 1.8454,
+      "step": 58198
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4325221238938056e-07,
+      "loss": 1.1395,
+      "step": 58199
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.43204804045512e-07,
+      "loss": 1.0536,
+      "step": 58200
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.431573957016435e-07,
+      "loss": 2.07,
+      "step": 58201
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.43109987357775e-07,
+      "loss": 2.2274,
+      "step": 58202
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4306257901390647e-07,
+      "loss": 2.1178,
+      "step": 58203
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.430151706700379e-07,
+      "loss": 1.8471,
+      "step": 58204
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.429677623261694e-07,
+      "loss": 2.1584,
+      "step": 58205
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.429203539823009e-07,
+      "loss": 0.7161,
+      "step": 58206
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.428729456384324e-07,
+      "loss": 3.2505,
+      "step": 58207
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4282553729456383e-07,
+      "loss": 1.165,
+      "step": 58208
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.427781289506953e-07,
+      "loss": 1.4882,
+      "step": 58209
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.427307206068268e-07,
+      "loss": 2.4021,
+      "step": 58210
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.426833122629583e-07,
+      "loss": 2.1568,
+      "step": 58211
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.426359039190898e-07,
+      "loss": 1.5232,
+      "step": 58212
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4258849557522124e-07,
+      "loss": 1.3802,
+      "step": 58213
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4254108723135273e-07,
+      "loss": 1.1637,
+      "step": 58214
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4249367888748417e-07,
+      "loss": 0.9469,
+      "step": 58215
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.424462705436157e-07,
+      "loss": 2.3457,
+      "step": 58216
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4239886219974715e-07,
+      "loss": 1.4741,
+      "step": 58217
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4235145385587864e-07,
+      "loss": 1.3096,
+      "step": 58218
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.423040455120101e-07,
+      "loss": 1.5625,
+      "step": 58219
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.422566371681416e-07,
+      "loss": 2.017,
+      "step": 58220
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.422092288242731e-07,
+      "loss": 1.8198,
+      "step": 58221
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4216182048040456e-07,
+      "loss": 1.3959,
+      "step": 58222
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4211441213653605e-07,
+      "loss": 1.2544,
+      "step": 58223
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.420670037926675e-07,
+      "loss": 3.4644,
+      "step": 58224
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.42019595448799e-07,
+      "loss": 0.723,
+      "step": 58225
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.419721871049305e-07,
+      "loss": 2.0495,
+      "step": 58226
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.4192477876106197e-07,
+      "loss": 0.6882,
+      "step": 58227
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.418773704171934e-07,
+      "loss": 1.1071,
+      "step": 58228
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.418299620733249e-07,
+      "loss": 0.856,
+      "step": 58229
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.417825537294564e-07,
+      "loss": 2.2374,
+      "step": 58230
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.417351453855879e-07,
+      "loss": 2.2363,
+      "step": 58231
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.416877370417193e-07,
+      "loss": 1.3024,
+      "step": 58232
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.416403286978508e-07,
+      "loss": 2.4922,
+      "step": 58233
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.415929203539823e-07,
+      "loss": 3.4922,
+      "step": 58234
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.415455120101138e-07,
+      "loss": 1.7993,
+      "step": 58235
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.414981036662453e-07,
+      "loss": 1.9342,
+      "step": 58236
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4145069532237673e-07,
+      "loss": 2.6921,
+      "step": 58237
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.414032869785082e-07,
+      "loss": 1.8783,
+      "step": 58238
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.413558786346397e-07,
+      "loss": 1.612,
+      "step": 58239
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.413084702907712e-07,
+      "loss": 2.3462,
+      "step": 58240
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4126106194690265e-07,
+      "loss": 3.0676,
+      "step": 58241
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4121365360303414e-07,
+      "loss": 1.3653,
+      "step": 58242
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.411662452591656e-07,
+      "loss": 3.1884,
+      "step": 58243
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.411188369152971e-07,
+      "loss": 0.4924,
+      "step": 58244
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.410714285714286e-07,
+      "loss": 1.2104,
+      "step": 58245
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4102402022756005e-07,
+      "loss": 1.3369,
+      "step": 58246
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4097661188369155e-07,
+      "loss": 3.5225,
+      "step": 58247
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.40929203539823e-07,
+      "loss": 1.2086,
+      "step": 58248
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.408817951959545e-07,
+      "loss": 1.2971,
+      "step": 58249
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4083438685208597e-07,
+      "loss": 0.2,
+      "step": 58250
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4078697850821746e-07,
+      "loss": 1.2698,
+      "step": 58251
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.407395701643489e-07,
+      "loss": 3.0812,
+      "step": 58252
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.406921618204804e-07,
+      "loss": 1.3212,
+      "step": 58253
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.406447534766119e-07,
+      "loss": 1.4547,
+      "step": 58254
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.405973451327434e-07,
+      "loss": 2.09,
+      "step": 58255
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4054993678887487e-07,
+      "loss": 1.4871,
+      "step": 58256
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.405025284450063e-07,
+      "loss": 1.6193,
+      "step": 58257
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.404551201011378e-07,
+      "loss": 4.1534,
+      "step": 58258
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.404077117572693e-07,
+      "loss": 1.1479,
+      "step": 58259
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.403603034134008e-07,
+      "loss": 3.9184,
+      "step": 58260
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.403128950695322e-07,
+      "loss": 1.5312,
+      "step": 58261
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.402654867256637e-07,
+      "loss": 1.9504,
+      "step": 58262
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.402180783817952e-07,
+      "loss": 0.7355,
+      "step": 58263
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.401706700379267e-07,
+      "loss": 0.5432,
+      "step": 58264
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4012326169405814e-07,
+      "loss": 1.8908,
+      "step": 58265
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.4007585335018963e-07,
+      "loss": 1.0556,
+      "step": 58266
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.400284450063211e-07,
+      "loss": 1.9649,
+      "step": 58267
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.399810366624526e-07,
+      "loss": 1.2422,
+      "step": 58268
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.399336283185841e-07,
+      "loss": 1.7263,
+      "step": 58269
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3988621997471555e-07,
+      "loss": 2.7831,
+      "step": 58270
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3983881163084704e-07,
+      "loss": 2.1751,
+      "step": 58271
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.397914032869785e-07,
+      "loss": 2.0765,
+      "step": 58272
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3974399494311e-07,
+      "loss": 2.9455,
+      "step": 58273
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3969658659924146e-07,
+      "loss": 0.4036,
+      "step": 58274
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3964917825537296e-07,
+      "loss": 0.7928,
+      "step": 58275
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.396017699115044e-07,
+      "loss": 2.7817,
+      "step": 58276
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.395543615676359e-07,
+      "loss": 1.9713,
+      "step": 58277
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3950695322376743e-07,
+      "loss": 1.7711,
+      "step": 58278
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3945954487989887e-07,
+      "loss": 2.5372,
+      "step": 58279
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3941213653603037e-07,
+      "loss": 2.7805,
+      "step": 58280
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.393647281921618e-07,
+      "loss": 1.049,
+      "step": 58281
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.393173198482933e-07,
+      "loss": 1.9267,
+      "step": 58282
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.392699115044248e-07,
+      "loss": 1.3774,
+      "step": 58283
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.392225031605563e-07,
+      "loss": 2.2916,
+      "step": 58284
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.391750948166877e-07,
+      "loss": 1.0282,
+      "step": 58285
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.391276864728192e-07,
+      "loss": 1.984,
+      "step": 58286
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.390802781289507e-07,
+      "loss": 2.7835,
+      "step": 58287
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.390328697850822e-07,
+      "loss": 3.0039,
+      "step": 58288
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.389854614412137e-07,
+      "loss": 0.9553,
+      "step": 58289
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3893805309734513e-07,
+      "loss": 1.7294,
+      "step": 58290
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.388906447534766e-07,
+      "loss": 0.9384,
+      "step": 58291
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.388432364096081e-07,
+      "loss": 2.0295,
+      "step": 58292
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.387958280657396e-07,
+      "loss": 1.6672,
+      "step": 58293
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3874841972187104e-07,
+      "loss": 2.7691,
+      "step": 58294
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.3870101137800254e-07,
+      "loss": 1.7638,
+      "step": 58295
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.38653603034134e-07,
+      "loss": 3.2873,
+      "step": 58296
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.386061946902655e-07,
+      "loss": 2.3811,
+      "step": 58297
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3855878634639696e-07,
+      "loss": 1.9431,
+      "step": 58298
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3851137800252845e-07,
+      "loss": 1.9115,
+      "step": 58299
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3846396965865994e-07,
+      "loss": 0.9777,
+      "step": 58300
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.384165613147914e-07,
+      "loss": 1.5415,
+      "step": 58301
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.383691529709229e-07,
+      "loss": 1.2833,
+      "step": 58302
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3832174462705437e-07,
+      "loss": 2.3262,
+      "step": 58303
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3827433628318586e-07,
+      "loss": 2.0614,
+      "step": 58304
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3822692793931733e-07,
+      "loss": 1.3439,
+      "step": 58305
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3817951959544882e-07,
+      "loss": 1.2214,
+      "step": 58306
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3813211125158028e-07,
+      "loss": 2.3539,
+      "step": 58307
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3808470290771178e-07,
+      "loss": 1.6358,
+      "step": 58308
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3803729456384321e-07,
+      "loss": 0.5143,
+      "step": 58309
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3798988621997473e-07,
+      "loss": 2.2097,
+      "step": 58310
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3794247787610623e-07,
+      "loss": 1.5979,
+      "step": 58311
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3789506953223766e-07,
+      "loss": 2.9541,
+      "step": 58312
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3784766118836918e-07,
+      "loss": 1.52,
+      "step": 58313
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3780025284450062e-07,
+      "loss": 2.5226,
+      "step": 58314
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3775284450063214e-07,
+      "loss": 2.2062,
+      "step": 58315
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3770543615676358e-07,
+      "loss": 1.9392,
+      "step": 58316
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3765802781289507e-07,
+      "loss": 1.262,
+      "step": 58317
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3761061946902654e-07,
+      "loss": 2.0126,
+      "step": 58318
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3756321112515803e-07,
+      "loss": 3.0109,
+      "step": 58319
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.375158027812895e-07,
+      "loss": 2.3492,
+      "step": 58320
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.37468394437421e-07,
+      "loss": 2.9423,
+      "step": 58321
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3742098609355248e-07,
+      "loss": 1.4132,
+      "step": 58322
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3737357774968395e-07,
+      "loss": 1.7917,
+      "step": 58323
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3732616940581544e-07,
+      "loss": 0.9927,
+      "step": 58324
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.372787610619469e-07,
+      "loss": 0.7692,
+      "step": 58325
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.372313527180784e-07,
+      "loss": 2.4233,
+      "step": 58326
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3718394437420986e-07,
+      "loss": 0.5011,
+      "step": 58327
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3713653603034135e-07,
+      "loss": 1.8239,
+      "step": 58328
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3708912768647282e-07,
+      "loss": 2.056,
+      "step": 58329
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.370417193426043e-07,
+      "loss": 1.9717,
+      "step": 58330
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3699431099873578e-07,
+      "loss": 2.2971,
+      "step": 58331
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3694690265486727e-07,
+      "loss": 1.4674,
+      "step": 58332
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.368994943109987e-07,
+      "loss": 2.2834,
+      "step": 58333
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3685208596713023e-07,
+      "loss": 1.9992,
+      "step": 58334
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3680467762326172e-07,
+      "loss": 2.5603,
+      "step": 58335
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3675726927939319e-07,
+      "loss": 1.6746,
+      "step": 58336
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3670986093552468e-07,
+      "loss": 0.5696,
+      "step": 58337
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3666245259165612e-07,
+      "loss": 1.9386,
+      "step": 58338
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3661504424778764e-07,
+      "loss": 2.6676,
+      "step": 58339
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3656763590391908e-07,
+      "loss": 1.5205,
+      "step": 58340
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.365202275600506e-07,
+      "loss": 1.6821,
+      "step": 58341
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3647281921618203e-07,
+      "loss": 1.865,
+      "step": 58342
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3642541087231353e-07,
+      "loss": 1.568,
+      "step": 58343
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.36378002528445e-07,
+      "loss": 3.0363,
+      "step": 58344
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3633059418457648e-07,
+      "loss": 2.466,
+      "step": 58345
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3628318584070798e-07,
+      "loss": 2.2106,
+      "step": 58346
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3623577749683944e-07,
+      "loss": 2.5911,
+      "step": 58347
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3618836915297093e-07,
+      "loss": 1.4385,
+      "step": 58348
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.361409608091024e-07,
+      "loss": 2.2448,
+      "step": 58349
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.360935524652339e-07,
+      "loss": 1.4758,
+      "step": 58350
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3604614412136536e-07,
+      "loss": 2.5113,
+      "step": 58351
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3599873577749685e-07,
+      "loss": 1.5431,
+      "step": 58352
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3595132743362831e-07,
+      "loss": 1.8212,
+      "step": 58353
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.359039190897598e-07,
+      "loss": 1.9628,
+      "step": 58354
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3585651074589127e-07,
+      "loss": 1.3407,
+      "step": 58355
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3580910240202276e-07,
+      "loss": 2.7555,
+      "step": 58356
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3576169405815426e-07,
+      "loss": 2.0153,
+      "step": 58357
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.3571428571428572e-07,
+      "loss": 2.1135,
+      "step": 58358
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3566687737041721e-07,
+      "loss": 0.9267,
+      "step": 58359
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3561946902654868e-07,
+      "loss": 1.9839,
+      "step": 58360
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3557206068268017e-07,
+      "loss": 2.3293,
+      "step": 58361
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.355246523388116e-07,
+      "loss": 1.0047,
+      "step": 58362
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3547724399494313e-07,
+      "loss": 1.0439,
+      "step": 58363
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3542983565107457e-07,
+      "loss": 0.8864,
+      "step": 58364
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.353824273072061e-07,
+      "loss": 1.8509,
+      "step": 58365
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3533501896333753e-07,
+      "loss": 1.1249,
+      "step": 58366
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3528761061946902e-07,
+      "loss": 0.8312,
+      "step": 58367
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3524020227560054e-07,
+      "loss": 1.3677,
+      "step": 58368
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3519279393173198e-07,
+      "loss": 1.6006,
+      "step": 58369
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.351453855878635e-07,
+      "loss": 1.7154,
+      "step": 58370
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3509797724399494e-07,
+      "loss": 2.4793,
+      "step": 58371
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3505056890012643e-07,
+      "loss": 2.4345,
+      "step": 58372
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.350031605562579e-07,
+      "loss": 1.3631,
+      "step": 58373
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3495575221238939e-07,
+      "loss": 0.4068,
+      "step": 58374
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3490834386852085e-07,
+      "loss": 2.9354,
+      "step": 58375
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3486093552465234e-07,
+      "loss": 2.7388,
+      "step": 58376
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.348135271807838e-07,
+      "loss": 1.2227,
+      "step": 58377
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.347661188369153e-07,
+      "loss": 0.7657,
+      "step": 58378
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.347187104930468e-07,
+      "loss": 2.1646,
+      "step": 58379
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3467130214917826e-07,
+      "loss": 1.9862,
+      "step": 58380
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3462389380530975e-07,
+      "loss": 0.3154,
+      "step": 58381
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3457648546144122e-07,
+      "loss": 1.2672,
+      "step": 58382
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.345290771175727e-07,
+      "loss": 1.4674,
+      "step": 58383
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3448166877370418e-07,
+      "loss": 1.9429,
+      "step": 58384
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3443426042983567e-07,
+      "loss": 2.54,
+      "step": 58385
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3438685208596713e-07,
+      "loss": 2.6301,
+      "step": 58386
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3433944374209863e-07,
+      "loss": 1.4661,
+      "step": 58387
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3429203539823006e-07,
+      "loss": 2.3151,
+      "step": 58388
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3424462705436158e-07,
+      "loss": 1.0619,
+      "step": 58389
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3419721871049308e-07,
+      "loss": 1.7951,
+      "step": 58390
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3414981036662454e-07,
+      "loss": 0.8692,
+      "step": 58391
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3410240202275603e-07,
+      "loss": 1.2572,
+      "step": 58392
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3405499367888747e-07,
+      "loss": 0.7474,
+      "step": 58393
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.34007585335019e-07,
+      "loss": 1.8081,
+      "step": 58394
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3396017699115043e-07,
+      "loss": 1.2865,
+      "step": 58395
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3391276864728192e-07,
+      "loss": 2.0545,
+      "step": 58396
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.338653603034134e-07,
+      "loss": 0.9534,
+      "step": 58397
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3381795195954488e-07,
+      "loss": 1.8723,
+      "step": 58398
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3377054361567635e-07,
+      "loss": 0.2194,
+      "step": 58399
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3372313527180784e-07,
+      "loss": 1.1128,
+      "step": 58400
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3367572692793933e-07,
+      "loss": 1.9601,
+      "step": 58401
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.336283185840708e-07,
+      "loss": 1.3729,
+      "step": 58402
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.335809102402023e-07,
+      "loss": 0.8261,
+      "step": 58403
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3353350189633375e-07,
+      "loss": 1.7769,
+      "step": 58404
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3348609355246525e-07,
+      "loss": 2.0623,
+      "step": 58405
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.334386852085967e-07,
+      "loss": 1.9905,
+      "step": 58406
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.333912768647282e-07,
+      "loss": 1.5559,
+      "step": 58407
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3334386852085967e-07,
+      "loss": 2.5498,
+      "step": 58408
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3329646017699116e-07,
+      "loss": 2.0514,
+      "step": 58409
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3324905183312263e-07,
+      "loss": 2.6043,
+      "step": 58410
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3320164348925412e-07,
+      "loss": 1.2251,
+      "step": 58411
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.331542351453856e-07,
+      "loss": 1.5327,
+      "step": 58412
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3310682680151708e-07,
+      "loss": 0.9351,
+      "step": 58413
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3305941845764857e-07,
+      "loss": 2.5229,
+      "step": 58414
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3301201011378004e-07,
+      "loss": 1.4517,
+      "step": 58415
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3296460176991153e-07,
+      "loss": 1.3782,
+      "step": 58416
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3291719342604297e-07,
+      "loss": 1.2231,
+      "step": 58417
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3286978508217449e-07,
+      "loss": 1.2581,
+      "step": 58418
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3282237673830592e-07,
+      "loss": 1.9318,
+      "step": 58419
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3277496839443744e-07,
+      "loss": 1.6301,
+      "step": 58420
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.3272756005056888e-07,
+      "loss": 3.5231,
+      "step": 58421
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3268015170670037e-07,
+      "loss": 1.1874,
+      "step": 58422
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.326327433628319e-07,
+      "loss": 1.5139,
+      "step": 58423
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3258533501896333e-07,
+      "loss": 1.7038,
+      "step": 58424
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3253792667509485e-07,
+      "loss": 1.4438,
+      "step": 58425
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.324905183312263e-07,
+      "loss": 2.7908,
+      "step": 58426
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3244310998735778e-07,
+      "loss": 0.8693,
+      "step": 58427
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3239570164348925e-07,
+      "loss": 0.6773,
+      "step": 58428
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3234829329962074e-07,
+      "loss": 1.544,
+      "step": 58429
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.323008849557522e-07,
+      "loss": 3.1103,
+      "step": 58430
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.322534766118837e-07,
+      "loss": 2.6501,
+      "step": 58431
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3220606826801516e-07,
+      "loss": 3.6406,
+      "step": 58432
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3215865992414666e-07,
+      "loss": 1.0764,
+      "step": 58433
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3211125158027815e-07,
+      "loss": 1.0984,
+      "step": 58434
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3206384323640961e-07,
+      "loss": 1.3465,
+      "step": 58435
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.320164348925411e-07,
+      "loss": 1.7879,
+      "step": 58436
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3196902654867257e-07,
+      "loss": 1.8937,
+      "step": 58437
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3192161820480406e-07,
+      "loss": 2.0466,
+      "step": 58438
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3187420986093553e-07,
+      "loss": 1.6726,
+      "step": 58439
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3182680151706702e-07,
+      "loss": 0.8148,
+      "step": 58440
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.317793931731985e-07,
+      "loss": 2.4176,
+      "step": 58441
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3173198482932998e-07,
+      "loss": 1.4888,
+      "step": 58442
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3168457648546142e-07,
+      "loss": 2.0266,
+      "step": 58443
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3163716814159294e-07,
+      "loss": 2.9476,
+      "step": 58444
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3158975979772438e-07,
+      "loss": 1.771,
+      "step": 58445
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.315423514538559e-07,
+      "loss": 2.208,
+      "step": 58446
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.314949431099874e-07,
+      "loss": 1.5713,
+      "step": 58447
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3144753476611883e-07,
+      "loss": 0.4322,
+      "step": 58448
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3140012642225035e-07,
+      "loss": 4.0157,
+      "step": 58449
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3135271807838179e-07,
+      "loss": 1.7201,
+      "step": 58450
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3130530973451328e-07,
+      "loss": 2.699,
+      "step": 58451
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3125790139064474e-07,
+      "loss": 2.6268,
+      "step": 58452
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3121049304677624e-07,
+      "loss": 0.9105,
+      "step": 58453
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.311630847029077e-07,
+      "loss": 1.6019,
+      "step": 58454
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.311156763590392e-07,
+      "loss": 3.5793,
+      "step": 58455
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3106826801517066e-07,
+      "loss": 2.1804,
+      "step": 58456
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3102085967130215e-07,
+      "loss": 0.1619,
+      "step": 58457
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3097345132743364e-07,
+      "loss": 1.6399,
+      "step": 58458
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.309260429835651e-07,
+      "loss": 1.1374,
+      "step": 58459
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.308786346396966e-07,
+      "loss": 3.0965,
+      "step": 58460
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3083122629582807e-07,
+      "loss": 1.1657,
+      "step": 58461
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3078381795195956e-07,
+      "loss": 0.9168,
+      "step": 58462
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3073640960809102e-07,
+      "loss": 1.018,
+      "step": 58463
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3068900126422252e-07,
+      "loss": 2.558,
+      "step": 58464
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3064159292035398e-07,
+      "loss": 1.7878,
+      "step": 58465
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3059418457648547e-07,
+      "loss": 1.7692,
+      "step": 58466
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3054677623261691e-07,
+      "loss": 1.6716,
+      "step": 58467
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3049936788874843e-07,
+      "loss": 1.6058,
+      "step": 58468
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3045195954487992e-07,
+      "loss": 2.0104,
+      "step": 58469
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.304045512010114e-07,
+      "loss": 3.04,
+      "step": 58470
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3035714285714288e-07,
+      "loss": 0.9328,
+      "step": 58471
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3030973451327432e-07,
+      "loss": 1.4369,
+      "step": 58472
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3026232616940584e-07,
+      "loss": 1.8663,
+      "step": 58473
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3021491782553728e-07,
+      "loss": 2.9566,
+      "step": 58474
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.301675094816688e-07,
+      "loss": 1.9045,
+      "step": 58475
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3012010113780024e-07,
+      "loss": 2.5836,
+      "step": 58476
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.3007269279393173e-07,
+      "loss": 0.783,
+      "step": 58477
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.300252844500632e-07,
+      "loss": 1.3438,
+      "step": 58478
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.299778761061947e-07,
+      "loss": 2.2083,
+      "step": 58479
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.299304677623262e-07,
+      "loss": 2.2776,
+      "step": 58480
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.2988305941845765e-07,
+      "loss": 0.9976,
+      "step": 58481
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.2983565107458914e-07,
+      "loss": 1.8465,
+      "step": 58482
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.297882427307206e-07,
+      "loss": 1.8561,
+      "step": 58483
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.297408343868521e-07,
+      "loss": 2.9304,
+      "step": 58484
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.2969342604298356e-07,
+      "loss": 2.4394,
+      "step": 58485
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2964601769911505e-07,
+      "loss": 2.0118,
+      "step": 58486
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2959860935524652e-07,
+      "loss": 0.8917,
+      "step": 58487
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.29551201011378e-07,
+      "loss": 1.7916,
+      "step": 58488
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2950379266750948e-07,
+      "loss": 2.1823,
+      "step": 58489
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2945638432364097e-07,
+      "loss": 0.9508,
+      "step": 58490
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2940897597977246e-07,
+      "loss": 1.6625,
+      "step": 58491
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2936156763590393e-07,
+      "loss": 2.6101,
+      "step": 58492
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2931415929203542e-07,
+      "loss": 2.752,
+      "step": 58493
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2926675094816689e-07,
+      "loss": 1.5501,
+      "step": 58494
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2921934260429838e-07,
+      "loss": 0.6216,
+      "step": 58495
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2917193426042984e-07,
+      "loss": 1.1165,
+      "step": 58496
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2912452591656134e-07,
+      "loss": 1.5127,
+      "step": 58497
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2907711757269277e-07,
+      "loss": 2.3215,
+      "step": 58498
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.290297092288243e-07,
+      "loss": 0.4389,
+      "step": 58499
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2898230088495573e-07,
+      "loss": 2.1949,
+      "step": 58500
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2893489254108722e-07,
+      "loss": 1.8983,
+      "step": 58501
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2888748419721874e-07,
+      "loss": 2.1112,
+      "step": 58502
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2884007585335018e-07,
+      "loss": 1.2141,
+      "step": 58503
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.287926675094817e-07,
+      "loss": 1.8294,
+      "step": 58504
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2874525916561314e-07,
+      "loss": 2.5392,
+      "step": 58505
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2869785082174463e-07,
+      "loss": 0.6704,
+      "step": 58506
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.286504424778761e-07,
+      "loss": 1.3854,
+      "step": 58507
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.286030341340076e-07,
+      "loss": 2.4197,
+      "step": 58508
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2855562579013906e-07,
+      "loss": 2.0901,
+      "step": 58509
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2850821744627055e-07,
+      "loss": 2.3638,
+      "step": 58510
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2846080910240201e-07,
+      "loss": 0.9535,
+      "step": 58511
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.284134007585335e-07,
+      "loss": 2.7524,
+      "step": 58512
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.28365992414665e-07,
+      "loss": 1.9487,
+      "step": 58513
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2831858407079646e-07,
+      "loss": 1.0729,
+      "step": 58514
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2827117572692796e-07,
+      "loss": 1.6183,
+      "step": 58515
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2822376738305942e-07,
+      "loss": 1.1279,
+      "step": 58516
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2817635903919091e-07,
+      "loss": 1.4565,
+      "step": 58517
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2812895069532238e-07,
+      "loss": 2.9807,
+      "step": 58518
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2808154235145387e-07,
+      "loss": 2.7863,
+      "step": 58519
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2803413400758534e-07,
+      "loss": 0.1403,
+      "step": 58520
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2798672566371683e-07,
+      "loss": 3.1094,
+      "step": 58521
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2793931731984827e-07,
+      "loss": 1.6393,
+      "step": 58522
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.278919089759798e-07,
+      "loss": 2.2042,
+      "step": 58523
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2784450063211128e-07,
+      "loss": 2.1844,
+      "step": 58524
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2779709228824275e-07,
+      "loss": 2.4786,
+      "step": 58525
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2774968394437424e-07,
+      "loss": 2.5664,
+      "step": 58526
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2770227560050568e-07,
+      "loss": 1.6862,
+      "step": 58527
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.276548672566372e-07,
+      "loss": 2.3126,
+      "step": 58528
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2760745891276863e-07,
+      "loss": 1.5952,
+      "step": 58529
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2756005056890015e-07,
+      "loss": 1.7364,
+      "step": 58530
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.275126422250316e-07,
+      "loss": 1.9583,
+      "step": 58531
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2746523388116308e-07,
+      "loss": 2.8582,
+      "step": 58532
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2741782553729455e-07,
+      "loss": 1.778,
+      "step": 58533
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2737041719342604e-07,
+      "loss": 1.3601,
+      "step": 58534
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2732300884955756e-07,
+      "loss": 2.2327,
+      "step": 58535
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.27275600505689e-07,
+      "loss": 1.6002,
+      "step": 58536
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.272281921618205e-07,
+      "loss": 0.6931,
+      "step": 58537
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2718078381795196e-07,
+      "loss": 2.0656,
+      "step": 58538
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2713337547408345e-07,
+      "loss": 3.2577,
+      "step": 58539
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2708596713021492e-07,
+      "loss": 1.7883,
+      "step": 58540
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.270385587863464e-07,
+      "loss": 1.3152,
+      "step": 58541
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2699115044247787e-07,
+      "loss": 1.4172,
+      "step": 58542
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2694374209860937e-07,
+      "loss": 1.7468,
+      "step": 58543
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2689633375474083e-07,
+      "loss": 3.5447,
+      "step": 58544
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2684892541087232e-07,
+      "loss": 3.0423,
+      "step": 58545
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2680151706700382e-07,
+      "loss": 0.8134,
+      "step": 58546
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2675410872313528e-07,
+      "loss": 1.9035,
+      "step": 58547
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.2670670037926677e-07,
+      "loss": 2.2032,
+      "step": 58548
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2665929203539824e-07,
+      "loss": 1.8484,
+      "step": 58549
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2661188369152973e-07,
+      "loss": 1.9377,
+      "step": 58550
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2656447534766117e-07,
+      "loss": 1.0859,
+      "step": 58551
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.265170670037927e-07,
+      "loss": 0.807,
+      "step": 58552
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2646965865992413e-07,
+      "loss": 1.0493,
+      "step": 58553
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2642225031605565e-07,
+      "loss": 2.7944,
+      "step": 58554
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.263748419721871e-07,
+      "loss": 1.9794,
+      "step": 58555
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2632743362831858e-07,
+      "loss": 0.842,
+      "step": 58556
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2628002528445005e-07,
+      "loss": 2.2396,
+      "step": 58557
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2623261694058154e-07,
+      "loss": 1.7369,
+      "step": 58558
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2618520859671306e-07,
+      "loss": 1.627,
+      "step": 58559
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.261378002528445e-07,
+      "loss": 0.8111,
+      "step": 58560
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.26090391908976e-07,
+      "loss": 1.8397,
+      "step": 58561
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2604298356510745e-07,
+      "loss": 3.4166,
+      "step": 58562
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2599557522123895e-07,
+      "loss": 1.4874,
+      "step": 58563
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.259481668773704e-07,
+      "loss": 0.8757,
+      "step": 58564
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.259007585335019e-07,
+      "loss": 2.8401,
+      "step": 58565
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2585335018963337e-07,
+      "loss": 1.328,
+      "step": 58566
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2580594184576486e-07,
+      "loss": 1.09,
+      "step": 58567
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2575853350189633e-07,
+      "loss": 1.7561,
+      "step": 58568
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2571112515802782e-07,
+      "loss": 1.8301,
+      "step": 58569
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.256637168141593e-07,
+      "loss": 2.2653,
+      "step": 58570
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2561630847029078e-07,
+      "loss": 1.8395,
+      "step": 58571
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2556890012642227e-07,
+      "loss": 2.1112,
+      "step": 58572
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2552149178255373e-07,
+      "loss": 1.984,
+      "step": 58573
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2547408343868523e-07,
+      "loss": 2.5218,
+      "step": 58574
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.254266750948167e-07,
+      "loss": 3.5605,
+      "step": 58575
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2537926675094818e-07,
+      "loss": 2.7884,
+      "step": 58576
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2533185840707962e-07,
+      "loss": 2.2405,
+      "step": 58577
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2528445006321114e-07,
+      "loss": 2.5465,
+      "step": 58578
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2523704171934258e-07,
+      "loss": 2.9801,
+      "step": 58579
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.251896333754741e-07,
+      "loss": 1.667,
+      "step": 58580
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.251422250316056e-07,
+      "loss": 0.3664,
+      "step": 58581
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2509481668773703e-07,
+      "loss": 1.617,
+      "step": 58582
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2504740834386855e-07,
+      "loss": 1.0869,
+      "step": 58583
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.25e-07,
+      "loss": 2.3768,
+      "step": 58584
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.249525916561315e-07,
+      "loss": 1.4386,
+      "step": 58585
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2490518331226295e-07,
+      "loss": 0.8737,
+      "step": 58586
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2485777496839444e-07,
+      "loss": 2.7194,
+      "step": 58587
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.248103666245259e-07,
+      "loss": 0.5155,
+      "step": 58588
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.247629582806574e-07,
+      "loss": 2.3428,
+      "step": 58589
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2471554993678886e-07,
+      "loss": 1.5092,
+      "step": 58590
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2466814159292036e-07,
+      "loss": 1.6807,
+      "step": 58591
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2462073324905185e-07,
+      "loss": 0.7928,
+      "step": 58592
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2457332490518331e-07,
+      "loss": 1.1911,
+      "step": 58593
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.245259165613148e-07,
+      "loss": 0.9876,
+      "step": 58594
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2447850821744627e-07,
+      "loss": 2.3994,
+      "step": 58595
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2443109987357776e-07,
+      "loss": 1.6615,
+      "step": 58596
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2438369152970923e-07,
+      "loss": 2.6378,
+      "step": 58597
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2433628318584072e-07,
+      "loss": 3.1758,
+      "step": 58598
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.242888748419722e-07,
+      "loss": 1.8036,
+      "step": 58599
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2424146649810368e-07,
+      "loss": 1.6181,
+      "step": 58600
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2419405815423515e-07,
+      "loss": 1.7586,
+      "step": 58601
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2414664981036664e-07,
+      "loss": 2.2972,
+      "step": 58602
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2409924146649813e-07,
+      "loss": 0.9182,
+      "step": 58603
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.240518331226296e-07,
+      "loss": 2.4269,
+      "step": 58604
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.240044247787611e-07,
+      "loss": 2.4341,
+      "step": 58605
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2395701643489253e-07,
+      "loss": 0.9989,
+      "step": 58606
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2390960809102405e-07,
+      "loss": 1.0139,
+      "step": 58607
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2386219974715548e-07,
+      "loss": 1.2482,
+      "step": 58608
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.23814791403287e-07,
+      "loss": 2.1175,
+      "step": 58609
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2376738305941844e-07,
+      "loss": 2.5065,
+      "step": 58610
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.2371997471554993e-07,
+      "loss": 1.6708,
+      "step": 58611
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.236725663716814e-07,
+      "loss": 2.4277,
+      "step": 58612
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.236251580278129e-07,
+      "loss": 2.4367,
+      "step": 58613
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.235777496839444e-07,
+      "loss": 2.791,
+      "step": 58614
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2353034134007585e-07,
+      "loss": 2.2112,
+      "step": 58615
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2348293299620734e-07,
+      "loss": 1.6632,
+      "step": 58616
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.234355246523388e-07,
+      "loss": 0.6044,
+      "step": 58617
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.233881163084703e-07,
+      "loss": 2.2112,
+      "step": 58618
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2334070796460177e-07,
+      "loss": 1.7638,
+      "step": 58619
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2329329962073326e-07,
+      "loss": 0.8162,
+      "step": 58620
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2324589127686472e-07,
+      "loss": 1.6497,
+      "step": 58621
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2319848293299622e-07,
+      "loss": 2.2351,
+      "step": 58622
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2315107458912768e-07,
+      "loss": 3.2327,
+      "step": 58623
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2310366624525917e-07,
+      "loss": 4.1694,
+      "step": 58624
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2305625790139067e-07,
+      "loss": 1.4883,
+      "step": 58625
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2300884955752213e-07,
+      "loss": 2.7824,
+      "step": 58626
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2296144121365362e-07,
+      "loss": 1.2188,
+      "step": 58627
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.229140328697851e-07,
+      "loss": 2.0024,
+      "step": 58628
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2286662452591658e-07,
+      "loss": 1.999,
+      "step": 58629
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2281921618204805e-07,
+      "loss": 1.8359,
+      "step": 58630
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2277180783817954e-07,
+      "loss": 1.5572,
+      "step": 58631
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2272439949431098e-07,
+      "loss": 2.4268,
+      "step": 58632
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.226769911504425e-07,
+      "loss": 0.951,
+      "step": 58633
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2262958280657394e-07,
+      "loss": 3.2194,
+      "step": 58634
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2258217446270546e-07,
+      "loss": 1.8191,
+      "step": 58635
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2253476611883695e-07,
+      "loss": 0.5994,
+      "step": 58636
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.224873577749684e-07,
+      "loss": 1.8089,
+      "step": 58637
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.224399494310999e-07,
+      "loss": 1.5493,
+      "step": 58638
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2239254108723134e-07,
+      "loss": 2.7147,
+      "step": 58639
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2234513274336284e-07,
+      "loss": 0.941,
+      "step": 58640
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.222977243994943e-07,
+      "loss": 0.4754,
+      "step": 58641
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.222503160556258e-07,
+      "loss": 0.9447,
+      "step": 58642
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2220290771175726e-07,
+      "loss": 1.3575,
+      "step": 58643
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2215549936788875e-07,
+      "loss": 1.4138,
+      "step": 58644
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2210809102402022e-07,
+      "loss": 1.291,
+      "step": 58645
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.220606826801517e-07,
+      "loss": 2.0796,
+      "step": 58646
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.220132743362832e-07,
+      "loss": 0.715,
+      "step": 58647
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2196586599241467e-07,
+      "loss": 1.9534,
+      "step": 58648
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2191845764854616e-07,
+      "loss": 1.4486,
+      "step": 58649
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2187104930467763e-07,
+      "loss": 1.5068,
+      "step": 58650
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2182364096080912e-07,
+      "loss": 1.2569,
+      "step": 58651
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2177623261694058e-07,
+      "loss": 1.1673,
+      "step": 58652
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2172882427307208e-07,
+      "loss": 1.5266,
+      "step": 58653
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2168141592920354e-07,
+      "loss": 2.4805,
+      "step": 58654
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2163400758533503e-07,
+      "loss": 2.7554,
+      "step": 58655
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2158659924146647e-07,
+      "loss": 1.4903,
+      "step": 58656
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.21539190897598e-07,
+      "loss": 2.048,
+      "step": 58657
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2149178255372948e-07,
+      "loss": 2.466,
+      "step": 58658
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2144437420986095e-07,
+      "loss": 0.6219,
+      "step": 58659
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2139696586599244e-07,
+      "loss": 1.5859,
+      "step": 58660
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2134955752212388e-07,
+      "loss": 1.6172,
+      "step": 58661
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.213021491782554e-07,
+      "loss": 2.6661,
+      "step": 58662
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2125474083438684e-07,
+      "loss": 1.7845,
+      "step": 58663
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2120733249051836e-07,
+      "loss": 1.6031,
+      "step": 58664
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.211599241466498e-07,
+      "loss": 1.6903,
+      "step": 58665
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.211125158027813e-07,
+      "loss": 1.0692,
+      "step": 58666
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2106510745891276e-07,
+      "loss": 3.0433,
+      "step": 58667
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2101769911504425e-07,
+      "loss": 2.2355,
+      "step": 58668
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.209702907711757e-07,
+      "loss": 1.7775,
+      "step": 58669
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.209228824273072e-07,
+      "loss": 3.4276,
+      "step": 58670
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.208754740834387e-07,
+      "loss": 1.7632,
+      "step": 58671
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2082806573957016e-07,
+      "loss": 1.6499,
+      "step": 58672
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2078065739570166e-07,
+      "loss": 1.578,
+      "step": 58673
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.2073324905183312e-07,
+      "loss": 1.3435,
+      "step": 58674
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.206858407079646e-07,
+      "loss": 2.3167,
+      "step": 58675
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2063843236409608e-07,
+      "loss": 1.3119,
+      "step": 58676
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2059102402022757e-07,
+      "loss": 2.1697,
+      "step": 58677
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2054361567635904e-07,
+      "loss": 1.7929,
+      "step": 58678
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2049620733249053e-07,
+      "loss": 0.8952,
+      "step": 58679
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.20448798988622e-07,
+      "loss": 1.7174,
+      "step": 58680
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.204013906447535e-07,
+      "loss": 1.5315,
+      "step": 58681
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2035398230088498e-07,
+      "loss": 0.9816,
+      "step": 58682
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2030657395701644e-07,
+      "loss": 0.9176,
+      "step": 58683
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2025916561314794e-07,
+      "loss": 2.5746,
+      "step": 58684
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.202117572692794e-07,
+      "loss": 1.3616,
+      "step": 58685
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.201643489254109e-07,
+      "loss": 0.501,
+      "step": 58686
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2011694058154233e-07,
+      "loss": 1.8534,
+      "step": 58687
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.2006953223767385e-07,
+      "loss": 0.6384,
+      "step": 58688
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.200221238938053e-07,
+      "loss": 0.7874,
+      "step": 58689
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.199747155499368e-07,
+      "loss": 1.4486,
+      "step": 58690
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1992730720606825e-07,
+      "loss": 1.6622,
+      "step": 58691
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1987989886219974e-07,
+      "loss": 2.4279,
+      "step": 58692
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1983249051833126e-07,
+      "loss": 2.2449,
+      "step": 58693
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.197850821744627e-07,
+      "loss": 1.8699,
+      "step": 58694
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.197376738305942e-07,
+      "loss": 1.5401,
+      "step": 58695
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1969026548672566e-07,
+      "loss": 2.2927,
+      "step": 58696
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1964285714285715e-07,
+      "loss": 1.519,
+      "step": 58697
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1959544879898862e-07,
+      "loss": 1.9274,
+      "step": 58698
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.195480404551201e-07,
+      "loss": 1.268,
+      "step": 58699
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1950063211125157e-07,
+      "loss": 2.0122,
+      "step": 58700
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1945322376738307e-07,
+      "loss": 1.1166,
+      "step": 58701
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1940581542351453e-07,
+      "loss": 1.3572,
+      "step": 58702
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1935840707964602e-07,
+      "loss": 1.8912,
+      "step": 58703
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1931099873577752e-07,
+      "loss": 2.4326,
+      "step": 58704
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1926359039190898e-07,
+      "loss": 1.6962,
+      "step": 58705
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1921618204804047e-07,
+      "loss": 1.2641,
+      "step": 58706
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1916877370417194e-07,
+      "loss": 1.963,
+      "step": 58707
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1912136536030343e-07,
+      "loss": 3.1005,
+      "step": 58708
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.190739570164349e-07,
+      "loss": 2.6772,
+      "step": 58709
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.190265486725664e-07,
+      "loss": 1.9056,
+      "step": 58710
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1897914032869783e-07,
+      "loss": 1.2764,
+      "step": 58711
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1893173198482935e-07,
+      "loss": 2.2882,
+      "step": 58712
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1888432364096079e-07,
+      "loss": 1.5066,
+      "step": 58713
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.188369152970923e-07,
+      "loss": 3.1486,
+      "step": 58714
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.187895069532238e-07,
+      "loss": 0.1634,
+      "step": 58715
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1874209860935524e-07,
+      "loss": 3.2077,
+      "step": 58716
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1869469026548676e-07,
+      "loss": 3.3546,
+      "step": 58717
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.186472819216182e-07,
+      "loss": 1.4943,
+      "step": 58718
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.185998735777497e-07,
+      "loss": 0.9531,
+      "step": 58719
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1855246523388115e-07,
+      "loss": 2.5447,
+      "step": 58720
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1850505689001264e-07,
+      "loss": 0.9016,
+      "step": 58721
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.184576485461441e-07,
+      "loss": 2.0921,
+      "step": 58722
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.184102402022756e-07,
+      "loss": 1.4274,
+      "step": 58723
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1836283185840707e-07,
+      "loss": 0.4114,
+      "step": 58724
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1831542351453856e-07,
+      "loss": 3.2695,
+      "step": 58725
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1826801517067005e-07,
+      "loss": 0.8927,
+      "step": 58726
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1822060682680152e-07,
+      "loss": 1.1525,
+      "step": 58727
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.18173198482933e-07,
+      "loss": 2.984,
+      "step": 58728
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1812579013906448e-07,
+      "loss": 2.2987,
+      "step": 58729
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1807838179519597e-07,
+      "loss": 2.632,
+      "step": 58730
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1803097345132743e-07,
+      "loss": 2.1151,
+      "step": 58731
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1798356510745893e-07,
+      "loss": 2.5743,
+      "step": 58732
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.179361567635904e-07,
+      "loss": 3.4334,
+      "step": 58733
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1788874841972188e-07,
+      "loss": 1.2441,
+      "step": 58734
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1784134007585335e-07,
+      "loss": 1.782,
+      "step": 58735
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1779393173198484e-07,
+      "loss": 0.4542,
+      "step": 58736
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.1774652338811633e-07,
+      "loss": 1.1764,
+      "step": 58737
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.176991150442478e-07,
+      "loss": 1.8809,
+      "step": 58738
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.176517067003793e-07,
+      "loss": 1.7441,
+      "step": 58739
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1760429835651076e-07,
+      "loss": 0.4447,
+      "step": 58740
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1755689001264225e-07,
+      "loss": 1.272,
+      "step": 58741
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.175094816687737e-07,
+      "loss": 0.7552,
+      "step": 58742
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.174620733249052e-07,
+      "loss": 1.0127,
+      "step": 58743
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1741466498103665e-07,
+      "loss": 1.7379,
+      "step": 58744
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1736725663716814e-07,
+      "loss": 1.9544,
+      "step": 58745
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.173198482932996e-07,
+      "loss": 2.5291,
+      "step": 58746
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.172724399494311e-07,
+      "loss": 1.71,
+      "step": 58747
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1722503160556262e-07,
+      "loss": 1.8506,
+      "step": 58748
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1717762326169405e-07,
+      "loss": 1.0307,
+      "step": 58749
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1713021491782555e-07,
+      "loss": 0.3439,
+      "step": 58750
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.17082806573957e-07,
+      "loss": 1.0659,
+      "step": 58751
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.170353982300885e-07,
+      "loss": 0.8691,
+      "step": 58752
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1698798988621997e-07,
+      "loss": 2.9454,
+      "step": 58753
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1694058154235146e-07,
+      "loss": 4.0913,
+      "step": 58754
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1689317319848293e-07,
+      "loss": 1.4866,
+      "step": 58755
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1684576485461442e-07,
+      "loss": 2.1365,
+      "step": 58756
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1679835651074589e-07,
+      "loss": 0.2541,
+      "step": 58757
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1675094816687738e-07,
+      "loss": 0.8259,
+      "step": 58758
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1670353982300887e-07,
+      "loss": 1.9752,
+      "step": 58759
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1665613147914034e-07,
+      "loss": 2.0222,
+      "step": 58760
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1660872313527183e-07,
+      "loss": 1.6917,
+      "step": 58761
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.165613147914033e-07,
+      "loss": 2.1295,
+      "step": 58762
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1651390644753479e-07,
+      "loss": 1.6913,
+      "step": 58763
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1646649810366625e-07,
+      "loss": 1.5314,
+      "step": 58764
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1641908975979774e-07,
+      "loss": 2.1604,
+      "step": 58765
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1637168141592918e-07,
+      "loss": 2.2913,
+      "step": 58766
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.163242730720607e-07,
+      "loss": 2.5529,
+      "step": 58767
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1627686472819214e-07,
+      "loss": 1.8766,
+      "step": 58768
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1622945638432366e-07,
+      "loss": 0.6818,
+      "step": 58769
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1618204804045515e-07,
+      "loss": 2.6808,
+      "step": 58770
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.161346396965866e-07,
+      "loss": 3.1389,
+      "step": 58771
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.160872313527181e-07,
+      "loss": 2.1883,
+      "step": 58772
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1603982300884955e-07,
+      "loss": 0.6673,
+      "step": 58773
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1599241466498107e-07,
+      "loss": 1.1424,
+      "step": 58774
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.159450063211125e-07,
+      "loss": 1.369,
+      "step": 58775
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.15897597977244e-07,
+      "loss": 2.5875,
+      "step": 58776
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1585018963337547e-07,
+      "loss": 1.4341,
+      "step": 58777
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1580278128950696e-07,
+      "loss": 2.5061,
+      "step": 58778
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1575537294563842e-07,
+      "loss": 1.5363,
+      "step": 58779
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1570796460176992e-07,
+      "loss": 3.6159,
+      "step": 58780
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1566055625790138e-07,
+      "loss": 1.1759,
+      "step": 58781
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1561314791403287e-07,
+      "loss": 1.6279,
+      "step": 58782
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1556573957016437e-07,
+      "loss": 2.04,
+      "step": 58783
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1551833122629583e-07,
+      "loss": 2.7239,
+      "step": 58784
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1547092288242732e-07,
+      "loss": 1.3678,
+      "step": 58785
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.154235145385588e-07,
+      "loss": 1.0506,
+      "step": 58786
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1537610619469028e-07,
+      "loss": 2.3162,
+      "step": 58787
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1532869785082175e-07,
+      "loss": 1.9704,
+      "step": 58788
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1528128950695324e-07,
+      "loss": 1.527,
+      "step": 58789
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.152338811630847e-07,
+      "loss": 2.9737,
+      "step": 58790
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.151864728192162e-07,
+      "loss": 2.1661,
+      "step": 58791
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1513906447534764e-07,
+      "loss": 1.8092,
+      "step": 58792
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1509165613147915e-07,
+      "loss": 1.2411,
+      "step": 58793
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1504424778761065e-07,
+      "loss": 1.8254,
+      "step": 58794
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1499683944374209e-07,
+      "loss": 1.3831,
+      "step": 58795
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.149494310998736e-07,
+      "loss": 0.5828,
+      "step": 58796
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1490202275600504e-07,
+      "loss": 2.1179,
+      "step": 58797
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1485461441213656e-07,
+      "loss": 1.8887,
+      "step": 58798
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.14807206068268e-07,
+      "loss": 0.2095,
+      "step": 58799
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.147597977243995e-07,
+      "loss": 2.2503,
+      "step": 58800
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.1471238938053096e-07,
+      "loss": 2.9969,
+      "step": 58801
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1466498103666245e-07,
+      "loss": 2.2122,
+      "step": 58802
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1461757269279392e-07,
+      "loss": 2.6482,
+      "step": 58803
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.145701643489254e-07,
+      "loss": 0.889,
+      "step": 58804
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.145227560050569e-07,
+      "loss": 0.9737,
+      "step": 58805
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1447534766118837e-07,
+      "loss": 2.1342,
+      "step": 58806
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1442793931731986e-07,
+      "loss": 1.0589,
+      "step": 58807
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1438053097345133e-07,
+      "loss": 1.8275,
+      "step": 58808
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1433312262958282e-07,
+      "loss": 1.2294,
+      "step": 58809
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1428571428571428e-07,
+      "loss": 1.8818,
+      "step": 58810
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1423830594184578e-07,
+      "loss": 1.9778,
+      "step": 58811
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1419089759797724e-07,
+      "loss": 2.2136,
+      "step": 58812
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1414348925410873e-07,
+      "loss": 2.04,
+      "step": 58813
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.140960809102402e-07,
+      "loss": 1.2761,
+      "step": 58814
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.140486725663717e-07,
+      "loss": 1.6652,
+      "step": 58815
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1400126422250318e-07,
+      "loss": 1.1859,
+      "step": 58816
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1395385587863465e-07,
+      "loss": 2.0735,
+      "step": 58817
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1390644753476614e-07,
+      "loss": 1.8784,
+      "step": 58818
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.138590391908976e-07,
+      "loss": 3.5865,
+      "step": 58819
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.138116308470291e-07,
+      "loss": 3.4816,
+      "step": 58820
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1376422250316054e-07,
+      "loss": 3.102,
+      "step": 58821
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1371681415929206e-07,
+      "loss": 0.7283,
+      "step": 58822
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.136694058154235e-07,
+      "loss": 0.7148,
+      "step": 58823
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1362199747155502e-07,
+      "loss": 0.911,
+      "step": 58824
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1357458912768645e-07,
+      "loss": 0.5679,
+      "step": 58825
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1352718078381795e-07,
+      "loss": 1.5111,
+      "step": 58826
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1347977243994947e-07,
+      "loss": 0.7172,
+      "step": 58827
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.134323640960809e-07,
+      "loss": 3.2971,
+      "step": 58828
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1338495575221242e-07,
+      "loss": 2.8444,
+      "step": 58829
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1333754740834386e-07,
+      "loss": 2.0574,
+      "step": 58830
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1329013906447535e-07,
+      "loss": 1.3814,
+      "step": 58831
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1324273072060682e-07,
+      "loss": 1.3791,
+      "step": 58832
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.131953223767383e-07,
+      "loss": 1.1714,
+      "step": 58833
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1314791403286978e-07,
+      "loss": 3.0965,
+      "step": 58834
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1310050568900127e-07,
+      "loss": 2.7954,
+      "step": 58835
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1305309734513274e-07,
+      "loss": 1.6143,
+      "step": 58836
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1300568900126423e-07,
+      "loss": 1.6943,
+      "step": 58837
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1295828065739572e-07,
+      "loss": 2.514,
+      "step": 58838
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1291087231352719e-07,
+      "loss": 0.6949,
+      "step": 58839
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1286346396965868e-07,
+      "loss": 2.6846,
+      "step": 58840
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1281605562579014e-07,
+      "loss": 0.9392,
+      "step": 58841
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1276864728192164e-07,
+      "loss": 1.609,
+      "step": 58842
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.127212389380531e-07,
+      "loss": 1.817,
+      "step": 58843
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.126738305941846e-07,
+      "loss": 0.9236,
+      "step": 58844
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1262642225031603e-07,
+      "loss": 1.5813,
+      "step": 58845
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1257901390644755e-07,
+      "loss": 0.4588,
+      "step": 58846
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.12531605562579e-07,
+      "loss": 1.1228,
+      "step": 58847
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.124841972187105e-07,
+      "loss": 1.2788,
+      "step": 58848
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.12436788874842e-07,
+      "loss": 1.5573,
+      "step": 58849
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1238938053097344e-07,
+      "loss": 1.5241,
+      "step": 58850
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1234197218710496e-07,
+      "loss": 3.4356,
+      "step": 58851
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.122945638432364e-07,
+      "loss": 1.5727,
+      "step": 58852
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1224715549936792e-07,
+      "loss": 1.5838,
+      "step": 58853
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1219974715549936e-07,
+      "loss": 3.104,
+      "step": 58854
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1215233881163085e-07,
+      "loss": 0.4021,
+      "step": 58855
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1210493046776231e-07,
+      "loss": 2.0305,
+      "step": 58856
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.120575221238938e-07,
+      "loss": 2.3701,
+      "step": 58857
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1201011378002527e-07,
+      "loss": 1.136,
+      "step": 58858
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1196270543615676e-07,
+      "loss": 1.9573,
+      "step": 58859
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1191529709228826e-07,
+      "loss": 1.4211,
+      "step": 58860
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1186788874841972e-07,
+      "loss": 3.7525,
+      "step": 58861
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1182048040455122e-07,
+      "loss": 2.4117,
+      "step": 58862
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1177307206068268e-07,
+      "loss": 1.8398,
+      "step": 58863
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1172566371681417e-07,
+      "loss": 2.6368,
+      "step": 58864
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.1167825537294564e-07,
+      "loss": 2.0258,
+      "step": 58865
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1163084702907713e-07,
+      "loss": 2.7049,
+      "step": 58866
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.115834386852086e-07,
+      "loss": 1.8084,
+      "step": 58867
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.115360303413401e-07,
+      "loss": 1.9341,
+      "step": 58868
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1148862199747155e-07,
+      "loss": 1.7127,
+      "step": 58869
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1144121365360305e-07,
+      "loss": 2.172,
+      "step": 58870
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1139380530973454e-07,
+      "loss": 2.2535,
+      "step": 58871
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.11346396965866e-07,
+      "loss": 2.4466,
+      "step": 58872
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.112989886219975e-07,
+      "loss": 0.2442,
+      "step": 58873
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1125158027812896e-07,
+      "loss": 1.1194,
+      "step": 58874
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1120417193426045e-07,
+      "loss": 0.9849,
+      "step": 58875
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.111567635903919e-07,
+      "loss": 1.9702,
+      "step": 58876
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.111093552465234e-07,
+      "loss": 2.2454,
+      "step": 58877
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1106194690265485e-07,
+      "loss": 1.228,
+      "step": 58878
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1101453855878637e-07,
+      "loss": 2.3141,
+      "step": 58879
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.109671302149178e-07,
+      "loss": 1.3539,
+      "step": 58880
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.109197218710493e-07,
+      "loss": 1.2291,
+      "step": 58881
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1087231352718077e-07,
+      "loss": 1.4955,
+      "step": 58882
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1082490518331226e-07,
+      "loss": 0.6685,
+      "step": 58883
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1077749683944375e-07,
+      "loss": 2.9562,
+      "step": 58884
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1073008849557522e-07,
+      "loss": 3.153,
+      "step": 58885
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.106826801517067e-07,
+      "loss": 2.6988,
+      "step": 58886
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1063527180783818e-07,
+      "loss": 2.4374,
+      "step": 58887
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1058786346396967e-07,
+      "loss": 1.6113,
+      "step": 58888
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1054045512010113e-07,
+      "loss": 2.3582,
+      "step": 58889
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1049304677623263e-07,
+      "loss": 3.6629,
+      "step": 58890
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.104456384323641e-07,
+      "loss": 2.3878,
+      "step": 58891
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1039823008849558e-07,
+      "loss": 2.4271,
+      "step": 58892
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1035082174462705e-07,
+      "loss": 2.3274,
+      "step": 58893
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1030341340075854e-07,
+      "loss": 1.7144,
+      "step": 58894
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1025600505689003e-07,
+      "loss": 0.783,
+      "step": 58895
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.102085967130215e-07,
+      "loss": 2.7678,
+      "step": 58896
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.10161188369153e-07,
+      "loss": 1.4779,
+      "step": 58897
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1011378002528446e-07,
+      "loss": 2.221,
+      "step": 58898
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.1006637168141595e-07,
+      "loss": 2.6322,
+      "step": 58899
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.100189633375474e-07,
+      "loss": 2.4751,
+      "step": 58900
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.099715549936789e-07,
+      "loss": 3.5805,
+      "step": 58901
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0992414664981035e-07,
+      "loss": 1.1735,
+      "step": 58902
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0987673830594186e-07,
+      "loss": 2.1865,
+      "step": 58903
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.098293299620733e-07,
+      "loss": 0.6198,
+      "step": 58904
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.097819216182048e-07,
+      "loss": 2.1122,
+      "step": 58905
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0973451327433631e-07,
+      "loss": 3.1978,
+      "step": 58906
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0968710493046775e-07,
+      "loss": 2.7992,
+      "step": 58907
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0963969658659927e-07,
+      "loss": 1.9095,
+      "step": 58908
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.095922882427307e-07,
+      "loss": 1.9731,
+      "step": 58909
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.095448798988622e-07,
+      "loss": 2.4733,
+      "step": 58910
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0949747155499367e-07,
+      "loss": 2.1903,
+      "step": 58911
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0945006321112516e-07,
+      "loss": 1.131,
+      "step": 58912
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0940265486725663e-07,
+      "loss": 0.989,
+      "step": 58913
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0935524652338812e-07,
+      "loss": 1.2705,
+      "step": 58914
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0930783817951959e-07,
+      "loss": 2.1893,
+      "step": 58915
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0926042983565108e-07,
+      "loss": 1.7438,
+      "step": 58916
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0921302149178257e-07,
+      "loss": 1.3649,
+      "step": 58917
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0916561314791404e-07,
+      "loss": 2.2635,
+      "step": 58918
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0911820480404553e-07,
+      "loss": 0.855,
+      "step": 58919
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.09070796460177e-07,
+      "loss": 1.9998,
+      "step": 58920
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0902338811630849e-07,
+      "loss": 1.613,
+      "step": 58921
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0897597977243995e-07,
+      "loss": 3.1636,
+      "step": 58922
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0892857142857144e-07,
+      "loss": 1.3937,
+      "step": 58923
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.088811630847029e-07,
+      "loss": 4.5883,
+      "step": 58924
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.088337547408344e-07,
+      "loss": 0.6343,
+      "step": 58925
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0878634639696584e-07,
+      "loss": 1.2251,
+      "step": 58926
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0873893805309736e-07,
+      "loss": 0.8214,
+      "step": 58927
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.0869152970922885e-07,
+      "loss": 1.5032,
+      "step": 58928
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0864412136536032e-07,
+      "loss": 0.3612,
+      "step": 58929
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.085967130214918e-07,
+      "loss": 2.0604,
+      "step": 58930
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0854930467762325e-07,
+      "loss": 2.8544,
+      "step": 58931
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0850189633375477e-07,
+      "loss": 2.2051,
+      "step": 58932
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.084544879898862e-07,
+      "loss": 2.4757,
+      "step": 58933
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.084070796460177e-07,
+      "loss": 1.656,
+      "step": 58934
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0835967130214916e-07,
+      "loss": 1.637,
+      "step": 58935
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0831226295828066e-07,
+      "loss": 2.2079,
+      "step": 58936
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0826485461441212e-07,
+      "loss": 1.1831,
+      "step": 58937
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0821744627054361e-07,
+      "loss": 1.2231,
+      "step": 58938
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.081700379266751e-07,
+      "loss": 2.0323,
+      "step": 58939
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0812262958280657e-07,
+      "loss": 1.9966,
+      "step": 58940
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0807522123893806e-07,
+      "loss": 2.0975,
+      "step": 58941
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0802781289506953e-07,
+      "loss": 3.3173,
+      "step": 58942
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0798040455120102e-07,
+      "loss": 1.945,
+      "step": 58943
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.079329962073325e-07,
+      "loss": 1.3235,
+      "step": 58944
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0788558786346398e-07,
+      "loss": 2.0193,
+      "step": 58945
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0783817951959545e-07,
+      "loss": 0.9914,
+      "step": 58946
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0779077117572694e-07,
+      "loss": 2.4028,
+      "step": 58947
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.077433628318584e-07,
+      "loss": 1.9509,
+      "step": 58948
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.076959544879899e-07,
+      "loss": 1.6282,
+      "step": 58949
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.076485461441214e-07,
+      "loss": 1.4534,
+      "step": 58950
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0760113780025285e-07,
+      "loss": 1.9166,
+      "step": 58951
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0755372945638435e-07,
+      "loss": 1.3933,
+      "step": 58952
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.075063211125158e-07,
+      "loss": 2.2576,
+      "step": 58953
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.074589127686473e-07,
+      "loss": 0.7402,
+      "step": 58954
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0741150442477874e-07,
+      "loss": 2.2573,
+      "step": 58955
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0736409608091026e-07,
+      "loss": 3.0892,
+      "step": 58956
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.073166877370417e-07,
+      "loss": 2.4224,
+      "step": 58957
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0726927939317322e-07,
+      "loss": 2.3789,
+      "step": 58958
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0722187104930466e-07,
+      "loss": 1.2602,
+      "step": 58959
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0717446270543615e-07,
+      "loss": 2.1079,
+      "step": 58960
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0712705436156767e-07,
+      "loss": 1.1927,
+      "step": 58961
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.070796460176991e-07,
+      "loss": 0.3022,
+      "step": 58962
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0703223767383063e-07,
+      "loss": 2.6139,
+      "step": 58963
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0698482932996207e-07,
+      "loss": 1.5453,
+      "step": 58964
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0693742098609356e-07,
+      "loss": 2.3603,
+      "step": 58965
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0689001264222502e-07,
+      "loss": 1.3708,
+      "step": 58966
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0684260429835652e-07,
+      "loss": 1.4529,
+      "step": 58967
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0679519595448798e-07,
+      "loss": 2.3065,
+      "step": 58968
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0674778761061948e-07,
+      "loss": 1.3547,
+      "step": 58969
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0670037926675094e-07,
+      "loss": 0.9369,
+      "step": 58970
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0665297092288243e-07,
+      "loss": 2.5948,
+      "step": 58971
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0660556257901393e-07,
+      "loss": 1.2019,
+      "step": 58972
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.065581542351454e-07,
+      "loss": 2.4318,
+      "step": 58973
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0651074589127688e-07,
+      "loss": 1.972,
+      "step": 58974
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0646333754740835e-07,
+      "loss": 2.6188,
+      "step": 58975
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0641592920353984e-07,
+      "loss": 2.4068,
+      "step": 58976
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.063685208596713e-07,
+      "loss": 1.2633,
+      "step": 58977
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.063211125158028e-07,
+      "loss": 1.9581,
+      "step": 58978
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0627370417193426e-07,
+      "loss": 1.8568,
+      "step": 58979
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0622629582806576e-07,
+      "loss": 0.9922,
+      "step": 58980
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.061788874841972e-07,
+      "loss": 2.8014,
+      "step": 58981
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0613147914032871e-07,
+      "loss": 1.5192,
+      "step": 58982
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.060840707964602e-07,
+      "loss": 2.4949,
+      "step": 58983
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0603666245259167e-07,
+      "loss": 1.3019,
+      "step": 58984
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0598925410872316e-07,
+      "loss": 2.0956,
+      "step": 58985
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.059418457648546e-07,
+      "loss": 1.8657,
+      "step": 58986
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0589443742098612e-07,
+      "loss": 2.2964,
+      "step": 58987
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0584702907711756e-07,
+      "loss": 2.461,
+      "step": 58988
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0579962073324905e-07,
+      "loss": 2.9343,
+      "step": 58989
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.0575221238938052e-07,
+      "loss": 2.149,
+      "step": 58990
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.05704804045512e-07,
+      "loss": 2.2156,
+      "step": 58991
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0565739570164348e-07,
+      "loss": 1.8955,
+      "step": 58992
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0560998735777497e-07,
+      "loss": 2.5131,
+      "step": 58993
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0556257901390644e-07,
+      "loss": 2.3453,
+      "step": 58994
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0551517067003793e-07,
+      "loss": 2.4782,
+      "step": 58995
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0546776232616942e-07,
+      "loss": 2.2861,
+      "step": 58996
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0542035398230089e-07,
+      "loss": 0.7486,
+      "step": 58997
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0537294563843238e-07,
+      "loss": 2.3557,
+      "step": 58998
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0532553729456384e-07,
+      "loss": 1.461,
+      "step": 58999
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0527812895069534e-07,
+      "loss": 1.6173,
+      "step": 59000
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.052307206068268e-07,
+      "loss": 1.1054,
+      "step": 59001
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.051833122629583e-07,
+      "loss": 3.2041,
+      "step": 59002
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0513590391908976e-07,
+      "loss": 1.9107,
+      "step": 59003
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0508849557522125e-07,
+      "loss": 2.343,
+      "step": 59004
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.050410872313527e-07,
+      "loss": 0.6496,
+      "step": 59005
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.049936788874842e-07,
+      "loss": 0.6679,
+      "step": 59006
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.049462705436157e-07,
+      "loss": 1.6939,
+      "step": 59007
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0489886219974717e-07,
+      "loss": 2.0064,
+      "step": 59008
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0485145385587866e-07,
+      "loss": 1.2297,
+      "step": 59009
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.048040455120101e-07,
+      "loss": 1.5127,
+      "step": 59010
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0475663716814162e-07,
+      "loss": 1.6128,
+      "step": 59011
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0470922882427306e-07,
+      "loss": 1.4554,
+      "step": 59012
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0466182048040457e-07,
+      "loss": 1.4162,
+      "step": 59013
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0461441213653601e-07,
+      "loss": 1.5888,
+      "step": 59014
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.045670037926675e-07,
+      "loss": 3.1055,
+      "step": 59015
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0451959544879897e-07,
+      "loss": 1.5086,
+      "step": 59016
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0447218710493046e-07,
+      "loss": 2.0147,
+      "step": 59017
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0442477876106198e-07,
+      "loss": 3.7838,
+      "step": 59018
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0437737041719342e-07,
+      "loss": 2.2464,
+      "step": 59019
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0432996207332491e-07,
+      "loss": 2.0542,
+      "step": 59020
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0428255372945638e-07,
+      "loss": 2.4672,
+      "step": 59021
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0423514538558787e-07,
+      "loss": 2.6347,
+      "step": 59022
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0418773704171934e-07,
+      "loss": 2.0173,
+      "step": 59023
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0414032869785083e-07,
+      "loss": 2.4149,
+      "step": 59024
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.040929203539823e-07,
+      "loss": 2.0582,
+      "step": 59025
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.040455120101138e-07,
+      "loss": 1.0684,
+      "step": 59026
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0399810366624525e-07,
+      "loss": 1.3848,
+      "step": 59027
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0395069532237675e-07,
+      "loss": 2.1336,
+      "step": 59028
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0390328697850824e-07,
+      "loss": 2.253,
+      "step": 59029
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.038558786346397e-07,
+      "loss": 2.1644,
+      "step": 59030
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.038084702907712e-07,
+      "loss": 0.5218,
+      "step": 59031
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0376106194690266e-07,
+      "loss": 2.2534,
+      "step": 59032
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0371365360303415e-07,
+      "loss": 0.7633,
+      "step": 59033
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0366624525916562e-07,
+      "loss": 2.5884,
+      "step": 59034
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.036188369152971e-07,
+      "loss": 2.4038,
+      "step": 59035
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0357142857142855e-07,
+      "loss": 1.49,
+      "step": 59036
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0352402022756007e-07,
+      "loss": 2.2863,
+      "step": 59037
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.034766118836915e-07,
+      "loss": 1.9177,
+      "step": 59038
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.03429203539823e-07,
+      "loss": 1.5479,
+      "step": 59039
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0338179519595452e-07,
+      "loss": 1.9734,
+      "step": 59040
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0333438685208596e-07,
+      "loss": 1.2193,
+      "step": 59041
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0328697850821748e-07,
+      "loss": 4.3595,
+      "step": 59042
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0323957016434892e-07,
+      "loss": 0.5011,
+      "step": 59043
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.031921618204804e-07,
+      "loss": 1.7143,
+      "step": 59044
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0314475347661187e-07,
+      "loss": 2.2227,
+      "step": 59045
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0309734513274337e-07,
+      "loss": 1.8044,
+      "step": 59046
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0304993678887483e-07,
+      "loss": 1.6342,
+      "step": 59047
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0300252844500632e-07,
+      "loss": 1.7568,
+      "step": 59048
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.029551201011378e-07,
+      "loss": 2.3992,
+      "step": 59049
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0290771175726928e-07,
+      "loss": 1.4391,
+      "step": 59050
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0286030341340077e-07,
+      "loss": 2.2162,
+      "step": 59051
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0281289506953224e-07,
+      "loss": 0.654,
+      "step": 59052
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.0276548672566373e-07,
+      "loss": 0.9634,
+      "step": 59053
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.027180783817952e-07,
+      "loss": 1.9197,
+      "step": 59054
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.026706700379267e-07,
+      "loss": 1.8266,
+      "step": 59055
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0262326169405816e-07,
+      "loss": 0.6992,
+      "step": 59056
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0257585335018965e-07,
+      "loss": 0.4895,
+      "step": 59057
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0252844500632111e-07,
+      "loss": 1.6841,
+      "step": 59058
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.024810366624526e-07,
+      "loss": 3.8006,
+      "step": 59059
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0243362831858405e-07,
+      "loss": 2.8284,
+      "step": 59060
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0238621997471556e-07,
+      "loss": 2.4596,
+      "step": 59061
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0233881163084706e-07,
+      "loss": 1.5502,
+      "step": 59062
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0229140328697852e-07,
+      "loss": 1.4783,
+      "step": 59063
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0224399494311001e-07,
+      "loss": 1.2793,
+      "step": 59064
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0219658659924145e-07,
+      "loss": 0.8961,
+      "step": 59065
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0214917825537297e-07,
+      "loss": 1.2093,
+      "step": 59066
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.021017699115044e-07,
+      "loss": 3.5974,
+      "step": 59067
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0205436156763593e-07,
+      "loss": 0.6085,
+      "step": 59068
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0200695322376737e-07,
+      "loss": 2.0805,
+      "step": 59069
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0195954487989886e-07,
+      "loss": 1.3073,
+      "step": 59070
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0191213653603033e-07,
+      "loss": 1.9107,
+      "step": 59071
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0186472819216182e-07,
+      "loss": 3.3086,
+      "step": 59072
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.018173198482933e-07,
+      "loss": 1.1043,
+      "step": 59073
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0176991150442478e-07,
+      "loss": 2.7444,
+      "step": 59074
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0172250316055627e-07,
+      "loss": 1.1815,
+      "step": 59075
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0167509481668774e-07,
+      "loss": 1.737,
+      "step": 59076
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0162768647281923e-07,
+      "loss": 1.6957,
+      "step": 59077
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.015802781289507e-07,
+      "loss": 1.1018,
+      "step": 59078
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0153286978508219e-07,
+      "loss": 1.6686,
+      "step": 59079
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0148546144121365e-07,
+      "loss": 2.2829,
+      "step": 59080
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0143805309734514e-07,
+      "loss": 0.9785,
+      "step": 59081
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.013906447534766e-07,
+      "loss": 2.2986,
+      "step": 59082
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.013432364096081e-07,
+      "loss": 1.4046,
+      "step": 59083
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.012958280657396e-07,
+      "loss": 0.7884,
+      "step": 59084
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0124841972187106e-07,
+      "loss": 1.4409,
+      "step": 59085
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0120101137800255e-07,
+      "loss": 1.8426,
+      "step": 59086
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0115360303413402e-07,
+      "loss": 1.2425,
+      "step": 59087
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.011061946902655e-07,
+      "loss": 2.008,
+      "step": 59088
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0105878634639695e-07,
+      "loss": 1.1782,
+      "step": 59089
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0101137800252847e-07,
+      "loss": 1.1423,
+      "step": 59090
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.009639696586599e-07,
+      "loss": 1.9414,
+      "step": 59091
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0091656131479142e-07,
+      "loss": 0.67,
+      "step": 59092
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0086915297092286e-07,
+      "loss": 0.9901,
+      "step": 59093
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0082174462705436e-07,
+      "loss": 2.3767,
+      "step": 59094
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0077433628318587e-07,
+      "loss": 2.4946,
+      "step": 59095
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0072692793931731e-07,
+      "loss": 3.1469,
+      "step": 59096
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0067951959544883e-07,
+      "loss": 2.514,
+      "step": 59097
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0063211125158027e-07,
+      "loss": 1.486,
+      "step": 59098
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0058470290771176e-07,
+      "loss": 1.818,
+      "step": 59099
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0053729456384323e-07,
+      "loss": 3.5665,
+      "step": 59100
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0048988621997472e-07,
+      "loss": 2.3809,
+      "step": 59101
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.004424778761062e-07,
+      "loss": 1.4771,
+      "step": 59102
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0039506953223768e-07,
+      "loss": 1.6353,
+      "step": 59103
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0034766118836915e-07,
+      "loss": 1.567,
+      "step": 59104
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0030025284450064e-07,
+      "loss": 2.7495,
+      "step": 59105
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.002528445006321e-07,
+      "loss": 0.8271,
+      "step": 59106
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.002054361567636e-07,
+      "loss": 1.3551,
+      "step": 59107
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.001580278128951e-07,
+      "loss": 1.3315,
+      "step": 59108
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0011061946902655e-07,
+      "loss": 3.8085,
+      "step": 59109
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.0006321112515805e-07,
+      "loss": 1.8917,
+      "step": 59110
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.000158027812895e-07,
+      "loss": 1.7553,
+      "step": 59111
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.99968394437421e-07,
+      "loss": 1.5835,
+      "step": 59112
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.9992098609355247e-07,
+      "loss": 1.6598,
+      "step": 59113
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.9987357774968396e-07,
+      "loss": 2.4385,
+      "step": 59114
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.998261694058154e-07,
+      "loss": 2.1158,
+      "step": 59115
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.9977876106194692e-07,
+      "loss": 1.3676,
+      "step": 59116
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.9973135271807836e-07,
+      "loss": 1.1379,
+      "step": 59117
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.9968394437420988e-07,
+      "loss": 1.4153,
+      "step": 59118
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9963653603034137e-07,
+      "loss": 2.7315,
+      "step": 59119
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.995891276864728e-07,
+      "loss": 1.0073,
+      "step": 59120
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9954171934260433e-07,
+      "loss": 1.3993,
+      "step": 59121
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9949431099873577e-07,
+      "loss": 1.9907,
+      "step": 59122
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9944690265486728e-07,
+      "loss": 1.9878,
+      "step": 59123
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9939949431099872e-07,
+      "loss": 2.2618,
+      "step": 59124
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9935208596713022e-07,
+      "loss": 1.6184,
+      "step": 59125
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9930467762326168e-07,
+      "loss": 1.5634,
+      "step": 59126
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9925726927939317e-07,
+      "loss": 1.3437,
+      "step": 59127
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9920986093552464e-07,
+      "loss": 3.5024,
+      "step": 59128
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9916245259165613e-07,
+      "loss": 2.1588,
+      "step": 59129
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9911504424778762e-07,
+      "loss": 2.8394,
+      "step": 59130
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.990676359039191e-07,
+      "loss": 0.6057,
+      "step": 59131
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9902022756005058e-07,
+      "loss": 1.2704,
+      "step": 59132
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9897281921618205e-07,
+      "loss": 1.7195,
+      "step": 59133
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9892541087231354e-07,
+      "loss": 1.0152,
+      "step": 59134
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.98878002528445e-07,
+      "loss": 1.8192,
+      "step": 59135
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.988305941845765e-07,
+      "loss": 1.2595,
+      "step": 59136
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9878318584070796e-07,
+      "loss": 0.3037,
+      "step": 59137
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9873577749683946e-07,
+      "loss": 2.1389,
+      "step": 59138
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9868836915297092e-07,
+      "loss": 2.6071,
+      "step": 59139
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9864096080910241e-07,
+      "loss": 1.7738,
+      "step": 59140
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.985935524652339e-07,
+      "loss": 2.2778,
+      "step": 59141
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9854614412136537e-07,
+      "loss": 1.8309,
+      "step": 59142
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9849873577749686e-07,
+      "loss": 1.9497,
+      "step": 59143
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.984513274336283e-07,
+      "loss": 1.3913,
+      "step": 59144
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9840391908975982e-07,
+      "loss": 1.1395,
+      "step": 59145
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9835651074589126e-07,
+      "loss": 1.2692,
+      "step": 59146
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9830910240202278e-07,
+      "loss": 1.1827,
+      "step": 59147
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9826169405815422e-07,
+      "loss": 1.9696,
+      "step": 59148
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.982142857142857e-07,
+      "loss": 1.2861,
+      "step": 59149
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9816687737041718e-07,
+      "loss": 1.9541,
+      "step": 59150
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9811946902654867e-07,
+      "loss": 2.6272,
+      "step": 59151
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.980720606826802e-07,
+      "loss": 0.8948,
+      "step": 59152
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9802465233881163e-07,
+      "loss": 2.7283,
+      "step": 59153
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9797724399494312e-07,
+      "loss": 1.4933,
+      "step": 59154
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9792983565107458e-07,
+      "loss": 2.6821,
+      "step": 59155
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9788242730720608e-07,
+      "loss": 3.0115,
+      "step": 59156
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9783501896333754e-07,
+      "loss": 2.6999,
+      "step": 59157
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9778761061946903e-07,
+      "loss": 1.3562,
+      "step": 59158
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.977402022756005e-07,
+      "loss": 1.4288,
+      "step": 59159
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.97692793931732e-07,
+      "loss": 2.4404,
+      "step": 59160
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9764538558786346e-07,
+      "loss": 1.2758,
+      "step": 59161
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9759797724399495e-07,
+      "loss": 0.8767,
+      "step": 59162
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9755056890012644e-07,
+      "loss": 1.2761,
+      "step": 59163
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.975031605562579e-07,
+      "loss": 2.2331,
+      "step": 59164
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.974557522123894e-07,
+      "loss": 1.474,
+      "step": 59165
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9740834386852087e-07,
+      "loss": 1.8343,
+      "step": 59166
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9736093552465236e-07,
+      "loss": 1.2677,
+      "step": 59167
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9731352718078382e-07,
+      "loss": 2.9969,
+      "step": 59168
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9726611883691532e-07,
+      "loss": 1.6137,
+      "step": 59169
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9721871049304676e-07,
+      "loss": 1.6241,
+      "step": 59170
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9717130214917827e-07,
+      "loss": 0.4715,
+      "step": 59171
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9712389380530971e-07,
+      "loss": 3.4065,
+      "step": 59172
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9707648546144123e-07,
+      "loss": 1.4623,
+      "step": 59173
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9702907711757272e-07,
+      "loss": 0.9288,
+      "step": 59174
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9698166877370416e-07,
+      "loss": 2.1551,
+      "step": 59175
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9693426042983568e-07,
+      "loss": 1.516,
+      "step": 59176
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9688685208596712e-07,
+      "loss": 3.8666,
+      "step": 59177
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9683944374209861e-07,
+      "loss": 1.1471,
+      "step": 59178
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9679203539823008e-07,
+      "loss": 1.6023,
+      "step": 59179
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9674462705436157e-07,
+      "loss": 1.6063,
+      "step": 59180
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.9669721871049304e-07,
+      "loss": 2.3135,
+      "step": 59181
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9664981036662453e-07,
+      "loss": 1.46,
+      "step": 59182
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.96602402022756e-07,
+      "loss": 1.6941,
+      "step": 59183
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.965549936788875e-07,
+      "loss": 2.51,
+      "step": 59184
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9650758533501898e-07,
+      "loss": 0.9864,
+      "step": 59185
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9646017699115045e-07,
+      "loss": 1.1941,
+      "step": 59186
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9641276864728194e-07,
+      "loss": 1.867,
+      "step": 59187
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.963653603034134e-07,
+      "loss": 1.6345,
+      "step": 59188
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.963179519595449e-07,
+      "loss": 1.6359,
+      "step": 59189
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9627054361567636e-07,
+      "loss": 1.616,
+      "step": 59190
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9622313527180785e-07,
+      "loss": 0.7282,
+      "step": 59191
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9617572692793932e-07,
+      "loss": 2.1684,
+      "step": 59192
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.961283185840708e-07,
+      "loss": 3.1462,
+      "step": 59193
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9608091024020225e-07,
+      "loss": 0.8431,
+      "step": 59194
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9603350189633377e-07,
+      "loss": 2.0026,
+      "step": 59195
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9598609355246526e-07,
+      "loss": 2.445,
+      "step": 59196
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9593868520859673e-07,
+      "loss": 1.2199,
+      "step": 59197
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9589127686472822e-07,
+      "loss": 1.2246,
+      "step": 59198
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9584386852085966e-07,
+      "loss": 1.7518,
+      "step": 59199
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9579646017699118e-07,
+      "loss": 1.836,
+      "step": 59200
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9574905183312262e-07,
+      "loss": 0.6301,
+      "step": 59201
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9570164348925413e-07,
+      "loss": 1.5953,
+      "step": 59202
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9565423514538557e-07,
+      "loss": 2.2112,
+      "step": 59203
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9560682680151707e-07,
+      "loss": 2.711,
+      "step": 59204
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9555941845764853e-07,
+      "loss": 1.0433,
+      "step": 59205
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9551201011378002e-07,
+      "loss": 1.9662,
+      "step": 59206
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9546460176991154e-07,
+      "loss": 2.272,
+      "step": 59207
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9541719342604298e-07,
+      "loss": 3.0207,
+      "step": 59208
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9536978508217447e-07,
+      "loss": 2.2359,
+      "step": 59209
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9532237673830594e-07,
+      "loss": 2.298,
+      "step": 59210
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9527496839443743e-07,
+      "loss": 1.5588,
+      "step": 59211
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.952275600505689e-07,
+      "loss": 2.0623,
+      "step": 59212
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.951801517067004e-07,
+      "loss": 1.6349,
+      "step": 59213
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9513274336283186e-07,
+      "loss": 1.608,
+      "step": 59214
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9508533501896335e-07,
+      "loss": 0.7334,
+      "step": 59215
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9503792667509481e-07,
+      "loss": 1.0692,
+      "step": 59216
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.949905183312263e-07,
+      "loss": 1.4339,
+      "step": 59217
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9494310998735777e-07,
+      "loss": 2.3747,
+      "step": 59218
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9489570164348926e-07,
+      "loss": 1.6118,
+      "step": 59219
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9484829329962076e-07,
+      "loss": 2.1374,
+      "step": 59220
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9480088495575222e-07,
+      "loss": 0.5745,
+      "step": 59221
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9475347661188371e-07,
+      "loss": 1.7672,
+      "step": 59222
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9470606826801518e-07,
+      "loss": 1.8281,
+      "step": 59223
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9465865992414667e-07,
+      "loss": 2.1733,
+      "step": 59224
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.946112515802781e-07,
+      "loss": 1.4385,
+      "step": 59225
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9456384323640963e-07,
+      "loss": 1.4741,
+      "step": 59226
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9451643489254107e-07,
+      "loss": 2.8673,
+      "step": 59227
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9446902654867256e-07,
+      "loss": 2.3808,
+      "step": 59228
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9442161820480403e-07,
+      "loss": 1.9247,
+      "step": 59229
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9437420986093552e-07,
+      "loss": 2.4113,
+      "step": 59230
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9432680151706704e-07,
+      "loss": 2.3586,
+      "step": 59231
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9427939317319848e-07,
+      "loss": 1.3996,
+      "step": 59232
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9423198482932997e-07,
+      "loss": 2.2499,
+      "step": 59233
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9418457648546143e-07,
+      "loss": 1.6268,
+      "step": 59234
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9413716814159293e-07,
+      "loss": 3.7688,
+      "step": 59235
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.940897597977244e-07,
+      "loss": 1.8367,
+      "step": 59236
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9404235145385588e-07,
+      "loss": 1.633,
+      "step": 59237
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9399494310998735e-07,
+      "loss": 0.9031,
+      "step": 59238
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9394753476611884e-07,
+      "loss": 1.6171,
+      "step": 59239
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.939001264222503e-07,
+      "loss": 2.2816,
+      "step": 59240
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.938527180783818e-07,
+      "loss": 0.9495,
+      "step": 59241
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.938053097345133e-07,
+      "loss": 0.1274,
+      "step": 59242
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9375790139064476e-07,
+      "loss": 0.8517,
+      "step": 59243
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9371049304677625e-07,
+      "loss": 2.253,
+      "step": 59244
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.9366308470290772e-07,
+      "loss": 0.6477,
+      "step": 59245
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.936156763590392e-07,
+      "loss": 1.0722,
+      "step": 59246
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9356826801517067e-07,
+      "loss": 2.8353,
+      "step": 59247
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9352085967130217e-07,
+      "loss": 2.4965,
+      "step": 59248
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.934734513274336e-07,
+      "loss": 1.1448,
+      "step": 59249
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9342604298356512e-07,
+      "loss": 0.5942,
+      "step": 59250
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9337863463969656e-07,
+      "loss": 1.2132,
+      "step": 59251
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9333122629582808e-07,
+      "loss": 1.773,
+      "step": 59252
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9328381795195957e-07,
+      "loss": 1.7201,
+      "step": 59253
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.93236409608091e-07,
+      "loss": 0.8802,
+      "step": 59254
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9318900126422253e-07,
+      "loss": 1.5209,
+      "step": 59255
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9314159292035397e-07,
+      "loss": 1.2612,
+      "step": 59256
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.930941845764855e-07,
+      "loss": 2.6024,
+      "step": 59257
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9304677623261693e-07,
+      "loss": 1.194,
+      "step": 59258
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9299936788874842e-07,
+      "loss": 1.5267,
+      "step": 59259
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9295195954487989e-07,
+      "loss": 1.109,
+      "step": 59260
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9290455120101138e-07,
+      "loss": 1.2216,
+      "step": 59261
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9285714285714284e-07,
+      "loss": 2.5915,
+      "step": 59262
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9280973451327434e-07,
+      "loss": 2.1356,
+      "step": 59263
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9276232616940583e-07,
+      "loss": 1.8856,
+      "step": 59264
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.927149178255373e-07,
+      "loss": 1.236,
+      "step": 59265
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.926675094816688e-07,
+      "loss": 2.6135,
+      "step": 59266
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9262010113780025e-07,
+      "loss": 1.533,
+      "step": 59267
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9257269279393174e-07,
+      "loss": 1.408,
+      "step": 59268
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.925252844500632e-07,
+      "loss": 1.8431,
+      "step": 59269
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.924778761061947e-07,
+      "loss": 0.4175,
+      "step": 59270
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9243046776232617e-07,
+      "loss": 1.2036,
+      "step": 59271
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9238305941845766e-07,
+      "loss": 2.7535,
+      "step": 59272
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9233565107458913e-07,
+      "loss": 2.4981,
+      "step": 59273
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9228824273072062e-07,
+      "loss": 2.0119,
+      "step": 59274
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.922408343868521e-07,
+      "loss": 2.2632,
+      "step": 59275
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9219342604298358e-07,
+      "loss": 1.5366,
+      "step": 59276
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9214601769911507e-07,
+      "loss": 1.4823,
+      "step": 59277
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9209860935524653e-07,
+      "loss": 4.3168,
+      "step": 59278
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9205120101137803e-07,
+      "loss": 3.0509,
+      "step": 59279
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9200379266750947e-07,
+      "loss": 0.6879,
+      "step": 59280
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9195638432364098e-07,
+      "loss": 3.1899,
+      "step": 59281
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9190897597977242e-07,
+      "loss": 1.64,
+      "step": 59282
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9186156763590392e-07,
+      "loss": 1.1946,
+      "step": 59283
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9181415929203538e-07,
+      "loss": 0.854,
+      "step": 59284
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9176675094816687e-07,
+      "loss": 2.1951,
+      "step": 59285
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.917193426042984e-07,
+      "loss": 3.0564,
+      "step": 59286
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9167193426042983e-07,
+      "loss": 0.779,
+      "step": 59287
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9162452591656132e-07,
+      "loss": 0.9213,
+      "step": 59288
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.915771175726928e-07,
+      "loss": 2.5815,
+      "step": 59289
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9152970922882428e-07,
+      "loss": 1.7723,
+      "step": 59290
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9148230088495575e-07,
+      "loss": 1.7315,
+      "step": 59291
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9143489254108724e-07,
+      "loss": 0.9636,
+      "step": 59292
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.913874841972187e-07,
+      "loss": 3.0958,
+      "step": 59293
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.913400758533502e-07,
+      "loss": 0.7753,
+      "step": 59294
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9129266750948166e-07,
+      "loss": 2.0655,
+      "step": 59295
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9124525916561316e-07,
+      "loss": 1.5714,
+      "step": 59296
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9119785082174465e-07,
+      "loss": 0.974,
+      "step": 59297
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.911504424778761e-07,
+      "loss": 1.4377,
+      "step": 59298
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.911030341340076e-07,
+      "loss": 1.7879,
+      "step": 59299
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9105562579013907e-07,
+      "loss": 3.0502,
+      "step": 59300
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9100821744627056e-07,
+      "loss": 1.3301,
+      "step": 59301
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9096080910240203e-07,
+      "loss": 1.5326,
+      "step": 59302
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9091340075853352e-07,
+      "loss": 0.5318,
+      "step": 59303
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9086599241466496e-07,
+      "loss": 1.6407,
+      "step": 59304
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9081858407079648e-07,
+      "loss": 2.6144,
+      "step": 59305
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9077117572692792e-07,
+      "loss": 2.1398,
+      "step": 59306
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9072376738305944e-07,
+      "loss": 0.9505,
+      "step": 59307
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.9067635903919093e-07,
+      "loss": 1.7835,
+      "step": 59308
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9062895069532237e-07,
+      "loss": 2.2169,
+      "step": 59309
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.905815423514539e-07,
+      "loss": 2.155,
+      "step": 59310
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9053413400758533e-07,
+      "loss": 2.1792,
+      "step": 59311
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9048672566371684e-07,
+      "loss": 1.3671,
+      "step": 59312
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9043931731984828e-07,
+      "loss": 0.9259,
+      "step": 59313
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9039190897597978e-07,
+      "loss": 1.3699,
+      "step": 59314
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9034450063211124e-07,
+      "loss": 2.5866,
+      "step": 59315
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9029709228824273e-07,
+      "loss": 1.5947,
+      "step": 59316
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.902496839443742e-07,
+      "loss": 3.1964,
+      "step": 59317
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.902022756005057e-07,
+      "loss": 2.1484,
+      "step": 59318
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9015486725663718e-07,
+      "loss": 2.5675,
+      "step": 59319
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9010745891276865e-07,
+      "loss": 1.6993,
+      "step": 59320
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.9006005056890014e-07,
+      "loss": 1.2549,
+      "step": 59321
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.900126422250316e-07,
+      "loss": 2.0748,
+      "step": 59322
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.899652338811631e-07,
+      "loss": 1.8883,
+      "step": 59323
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8991782553729457e-07,
+      "loss": 1.0729,
+      "step": 59324
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8987041719342606e-07,
+      "loss": 3.2763,
+      "step": 59325
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8982300884955752e-07,
+      "loss": 1.1203,
+      "step": 59326
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8977560050568902e-07,
+      "loss": 2.0576,
+      "step": 59327
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8972819216182048e-07,
+      "loss": 2.855,
+      "step": 59328
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8968078381795197e-07,
+      "loss": 0.8127,
+      "step": 59329
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.896333754740834e-07,
+      "loss": 2.6292,
+      "step": 59330
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8958596713021493e-07,
+      "loss": 0.7612,
+      "step": 59331
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8953855878634642e-07,
+      "loss": 3.0595,
+      "step": 59332
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8949115044247786e-07,
+      "loss": 0.4268,
+      "step": 59333
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8944374209860938e-07,
+      "loss": 1.2348,
+      "step": 59334
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8939633375474082e-07,
+      "loss": 1.6478,
+      "step": 59335
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8934892541087234e-07,
+      "loss": 2.6903,
+      "step": 59336
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8930151706700378e-07,
+      "loss": 2.1339,
+      "step": 59337
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8925410872313527e-07,
+      "loss": 1.5603,
+      "step": 59338
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8920670037926674e-07,
+      "loss": 0.8479,
+      "step": 59339
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8915929203539823e-07,
+      "loss": 2.1959,
+      "step": 59340
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.891118836915297e-07,
+      "loss": 2.2936,
+      "step": 59341
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8906447534766119e-07,
+      "loss": 1.6793,
+      "step": 59342
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8901706700379268e-07,
+      "loss": 2.1221,
+      "step": 59343
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8896965865992414e-07,
+      "loss": 1.3562,
+      "step": 59344
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8892225031605564e-07,
+      "loss": 3.5517,
+      "step": 59345
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.888748419721871e-07,
+      "loss": 1.9282,
+      "step": 59346
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.888274336283186e-07,
+      "loss": 1.5924,
+      "step": 59347
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8878002528445006e-07,
+      "loss": 2.1174,
+      "step": 59348
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8873261694058155e-07,
+      "loss": 1.7254,
+      "step": 59349
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8868520859671302e-07,
+      "loss": 1.2348,
+      "step": 59350
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.886378002528445e-07,
+      "loss": 2.3862,
+      "step": 59351
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8859039190897598e-07,
+      "loss": 1.7481,
+      "step": 59352
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8854298356510747e-07,
+      "loss": 2.7139,
+      "step": 59353
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8849557522123896e-07,
+      "loss": 2.3697,
+      "step": 59354
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8844816687737043e-07,
+      "loss": 1.0417,
+      "step": 59355
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8840075853350192e-07,
+      "loss": 2.1961,
+      "step": 59356
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8835335018963338e-07,
+      "loss": 2.3209,
+      "step": 59357
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8830594184576488e-07,
+      "loss": 1.4198,
+      "step": 59358
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8825853350189632e-07,
+      "loss": 0.8563,
+      "step": 59359
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8821112515802783e-07,
+      "loss": 2.1768,
+      "step": 59360
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8816371681415927e-07,
+      "loss": 1.4464,
+      "step": 59361
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.881163084702908e-07,
+      "loss": 2.8948,
+      "step": 59362
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8806890012642223e-07,
+      "loss": 2.0501,
+      "step": 59363
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8802149178255372e-07,
+      "loss": 1.613,
+      "step": 59364
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8797408343868524e-07,
+      "loss": 3.0812,
+      "step": 59365
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8792667509481668e-07,
+      "loss": 1.3891,
+      "step": 59366
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.878792667509482e-07,
+      "loss": 1.4215,
+      "step": 59367
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8783185840707964e-07,
+      "loss": 4.1583,
+      "step": 59368
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.8778445006321113e-07,
+      "loss": 3.0973,
+      "step": 59369
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.877370417193426e-07,
+      "loss": 3.0924,
+      "step": 59370
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.876896333754741e-07,
+      "loss": 3.1566,
+      "step": 59371
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8764222503160555e-07,
+      "loss": 3.5849,
+      "step": 59372
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8759481668773705e-07,
+      "loss": 2.3015,
+      "step": 59373
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.875474083438685e-07,
+      "loss": 3.1275,
+      "step": 59374
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.875e-07,
+      "loss": 2.9203,
+      "step": 59375
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8745259165613147e-07,
+      "loss": 2.1752,
+      "step": 59376
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8740518331226296e-07,
+      "loss": 2.2423,
+      "step": 59377
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8735777496839443e-07,
+      "loss": 1.8207,
+      "step": 59378
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8731036662452592e-07,
+      "loss": 1.7058,
+      "step": 59379
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8726295828065739e-07,
+      "loss": 1.5369,
+      "step": 59380
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8721554993678888e-07,
+      "loss": 2.9675,
+      "step": 59381
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8716814159292037e-07,
+      "loss": 1.2483,
+      "step": 59382
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8712073324905184e-07,
+      "loss": 2.1497,
+      "step": 59383
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8707332490518333e-07,
+      "loss": 1.6837,
+      "step": 59384
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.870259165613148e-07,
+      "loss": 2.4624,
+      "step": 59385
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8697850821744629e-07,
+      "loss": 2.586,
+      "step": 59386
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8693109987357775e-07,
+      "loss": 1.9162,
+      "step": 59387
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8688369152970922e-07,
+      "loss": 2.4726,
+      "step": 59388
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.868362831858407e-07,
+      "loss": 2.3005,
+      "step": 59389
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8678887484197218e-07,
+      "loss": 2.0867,
+      "step": 59390
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8674146649810367e-07,
+      "loss": 1.4944,
+      "step": 59391
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8669405815423516e-07,
+      "loss": 2.3901,
+      "step": 59392
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8664664981036663e-07,
+      "loss": 1.586,
+      "step": 59393
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8659924146649812e-07,
+      "loss": 1.8037,
+      "step": 59394
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8655183312262958e-07,
+      "loss": 1.8037,
+      "step": 59395
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8650442477876108e-07,
+      "loss": 0.5957,
+      "step": 59396
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8645701643489254e-07,
+      "loss": 2.3008,
+      "step": 59397
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8640960809102403e-07,
+      "loss": 0.3425,
+      "step": 59398
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.863621997471555e-07,
+      "loss": 0.838,
+      "step": 59399
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.86314791403287e-07,
+      "loss": 2.0832,
+      "step": 59400
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8626738305941846e-07,
+      "loss": 3.156,
+      "step": 59401
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8621997471554992e-07,
+      "loss": 2.7571,
+      "step": 59402
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8617256637168142e-07,
+      "loss": 2.2659,
+      "step": 59403
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.861251580278129e-07,
+      "loss": 1.2691,
+      "step": 59404
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8607774968394437e-07,
+      "loss": 2.121,
+      "step": 59405
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8603034134007587e-07,
+      "loss": 1.3274,
+      "step": 59406
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8598293299620733e-07,
+      "loss": 2.5868,
+      "step": 59407
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8593552465233882e-07,
+      "loss": 1.5358,
+      "step": 59408
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.858881163084703e-07,
+      "loss": 0.7954,
+      "step": 59409
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8584070796460178e-07,
+      "loss": 3.6282,
+      "step": 59410
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8579329962073325e-07,
+      "loss": 0.4932,
+      "step": 59411
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8574589127686474e-07,
+      "loss": 2.3187,
+      "step": 59412
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.856984829329962e-07,
+      "loss": 1.4053,
+      "step": 59413
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8565107458912767e-07,
+      "loss": 1.0111,
+      "step": 59414
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.856036662452592e-07,
+      "loss": 3.0981,
+      "step": 59415
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8555625790139065e-07,
+      "loss": 2.8208,
+      "step": 59416
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8550884955752215e-07,
+      "loss": 0.9714,
+      "step": 59417
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.854614412136536e-07,
+      "loss": 1.9755,
+      "step": 59418
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8541403286978508e-07,
+      "loss": 1.7827,
+      "step": 59419
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8536662452591657e-07,
+      "loss": 1.3631,
+      "step": 59420
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8531921618204804e-07,
+      "loss": 4.1317,
+      "step": 59421
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8527180783817953e-07,
+      "loss": 0.7006,
+      "step": 59422
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.85224399494311e-07,
+      "loss": 2.0653,
+      "step": 59423
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8517699115044249e-07,
+      "loss": 2.2909,
+      "step": 59424
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8512958280657395e-07,
+      "loss": 1.3321,
+      "step": 59425
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8508217446270544e-07,
+      "loss": 1.3915,
+      "step": 59426
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8503476611883694e-07,
+      "loss": 3.7911,
+      "step": 59427
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.849873577749684e-07,
+      "loss": 1.9364,
+      "step": 59428
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.849399494310999e-07,
+      "loss": 1.9417,
+      "step": 59429
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8489254108723136e-07,
+      "loss": 1.4923,
+      "step": 59430
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8484513274336283e-07,
+      "loss": 1.8923,
+      "step": 59431
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8479772439949432e-07,
+      "loss": 0.9199,
+      "step": 59432
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8475031605562578e-07,
+      "loss": 1.9443,
+      "step": 59433
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8470290771175728e-07,
+      "loss": 1.2546,
+      "step": 59434
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.8465549936788874e-07,
+      "loss": 0.8322,
+      "step": 59435
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8460809102402023e-07,
+      "loss": 1.453,
+      "step": 59436
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8456068268015173e-07,
+      "loss": 0.6911,
+      "step": 59437
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.845132743362832e-07,
+      "loss": 1.1257,
+      "step": 59438
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8446586599241468e-07,
+      "loss": 2.5298,
+      "step": 59439
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8441845764854615e-07,
+      "loss": 1.8049,
+      "step": 59440
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8437104930467764e-07,
+      "loss": 2.1026,
+      "step": 59441
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.843236409608091e-07,
+      "loss": 1.8725,
+      "step": 59442
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8427623261694057e-07,
+      "loss": 1.8352,
+      "step": 59443
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8422882427307206e-07,
+      "loss": 2.4656,
+      "step": 59444
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8418141592920353e-07,
+      "loss": 2.0952,
+      "step": 59445
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8413400758533502e-07,
+      "loss": 4.6183,
+      "step": 59446
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.840865992414665e-07,
+      "loss": 4.0352,
+      "step": 59447
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8403919089759798e-07,
+      "loss": 1.6037,
+      "step": 59448
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8399178255372947e-07,
+      "loss": 1.8937,
+      "step": 59449
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8394437420986094e-07,
+      "loss": 1.0958,
+      "step": 59450
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8389696586599243e-07,
+      "loss": 1.0756,
+      "step": 59451
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.838495575221239e-07,
+      "loss": 1.7515,
+      "step": 59452
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.838021491782554e-07,
+      "loss": 1.1462,
+      "step": 59453
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8375474083438685e-07,
+      "loss": 0.7889,
+      "step": 59454
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8370733249051832e-07,
+      "loss": 2.2595,
+      "step": 59455
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.836599241466498e-07,
+      "loss": 2.5922,
+      "step": 59456
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8361251580278128e-07,
+      "loss": 2.2468,
+      "step": 59457
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8356510745891277e-07,
+      "loss": 1.7359,
+      "step": 59458
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8351769911504424e-07,
+      "loss": 1.724,
+      "step": 59459
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8347029077117573e-07,
+      "loss": 2.0186,
+      "step": 59460
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8342288242730722e-07,
+      "loss": 2.329,
+      "step": 59461
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8337547408343869e-07,
+      "loss": 0.8568,
+      "step": 59462
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8332806573957018e-07,
+      "loss": 2.0771,
+      "step": 59463
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8328065739570164e-07,
+      "loss": 2.5459,
+      "step": 59464
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8323324905183314e-07,
+      "loss": 1.6137,
+      "step": 59465
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.831858407079646e-07,
+      "loss": 3.4497,
+      "step": 59466
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.831384323640961e-07,
+      "loss": 2.1246,
+      "step": 59467
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8309102402022756e-07,
+      "loss": 1.8409,
+      "step": 59468
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8304361567635903e-07,
+      "loss": 2.7654,
+      "step": 59469
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8299620733249052e-07,
+      "loss": 1.0849,
+      "step": 59470
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.82948798988622e-07,
+      "loss": 1.6989,
+      "step": 59471
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8290139064475348e-07,
+      "loss": 1.6929,
+      "step": 59472
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8285398230088497e-07,
+      "loss": 1.5748,
+      "step": 59473
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8280657395701643e-07,
+      "loss": 2.6853,
+      "step": 59474
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8275916561314793e-07,
+      "loss": 0.9819,
+      "step": 59475
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.827117572692794e-07,
+      "loss": 2.1442,
+      "step": 59476
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8266434892541088e-07,
+      "loss": 2.0728,
+      "step": 59477
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8261694058154235e-07,
+      "loss": 2.3077,
+      "step": 59478
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8256953223767384e-07,
+      "loss": 1.7204,
+      "step": 59479
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.825221238938053e-07,
+      "loss": 1.5037,
+      "step": 59480
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8247471554993677e-07,
+      "loss": 2.4134,
+      "step": 59481
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.824273072060683e-07,
+      "loss": 3.8643,
+      "step": 59482
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8237989886219976e-07,
+      "loss": 2.2391,
+      "step": 59483
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8233249051833125e-07,
+      "loss": 1.4953,
+      "step": 59484
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8228508217446271e-07,
+      "loss": 2.8114,
+      "step": 59485
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8223767383059418e-07,
+      "loss": 2.8043,
+      "step": 59486
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8219026548672567e-07,
+      "loss": 0.8924,
+      "step": 59487
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8214285714285714e-07,
+      "loss": 1.5133,
+      "step": 59488
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8209544879898863e-07,
+      "loss": 1.3631,
+      "step": 59489
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.820480404551201e-07,
+      "loss": 1.1913,
+      "step": 59490
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.820006321112516e-07,
+      "loss": 1.8321,
+      "step": 59491
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8195322376738305e-07,
+      "loss": 1.7905,
+      "step": 59492
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8190581542351455e-07,
+      "loss": 2.3247,
+      "step": 59493
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8185840707964604e-07,
+      "loss": 0.8139,
+      "step": 59494
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.818109987357775e-07,
+      "loss": 2.3286,
+      "step": 59495
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.81763590391909e-07,
+      "loss": 2.2884,
+      "step": 59496
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8171618204804046e-07,
+      "loss": 2.0772,
+      "step": 59497
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.8166877370417193e-07,
+      "loss": 1.7048,
+      "step": 59498
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8162136536030342e-07,
+      "loss": 1.4256,
+      "step": 59499
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8157395701643489e-07,
+      "loss": 2.4593,
+      "step": 59500
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8152654867256638e-07,
+      "loss": 0.6336,
+      "step": 59501
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8147914032869784e-07,
+      "loss": 1.6891,
+      "step": 59502
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8143173198482934e-07,
+      "loss": 1.6791,
+      "step": 59503
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8138432364096083e-07,
+      "loss": 4.197,
+      "step": 59504
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.813369152970923e-07,
+      "loss": 3.437,
+      "step": 59505
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8128950695322379e-07,
+      "loss": 3.0146,
+      "step": 59506
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8124209860935525e-07,
+      "loss": 1.3359,
+      "step": 59507
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8119469026548674e-07,
+      "loss": 0.9108,
+      "step": 59508
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.811472819216182e-07,
+      "loss": 0.7611,
+      "step": 59509
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8109987357774968e-07,
+      "loss": 1.1651,
+      "step": 59510
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8105246523388117e-07,
+      "loss": 1.43,
+      "step": 59511
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8100505689001263e-07,
+      "loss": 1.8515,
+      "step": 59512
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8095764854614413e-07,
+      "loss": 1.771,
+      "step": 59513
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.809102402022756e-07,
+      "loss": 2.8693,
+      "step": 59514
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8086283185840708e-07,
+      "loss": 1.6256,
+      "step": 59515
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8081542351453858e-07,
+      "loss": 1.1873,
+      "step": 59516
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8076801517067004e-07,
+      "loss": 2.3441,
+      "step": 59517
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8072060682680153e-07,
+      "loss": 2.0654,
+      "step": 59518
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.80673198482933e-07,
+      "loss": 1.1479,
+      "step": 59519
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.806257901390645e-07,
+      "loss": 2.4371,
+      "step": 59520
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8057838179519596e-07,
+      "loss": 2.8003,
+      "step": 59521
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8053097345132745e-07,
+      "loss": 2.587,
+      "step": 59522
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8048356510745891e-07,
+      "loss": 0.6636,
+      "step": 59523
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8043615676359038e-07,
+      "loss": 2.7936,
+      "step": 59524
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8038874841972187e-07,
+      "loss": 1.8752,
+      "step": 59525
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8034134007585334e-07,
+      "loss": 1.2543,
+      "step": 59526
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8029393173198483e-07,
+      "loss": 2.8044,
+      "step": 59527
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8024652338811632e-07,
+      "loss": 1.1568,
+      "step": 59528
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.801991150442478e-07,
+      "loss": 0.9386,
+      "step": 59529
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8015170670037928e-07,
+      "loss": 1.8851,
+      "step": 59530
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8010429835651075e-07,
+      "loss": 2.0667,
+      "step": 59531
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.8005689001264224e-07,
+      "loss": 1.3459,
+      "step": 59532
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.800094816687737e-07,
+      "loss": 2.0268,
+      "step": 59533
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.799620733249052e-07,
+      "loss": 0.5474,
+      "step": 59534
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7991466498103666e-07,
+      "loss": 1.5649,
+      "step": 59535
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7986725663716813e-07,
+      "loss": 2.353,
+      "step": 59536
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7981984829329962e-07,
+      "loss": 1.1773,
+      "step": 59537
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.797724399494311e-07,
+      "loss": 1.7725,
+      "step": 59538
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.797250316055626e-07,
+      "loss": 1.7458,
+      "step": 59539
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7967762326169407e-07,
+      "loss": 1.7233,
+      "step": 59540
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7963021491782554e-07,
+      "loss": 0.9242,
+      "step": 59541
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7958280657395703e-07,
+      "loss": 1.2783,
+      "step": 59542
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.795353982300885e-07,
+      "loss": 1.6309,
+      "step": 59543
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7948798988621999e-07,
+      "loss": 2.7618,
+      "step": 59544
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7944058154235145e-07,
+      "loss": 2.1603,
+      "step": 59545
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7939317319848294e-07,
+      "loss": 1.499,
+      "step": 59546
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.793457648546144e-07,
+      "loss": 1.9232,
+      "step": 59547
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7929835651074587e-07,
+      "loss": 1.2457,
+      "step": 59548
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.792509481668774e-07,
+      "loss": 2.2591,
+      "step": 59549
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7920353982300886e-07,
+      "loss": 1.3395,
+      "step": 59550
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7915613147914035e-07,
+      "loss": 2.6579,
+      "step": 59551
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7910872313527182e-07,
+      "loss": 0.8301,
+      "step": 59552
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7906131479140328e-07,
+      "loss": 1.0277,
+      "step": 59553
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7901390644753478e-07,
+      "loss": 1.1375,
+      "step": 59554
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7896649810366624e-07,
+      "loss": 1.272,
+      "step": 59555
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7891908975979773e-07,
+      "loss": 1.2413,
+      "step": 59556
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.788716814159292e-07,
+      "loss": 2.7564,
+      "step": 59557
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.788242730720607e-07,
+      "loss": 2.4364,
+      "step": 59558
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7877686472819216e-07,
+      "loss": 3.1593,
+      "step": 59559
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7872945638432365e-07,
+      "loss": 2.5196,
+      "step": 59560
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.7868204804045514e-07,
+      "loss": 2.2011,
+      "step": 59561
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.786346396965866e-07,
+      "loss": 2.312,
+      "step": 59562
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.785872313527181e-07,
+      "loss": 2.0731,
+      "step": 59563
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7853982300884956e-07,
+      "loss": 2.491,
+      "step": 59564
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7849241466498103e-07,
+      "loss": 1.0502,
+      "step": 59565
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7844500632111252e-07,
+      "loss": 3.7341,
+      "step": 59566
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.78397597977244e-07,
+      "loss": 1.2625,
+      "step": 59567
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7835018963337548e-07,
+      "loss": 2.1692,
+      "step": 59568
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7830278128950695e-07,
+      "loss": 3.166,
+      "step": 59569
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7825537294563844e-07,
+      "loss": 2.9789,
+      "step": 59570
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.782079646017699e-07,
+      "loss": 3.3023,
+      "step": 59571
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.781605562579014e-07,
+      "loss": 1.4222,
+      "step": 59572
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.781131479140329e-07,
+      "loss": 3.0923,
+      "step": 59573
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7806573957016435e-07,
+      "loss": 2.7269,
+      "step": 59574
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7801833122629585e-07,
+      "loss": 0.6292,
+      "step": 59575
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.779709228824273e-07,
+      "loss": 1.5908,
+      "step": 59576
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7792351453855878e-07,
+      "loss": 1.5005,
+      "step": 59577
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7787610619469027e-07,
+      "loss": 3.3645,
+      "step": 59578
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7782869785082174e-07,
+      "loss": 1.8916,
+      "step": 59579
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7778128950695323e-07,
+      "loss": 1.3601,
+      "step": 59580
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.777338811630847e-07,
+      "loss": 2.7259,
+      "step": 59581
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7768647281921619e-07,
+      "loss": 0.364,
+      "step": 59582
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7763906447534768e-07,
+      "loss": 1.5929,
+      "step": 59583
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7759165613147914e-07,
+      "loss": 1.5311,
+      "step": 59584
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7754424778761064e-07,
+      "loss": 2.2931,
+      "step": 59585
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.774968394437421e-07,
+      "loss": 1.0728,
+      "step": 59586
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.774494310998736e-07,
+      "loss": 1.684,
+      "step": 59587
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7740202275600506e-07,
+      "loss": 2.7776,
+      "step": 59588
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7735461441213655e-07,
+      "loss": 3.4694,
+      "step": 59589
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7730720606826802e-07,
+      "loss": 1.7448,
+      "step": 59590
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7725979772439948e-07,
+      "loss": 2.6743,
+      "step": 59591
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7721238938053097e-07,
+      "loss": 1.3168,
+      "step": 59592
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7716498103666244e-07,
+      "loss": 0.6168,
+      "step": 59593
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7711757269279393e-07,
+      "loss": 2.0048,
+      "step": 59594
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7707016434892542e-07,
+      "loss": 1.7697,
+      "step": 59595
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.770227560050569e-07,
+      "loss": 1.3913,
+      "step": 59596
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7697534766118838e-07,
+      "loss": 1.3369,
+      "step": 59597
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7692793931731985e-07,
+      "loss": 3.7376,
+      "step": 59598
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7688053097345134e-07,
+      "loss": 1.4846,
+      "step": 59599
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.768331226295828e-07,
+      "loss": 2.8199,
+      "step": 59600
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.767857142857143e-07,
+      "loss": 3.0672,
+      "step": 59601
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7673830594184576e-07,
+      "loss": 3.0287,
+      "step": 59602
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7669089759797723e-07,
+      "loss": 2.4234,
+      "step": 59603
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7664348925410872e-07,
+      "loss": 2.0225,
+      "step": 59604
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7659608091024021e-07,
+      "loss": 2.0018,
+      "step": 59605
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.765486725663717e-07,
+      "loss": 2.3622,
+      "step": 59606
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7650126422250317e-07,
+      "loss": 2.5609,
+      "step": 59607
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7645385587863464e-07,
+      "loss": 1.5404,
+      "step": 59608
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7640644753476613e-07,
+      "loss": 2.712,
+      "step": 59609
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.763590391908976e-07,
+      "loss": 0.8638,
+      "step": 59610
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.763116308470291e-07,
+      "loss": 1.9621,
+      "step": 59611
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7626422250316055e-07,
+      "loss": 1.2051,
+      "step": 59612
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7621681415929205e-07,
+      "loss": 1.9615,
+      "step": 59613
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.761694058154235e-07,
+      "loss": 1.5602,
+      "step": 59614
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7612199747155498e-07,
+      "loss": 2.3062,
+      "step": 59615
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.760745891276865e-07,
+      "loss": 1.4348,
+      "step": 59616
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7602718078381796e-07,
+      "loss": 2.6955,
+      "step": 59617
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7597977243994945e-07,
+      "loss": 0.3256,
+      "step": 59618
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7593236409608092e-07,
+      "loss": 2.2416,
+      "step": 59619
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7588495575221239e-07,
+      "loss": 2.397,
+      "step": 59620
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7583754740834388e-07,
+      "loss": 2.4559,
+      "step": 59621
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7579013906447534e-07,
+      "loss": 1.9627,
+      "step": 59622
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.7574273072060684e-07,
+      "loss": 0.8962,
+      "step": 59623
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.756953223767383e-07,
+      "loss": 1.8918,
+      "step": 59624
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.756479140328698e-07,
+      "loss": 1.5236,
+      "step": 59625
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7560050568900126e-07,
+      "loss": 3.1305,
+      "step": 59626
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7555309734513272e-07,
+      "loss": 1.8383,
+      "step": 59627
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7550568900126424e-07,
+      "loss": 1.6379,
+      "step": 59628
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.754582806573957e-07,
+      "loss": 2.1083,
+      "step": 59629
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.754108723135272e-07,
+      "loss": 3.08,
+      "step": 59630
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7536346396965867e-07,
+      "loss": 2.3609,
+      "step": 59631
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7531605562579013e-07,
+      "loss": 1.6131,
+      "step": 59632
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7526864728192162e-07,
+      "loss": 1.3092,
+      "step": 59633
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.752212389380531e-07,
+      "loss": 2.0947,
+      "step": 59634
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7517383059418458e-07,
+      "loss": 1.6012,
+      "step": 59635
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7512642225031605e-07,
+      "loss": 1.3296,
+      "step": 59636
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7507901390644754e-07,
+      "loss": 1.6826,
+      "step": 59637
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.75031605562579e-07,
+      "loss": 2.4722,
+      "step": 59638
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.749841972187105e-07,
+      "loss": 1.789,
+      "step": 59639
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.74936788874842e-07,
+      "loss": 2.181,
+      "step": 59640
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7488938053097346e-07,
+      "loss": 2.5167,
+      "step": 59641
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7484197218710495e-07,
+      "loss": 1.7501,
+      "step": 59642
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7479456384323641e-07,
+      "loss": 2.4135,
+      "step": 59643
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7474715549936788e-07,
+      "loss": 3.8538,
+      "step": 59644
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7469974715549937e-07,
+      "loss": 1.5758,
+      "step": 59645
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7465233881163084e-07,
+      "loss": 0.7814,
+      "step": 59646
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7460493046776233e-07,
+      "loss": 1.1693,
+      "step": 59647
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.745575221238938e-07,
+      "loss": 1.4861,
+      "step": 59648
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.745101137800253e-07,
+      "loss": 1.6329,
+      "step": 59649
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7446270543615678e-07,
+      "loss": 2.722,
+      "step": 59650
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7441529709228825e-07,
+      "loss": 1.7069,
+      "step": 59651
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7436788874841974e-07,
+      "loss": 1.3922,
+      "step": 59652
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.743204804045512e-07,
+      "loss": 0.5536,
+      "step": 59653
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.742730720606827e-07,
+      "loss": 1.4887,
+      "step": 59654
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7422566371681416e-07,
+      "loss": 1.9118,
+      "step": 59655
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7417825537294565e-07,
+      "loss": 1.9129,
+      "step": 59656
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7413084702907712e-07,
+      "loss": 2.6377,
+      "step": 59657
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7408343868520858e-07,
+      "loss": 2.6703,
+      "step": 59658
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7403603034134008e-07,
+      "loss": 2.8691,
+      "step": 59659
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7398862199747154e-07,
+      "loss": 2.3321,
+      "step": 59660
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7394121365360306e-07,
+      "loss": 1.4614,
+      "step": 59661
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7389380530973453e-07,
+      "loss": 0.7717,
+      "step": 59662
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.73846396965866e-07,
+      "loss": 1.4971,
+      "step": 59663
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7379898862199749e-07,
+      "loss": 0.7918,
+      "step": 59664
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7375158027812895e-07,
+      "loss": 1.4582,
+      "step": 59665
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7370417193426044e-07,
+      "loss": 2.2144,
+      "step": 59666
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.736567635903919e-07,
+      "loss": 1.7218,
+      "step": 59667
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.736093552465234e-07,
+      "loss": 4.2617,
+      "step": 59668
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7356194690265487e-07,
+      "loss": 1.8108,
+      "step": 59669
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7351453855878633e-07,
+      "loss": 2.4895,
+      "step": 59670
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7346713021491782e-07,
+      "loss": 1.1723,
+      "step": 59671
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7341972187104932e-07,
+      "loss": 2.2936,
+      "step": 59672
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.733723135271808e-07,
+      "loss": 1.7637,
+      "step": 59673
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7332490518331227e-07,
+      "loss": 1.228,
+      "step": 59674
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7327749683944374e-07,
+      "loss": 1.6053,
+      "step": 59675
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7323008849557523e-07,
+      "loss": 1.6801,
+      "step": 59676
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.731826801517067e-07,
+      "loss": 0.9523,
+      "step": 59677
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.731352718078382e-07,
+      "loss": 2.7459,
+      "step": 59678
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7308786346396966e-07,
+      "loss": 3.1676,
+      "step": 59679
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7304045512010115e-07,
+      "loss": 1.1906,
+      "step": 59680
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7299304677623261e-07,
+      "loss": 2.8146,
+      "step": 59681
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7294563843236408e-07,
+      "loss": 2.2541,
+      "step": 59682
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7289823008849557e-07,
+      "loss": 0.7431,
+      "step": 59683
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7285082174462706e-07,
+      "loss": 2.9955,
+      "step": 59684
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7280341340075856e-07,
+      "loss": 0.8779,
+      "step": 59685
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7275600505689002e-07,
+      "loss": 2.3032,
+      "step": 59686
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.727085967130215e-07,
+      "loss": 1.7453,
+      "step": 59687
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.7266118836915298e-07,
+      "loss": 1.8963,
+      "step": 59688
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7261378002528445e-07,
+      "loss": 1.5417,
+      "step": 59689
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7256637168141594e-07,
+      "loss": 2.3259,
+      "step": 59690
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.725189633375474e-07,
+      "loss": 2.4093,
+      "step": 59691
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.724715549936789e-07,
+      "loss": 2.3013,
+      "step": 59692
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7242414664981036e-07,
+      "loss": 1.32,
+      "step": 59693
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7237673830594185e-07,
+      "loss": 2.1539,
+      "step": 59694
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7232932996207335e-07,
+      "loss": 1.3748,
+      "step": 59695
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.722819216182048e-07,
+      "loss": 1.6509,
+      "step": 59696
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.722345132743363e-07,
+      "loss": 2.6383,
+      "step": 59697
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7218710493046777e-07,
+      "loss": 2.4989,
+      "step": 59698
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7213969658659923e-07,
+      "loss": 3.4678,
+      "step": 59699
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7209228824273073e-07,
+      "loss": 1.395,
+      "step": 59700
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.720448798988622e-07,
+      "loss": 2.0165,
+      "step": 59701
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7199747155499368e-07,
+      "loss": 1.9737,
+      "step": 59702
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7195006321112515e-07,
+      "loss": 3.7637,
+      "step": 59703
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7190265486725664e-07,
+      "loss": 0.8582,
+      "step": 59704
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.718552465233881e-07,
+      "loss": 1.3473,
+      "step": 59705
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.718078381795196e-07,
+      "loss": 3.13,
+      "step": 59706
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.717604298356511e-07,
+      "loss": 2.6075,
+      "step": 59707
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7171302149178256e-07,
+      "loss": 1.2333,
+      "step": 59708
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7166561314791405e-07,
+      "loss": 3.9022,
+      "step": 59709
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7161820480404552e-07,
+      "loss": 2.6119,
+      "step": 59710
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.71570796460177e-07,
+      "loss": 1.2654,
+      "step": 59711
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7152338811630847e-07,
+      "loss": 1.6544,
+      "step": 59712
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7147597977243994e-07,
+      "loss": 1.0014,
+      "step": 59713
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7142857142857143e-07,
+      "loss": 1.3396,
+      "step": 59714
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.713811630847029e-07,
+      "loss": 1.7153,
+      "step": 59715
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.713337547408344e-07,
+      "loss": 1.3353,
+      "step": 59716
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7128634639696588e-07,
+      "loss": 0.2527,
+      "step": 59717
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7123893805309735e-07,
+      "loss": 3.6398,
+      "step": 59718
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7119152970922884e-07,
+      "loss": 1.9307,
+      "step": 59719
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.711441213653603e-07,
+      "loss": 0.8684,
+      "step": 59720
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.710967130214918e-07,
+      "loss": 2.1973,
+      "step": 59721
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7104930467762326e-07,
+      "loss": 2.3072,
+      "step": 59722
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7100189633375476e-07,
+      "loss": 2.3102,
+      "step": 59723
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7095448798988622e-07,
+      "loss": 1.7151,
+      "step": 59724
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.709070796460177e-07,
+      "loss": 1.0576,
+      "step": 59725
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7085967130214918e-07,
+      "loss": 1.364,
+      "step": 59726
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7081226295828065e-07,
+      "loss": 1.3744,
+      "step": 59727
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7076485461441216e-07,
+      "loss": 3.0533,
+      "step": 59728
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7071744627054363e-07,
+      "loss": 1.5943,
+      "step": 59729
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.706700379266751e-07,
+      "loss": 1.0417,
+      "step": 59730
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.706226295828066e-07,
+      "loss": 1.7041,
+      "step": 59731
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7057522123893805e-07,
+      "loss": 1.6745,
+      "step": 59732
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7052781289506955e-07,
+      "loss": 1.3956,
+      "step": 59733
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.70480404551201e-07,
+      "loss": 2.4825,
+      "step": 59734
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.704329962073325e-07,
+      "loss": 1.3798,
+      "step": 59735
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7038558786346397e-07,
+      "loss": 2.956,
+      "step": 59736
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7033817951959543e-07,
+      "loss": 2.9343,
+      "step": 59737
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7029077117572693e-07,
+      "loss": 0.7014,
+      "step": 59738
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.702433628318584e-07,
+      "loss": 0.8313,
+      "step": 59739
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.701959544879899e-07,
+      "loss": 1.3364,
+      "step": 59740
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7014854614412138e-07,
+      "loss": 1.5489,
+      "step": 59741
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7010113780025284e-07,
+      "loss": 1.5494,
+      "step": 59742
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.7005372945638433e-07,
+      "loss": 2.3456,
+      "step": 59743
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.700063211125158e-07,
+      "loss": 2.8469,
+      "step": 59744
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.699589127686473e-07,
+      "loss": 0.6706,
+      "step": 59745
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.6991150442477876e-07,
+      "loss": 0.8597,
+      "step": 59746
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.6986409608091025e-07,
+      "loss": 2.3552,
+      "step": 59747
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.6981668773704172e-07,
+      "loss": 3.2404,
+      "step": 59748
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.6976927939317318e-07,
+      "loss": 2.4897,
+      "step": 59749
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.6972187104930467e-07,
+      "loss": 1.9542,
+      "step": 59750
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.6967446270543617e-07,
+      "loss": 1.2446,
+      "step": 59751
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6962705436156766e-07,
+      "loss": 1.8747,
+      "step": 59752
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6957964601769912e-07,
+      "loss": 2.6046,
+      "step": 59753
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.695322376738306e-07,
+      "loss": 2.9486,
+      "step": 59754
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6948482932996208e-07,
+      "loss": 2.2103,
+      "step": 59755
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6943742098609355e-07,
+      "loss": 1.6817,
+      "step": 59756
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6939001264222504e-07,
+      "loss": 0.9645,
+      "step": 59757
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.693426042983565e-07,
+      "loss": 2.2673,
+      "step": 59758
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.69295195954488e-07,
+      "loss": 2.4365,
+      "step": 59759
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6924778761061946e-07,
+      "loss": 2.3499,
+      "step": 59760
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6920037926675096e-07,
+      "loss": 1.1485,
+      "step": 59761
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6915297092288245e-07,
+      "loss": 4.0649,
+      "step": 59762
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6910556257901391e-07,
+      "loss": 0.3749,
+      "step": 59763
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.690581542351454e-07,
+      "loss": 1.4643,
+      "step": 59764
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6901074589127687e-07,
+      "loss": 2.5131,
+      "step": 59765
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6896333754740834e-07,
+      "loss": 1.1462,
+      "step": 59766
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6891592920353983e-07,
+      "loss": 2.5942,
+      "step": 59767
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.688685208596713e-07,
+      "loss": 2.4661,
+      "step": 59768
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.688211125158028e-07,
+      "loss": 1.3216,
+      "step": 59769
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6877370417193425e-07,
+      "loss": 1.2185,
+      "step": 59770
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6872629582806575e-07,
+      "loss": 2.0612,
+      "step": 59771
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.686788874841972e-07,
+      "loss": 0.9545,
+      "step": 59772
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.686314791403287e-07,
+      "loss": 1.0005,
+      "step": 59773
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.685840707964602e-07,
+      "loss": 2.978,
+      "step": 59774
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6853666245259166e-07,
+      "loss": 1.5257,
+      "step": 59775
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6848925410872315e-07,
+      "loss": 1.8397,
+      "step": 59776
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6844184576485462e-07,
+      "loss": 2.1139,
+      "step": 59777
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.683944374209861e-07,
+      "loss": 1.8717,
+      "step": 59778
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6834702907711758e-07,
+      "loss": 3.48,
+      "step": 59779
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6829962073324904e-07,
+      "loss": 2.4993,
+      "step": 59780
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6825221238938053e-07,
+      "loss": 2.2685,
+      "step": 59781
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.68204804045512e-07,
+      "loss": 0.8581,
+      "step": 59782
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.681573957016435e-07,
+      "loss": 1.207,
+      "step": 59783
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6810998735777498e-07,
+      "loss": 2.3808,
+      "step": 59784
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6806257901390645e-07,
+      "loss": 4.0382,
+      "step": 59785
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6801517067003794e-07,
+      "loss": 2.1898,
+      "step": 59786
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.679677623261694e-07,
+      "loss": 3.156,
+      "step": 59787
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.679203539823009e-07,
+      "loss": 1.1952,
+      "step": 59788
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6787294563843237e-07,
+      "loss": 2.4236,
+      "step": 59789
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6782553729456386e-07,
+      "loss": 1.7849,
+      "step": 59790
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6777812895069532e-07,
+      "loss": 2.1637,
+      "step": 59791
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.677307206068268e-07,
+      "loss": 0.5653,
+      "step": 59792
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6768331226295828e-07,
+      "loss": 2.251,
+      "step": 59793
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6763590391908975e-07,
+      "loss": 2.2626,
+      "step": 59794
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6758849557522124e-07,
+      "loss": 3.263,
+      "step": 59795
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6754108723135273e-07,
+      "loss": 0.7772,
+      "step": 59796
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.674936788874842e-07,
+      "loss": 1.9846,
+      "step": 59797
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.674462705436157e-07,
+      "loss": 1.3385,
+      "step": 59798
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6739886219974716e-07,
+      "loss": 2.6271,
+      "step": 59799
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6735145385587865e-07,
+      "loss": 2.4024,
+      "step": 59800
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6730404551201011e-07,
+      "loss": 1.4335,
+      "step": 59801
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.672566371681416e-07,
+      "loss": 1.5385,
+      "step": 59802
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6720922882427307e-07,
+      "loss": 0.9911,
+      "step": 59803
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6716182048040454e-07,
+      "loss": 1.7241,
+      "step": 59804
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6711441213653603e-07,
+      "loss": 1.067,
+      "step": 59805
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.670670037926675e-07,
+      "loss": 1.9892,
+      "step": 59806
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6701959544879901e-07,
+      "loss": 1.6612,
+      "step": 59807
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6697218710493048e-07,
+      "loss": 1.9234,
+      "step": 59808
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6692477876106194e-07,
+      "loss": 0.4716,
+      "step": 59809
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6687737041719344e-07,
+      "loss": 1.5154,
+      "step": 59810
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.668299620733249e-07,
+      "loss": 2.9626,
+      "step": 59811
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.667825537294564e-07,
+      "loss": 0.4999,
+      "step": 59812
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6673514538558786e-07,
+      "loss": 1.7864,
+      "step": 59813
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6668773704171935e-07,
+      "loss": 2.7991,
+      "step": 59814
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.6664032869785082e-07,
+      "loss": 2.0963,
+      "step": 59815
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.665929203539823e-07,
+      "loss": 1.6453,
+      "step": 59816
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6654551201011378e-07,
+      "loss": 2.6393,
+      "step": 59817
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6649810366624527e-07,
+      "loss": 1.6935,
+      "step": 59818
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6645069532237676e-07,
+      "loss": 1.1247,
+      "step": 59819
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6640328697850823e-07,
+      "loss": 2.9508,
+      "step": 59820
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.663558786346397e-07,
+      "loss": 2.0163,
+      "step": 59821
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6630847029077118e-07,
+      "loss": 3.2029,
+      "step": 59822
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6626106194690265e-07,
+      "loss": 0.493,
+      "step": 59823
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6621365360303414e-07,
+      "loss": 1.7217,
+      "step": 59824
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.661662452591656e-07,
+      "loss": 1.1066,
+      "step": 59825
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.661188369152971e-07,
+      "loss": 1.5085,
+      "step": 59826
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6607142857142857e-07,
+      "loss": 1.6787,
+      "step": 59827
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6602402022756006e-07,
+      "loss": 1.3514,
+      "step": 59828
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6597661188369155e-07,
+      "loss": 2.191,
+      "step": 59829
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6592920353982302e-07,
+      "loss": 2.8169,
+      "step": 59830
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.658817951959545e-07,
+      "loss": 2.7344,
+      "step": 59831
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6583438685208597e-07,
+      "loss": 2.5949,
+      "step": 59832
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6578697850821747e-07,
+      "loss": 2.1674,
+      "step": 59833
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6573957016434893e-07,
+      "loss": 1.6336,
+      "step": 59834
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.656921618204804e-07,
+      "loss": 1.2661,
+      "step": 59835
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.656447534766119e-07,
+      "loss": 1.4392,
+      "step": 59836
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6559734513274336e-07,
+      "loss": 1.0884,
+      "step": 59837
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6554993678887485e-07,
+      "loss": 2.0611,
+      "step": 59838
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.655025284450063e-07,
+      "loss": 2.1329,
+      "step": 59839
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.654551201011378e-07,
+      "loss": 2.3681,
+      "step": 59840
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.654077117572693e-07,
+      "loss": 1.7227,
+      "step": 59841
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6536030341340076e-07,
+      "loss": 2.6265,
+      "step": 59842
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6531289506953226e-07,
+      "loss": 2.0846,
+      "step": 59843
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6526548672566372e-07,
+      "loss": 0.69,
+      "step": 59844
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.652180783817952e-07,
+      "loss": 1.3718,
+      "step": 59845
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6517067003792668e-07,
+      "loss": 0.5129,
+      "step": 59846
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6512326169405814e-07,
+      "loss": 3.2877,
+      "step": 59847
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6507585335018964e-07,
+      "loss": 2.3552,
+      "step": 59848
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.650284450063211e-07,
+      "loss": 1.5653,
+      "step": 59849
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.649810366624526e-07,
+      "loss": 1.0358,
+      "step": 59850
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6493362831858406e-07,
+      "loss": 2.2304,
+      "step": 59851
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6488621997471555e-07,
+      "loss": 1.2507,
+      "step": 59852
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6483881163084704e-07,
+      "loss": 1.9411,
+      "step": 59853
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.647914032869785e-07,
+      "loss": 1.1574,
+      "step": 59854
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6474399494311e-07,
+      "loss": 2.2743,
+      "step": 59855
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6469658659924147e-07,
+      "loss": 1.4936,
+      "step": 59856
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6464917825537296e-07,
+      "loss": 2.7331,
+      "step": 59857
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6460176991150443e-07,
+      "loss": 2.186,
+      "step": 59858
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.645543615676359e-07,
+      "loss": 0.7033,
+      "step": 59859
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6450695322376738e-07,
+      "loss": 1.5219,
+      "step": 59860
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6445954487989885e-07,
+      "loss": 1.8508,
+      "step": 59861
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6441213653603034e-07,
+      "loss": 1.6223,
+      "step": 59862
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6436472819216183e-07,
+      "loss": 1.9605,
+      "step": 59863
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.643173198482933e-07,
+      "loss": 3.2584,
+      "step": 59864
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.642699115044248e-07,
+      "loss": 1.4985,
+      "step": 59865
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6422250316055626e-07,
+      "loss": 2.2539,
+      "step": 59866
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6417509481668775e-07,
+      "loss": 0.6652,
+      "step": 59867
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6412768647281922e-07,
+      "loss": 3.1077,
+      "step": 59868
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.640802781289507e-07,
+      "loss": 1.3122,
+      "step": 59869
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6403286978508217e-07,
+      "loss": 1.5108,
+      "step": 59870
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6398546144121364e-07,
+      "loss": 1.1471,
+      "step": 59871
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6393805309734513e-07,
+      "loss": 2.3204,
+      "step": 59872
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.638906447534766e-07,
+      "loss": 1.3971,
+      "step": 59873
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6384323640960812e-07,
+      "loss": 1.7403,
+      "step": 59874
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6379582806573958e-07,
+      "loss": 2.1278,
+      "step": 59875
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6374841972187105e-07,
+      "loss": 1.8903,
+      "step": 59876
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.6370101137800254e-07,
+      "loss": 1.9284,
+      "step": 59877
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.63653603034134e-07,
+      "loss": 1.4495,
+      "step": 59878
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.636061946902655e-07,
+      "loss": 2.576,
+      "step": 59879
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6355878634639696e-07,
+      "loss": 2.8675,
+      "step": 59880
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6351137800252846e-07,
+      "loss": 2.1014,
+      "step": 59881
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6346396965865992e-07,
+      "loss": 2.1639,
+      "step": 59882
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.634165613147914e-07,
+      "loss": 1.9514,
+      "step": 59883
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6336915297092288e-07,
+      "loss": 2.5471,
+      "step": 59884
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6332174462705437e-07,
+      "loss": 2.36,
+      "step": 59885
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6327433628318586e-07,
+      "loss": 2.2024,
+      "step": 59886
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6322692793931733e-07,
+      "loss": 4.0443,
+      "step": 59887
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.631795195954488e-07,
+      "loss": 1.9983,
+      "step": 59888
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6313211125158029e-07,
+      "loss": 2.2328,
+      "step": 59889
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6308470290771175e-07,
+      "loss": 2.7888,
+      "step": 59890
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6303729456384324e-07,
+      "loss": 2.1645,
+      "step": 59891
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.629898862199747e-07,
+      "loss": 1.1444,
+      "step": 59892
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.629424778761062e-07,
+      "loss": 1.908,
+      "step": 59893
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6289506953223767e-07,
+      "loss": 2.5547,
+      "step": 59894
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6284766118836916e-07,
+      "loss": 0.7525,
+      "step": 59895
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6280025284450065e-07,
+      "loss": 1.8539,
+      "step": 59896
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6275284450063212e-07,
+      "loss": 1.9754,
+      "step": 59897
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.627054361567636e-07,
+      "loss": 2.4005,
+      "step": 59898
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6265802781289508e-07,
+      "loss": 2.3744,
+      "step": 59899
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6261061946902657e-07,
+      "loss": 0.156,
+      "step": 59900
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6256321112515803e-07,
+      "loss": 0.9594,
+      "step": 59901
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.625158027812895e-07,
+      "loss": 3.1779,
+      "step": 59902
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.62468394437421e-07,
+      "loss": 3.8866,
+      "step": 59903
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6242098609355246e-07,
+      "loss": 0.7502,
+      "step": 59904
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6237357774968395e-07,
+      "loss": 1.914,
+      "step": 59905
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6232616940581542e-07,
+      "loss": 1.0719,
+      "step": 59906
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.622787610619469e-07,
+      "loss": 1.5624,
+      "step": 59907
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.622313527180784e-07,
+      "loss": 2.5893,
+      "step": 59908
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6218394437420987e-07,
+      "loss": 1.8004,
+      "step": 59909
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6213653603034136e-07,
+      "loss": 3.1941,
+      "step": 59910
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6208912768647282e-07,
+      "loss": 1.2968,
+      "step": 59911
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6204171934260432e-07,
+      "loss": 1.8414,
+      "step": 59912
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6199431099873578e-07,
+      "loss": 1.0161,
+      "step": 59913
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6194690265486725e-07,
+      "loss": 2.3681,
+      "step": 59914
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6189949431099874e-07,
+      "loss": 1.2934,
+      "step": 59915
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.618520859671302e-07,
+      "loss": 1.9449,
+      "step": 59916
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.618046776232617e-07,
+      "loss": 1.7606,
+      "step": 59917
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6175726927939316e-07,
+      "loss": 2.0573,
+      "step": 59918
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6170986093552465e-07,
+      "loss": 2.3986,
+      "step": 59919
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6166245259165615e-07,
+      "loss": 0.969,
+      "step": 59920
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.616150442477876e-07,
+      "loss": 0.5846,
+      "step": 59921
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.615676359039191e-07,
+      "loss": 1.3781,
+      "step": 59922
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6152022756005057e-07,
+      "loss": 1.9161,
+      "step": 59923
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6147281921618206e-07,
+      "loss": 3.0671,
+      "step": 59924
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6142541087231353e-07,
+      "loss": 1.2353,
+      "step": 59925
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.61378002528445e-07,
+      "loss": 1.8696,
+      "step": 59926
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6133059418457649e-07,
+      "loss": 1.1013,
+      "step": 59927
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6128318584070795e-07,
+      "loss": 1.4966,
+      "step": 59928
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6123577749683944e-07,
+      "loss": 1.7287,
+      "step": 59929
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6118836915297094e-07,
+      "loss": 2.3303,
+      "step": 59930
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.611409608091024e-07,
+      "loss": 1.5028,
+      "step": 59931
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.610935524652339e-07,
+      "loss": 2.3356,
+      "step": 59932
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6104614412136536e-07,
+      "loss": 1.8317,
+      "step": 59933
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6099873577749685e-07,
+      "loss": 1.7807,
+      "step": 59934
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6095132743362832e-07,
+      "loss": 1.1614,
+      "step": 59935
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.609039190897598e-07,
+      "loss": 1.6536,
+      "step": 59936
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6085651074589128e-07,
+      "loss": 1.0449,
+      "step": 59937
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6080910240202277e-07,
+      "loss": 1.4655,
+      "step": 59938
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6076169405815423e-07,
+      "loss": 1.3012,
+      "step": 59939
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.607142857142857e-07,
+      "loss": 1.5232,
+      "step": 59940
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.6066687737041722e-07,
+      "loss": 3.0839,
+      "step": 59941
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6061946902654868e-07,
+      "loss": 1.3062,
+      "step": 59942
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6057206068268015e-07,
+      "loss": 1.1947,
+      "step": 59943
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6052465233881164e-07,
+      "loss": 2.0971,
+      "step": 59944
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.604772439949431e-07,
+      "loss": 1.1521,
+      "step": 59945
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.604298356510746e-07,
+      "loss": 2.1438,
+      "step": 59946
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6038242730720607e-07,
+      "loss": 2.3929,
+      "step": 59947
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6033501896333756e-07,
+      "loss": 2.3849,
+      "step": 59948
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6028761061946902e-07,
+      "loss": 1.1616,
+      "step": 59949
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6024020227560052e-07,
+      "loss": 1.3743,
+      "step": 59950
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6019279393173198e-07,
+      "loss": 1.6244,
+      "step": 59951
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6014538558786345e-07,
+      "loss": 2.0232,
+      "step": 59952
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6009797724399497e-07,
+      "loss": 1.1761,
+      "step": 59953
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6005056890012643e-07,
+      "loss": 1.6505,
+      "step": 59954
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.6000316055625792e-07,
+      "loss": 1.358,
+      "step": 59955
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.599557522123894e-07,
+      "loss": 0.9059,
+      "step": 59956
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5990834386852085e-07,
+      "loss": 2.3325,
+      "step": 59957
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5986093552465235e-07,
+      "loss": 2.1442,
+      "step": 59958
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.598135271807838e-07,
+      "loss": 1.4649,
+      "step": 59959
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.597661188369153e-07,
+      "loss": 2.8582,
+      "step": 59960
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5971871049304677e-07,
+      "loss": 3.4244,
+      "step": 59961
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5967130214917826e-07,
+      "loss": 3.0233,
+      "step": 59962
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5962389380530973e-07,
+      "loss": 0.7929,
+      "step": 59963
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5957648546144122e-07,
+      "loss": 1.8239,
+      "step": 59964
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.595290771175727e-07,
+      "loss": 1.7089,
+      "step": 59965
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5948166877370418e-07,
+      "loss": 1.4371,
+      "step": 59966
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5943426042983567e-07,
+      "loss": 2.7828,
+      "step": 59967
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5938685208596714e-07,
+      "loss": 1.0162,
+      "step": 59968
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.593394437420986e-07,
+      "loss": 0.976,
+      "step": 59969
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.592920353982301e-07,
+      "loss": 2.0339,
+      "step": 59970
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5924462705436156e-07,
+      "loss": 1.5203,
+      "step": 59971
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5919721871049305e-07,
+      "loss": 1.4941,
+      "step": 59972
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5914981036662452e-07,
+      "loss": 1.2696,
+      "step": 59973
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.59102402022756e-07,
+      "loss": 2.4738,
+      "step": 59974
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.590549936788875e-07,
+      "loss": 1.9252,
+      "step": 59975
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5900758533501897e-07,
+      "loss": 2.8306,
+      "step": 59976
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5896017699115046e-07,
+      "loss": 2.6281,
+      "step": 59977
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5891276864728193e-07,
+      "loss": 0.8343,
+      "step": 59978
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5886536030341342e-07,
+      "loss": 1.96,
+      "step": 59979
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5881795195954488e-07,
+      "loss": 1.165,
+      "step": 59980
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5877054361567635e-07,
+      "loss": 1.8039,
+      "step": 59981
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5872313527180784e-07,
+      "loss": 0.3853,
+      "step": 59982
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.586757269279393e-07,
+      "loss": 0.6362,
+      "step": 59983
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.586283185840708e-07,
+      "loss": 1.1046,
+      "step": 59984
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5858091024020227e-07,
+      "loss": 2.3567,
+      "step": 59985
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5853350189633376e-07,
+      "loss": 0.7077,
+      "step": 59986
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5848609355246525e-07,
+      "loss": 1.6902,
+      "step": 59987
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5843868520859672e-07,
+      "loss": 1.784,
+      "step": 59988
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.583912768647282e-07,
+      "loss": 2.6669,
+      "step": 59989
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5834386852085967e-07,
+      "loss": 0.7699,
+      "step": 59990
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5829646017699117e-07,
+      "loss": 1.8649,
+      "step": 59991
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5824905183312263e-07,
+      "loss": 2.0407,
+      "step": 59992
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.582016434892541e-07,
+      "loss": 2.7379,
+      "step": 59993
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.581542351453856e-07,
+      "loss": 2.3379,
+      "step": 59994
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5810682680151705e-07,
+      "loss": 2.7683,
+      "step": 59995
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5805941845764855e-07,
+      "loss": 2.5511,
+      "step": 59996
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5801201011378004e-07,
+      "loss": 0.3518,
+      "step": 59997
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.579646017699115e-07,
+      "loss": 1.6233,
+      "step": 59998
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.57917193426043e-07,
+      "loss": 1.042,
+      "step": 59999
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5786978508217446e-07,
+      "loss": 1.4103,
+      "step": 60000
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5782237673830595e-07,
+      "loss": 0.2449,
+      "step": 60001
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5777496839443742e-07,
+      "loss": 1.4155,
+      "step": 60002
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.577275600505689e-07,
+      "loss": 1.3602,
+      "step": 60003
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5768015170670038e-07,
+      "loss": 0.88,
+      "step": 60004
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.5763274336283187e-07,
+      "loss": 1.7596,
+      "step": 60005
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5758533501896334e-07,
+      "loss": 0.4638,
+      "step": 60006
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.575379266750948e-07,
+      "loss": 3.8736,
+      "step": 60007
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.574905183312263e-07,
+      "loss": 2.1898,
+      "step": 60008
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5744310998735779e-07,
+      "loss": 1.5876,
+      "step": 60009
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5739570164348925e-07,
+      "loss": 2.3117,
+      "step": 60010
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5734829329962074e-07,
+      "loss": 1.5573,
+      "step": 60011
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.573008849557522e-07,
+      "loss": 1.0723,
+      "step": 60012
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.572534766118837e-07,
+      "loss": 2.1673,
+      "step": 60013
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5720606826801517e-07,
+      "loss": 1.1631,
+      "step": 60014
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5715865992414666e-07,
+      "loss": 1.7915,
+      "step": 60015
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5711125158027813e-07,
+      "loss": 3.6531,
+      "step": 60016
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5706384323640962e-07,
+      "loss": 2.021,
+      "step": 60017
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5701643489254108e-07,
+      "loss": 3.2236,
+      "step": 60018
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5696902654867255e-07,
+      "loss": 1.6203,
+      "step": 60019
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5692161820480407e-07,
+      "loss": 0.8995,
+      "step": 60020
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5687420986093553e-07,
+      "loss": 1.8967,
+      "step": 60021
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5682680151706703e-07,
+      "loss": 1.6592,
+      "step": 60022
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.567793931731985e-07,
+      "loss": 0.7369,
+      "step": 60023
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5673198482932996e-07,
+      "loss": 1.208,
+      "step": 60024
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5668457648546145e-07,
+      "loss": 3.4369,
+      "step": 60025
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5663716814159291e-07,
+      "loss": 0.9944,
+      "step": 60026
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.565897597977244e-07,
+      "loss": 1.4653,
+      "step": 60027
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5654235145385587e-07,
+      "loss": 2.2453,
+      "step": 60028
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5649494310998736e-07,
+      "loss": 0.4392,
+      "step": 60029
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5644753476611883e-07,
+      "loss": 1.4032,
+      "step": 60030
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5640012642225032e-07,
+      "loss": 1.2052,
+      "step": 60031
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5635271807838182e-07,
+      "loss": 1.3433,
+      "step": 60032
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5630530973451328e-07,
+      "loss": 2.7032,
+      "step": 60033
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5625790139064477e-07,
+      "loss": 1.201,
+      "step": 60034
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5621049304677624e-07,
+      "loss": 2.1448,
+      "step": 60035
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.561630847029077e-07,
+      "loss": 2.0295,
+      "step": 60036
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.561156763590392e-07,
+      "loss": 0.7799,
+      "step": 60037
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5606826801517066e-07,
+      "loss": 1.7259,
+      "step": 60038
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5602085967130215e-07,
+      "loss": 1.6726,
+      "step": 60039
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5597345132743362e-07,
+      "loss": 1.925,
+      "step": 60040
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.559260429835651e-07,
+      "loss": 1.3603,
+      "step": 60041
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.558786346396966e-07,
+      "loss": 1.522,
+      "step": 60042
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5583122629582807e-07,
+      "loss": 2.1942,
+      "step": 60043
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5578381795195956e-07,
+      "loss": 1.0565,
+      "step": 60044
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5573640960809103e-07,
+      "loss": 1.0553,
+      "step": 60045
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5568900126422252e-07,
+      "loss": 1.1051,
+      "step": 60046
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5564159292035399e-07,
+      "loss": 1.0721,
+      "step": 60047
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5559418457648545e-07,
+      "loss": 1.3892,
+      "step": 60048
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5554677623261694e-07,
+      "loss": 2.0507,
+      "step": 60049
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.554993678887484e-07,
+      "loss": 1.9092,
+      "step": 60050
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.554519595448799e-07,
+      "loss": 2.2721,
+      "step": 60051
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5540455120101137e-07,
+      "loss": 2.2052,
+      "step": 60052
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5535714285714286e-07,
+      "loss": 1.648,
+      "step": 60053
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5530973451327435e-07,
+      "loss": 1.0613,
+      "step": 60054
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5526232616940582e-07,
+      "loss": 1.8327,
+      "step": 60055
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.552149178255373e-07,
+      "loss": 3.284,
+      "step": 60056
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5516750948166878e-07,
+      "loss": 1.8925,
+      "step": 60057
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5512010113780027e-07,
+      "loss": 0.2506,
+      "step": 60058
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5507269279393173e-07,
+      "loss": 1.7799,
+      "step": 60059
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.550252844500632e-07,
+      "loss": 2.444,
+      "step": 60060
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.549778761061947e-07,
+      "loss": 2.3298,
+      "step": 60061
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5493046776232616e-07,
+      "loss": 1.2266,
+      "step": 60062
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5488305941845765e-07,
+      "loss": 2.3437,
+      "step": 60063
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5483565107458911e-07,
+      "loss": 3.1115,
+      "step": 60064
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.547882427307206e-07,
+      "loss": 3.3002,
+      "step": 60065
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.547408343868521e-07,
+      "loss": 3.7687,
+      "step": 60066
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5469342604298356e-07,
+      "loss": 2.6891,
+      "step": 60067
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.5464601769911506e-07,
+      "loss": 1.4182,
+      "step": 60068
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5459860935524652e-07,
+      "loss": 1.518,
+      "step": 60069
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5455120101137801e-07,
+      "loss": 1.0393,
+      "step": 60070
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5450379266750948e-07,
+      "loss": 2.3766,
+      "step": 60071
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5445638432364097e-07,
+      "loss": 2.8624,
+      "step": 60072
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5440897597977244e-07,
+      "loss": 1.7886,
+      "step": 60073
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.543615676359039e-07,
+      "loss": 0.5306,
+      "step": 60074
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.543141592920354e-07,
+      "loss": 2.0949,
+      "step": 60075
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.542667509481669e-07,
+      "loss": 1.4559,
+      "step": 60076
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5421934260429838e-07,
+      "loss": 2.5216,
+      "step": 60077
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5417193426042985e-07,
+      "loss": 1.5136,
+      "step": 60078
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.541245259165613e-07,
+      "loss": 1.7483,
+      "step": 60079
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.540771175726928e-07,
+      "loss": 2.4085,
+      "step": 60080
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5402970922882427e-07,
+      "loss": 0.9117,
+      "step": 60081
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5398230088495576e-07,
+      "loss": 1.8603,
+      "step": 60082
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5393489254108723e-07,
+      "loss": 1.8969,
+      "step": 60083
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5388748419721872e-07,
+      "loss": 0.7875,
+      "step": 60084
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5384007585335019e-07,
+      "loss": 2.3956,
+      "step": 60085
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5379266750948165e-07,
+      "loss": 2.414,
+      "step": 60086
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5374525916561317e-07,
+      "loss": 1.857,
+      "step": 60087
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5369785082174464e-07,
+      "loss": 2.5589,
+      "step": 60088
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5365044247787613e-07,
+      "loss": 1.0265,
+      "step": 60089
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.536030341340076e-07,
+      "loss": 1.0582,
+      "step": 60090
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5355562579013906e-07,
+      "loss": 0.9411,
+      "step": 60091
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5350821744627055e-07,
+      "loss": 1.4165,
+      "step": 60092
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5346080910240202e-07,
+      "loss": 0.4828,
+      "step": 60093
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.534134007585335e-07,
+      "loss": 2.3598,
+      "step": 60094
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5336599241466498e-07,
+      "loss": 3.8232,
+      "step": 60095
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5331858407079647e-07,
+      "loss": 0.3431,
+      "step": 60096
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5327117572692793e-07,
+      "loss": 1.2179,
+      "step": 60097
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5322376738305943e-07,
+      "loss": 2.2471,
+      "step": 60098
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5317635903919092e-07,
+      "loss": 0.8127,
+      "step": 60099
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5312895069532238e-07,
+      "loss": 2.0873,
+      "step": 60100
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5308154235145388e-07,
+      "loss": 0.8449,
+      "step": 60101
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5303413400758534e-07,
+      "loss": 1.9109,
+      "step": 60102
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.529867256637168e-07,
+      "loss": 1.0324,
+      "step": 60103
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.529393173198483e-07,
+      "loss": 2.2557,
+      "step": 60104
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5289190897597976e-07,
+      "loss": 3.0583,
+      "step": 60105
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5284450063211126e-07,
+      "loss": 1.0843,
+      "step": 60106
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5279709228824272e-07,
+      "loss": 1.7582,
+      "step": 60107
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5274968394437421e-07,
+      "loss": 2.8538,
+      "step": 60108
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.527022756005057e-07,
+      "loss": 1.4501,
+      "step": 60109
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5265486725663717e-07,
+      "loss": 2.2364,
+      "step": 60110
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5260745891276866e-07,
+      "loss": 1.3281,
+      "step": 60111
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5256005056890013e-07,
+      "loss": 2.6961,
+      "step": 60112
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5251264222503162e-07,
+      "loss": 2.4579,
+      "step": 60113
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.524652338811631e-07,
+      "loss": 2.1539,
+      "step": 60114
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5241782553729455e-07,
+      "loss": 1.3599,
+      "step": 60115
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5237041719342605e-07,
+      "loss": 1.9924,
+      "step": 60116
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.523230088495575e-07,
+      "loss": 1.5104,
+      "step": 60117
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.52275600505689e-07,
+      "loss": 2.3791,
+      "step": 60118
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5222819216182047e-07,
+      "loss": 1.6117,
+      "step": 60119
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5218078381795196e-07,
+      "loss": 1.2438,
+      "step": 60120
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5213337547408345e-07,
+      "loss": 0.8664,
+      "step": 60121
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5208596713021492e-07,
+      "loss": 1.9852,
+      "step": 60122
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.520385587863464e-07,
+      "loss": 1.2314,
+      "step": 60123
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5199115044247788e-07,
+      "loss": 2.0113,
+      "step": 60124
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5194374209860937e-07,
+      "loss": 1.0492,
+      "step": 60125
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5189633375474084e-07,
+      "loss": 0.9912,
+      "step": 60126
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5184892541087233e-07,
+      "loss": 1.7567,
+      "step": 60127
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.518015170670038e-07,
+      "loss": 1.6574,
+      "step": 60128
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5175410872313526e-07,
+      "loss": 2.3933,
+      "step": 60129
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5170670037926675e-07,
+      "loss": 1.2695,
+      "step": 60130
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.5165929203539822e-07,
+      "loss": 2.1877,
+      "step": 60131
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.516118836915297e-07,
+      "loss": 2.5177,
+      "step": 60132
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.515644753476612e-07,
+      "loss": 1.0806,
+      "step": 60133
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5151706700379267e-07,
+      "loss": 2.2972,
+      "step": 60134
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5146965865992416e-07,
+      "loss": 1.0621,
+      "step": 60135
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5142225031605562e-07,
+      "loss": 1.4726,
+      "step": 60136
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5137484197218712e-07,
+      "loss": 2.4452,
+      "step": 60137
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5132743362831858e-07,
+      "loss": 0.8874,
+      "step": 60138
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5128002528445008e-07,
+      "loss": 2.5221,
+      "step": 60139
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5123261694058154e-07,
+      "loss": 3.0522,
+      "step": 60140
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.51185208596713e-07,
+      "loss": 1.5865,
+      "step": 60141
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.511378002528445e-07,
+      "loss": 1.8818,
+      "step": 60142
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.51090391908976e-07,
+      "loss": 0.4167,
+      "step": 60143
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5104298356510748e-07,
+      "loss": 0.939,
+      "step": 60144
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5099557522123895e-07,
+      "loss": 1.5819,
+      "step": 60145
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5094816687737041e-07,
+      "loss": 1.483,
+      "step": 60146
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.509007585335019e-07,
+      "loss": 0.4254,
+      "step": 60147
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5085335018963337e-07,
+      "loss": 0.4658,
+      "step": 60148
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5080594184576486e-07,
+      "loss": 1.3915,
+      "step": 60149
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5075853350189633e-07,
+      "loss": 1.1906,
+      "step": 60150
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5071112515802782e-07,
+      "loss": 2.4353,
+      "step": 60151
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.506637168141593e-07,
+      "loss": 0.5546,
+      "step": 60152
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5061630847029075e-07,
+      "loss": 2.6058,
+      "step": 60153
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5056890012642227e-07,
+      "loss": 1.3284,
+      "step": 60154
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5052149178255374e-07,
+      "loss": 1.7361,
+      "step": 60155
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5047408343868523e-07,
+      "loss": 2.8528,
+      "step": 60156
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.504266750948167e-07,
+      "loss": 1.2162,
+      "step": 60157
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5037926675094816e-07,
+      "loss": 1.0385,
+      "step": 60158
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5033185840707965e-07,
+      "loss": 1.9583,
+      "step": 60159
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5028445006321112e-07,
+      "loss": 2.3707,
+      "step": 60160
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.502370417193426e-07,
+      "loss": 3.1779,
+      "step": 60161
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5018963337547408e-07,
+      "loss": 1.8108,
+      "step": 60162
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5014222503160557e-07,
+      "loss": 2.3604,
+      "step": 60163
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5009481668773704e-07,
+      "loss": 1.328,
+      "step": 60164
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5004740834386853e-07,
+      "loss": 1.8473,
+      "step": 60165
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.5000000000000002e-07,
+      "loss": 1.5584,
+      "step": 60166
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4995259165613149e-07,
+      "loss": 1.7033,
+      "step": 60167
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4990518331226298e-07,
+      "loss": 2.4829,
+      "step": 60168
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4985777496839444e-07,
+      "loss": 0.9372,
+      "step": 60169
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.498103666245259e-07,
+      "loss": 2.0373,
+      "step": 60170
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.497629582806574e-07,
+      "loss": 2.2578,
+      "step": 60171
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4971554993678887e-07,
+      "loss": 2.6177,
+      "step": 60172
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4966814159292036e-07,
+      "loss": 1.153,
+      "step": 60173
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4962073324905182e-07,
+      "loss": 0.6824,
+      "step": 60174
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4957332490518332e-07,
+      "loss": 2.0036,
+      "step": 60175
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4952591656131478e-07,
+      "loss": 1.4581,
+      "step": 60176
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4947850821744627e-07,
+      "loss": 2.8953,
+      "step": 60177
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4943109987357777e-07,
+      "loss": 0.5567,
+      "step": 60178
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4938369152970923e-07,
+      "loss": 1.6359,
+      "step": 60179
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4933628318584072e-07,
+      "loss": 1.7712,
+      "step": 60180
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.492888748419722e-07,
+      "loss": 1.5337,
+      "step": 60181
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4924146649810366e-07,
+      "loss": 1.0949,
+      "step": 60182
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4919405815423515e-07,
+      "loss": 1.0553,
+      "step": 60183
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4914664981036661e-07,
+      "loss": 3.9166,
+      "step": 60184
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.490992414664981e-07,
+      "loss": 3.7441,
+      "step": 60185
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4905183312262957e-07,
+      "loss": 3.7349,
+      "step": 60186
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4900442477876106e-07,
+      "loss": 1.6969,
+      "step": 60187
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4895701643489256e-07,
+      "loss": 1.1449,
+      "step": 60188
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4890960809102402e-07,
+      "loss": 0.8837,
+      "step": 60189
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4886219974715551e-07,
+      "loss": 1.9684,
+      "step": 60190
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4881479140328698e-07,
+      "loss": 1.5062,
+      "step": 60191
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4876738305941847e-07,
+      "loss": 2.5054,
+      "step": 60192
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4871997471554994e-07,
+      "loss": 1.5937,
+      "step": 60193
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.4867256637168143e-07,
+      "loss": 2.838,
+      "step": 60194
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.486251580278129e-07,
+      "loss": 2.3199,
+      "step": 60195
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4857774968394436e-07,
+      "loss": 1.8916,
+      "step": 60196
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4853034134007585e-07,
+      "loss": 2.1144,
+      "step": 60197
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4848293299620732e-07,
+      "loss": 1.6057,
+      "step": 60198
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4843552465233884e-07,
+      "loss": 2.5941,
+      "step": 60199
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.483881163084703e-07,
+      "loss": 3.1546,
+      "step": 60200
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4834070796460177e-07,
+      "loss": 2.0076,
+      "step": 60201
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4829329962073326e-07,
+      "loss": 0.9002,
+      "step": 60202
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4824589127686473e-07,
+      "loss": 3.373,
+      "step": 60203
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4819848293299622e-07,
+      "loss": 1.4254,
+      "step": 60204
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4815107458912769e-07,
+      "loss": 2.6165,
+      "step": 60205
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4810366624525918e-07,
+      "loss": 2.1106,
+      "step": 60206
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4805625790139064e-07,
+      "loss": 2.1329,
+      "step": 60207
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.480088495575221e-07,
+      "loss": 2.4848,
+      "step": 60208
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.479614412136536e-07,
+      "loss": 1.9842,
+      "step": 60209
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.479140328697851e-07,
+      "loss": 2.2226,
+      "step": 60210
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4786662452591659e-07,
+      "loss": 0.6778,
+      "step": 60211
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4781921618204805e-07,
+      "loss": 1.6967,
+      "step": 60212
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4777180783817952e-07,
+      "loss": 2.1025,
+      "step": 60213
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.47724399494311e-07,
+      "loss": 0.4368,
+      "step": 60214
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4767699115044247e-07,
+      "loss": 1.1525,
+      "step": 60215
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4762958280657397e-07,
+      "loss": 1.4338,
+      "step": 60216
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4758217446270543e-07,
+      "loss": 2.2162,
+      "step": 60217
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4753476611883692e-07,
+      "loss": 1.0856,
+      "step": 60218
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.474873577749684e-07,
+      "loss": 1.4733,
+      "step": 60219
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4743994943109986e-07,
+      "loss": 1.5693,
+      "step": 60220
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4739254108723137e-07,
+      "loss": 0.8445,
+      "step": 60221
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4734513274336284e-07,
+      "loss": 2.3108,
+      "step": 60222
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4729772439949433e-07,
+      "loss": 1.8423,
+      "step": 60223
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.472503160556258e-07,
+      "loss": 3.6036,
+      "step": 60224
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4720290771175726e-07,
+      "loss": 1.7862,
+      "step": 60225
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4715549936788876e-07,
+      "loss": 1.3534,
+      "step": 60226
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4710809102402022e-07,
+      "loss": 3.0395,
+      "step": 60227
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4706068268015171e-07,
+      "loss": 2.6297,
+      "step": 60228
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4701327433628318e-07,
+      "loss": 1.3925,
+      "step": 60229
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4696586599241467e-07,
+      "loss": 0.8173,
+      "step": 60230
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4691845764854614e-07,
+      "loss": 2.73,
+      "step": 60231
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4687104930467763e-07,
+      "loss": 1.0856,
+      "step": 60232
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4682364096080912e-07,
+      "loss": 2.3716,
+      "step": 60233
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.467762326169406e-07,
+      "loss": 0.994,
+      "step": 60234
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4672882427307208e-07,
+      "loss": 0.8568,
+      "step": 60235
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4668141592920355e-07,
+      "loss": 1.5833,
+      "step": 60236
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.46634007585335e-07,
+      "loss": 1.3729,
+      "step": 60237
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.465865992414665e-07,
+      "loss": 2.337,
+      "step": 60238
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4653919089759797e-07,
+      "loss": 0.462,
+      "step": 60239
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4649178255372946e-07,
+      "loss": 1.1477,
+      "step": 60240
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4644437420986093e-07,
+      "loss": 2.107,
+      "step": 60241
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4639696586599242e-07,
+      "loss": 2.0863,
+      "step": 60242
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4634955752212388e-07,
+      "loss": 2.0715,
+      "step": 60243
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4630214917825538e-07,
+      "loss": 1.0355,
+      "step": 60244
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4625474083438687e-07,
+      "loss": 1.893,
+      "step": 60245
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4620733249051834e-07,
+      "loss": 1.7558,
+      "step": 60246
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4615992414664983e-07,
+      "loss": 3.1959,
+      "step": 60247
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.461125158027813e-07,
+      "loss": 2.9719,
+      "step": 60248
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4606510745891279e-07,
+      "loss": 1.0728,
+      "step": 60249
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4601769911504425e-07,
+      "loss": 2.1025,
+      "step": 60250
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4597029077117572e-07,
+      "loss": 0.4063,
+      "step": 60251
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.459228824273072e-07,
+      "loss": 1.7934,
+      "step": 60252
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4587547408343867e-07,
+      "loss": 0.8806,
+      "step": 60253
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4582806573957017e-07,
+      "loss": 1.8668,
+      "step": 60254
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4578065739570166e-07,
+      "loss": 1.6586,
+      "step": 60255
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4573324905183312e-07,
+      "loss": 2.0917,
+      "step": 60256
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4568584070796462e-07,
+      "loss": 1.972,
+      "step": 60257
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 1.4563843236409608e-07,
+      "loss": 1.3336,
+      "step": 60258
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4559102402022757e-07,
+      "loss": 1.9723,
+      "step": 60259
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4554361567635904e-07,
+      "loss": 1.9735,
+      "step": 60260
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4549620733249053e-07,
+      "loss": 3.2043,
+      "step": 60261
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.45448798988622e-07,
+      "loss": 1.5677,
+      "step": 60262
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4540139064475346e-07,
+      "loss": 1.1711,
+      "step": 60263
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4535398230088496e-07,
+      "loss": 1.8941,
+      "step": 60264
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4530657395701642e-07,
+      "loss": 1.1729,
+      "step": 60265
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4525916561314794e-07,
+      "loss": 1.9418,
+      "step": 60266
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.452117572692794e-07,
+      "loss": 2.1768,
+      "step": 60267
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4516434892541087e-07,
+      "loss": 2.7221,
+      "step": 60268
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4511694058154236e-07,
+      "loss": 2.6556,
+      "step": 60269
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4506953223767383e-07,
+      "loss": 0.8622,
+      "step": 60270
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4502212389380532e-07,
+      "loss": 2.4596,
+      "step": 60271
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.449747155499368e-07,
+      "loss": 1.6236,
+      "step": 60272
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4492730720606828e-07,
+      "loss": 2.332,
+      "step": 60273
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4487989886219975e-07,
+      "loss": 1.5001,
+      "step": 60274
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.448324905183312e-07,
+      "loss": 1.1325,
+      "step": 60275
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.447850821744627e-07,
+      "loss": 4.0047,
+      "step": 60276
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.447376738305942e-07,
+      "loss": 1.0433,
+      "step": 60277
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.446902654867257e-07,
+      "loss": 1.3268,
+      "step": 60278
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4464285714285715e-07,
+      "loss": 2.4113,
+      "step": 60279
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4459544879898862e-07,
+      "loss": 0.5981,
+      "step": 60280
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.445480404551201e-07,
+      "loss": 0.8368,
+      "step": 60281
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4450063211125158e-07,
+      "loss": 2.5473,
+      "step": 60282
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4445322376738307e-07,
+      "loss": 1.3063,
+      "step": 60283
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4440581542351453e-07,
+      "loss": 1.2002,
+      "step": 60284
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4435840707964603e-07,
+      "loss": 3.2297,
+      "step": 60285
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.443109987357775e-07,
+      "loss": 1.331,
+      "step": 60286
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4426359039190896e-07,
+      "loss": 1.4519,
+      "step": 60287
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4421618204804045e-07,
+      "loss": 1.7191,
+      "step": 60288
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4416877370417194e-07,
+      "loss": 0.597,
+      "step": 60289
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4412136536030343e-07,
+      "loss": 2.1464,
+      "step": 60290
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.440739570164349e-07,
+      "loss": 2.2548,
+      "step": 60291
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4402654867256637e-07,
+      "loss": 3.1944,
+      "step": 60292
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4397914032869786e-07,
+      "loss": 1.2032,
+      "step": 60293
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4393173198482932e-07,
+      "loss": 5.3136,
+      "step": 60294
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4388432364096082e-07,
+      "loss": 2.2735,
+      "step": 60295
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4383691529709228e-07,
+      "loss": 2.0352,
+      "step": 60296
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4378950695322377e-07,
+      "loss": 3.3555,
+      "step": 60297
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4374209860935524e-07,
+      "loss": 2.5512,
+      "step": 60298
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4369469026548673e-07,
+      "loss": 1.2679,
+      "step": 60299
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4364728192161822e-07,
+      "loss": 1.3764,
+      "step": 60300
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.435998735777497e-07,
+      "loss": 1.549,
+      "step": 60301
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4355246523388118e-07,
+      "loss": 2.9747,
+      "step": 60302
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4350505689001265e-07,
+      "loss": 2.6551,
+      "step": 60303
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4345764854614411e-07,
+      "loss": 0.9341,
+      "step": 60304
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.434102402022756e-07,
+      "loss": 1.9416,
+      "step": 60305
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4336283185840707e-07,
+      "loss": 2.1438,
+      "step": 60306
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4331542351453856e-07,
+      "loss": 1.2501,
+      "step": 60307
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4326801517067003e-07,
+      "loss": 1.8422,
+      "step": 60308
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4322060682680152e-07,
+      "loss": 0.551,
+      "step": 60309
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.43173198482933e-07,
+      "loss": 0.8676,
+      "step": 60310
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4312579013906448e-07,
+      "loss": 0.5841,
+      "step": 60311
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4307838179519597e-07,
+      "loss": 1.9527,
+      "step": 60312
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4303097345132744e-07,
+      "loss": 1.2824,
+      "step": 60313
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4298356510745893e-07,
+      "loss": 2.3378,
+      "step": 60314
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.429361567635904e-07,
+      "loss": 2.7708,
+      "step": 60315
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.428887484197219e-07,
+      "loss": 2.3001,
+      "step": 60316
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4284134007585335e-07,
+      "loss": 0.4563,
+      "step": 60317
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4279393173198482e-07,
+      "loss": 2.0651,
+      "step": 60318
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.427465233881163e-07,
+      "loss": 2.8223,
+      "step": 60319
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4269911504424778e-07,
+      "loss": 1.7324,
+      "step": 60320
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 1.4265170670037927e-07,
+      "loss": 3.2812,
+      "step": 60321
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4260429835651076e-07,
+      "loss": 3.1787,
+      "step": 60322
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4255689001264223e-07,
+      "loss": 0.6183,
+      "step": 60323
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4250948166877372e-07,
+      "loss": 0.4501,
+      "step": 60324
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4246207332490518e-07,
+      "loss": 2.4647,
+      "step": 60325
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4241466498103668e-07,
+      "loss": 2.9533,
+      "step": 60326
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4236725663716814e-07,
+      "loss": 1.984,
+      "step": 60327
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4231984829329963e-07,
+      "loss": 2.6524,
+      "step": 60328
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.422724399494311e-07,
+      "loss": 1.3901,
+      "step": 60329
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4222503160556257e-07,
+      "loss": 2.453,
+      "step": 60330
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4217762326169406e-07,
+      "loss": 1.8423,
+      "step": 60331
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4213021491782552e-07,
+      "loss": 1.5512,
+      "step": 60332
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4208280657395704e-07,
+      "loss": 1.6258,
+      "step": 60333
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.420353982300885e-07,
+      "loss": 0.9256,
+      "step": 60334
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4198798988621997e-07,
+      "loss": 1.9805,
+      "step": 60335
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4194058154235147e-07,
+      "loss": 1.4838,
+      "step": 60336
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4189317319848293e-07,
+      "loss": 2.079,
+      "step": 60337
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4184576485461442e-07,
+      "loss": 1.9762,
+      "step": 60338
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.417983565107459e-07,
+      "loss": 0.4162,
+      "step": 60339
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4175094816687738e-07,
+      "loss": 3.0034,
+      "step": 60340
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4170353982300885e-07,
+      "loss": 1.6578,
+      "step": 60341
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4165613147914031e-07,
+      "loss": 1.2301,
+      "step": 60342
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.416087231352718e-07,
+      "loss": 1.1281,
+      "step": 60343
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4156131479140327e-07,
+      "loss": 1.1532,
+      "step": 60344
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.415139064475348e-07,
+      "loss": 3.5705,
+      "step": 60345
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4146649810366626e-07,
+      "loss": 1.3024,
+      "step": 60346
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4141908975979772e-07,
+      "loss": 2.5829,
+      "step": 60347
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4137168141592921e-07,
+      "loss": 0.7432,
+      "step": 60348
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4132427307206068e-07,
+      "loss": 2.5434,
+      "step": 60349
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4127686472819217e-07,
+      "loss": 2.1858,
+      "step": 60350
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4122945638432364e-07,
+      "loss": 1.4594,
+      "step": 60351
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4118204804045513e-07,
+      "loss": 1.8039,
+      "step": 60352
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.411346396965866e-07,
+      "loss": 2.3148,
+      "step": 60353
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.410872313527181e-07,
+      "loss": 1.4855,
+      "step": 60354
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4103982300884955e-07,
+      "loss": 1.9296,
+      "step": 60355
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4099241466498105e-07,
+      "loss": 0.9008,
+      "step": 60356
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4094500632111254e-07,
+      "loss": 1.4255,
+      "step": 60357
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.40897597977244e-07,
+      "loss": 1.119,
+      "step": 60358
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4085018963337547e-07,
+      "loss": 3.0271,
+      "step": 60359
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4080278128950696e-07,
+      "loss": 2.4571,
+      "step": 60360
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4075537294563843e-07,
+      "loss": 0.9807,
+      "step": 60361
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4070796460176992e-07,
+      "loss": 1.8906,
+      "step": 60362
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4066055625790138e-07,
+      "loss": 2.3692,
+      "step": 60363
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4061314791403288e-07,
+      "loss": 1.1955,
+      "step": 60364
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4056573957016434e-07,
+      "loss": 1.2139,
+      "step": 60365
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4051833122629583e-07,
+      "loss": 2.6126,
+      "step": 60366
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4047092288242733e-07,
+      "loss": 1.1676,
+      "step": 60367
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.404235145385588e-07,
+      "loss": 0.9094,
+      "step": 60368
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4037610619469028e-07,
+      "loss": 1.9543,
+      "step": 60369
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4032869785082175e-07,
+      "loss": 2.346,
+      "step": 60370
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4028128950695324e-07,
+      "loss": 1.0966,
+      "step": 60371
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.402338811630847e-07,
+      "loss": 2.1655,
+      "step": 60372
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4018647281921617e-07,
+      "loss": 2.2755,
+      "step": 60373
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4013906447534767e-07,
+      "loss": 0.9838,
+      "step": 60374
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4009165613147913e-07,
+      "loss": 2.6882,
+      "step": 60375
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.4004424778761062e-07,
+      "loss": 3.4975,
+      "step": 60376
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.399968394437421e-07,
+      "loss": 1.1985,
+      "step": 60377
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.3994943109987358e-07,
+      "loss": 2.7088,
+      "step": 60378
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.3990202275600507e-07,
+      "loss": 2.4946,
+      "step": 60379
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.3985461441213654e-07,
+      "loss": 1.9215,
+      "step": 60380
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.3980720606826803e-07,
+      "loss": 1.9823,
+      "step": 60381
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.397597977243995e-07,
+      "loss": 2.7606,
+      "step": 60382
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.39712389380531e-07,
+      "loss": 1.5684,
+      "step": 60383
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.3966498103666246e-07,
+      "loss": 2.8549,
+      "step": 60384
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 1.3961757269279392e-07,
+      "loss": 2.9026,
+      "step": 60385
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3957016434892541e-07,
+      "loss": 1.2574,
+      "step": 60386
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3952275600505688e-07,
+      "loss": 3.499,
+      "step": 60387
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3947534766118837e-07,
+      "loss": 1.9907,
+      "step": 60388
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3942793931731986e-07,
+      "loss": 1.5216,
+      "step": 60389
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3938053097345133e-07,
+      "loss": 1.348,
+      "step": 60390
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3933312262958282e-07,
+      "loss": 2.5401,
+      "step": 60391
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.392857142857143e-07,
+      "loss": 2.6187,
+      "step": 60392
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3923830594184578e-07,
+      "loss": 2.3306,
+      "step": 60393
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3919089759797724e-07,
+      "loss": 1.4639,
+      "step": 60394
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3914348925410874e-07,
+      "loss": 2.0388,
+      "step": 60395
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.390960809102402e-07,
+      "loss": 1.2511,
+      "step": 60396
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3904867256637167e-07,
+      "loss": 1.2316,
+      "step": 60397
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3900126422250316e-07,
+      "loss": 1.6322,
+      "step": 60398
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3895385587863463e-07,
+      "loss": 1.5997,
+      "step": 60399
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3890644753476612e-07,
+      "loss": 2.1373,
+      "step": 60400
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.388590391908976e-07,
+      "loss": 1.2718,
+      "step": 60401
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3881163084702908e-07,
+      "loss": 1.5196,
+      "step": 60402
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3876422250316057e-07,
+      "loss": 0.3084,
+      "step": 60403
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3871681415929203e-07,
+      "loss": 1.8055,
+      "step": 60404
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3866940581542353e-07,
+      "loss": 2.1524,
+      "step": 60405
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.38621997471555e-07,
+      "loss": 1.5754,
+      "step": 60406
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3857458912768648e-07,
+      "loss": 1.8233,
+      "step": 60407
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3852718078381795e-07,
+      "loss": 1.4655,
+      "step": 60408
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3847977243994942e-07,
+      "loss": 2.2497,
+      "step": 60409
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.384323640960809e-07,
+      "loss": 2.2264,
+      "step": 60410
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3838495575221237e-07,
+      "loss": 2.2423,
+      "step": 60411
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.383375474083439e-07,
+      "loss": 2.0582,
+      "step": 60412
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3829013906447536e-07,
+      "loss": 1.2924,
+      "step": 60413
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3824273072060682e-07,
+      "loss": 1.6362,
+      "step": 60414
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3819532237673832e-07,
+      "loss": 1.9018,
+      "step": 60415
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3814791403286978e-07,
+      "loss": 3.3179,
+      "step": 60416
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3810050568900127e-07,
+      "loss": 1.8735,
+      "step": 60417
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3805309734513274e-07,
+      "loss": 2.3524,
+      "step": 60418
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3800568900126423e-07,
+      "loss": 1.765,
+      "step": 60419
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.379582806573957e-07,
+      "loss": 1.0995,
+      "step": 60420
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.379108723135272e-07,
+      "loss": 1.0187,
+      "step": 60421
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3786346396965866e-07,
+      "loss": 0.851,
+      "step": 60422
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3781605562579015e-07,
+      "loss": 1.5461,
+      "step": 60423
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3776864728192164e-07,
+      "loss": 1.5735,
+      "step": 60424
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.377212389380531e-07,
+      "loss": 2.3901,
+      "step": 60425
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3767383059418457e-07,
+      "loss": 3.1132,
+      "step": 60426
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3762642225031606e-07,
+      "loss": 1.1542,
+      "step": 60427
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3757901390644753e-07,
+      "loss": 1.4646,
+      "step": 60428
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3753160556257902e-07,
+      "loss": 1.8274,
+      "step": 60429
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3748419721871049e-07,
+      "loss": 2.0323,
+      "step": 60430
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3743678887484198e-07,
+      "loss": 1.8359,
+      "step": 60431
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3738938053097344e-07,
+      "loss": 3.6407,
+      "step": 60432
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3734197218710494e-07,
+      "loss": 2.1224,
+      "step": 60433
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3729456384323643e-07,
+      "loss": 0.8206,
+      "step": 60434
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.372471554993679e-07,
+      "loss": 2.8154,
+      "step": 60435
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.371997471554994e-07,
+      "loss": 0.6516,
+      "step": 60436
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3715233881163085e-07,
+      "loss": 1.424,
+      "step": 60437
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3710493046776234e-07,
+      "loss": 2.0642,
+      "step": 60438
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.370575221238938e-07,
+      "loss": 1.5958,
+      "step": 60439
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3701011378002528e-07,
+      "loss": 1.8609,
+      "step": 60440
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3696270543615677e-07,
+      "loss": 2.9363,
+      "step": 60441
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3691529709228823e-07,
+      "loss": 2.5287,
+      "step": 60442
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3686788874841973e-07,
+      "loss": 1.8943,
+      "step": 60443
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.368204804045512e-07,
+      "loss": 1.606,
+      "step": 60444
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3677307206068268e-07,
+      "loss": 2.6175,
+      "step": 60445
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3672566371681418e-07,
+      "loss": 2.0296,
+      "step": 60446
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3667825537294564e-07,
+      "loss": 2.9963,
+      "step": 60447
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 1.3663084702907713e-07,
+      "loss": 1.6217,
+      "step": 60448
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.365834386852086e-07,
+      "loss": 2.8338,
+      "step": 60449
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.365360303413401e-07,
+      "loss": 1.8962,
+      "step": 60450
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3648862199747156e-07,
+      "loss": 1.8833,
+      "step": 60451
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3644121365360302e-07,
+      "loss": 0.1821,
+      "step": 60452
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3639380530973452e-07,
+      "loss": 1.1789,
+      "step": 60453
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3634639696586598e-07,
+      "loss": 1.1657,
+      "step": 60454
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3629898862199747e-07,
+      "loss": 0.6273,
+      "step": 60455
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3625158027812894e-07,
+      "loss": 1.0003,
+      "step": 60456
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3620417193426043e-07,
+      "loss": 1.0753,
+      "step": 60457
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3615676359039192e-07,
+      "loss": 1.9229,
+      "step": 60458
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.361093552465234e-07,
+      "loss": 1.7605,
+      "step": 60459
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3606194690265488e-07,
+      "loss": 1.0478,
+      "step": 60460
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3601453855878635e-07,
+      "loss": 2.0709,
+      "step": 60461
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3596713021491784e-07,
+      "loss": 0.5924,
+      "step": 60462
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.359197218710493e-07,
+      "loss": 2.8625,
+      "step": 60463
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3587231352718077e-07,
+      "loss": 0.6229,
+      "step": 60464
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3582490518331226e-07,
+      "loss": 2.5498,
+      "step": 60465
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3577749683944373e-07,
+      "loss": 0.5614,
+      "step": 60466
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3573008849557522e-07,
+      "loss": 2.1409,
+      "step": 60467
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.356826801517067e-07,
+      "loss": 2.6887,
+      "step": 60468
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3563527180783818e-07,
+      "loss": 1.8052,
+      "step": 60469
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3558786346396967e-07,
+      "loss": 1.9912,
+      "step": 60470
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3554045512010114e-07,
+      "loss": 1.8661,
+      "step": 60471
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3549304677623263e-07,
+      "loss": 1.2104,
+      "step": 60472
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.354456384323641e-07,
+      "loss": 2.3531,
+      "step": 60473
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3539823008849559e-07,
+      "loss": 1.3928,
+      "step": 60474
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3535082174462705e-07,
+      "loss": 1.7612,
+      "step": 60475
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3530341340075854e-07,
+      "loss": 0.7956,
+      "step": 60476
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3525600505689e-07,
+      "loss": 2.2409,
+      "step": 60477
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3520859671302148e-07,
+      "loss": 3.2155,
+      "step": 60478
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.35161188369153e-07,
+      "loss": 2.041,
+      "step": 60479
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3511378002528446e-07,
+      "loss": 1.5137,
+      "step": 60480
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3506637168141593e-07,
+      "loss": 2.4484,
+      "step": 60481
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3501896333754742e-07,
+      "loss": 1.3868,
+      "step": 60482
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3497155499367888e-07,
+      "loss": 1.3084,
+      "step": 60483
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3492414664981038e-07,
+      "loss": 0.4013,
+      "step": 60484
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3487673830594184e-07,
+      "loss": 1.4753,
+      "step": 60485
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3482932996207333e-07,
+      "loss": 2.3819,
+      "step": 60486
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.347819216182048e-07,
+      "loss": 0.956,
+      "step": 60487
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.347345132743363e-07,
+      "loss": 2.1788,
+      "step": 60488
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3468710493046776e-07,
+      "loss": 2.5105,
+      "step": 60489
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3463969658659925e-07,
+      "loss": 2.1059,
+      "step": 60490
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3459228824273074e-07,
+      "loss": 0.506,
+      "step": 60491
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.345448798988622e-07,
+      "loss": 1.4408,
+      "step": 60492
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.344974715549937e-07,
+      "loss": 2.4883,
+      "step": 60493
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3445006321112517e-07,
+      "loss": 1.7993,
+      "step": 60494
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3440265486725663e-07,
+      "loss": 1.6576,
+      "step": 60495
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3435524652338812e-07,
+      "loss": 1.571,
+      "step": 60496
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.343078381795196e-07,
+      "loss": 1.2932,
+      "step": 60497
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3426042983565108e-07,
+      "loss": 1.2388,
+      "step": 60498
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3421302149178255e-07,
+      "loss": 2.1058,
+      "step": 60499
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3416561314791404e-07,
+      "loss": 1.305,
+      "step": 60500
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3411820480404553e-07,
+      "loss": 2.0475,
+      "step": 60501
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.34070796460177e-07,
+      "loss": 2.6935,
+      "step": 60502
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.340233881163085e-07,
+      "loss": 0.9529,
+      "step": 60503
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3397597977243995e-07,
+      "loss": 2.9433,
+      "step": 60504
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3392857142857145e-07,
+      "loss": 1.7681,
+      "step": 60505
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.338811630847029e-07,
+      "loss": 2.2634,
+      "step": 60506
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3383375474083438e-07,
+      "loss": 2.5014,
+      "step": 60507
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3378634639696587e-07,
+      "loss": 1.8334,
+      "step": 60508
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3373893805309734e-07,
+      "loss": 3.6831,
+      "step": 60509
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.3369152970922883e-07,
+      "loss": 1.6726,
+      "step": 60510
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 1.336441213653603e-07,
+      "loss": 2.8725,
+      "step": 60511
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3359671302149179e-07,
+      "loss": 1.8695,
+      "step": 60512
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3354930467762328e-07,
+      "loss": 1.0247,
+      "step": 60513
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3350189633375474e-07,
+      "loss": 1.8885,
+      "step": 60514
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3345448798988624e-07,
+      "loss": 2.7454,
+      "step": 60515
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.334070796460177e-07,
+      "loss": 2.7856,
+      "step": 60516
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.333596713021492e-07,
+      "loss": 1.9762,
+      "step": 60517
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3331226295828066e-07,
+      "loss": 4.2054,
+      "step": 60518
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3326485461441213e-07,
+      "loss": 1.778,
+      "step": 60519
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3321744627054362e-07,
+      "loss": 1.8507,
+      "step": 60520
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3317003792667508e-07,
+      "loss": 1.2208,
+      "step": 60521
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3312262958280658e-07,
+      "loss": 1.673,
+      "step": 60522
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3307522123893804e-07,
+      "loss": 3.1319,
+      "step": 60523
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3302781289506953e-07,
+      "loss": 2.2201,
+      "step": 60524
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3298040455120103e-07,
+      "loss": 0.4846,
+      "step": 60525
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.329329962073325e-07,
+      "loss": 0.6318,
+      "step": 60526
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3288558786346398e-07,
+      "loss": 1.3802,
+      "step": 60527
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3283817951959545e-07,
+      "loss": 1.0269,
+      "step": 60528
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3279077117572694e-07,
+      "loss": 1.8036,
+      "step": 60529
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.327433628318584e-07,
+      "loss": 3.4689,
+      "step": 60530
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3269595448798987e-07,
+      "loss": 4.0485,
+      "step": 60531
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3264854614412137e-07,
+      "loss": 2.2839,
+      "step": 60532
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3260113780025283e-07,
+      "loss": 1.4338,
+      "step": 60533
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3255372945638432e-07,
+      "loss": 1.3589,
+      "step": 60534
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3250632111251582e-07,
+      "loss": 0.6953,
+      "step": 60535
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3245891276864728e-07,
+      "loss": 2.6763,
+      "step": 60536
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3241150442477877e-07,
+      "loss": 2.3887,
+      "step": 60537
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3236409608091024e-07,
+      "loss": 3.7103,
+      "step": 60538
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3231668773704173e-07,
+      "loss": 1.339,
+      "step": 60539
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.322692793931732e-07,
+      "loss": 3.679,
+      "step": 60540
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.322218710493047e-07,
+      "loss": 1.4053,
+      "step": 60541
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3217446270543615e-07,
+      "loss": 2.5333,
+      "step": 60542
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3212705436156765e-07,
+      "loss": 1.442,
+      "step": 60543
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.320796460176991e-07,
+      "loss": 0.7177,
+      "step": 60544
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3203223767383058e-07,
+      "loss": 1.2204,
+      "step": 60545
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.319848293299621e-07,
+      "loss": 1.8767,
+      "step": 60546
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3193742098609356e-07,
+      "loss": 2.389,
+      "step": 60547
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3189001264222503e-07,
+      "loss": 2.4816,
+      "step": 60548
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3184260429835652e-07,
+      "loss": 2.6395,
+      "step": 60549
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3179519595448799e-07,
+      "loss": 1.2828,
+      "step": 60550
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3174778761061948e-07,
+      "loss": 1.8433,
+      "step": 60551
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3170037926675094e-07,
+      "loss": 2.5149,
+      "step": 60552
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3165297092288244e-07,
+      "loss": 1.4043,
+      "step": 60553
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.316055625790139e-07,
+      "loss": 2.6265,
+      "step": 60554
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.315581542351454e-07,
+      "loss": 1.2187,
+      "step": 60555
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3151074589127686e-07,
+      "loss": 1.8941,
+      "step": 60556
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3146333754740833e-07,
+      "loss": 2.2223,
+      "step": 60557
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3141592920353984e-07,
+      "loss": 1.2925,
+      "step": 60558
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.313685208596713e-07,
+      "loss": 3.3727,
+      "step": 60559
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.313211125158028e-07,
+      "loss": 1.3477,
+      "step": 60560
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3127370417193427e-07,
+      "loss": 1.3236,
+      "step": 60561
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3122629582806573e-07,
+      "loss": 1.0375,
+      "step": 60562
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3117888748419723e-07,
+      "loss": 2.8031,
+      "step": 60563
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.311314791403287e-07,
+      "loss": 0.8951,
+      "step": 60564
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3108407079646018e-07,
+      "loss": 1.9018,
+      "step": 60565
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3103666245259165e-07,
+      "loss": 1.5018,
+      "step": 60566
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3098925410872314e-07,
+      "loss": 1.5985,
+      "step": 60567
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.309418457648546e-07,
+      "loss": 1.1548,
+      "step": 60568
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.308944374209861e-07,
+      "loss": 1.0897,
+      "step": 60569
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.308470290771176e-07,
+      "loss": 1.3221,
+      "step": 60570
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3079962073324906e-07,
+      "loss": 2.5707,
+      "step": 60571
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3075221238938055e-07,
+      "loss": 1.2343,
+      "step": 60572
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3070480404551202e-07,
+      "loss": 1.7489,
+      "step": 60573
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3065739570164348e-07,
+      "loss": 2.9389,
+      "step": 60574
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 1.3060998735777497e-07,
+      "loss": 1.5742,
+      "step": 60575
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3056257901390644e-07,
+      "loss": 1.5171,
+      "step": 60576
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3051517067003793e-07,
+      "loss": 1.8537,
+      "step": 60577
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.304677623261694e-07,
+      "loss": 1.4832,
+      "step": 60578
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.304203539823009e-07,
+      "loss": 1.6263,
+      "step": 60579
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3037294563843238e-07,
+      "loss": 1.5157,
+      "step": 60580
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3032553729456385e-07,
+      "loss": 2.1725,
+      "step": 60581
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3027812895069534e-07,
+      "loss": 2.3468,
+      "step": 60582
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.302307206068268e-07,
+      "loss": 1.93,
+      "step": 60583
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.301833122629583e-07,
+      "loss": 2.0056,
+      "step": 60584
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3013590391908976e-07,
+      "loss": 1.0058,
+      "step": 60585
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3008849557522123e-07,
+      "loss": 0.5124,
+      "step": 60586
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.3004108723135272e-07,
+      "loss": 1.8998,
+      "step": 60587
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2999367888748419e-07,
+      "loss": 1.8626,
+      "step": 60588
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2994627054361568e-07,
+      "loss": 0.8767,
+      "step": 60589
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2989886219974714e-07,
+      "loss": 2.6808,
+      "step": 60590
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2985145385587864e-07,
+      "loss": 1.4772,
+      "step": 60591
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2980404551201013e-07,
+      "loss": 2.2882,
+      "step": 60592
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.297566371681416e-07,
+      "loss": 3.6613,
+      "step": 60593
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2970922882427309e-07,
+      "loss": 2.2713,
+      "step": 60594
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2966182048040455e-07,
+      "loss": 3.4775,
+      "step": 60595
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2961441213653604e-07,
+      "loss": 3.5708,
+      "step": 60596
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.295670037926675e-07,
+      "loss": 1.684,
+      "step": 60597
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2951959544879898e-07,
+      "loss": 1.105,
+      "step": 60598
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2947218710493047e-07,
+      "loss": 1.0611,
+      "step": 60599
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2942477876106193e-07,
+      "loss": 3.0232,
+      "step": 60600
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2937737041719343e-07,
+      "loss": 2.8022,
+      "step": 60601
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2932996207332492e-07,
+      "loss": 1.4571,
+      "step": 60602
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2928255372945638e-07,
+      "loss": 1.506,
+      "step": 60603
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2923514538558788e-07,
+      "loss": 1.3805,
+      "step": 60604
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2918773704171934e-07,
+      "loss": 3.7326,
+      "step": 60605
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2914032869785083e-07,
+      "loss": 1.3984,
+      "step": 60606
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.290929203539823e-07,
+      "loss": 1.7532,
+      "step": 60607
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.290455120101138e-07,
+      "loss": 2.8031,
+      "step": 60608
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2899810366624526e-07,
+      "loss": 1.3544,
+      "step": 60609
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2895069532237675e-07,
+      "loss": 1.5164,
+      "step": 60610
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2890328697850821e-07,
+      "loss": 1.7436,
+      "step": 60611
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2885587863463968e-07,
+      "loss": 1.8825,
+      "step": 60612
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2880847029077117e-07,
+      "loss": 0.6582,
+      "step": 60613
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2876106194690266e-07,
+      "loss": 2.6122,
+      "step": 60614
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2871365360303416e-07,
+      "loss": 0.9657,
+      "step": 60615
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2866624525916562e-07,
+      "loss": 1.6189,
+      "step": 60616
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.286188369152971e-07,
+      "loss": 2.4193,
+      "step": 60617
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2857142857142858e-07,
+      "loss": 0.8442,
+      "step": 60618
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2852402022756005e-07,
+      "loss": 1.7384,
+      "step": 60619
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2847661188369154e-07,
+      "loss": 2.1661,
+      "step": 60620
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.28429203539823e-07,
+      "loss": 2.3242,
+      "step": 60621
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.283817951959545e-07,
+      "loss": 1.3552,
+      "step": 60622
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2833438685208596e-07,
+      "loss": 2.2795,
+      "step": 60623
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2828697850821743e-07,
+      "loss": 1.307,
+      "step": 60624
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2823957016434895e-07,
+      "loss": 1.8067,
+      "step": 60625
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.281921618204804e-07,
+      "loss": 1.22,
+      "step": 60626
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.281447534766119e-07,
+      "loss": 0.7101,
+      "step": 60627
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2809734513274337e-07,
+      "loss": 1.1014,
+      "step": 60628
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2804993678887484e-07,
+      "loss": 0.976,
+      "step": 60629
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2800252844500633e-07,
+      "loss": 0.6956,
+      "step": 60630
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.279551201011378e-07,
+      "loss": 2.7355,
+      "step": 60631
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2790771175726929e-07,
+      "loss": 2.2734,
+      "step": 60632
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2786030341340075e-07,
+      "loss": 1.6558,
+      "step": 60633
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2781289506953224e-07,
+      "loss": 1.8042,
+      "step": 60634
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.277654867256637e-07,
+      "loss": 2.4331,
+      "step": 60635
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.277180783817952e-07,
+      "loss": 1.4413,
+      "step": 60636
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.276706700379267e-07,
+      "loss": 1.1941,
+      "step": 60637
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 1.2762326169405816e-07,
+      "loss": 1.9909,
+      "step": 60638
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2757585335018965e-07,
+      "loss": 2.2118,
+      "step": 60639
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2752844500632112e-07,
+      "loss": 1.4846,
+      "step": 60640
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2748103666245258e-07,
+      "loss": 1.684,
+      "step": 60641
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2743362831858408e-07,
+      "loss": 2.1924,
+      "step": 60642
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2738621997471554e-07,
+      "loss": 1.8381,
+      "step": 60643
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2733881163084703e-07,
+      "loss": 2.5096,
+      "step": 60644
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.272914032869785e-07,
+      "loss": 2.0202,
+      "step": 60645
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2724399494311e-07,
+      "loss": 3.0612,
+      "step": 60646
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2719658659924148e-07,
+      "loss": 0.5175,
+      "step": 60647
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2714917825537295e-07,
+      "loss": 1.4843,
+      "step": 60648
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2710176991150444e-07,
+      "loss": 1.9702,
+      "step": 60649
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.270543615676359e-07,
+      "loss": 0.9928,
+      "step": 60650
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.270069532237674e-07,
+      "loss": 1.3479,
+      "step": 60651
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2695954487989886e-07,
+      "loss": 1.691,
+      "step": 60652
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2691213653603033e-07,
+      "loss": 2.833,
+      "step": 60653
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2686472819216182e-07,
+      "loss": 2.5262,
+      "step": 60654
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.268173198482933e-07,
+      "loss": 0.667,
+      "step": 60655
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2676991150442478e-07,
+      "loss": 1.9453,
+      "step": 60656
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2672250316055625e-07,
+      "loss": 1.8062,
+      "step": 60657
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2667509481668774e-07,
+      "loss": 1.7889,
+      "step": 60658
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2662768647281923e-07,
+      "loss": 2.7878,
+      "step": 60659
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.265802781289507e-07,
+      "loss": 0.6228,
+      "step": 60660
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.265328697850822e-07,
+      "loss": 0.2608,
+      "step": 60661
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2648546144121365e-07,
+      "loss": 1.7787,
+      "step": 60662
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2643805309734515e-07,
+      "loss": 1.2646,
+      "step": 60663
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.263906447534766e-07,
+      "loss": 1.8172,
+      "step": 60664
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.263432364096081e-07,
+      "loss": 2.1012,
+      "step": 60665
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2629582806573957e-07,
+      "loss": 1.1551,
+      "step": 60666
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2624841972187104e-07,
+      "loss": 2.3573,
+      "step": 60667
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2620101137800253e-07,
+      "loss": 1.4837,
+      "step": 60668
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.26153603034134e-07,
+      "loss": 1.0155,
+      "step": 60669
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2610619469026549e-07,
+      "loss": 2.0633,
+      "step": 60670
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2605878634639698e-07,
+      "loss": 1.3394,
+      "step": 60671
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2601137800252844e-07,
+      "loss": 1.7431,
+      "step": 60672
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2596396965865994e-07,
+      "loss": 2.2293,
+      "step": 60673
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.259165613147914e-07,
+      "loss": 2.015,
+      "step": 60674
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.258691529709229e-07,
+      "loss": 1.7359,
+      "step": 60675
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2582174462705436e-07,
+      "loss": 1.6318,
+      "step": 60676
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2577433628318585e-07,
+      "loss": 2.336,
+      "step": 60677
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2572692793931732e-07,
+      "loss": 1.4581,
+      "step": 60678
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2567951959544878e-07,
+      "loss": 1.1282,
+      "step": 60679
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2563211125158028e-07,
+      "loss": 1.651,
+      "step": 60680
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2558470290771177e-07,
+      "loss": 2.3451,
+      "step": 60681
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2553729456384326e-07,
+      "loss": 1.7414,
+      "step": 60682
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2548988621997473e-07,
+      "loss": 2.502,
+      "step": 60683
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.254424778761062e-07,
+      "loss": 1.0973,
+      "step": 60684
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2539506953223768e-07,
+      "loss": 2.7691,
+      "step": 60685
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2534766118836915e-07,
+      "loss": 2.905,
+      "step": 60686
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2530025284450064e-07,
+      "loss": 1.3319,
+      "step": 60687
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.252528445006321e-07,
+      "loss": 2.8012,
+      "step": 60688
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.252054361567636e-07,
+      "loss": 1.3361,
+      "step": 60689
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2515802781289506e-07,
+      "loss": 3.2101,
+      "step": 60690
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2511061946902653e-07,
+      "loss": 2.2459,
+      "step": 60691
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2506321112515805e-07,
+      "loss": 1.4589,
+      "step": 60692
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2501580278128951e-07,
+      "loss": 1.6713,
+      "step": 60693
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.24968394437421e-07,
+      "loss": 2.3546,
+      "step": 60694
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2492098609355247e-07,
+      "loss": 2.3307,
+      "step": 60695
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2487357774968394e-07,
+      "loss": 1.4707,
+      "step": 60696
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2482616940581543e-07,
+      "loss": 1.0006,
+      "step": 60697
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.247787610619469e-07,
+      "loss": 2.0778,
+      "step": 60698
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.247313527180784e-07,
+      "loss": 0.8806,
+      "step": 60699
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2468394437420985e-07,
+      "loss": 1.4151,
+      "step": 60700
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 1.2463653603034135e-07,
+      "loss": 2.5368,
+      "step": 60701
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.245891276864728e-07,
+      "loss": 1.7222,
+      "step": 60702
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.245417193426043e-07,
+      "loss": 0.9767,
+      "step": 60703
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.244943109987358e-07,
+      "loss": 1.5398,
+      "step": 60704
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2444690265486726e-07,
+      "loss": 2.2037,
+      "step": 60705
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2439949431099875e-07,
+      "loss": 2.3956,
+      "step": 60706
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2435208596713022e-07,
+      "loss": 1.6373,
+      "step": 60707
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2430467762326169e-07,
+      "loss": 1.2517,
+      "step": 60708
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2425726927939318e-07,
+      "loss": 1.0752,
+      "step": 60709
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2420986093552464e-07,
+      "loss": 1.8459,
+      "step": 60710
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2416245259165614e-07,
+      "loss": 1.689,
+      "step": 60711
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.241150442477876e-07,
+      "loss": 1.9232,
+      "step": 60712
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.240676359039191e-07,
+      "loss": 0.5932,
+      "step": 60713
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2402022756005059e-07,
+      "loss": 1.0701,
+      "step": 60714
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2397281921618205e-07,
+      "loss": 1.6349,
+      "step": 60715
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2392541087231354e-07,
+      "loss": 0.913,
+      "step": 60716
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.23878002528445e-07,
+      "loss": 2.7995,
+      "step": 60717
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.238305941845765e-07,
+      "loss": 1.9161,
+      "step": 60718
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2378318584070797e-07,
+      "loss": 1.7156,
+      "step": 60719
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2373577749683943e-07,
+      "loss": 1.8365,
+      "step": 60720
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2368836915297092e-07,
+      "loss": 0.4292,
+      "step": 60721
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.236409608091024e-07,
+      "loss": 1.9136,
+      "step": 60722
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2359355246523388e-07,
+      "loss": 1.5458,
+      "step": 60723
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2354614412136535e-07,
+      "loss": 1.268,
+      "step": 60724
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2349873577749684e-07,
+      "loss": 2.2512,
+      "step": 60725
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2345132743362833e-07,
+      "loss": 2.3847,
+      "step": 60726
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.234039190897598e-07,
+      "loss": 3.889,
+      "step": 60727
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.233565107458913e-07,
+      "loss": 1.837,
+      "step": 60728
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2330910240202276e-07,
+      "loss": 1.6391,
+      "step": 60729
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2326169405815425e-07,
+      "loss": 1.8131,
+      "step": 60730
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2321428571428571e-07,
+      "loss": 2.6224,
+      "step": 60731
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.231668773704172e-07,
+      "loss": 0.8642,
+      "step": 60732
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2311946902654867e-07,
+      "loss": 1.9008,
+      "step": 60733
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2307206068268014e-07,
+      "loss": 1.6229,
+      "step": 60734
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2302465233881163e-07,
+      "loss": 0.7176,
+      "step": 60735
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.229772439949431e-07,
+      "loss": 1.3205,
+      "step": 60736
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2292983565107461e-07,
+      "loss": 3.1585,
+      "step": 60737
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2288242730720608e-07,
+      "loss": 1.6448,
+      "step": 60738
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2283501896333755e-07,
+      "loss": 3.3139,
+      "step": 60739
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2278761061946904e-07,
+      "loss": 2.1069,
+      "step": 60740
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.227402022756005e-07,
+      "loss": 1.4194,
+      "step": 60741
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.22692793931732e-07,
+      "loss": 2.9703,
+      "step": 60742
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2264538558786346e-07,
+      "loss": 2.4964,
+      "step": 60743
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2259797724399495e-07,
+      "loss": 2.2944,
+      "step": 60744
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2255056890012642e-07,
+      "loss": 1.095,
+      "step": 60745
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2250316055625789e-07,
+      "loss": 1.4397,
+      "step": 60746
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2245575221238938e-07,
+      "loss": 1.9753,
+      "step": 60747
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2240834386852087e-07,
+      "loss": 3.1585,
+      "step": 60748
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2236093552465236e-07,
+      "loss": 2.5319,
+      "step": 60749
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2231352718078383e-07,
+      "loss": 1.7601,
+      "step": 60750
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.222661188369153e-07,
+      "loss": 1.2935,
+      "step": 60751
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2221871049304679e-07,
+      "loss": 0.572,
+      "step": 60752
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2217130214917825e-07,
+      "loss": 2.9639,
+      "step": 60753
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2212389380530974e-07,
+      "loss": 1.1414,
+      "step": 60754
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.220764854614412e-07,
+      "loss": 3.2091,
+      "step": 60755
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.220290771175727e-07,
+      "loss": 4.3423,
+      "step": 60756
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2198166877370417e-07,
+      "loss": 1.8799,
+      "step": 60757
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2193426042983563e-07,
+      "loss": 1.3624,
+      "step": 60758
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2188685208596715e-07,
+      "loss": 1.6523,
+      "step": 60759
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2183944374209862e-07,
+      "loss": 1.8584,
+      "step": 60760
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.217920353982301e-07,
+      "loss": 2.4847,
+      "step": 60761
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2174462705436157e-07,
+      "loss": 1.7059,
+      "step": 60762
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2169721871049304e-07,
+      "loss": 2.9714,
+      "step": 60763
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.2164981036662453e-07,
+      "loss": 1.2585,
+      "step": 60764
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 1.21602402022756e-07,
+      "loss": 1.7354,
+      "step": 60765
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.215549936788875e-07,
+      "loss": 1.8424,
+      "step": 60766
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2150758533501896e-07,
+      "loss": 1.809,
+      "step": 60767
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2146017699115045e-07,
+      "loss": 1.8222,
+      "step": 60768
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2141276864728191e-07,
+      "loss": 1.7599,
+      "step": 60769
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.213653603034134e-07,
+      "loss": 3.3926,
+      "step": 60770
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.213179519595449e-07,
+      "loss": 1.0855,
+      "step": 60771
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2127054361567636e-07,
+      "loss": 1.1458,
+      "step": 60772
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2122313527180786e-07,
+      "loss": 0.9656,
+      "step": 60773
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2117572692793932e-07,
+      "loss": 2.6462,
+      "step": 60774
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.211283185840708e-07,
+      "loss": 1.2301,
+      "step": 60775
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2108091024020228e-07,
+      "loss": 2.0433,
+      "step": 60776
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2103350189633375e-07,
+      "loss": 2.3608,
+      "step": 60777
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2098609355246524e-07,
+      "loss": 1.62,
+      "step": 60778
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.209386852085967e-07,
+      "loss": 1.4472,
+      "step": 60779
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.208912768647282e-07,
+      "loss": 1.5532,
+      "step": 60780
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2084386852085966e-07,
+      "loss": 2.9548,
+      "step": 60781
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2079646017699115e-07,
+      "loss": 2.3465,
+      "step": 60782
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2074905183312265e-07,
+      "loss": 0.7571,
+      "step": 60783
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.207016434892541e-07,
+      "loss": 2.6367,
+      "step": 60784
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.206542351453856e-07,
+      "loss": 2.7031,
+      "step": 60785
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2060682680151707e-07,
+      "loss": 1.0876,
+      "step": 60786
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2055941845764856e-07,
+      "loss": 2.1575,
+      "step": 60787
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2051201011378003e-07,
+      "loss": 2.2253,
+      "step": 60788
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.204646017699115e-07,
+      "loss": 1.8274,
+      "step": 60789
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2041719342604299e-07,
+      "loss": 1.0719,
+      "step": 60790
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2036978508217445e-07,
+      "loss": 2.8946,
+      "step": 60791
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2032237673830594e-07,
+      "loss": 1.9178,
+      "step": 60792
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2027496839443744e-07,
+      "loss": 2.3968,
+      "step": 60793
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.202275600505689e-07,
+      "loss": 2.9767,
+      "step": 60794
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.201801517067004e-07,
+      "loss": 1.3458,
+      "step": 60795
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2013274336283186e-07,
+      "loss": 1.3667,
+      "step": 60796
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2008533501896335e-07,
+      "loss": 1.2232,
+      "step": 60797
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.2003792667509482e-07,
+      "loss": 2.8274,
+      "step": 60798
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.199905183312263e-07,
+      "loss": 1.8767,
+      "step": 60799
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1994310998735777e-07,
+      "loss": 2.4291,
+      "step": 60800
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1989570164348924e-07,
+      "loss": 2.8312,
+      "step": 60801
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1984829329962073e-07,
+      "loss": 1.957,
+      "step": 60802
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.198008849557522e-07,
+      "loss": 1.6189,
+      "step": 60803
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1975347661188372e-07,
+      "loss": 1.7247,
+      "step": 60804
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1970606826801518e-07,
+      "loss": 1.9283,
+      "step": 60805
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1965865992414665e-07,
+      "loss": 0.1956,
+      "step": 60806
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1961125158027814e-07,
+      "loss": 1.0606,
+      "step": 60807
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.195638432364096e-07,
+      "loss": 1.3912,
+      "step": 60808
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.195164348925411e-07,
+      "loss": 0.7436,
+      "step": 60809
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1946902654867256e-07,
+      "loss": 3.0342,
+      "step": 60810
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1942161820480406e-07,
+      "loss": 1.7836,
+      "step": 60811
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1937420986093552e-07,
+      "loss": 2.5384,
+      "step": 60812
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.19326801517067e-07,
+      "loss": 0.6035,
+      "step": 60813
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1927939317319848e-07,
+      "loss": 1.9438,
+      "step": 60814
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1923198482932997e-07,
+      "loss": 2.2625,
+      "step": 60815
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1918457648546145e-07,
+      "loss": 0.3021,
+      "step": 60816
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1913716814159293e-07,
+      "loss": 1.0822,
+      "step": 60817
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1908975979772441e-07,
+      "loss": 1.0279,
+      "step": 60818
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1904235145385589e-07,
+      "loss": 1.3582,
+      "step": 60819
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1899494310998737e-07,
+      "loss": 2.5238,
+      "step": 60820
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1894753476611883e-07,
+      "loss": 1.9901,
+      "step": 60821
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1890012642225031e-07,
+      "loss": 0.9728,
+      "step": 60822
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1885271807838179e-07,
+      "loss": 1.8499,
+      "step": 60823
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1880530973451327e-07,
+      "loss": 1.8883,
+      "step": 60824
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1875790139064475e-07,
+      "loss": 0.4034,
+      "step": 60825
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1871049304677624e-07,
+      "loss": 1.5538,
+      "step": 60826
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.1866308470290772e-07,
+      "loss": 1.8651,
+      "step": 60827
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1.186156763590392e-07,
+      "loss": 1.7816,
+      "step": 60828
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1856826801517068e-07,
+      "loss": 2.0161,
+      "step": 60829
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1852085967130216e-07,
+      "loss": 2.932,
+      "step": 60830
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1847345132743364e-07,
+      "loss": 2.1573,
+      "step": 60831
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1842604298356511e-07,
+      "loss": 0.8333,
+      "step": 60832
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1837863463969659e-07,
+      "loss": 2.5972,
+      "step": 60833
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1833122629582806e-07,
+      "loss": 1.7761,
+      "step": 60834
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1828381795195954e-07,
+      "loss": 1.2443,
+      "step": 60835
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1823640960809102e-07,
+      "loss": 1.5465,
+      "step": 60836
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.181890012642225e-07,
+      "loss": 2.0765,
+      "step": 60837
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1814159292035399e-07,
+      "loss": 2.0488,
+      "step": 60838
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1809418457648547e-07,
+      "loss": 2.0073,
+      "step": 60839
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1804677623261695e-07,
+      "loss": 2.0855,
+      "step": 60840
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1799936788874842e-07,
+      "loss": 3.1497,
+      "step": 60841
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.179519595448799e-07,
+      "loss": 1.1713,
+      "step": 60842
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1790455120101138e-07,
+      "loss": 0.6788,
+      "step": 60843
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1785714285714286e-07,
+      "loss": 3.442,
+      "step": 60844
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1780973451327434e-07,
+      "loss": 0.9235,
+      "step": 60845
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.177623261694058e-07,
+      "loss": 0.732,
+      "step": 60846
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1771491782553728e-07,
+      "loss": 1.9897,
+      "step": 60847
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1766750948166876e-07,
+      "loss": 0.7184,
+      "step": 60848
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1762010113780027e-07,
+      "loss": 0.7125,
+      "step": 60849
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1757269279393175e-07,
+      "loss": 1.7026,
+      "step": 60850
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1752528445006321e-07,
+      "loss": 1.8579,
+      "step": 60851
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1747787610619469e-07,
+      "loss": 0.6989,
+      "step": 60852
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1743046776232617e-07,
+      "loss": 1.0209,
+      "step": 60853
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1738305941845765e-07,
+      "loss": 2.4006,
+      "step": 60854
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1733565107458913e-07,
+      "loss": 1.2401,
+      "step": 60855
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1728824273072061e-07,
+      "loss": 1.8932,
+      "step": 60856
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1724083438685209e-07,
+      "loss": 2.0935,
+      "step": 60857
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1719342604298357e-07,
+      "loss": 2.9272,
+      "step": 60858
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1714601769911503e-07,
+      "loss": 1.9878,
+      "step": 60859
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1709860935524654e-07,
+      "loss": 3.4344,
+      "step": 60860
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1705120101137802e-07,
+      "loss": 2.1315,
+      "step": 60861
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.170037926675095e-07,
+      "loss": 2.4396,
+      "step": 60862
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1695638432364096e-07,
+      "loss": 0.6254,
+      "step": 60863
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1690897597977244e-07,
+      "loss": 1.5238,
+      "step": 60864
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1686156763590392e-07,
+      "loss": 1.9123,
+      "step": 60865
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.168141592920354e-07,
+      "loss": 0.8543,
+      "step": 60866
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1676675094816688e-07,
+      "loss": 2.0377,
+      "step": 60867
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1671934260429836e-07,
+      "loss": 1.886,
+      "step": 60868
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1667193426042983e-07,
+      "loss": 1.327,
+      "step": 60869
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1662452591656131e-07,
+      "loss": 2.8922,
+      "step": 60870
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.165771175726928e-07,
+      "loss": 2.1323,
+      "step": 60871
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1652970922882428e-07,
+      "loss": 1.3829,
+      "step": 60872
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1648230088495576e-07,
+      "loss": 1.839,
+      "step": 60873
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1643489254108724e-07,
+      "loss": 2.0024,
+      "step": 60874
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1638748419721872e-07,
+      "loss": 2.5466,
+      "step": 60875
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1634007585335019e-07,
+      "loss": 2.0071,
+      "step": 60876
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1629266750948167e-07,
+      "loss": 3.0647,
+      "step": 60877
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1624525916561315e-07,
+      "loss": 1.012,
+      "step": 60878
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1619785082174462e-07,
+      "loss": 1.169,
+      "step": 60879
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.161504424778761e-07,
+      "loss": 2.0069,
+      "step": 60880
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1610303413400758e-07,
+      "loss": 0.6608,
+      "step": 60881
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1605562579013907e-07,
+      "loss": 2.1002,
+      "step": 60882
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1600821744627055e-07,
+      "loss": 1.2099,
+      "step": 60883
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1596080910240203e-07,
+      "loss": 0.9784,
+      "step": 60884
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1591340075853351e-07,
+      "loss": 1.6601,
+      "step": 60885
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1586599241466499e-07,
+      "loss": 1.6479,
+      "step": 60886
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1581858407079647e-07,
+      "loss": 1.9263,
+      "step": 60887
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1577117572692795e-07,
+      "loss": 1.1985,
+      "step": 60888
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1572376738305941e-07,
+      "loss": 1.5915,
+      "step": 60889
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1567635903919089e-07,
+      "loss": 3.1173,
+      "step": 60890
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 1.1562895069532237e-07,
+      "loss": 2.0171,
+      "step": 60891
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1558154235145385e-07,
+      "loss": 2.0136,
+      "step": 60892
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1553413400758533e-07,
+      "loss": 1.0906,
+      "step": 60893
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1548672566371682e-07,
+      "loss": 2.8968,
+      "step": 60894
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.154393173198483e-07,
+      "loss": 0.6322,
+      "step": 60895
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1539190897597978e-07,
+      "loss": 1.9953,
+      "step": 60896
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1534450063211126e-07,
+      "loss": 2.5153,
+      "step": 60897
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1529709228824274e-07,
+      "loss": 0.9774,
+      "step": 60898
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1524968394437422e-07,
+      "loss": 0.7443,
+      "step": 60899
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.152022756005057e-07,
+      "loss": 2.0772,
+      "step": 60900
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1515486725663716e-07,
+      "loss": 1.2676,
+      "step": 60901
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1510745891276864e-07,
+      "loss": 1.6087,
+      "step": 60902
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1506005056890012e-07,
+      "loss": 1.8621,
+      "step": 60903
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.150126422250316e-07,
+      "loss": 2.4836,
+      "step": 60904
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.149652338811631e-07,
+      "loss": 1.1954,
+      "step": 60905
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1491782553729457e-07,
+      "loss": 0.5781,
+      "step": 60906
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1487041719342605e-07,
+      "loss": 1.4444,
+      "step": 60907
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1482300884955753e-07,
+      "loss": 2.1889,
+      "step": 60908
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.14775600505689e-07,
+      "loss": 3.1555,
+      "step": 60909
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1472819216182048e-07,
+      "loss": 4.564,
+      "step": 60910
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1468078381795196e-07,
+      "loss": 0.7339,
+      "step": 60911
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1463337547408344e-07,
+      "loss": 1.8808,
+      "step": 60912
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1458596713021492e-07,
+      "loss": 0.7486,
+      "step": 60913
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1453855878634639e-07,
+      "loss": 1.1463,
+      "step": 60914
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1449115044247787e-07,
+      "loss": 2.3009,
+      "step": 60915
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1444374209860937e-07,
+      "loss": 0.8698,
+      "step": 60916
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1439633375474085e-07,
+      "loss": 2.0267,
+      "step": 60917
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1434892541087232e-07,
+      "loss": 0.4787,
+      "step": 60918
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.143015170670038e-07,
+      "loss": 0.5958,
+      "step": 60919
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1425410872313527e-07,
+      "loss": 3.3284,
+      "step": 60920
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1420670037926675e-07,
+      "loss": 1.849,
+      "step": 60921
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1415929203539823e-07,
+      "loss": 0.4647,
+      "step": 60922
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1411188369152971e-07,
+      "loss": 2.6234,
+      "step": 60923
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1406447534766119e-07,
+      "loss": 0.8262,
+      "step": 60924
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1401706700379267e-07,
+      "loss": 0.4422,
+      "step": 60925
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1396965865992413e-07,
+      "loss": 3.2688,
+      "step": 60926
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1392225031605564e-07,
+      "loss": 1.7361,
+      "step": 60927
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1387484197218712e-07,
+      "loss": 1.9157,
+      "step": 60928
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.138274336283186e-07,
+      "loss": 1.5277,
+      "step": 60929
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1378002528445008e-07,
+      "loss": 1.3118,
+      "step": 60930
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1373261694058154e-07,
+      "loss": 1.3829,
+      "step": 60931
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1368520859671302e-07,
+      "loss": 1.829,
+      "step": 60932
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.136378002528445e-07,
+      "loss": 3.9679,
+      "step": 60933
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1359039190897598e-07,
+      "loss": 2.0455,
+      "step": 60934
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1354298356510746e-07,
+      "loss": 1.3722,
+      "step": 60935
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1349557522123894e-07,
+      "loss": 1.3726,
+      "step": 60936
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1344816687737042e-07,
+      "loss": 1.6133,
+      "step": 60937
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1340075853350191e-07,
+      "loss": 0.9993,
+      "step": 60938
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1335335018963339e-07,
+      "loss": 2.2294,
+      "step": 60939
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1330594184576487e-07,
+      "loss": 3.0321,
+      "step": 60940
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1325853350189635e-07,
+      "loss": 0.6778,
+      "step": 60941
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1321112515802782e-07,
+      "loss": 2.0827,
+      "step": 60942
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1316371681415929e-07,
+      "loss": 2.3101,
+      "step": 60943
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1311630847029077e-07,
+      "loss": 2.0988,
+      "step": 60944
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1306890012642225e-07,
+      "loss": 2.7194,
+      "step": 60945
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1302149178255373e-07,
+      "loss": 3.0153,
+      "step": 60946
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.129740834386852e-07,
+      "loss": 2.0214,
+      "step": 60947
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1292667509481668e-07,
+      "loss": 2.5812,
+      "step": 60948
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1287926675094816e-07,
+      "loss": 1.6482,
+      "step": 60949
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1283185840707966e-07,
+      "loss": 1.6315,
+      "step": 60950
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1278445006321113e-07,
+      "loss": 2.4247,
+      "step": 60951
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1273704171934261e-07,
+      "loss": 1.0193,
+      "step": 60952
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1268963337547409e-07,
+      "loss": 1.5621,
+      "step": 60953
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1264222503160557e-07,
+      "loss": 0.7381,
+      "step": 60954
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 1.1259481668773705e-07,
+      "loss": 2.0853,
+      "step": 60955
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1254740834386852e-07,
+      "loss": 1.8078,
+      "step": 60956
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.125e-07,
+      "loss": 2.2556,
+      "step": 60957
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1245259165613147e-07,
+      "loss": 2.4235,
+      "step": 60958
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1240518331226295e-07,
+      "loss": 1.6605,
+      "step": 60959
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1235777496839443e-07,
+      "loss": 3.258,
+      "step": 60960
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1231036662452592e-07,
+      "loss": 1.6138,
+      "step": 60961
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.122629582806574e-07,
+      "loss": 1.8252,
+      "step": 60962
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1221554993678888e-07,
+      "loss": 1.6196,
+      "step": 60963
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1216814159292036e-07,
+      "loss": 3.1235,
+      "step": 60964
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1212073324905184e-07,
+      "loss": 3.0902,
+      "step": 60965
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1207332490518332e-07,
+      "loss": 1.056,
+      "step": 60966
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.120259165613148e-07,
+      "loss": 1.6285,
+      "step": 60967
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1197850821744626e-07,
+      "loss": 1.448,
+      "step": 60968
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1193109987357774e-07,
+      "loss": 1.6767,
+      "step": 60969
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1188369152970922e-07,
+      "loss": 0.8105,
+      "step": 60970
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.118362831858407e-07,
+      "loss": 1.7817,
+      "step": 60971
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.117888748419722e-07,
+      "loss": 0.4625,
+      "step": 60972
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1174146649810367e-07,
+      "loss": 1.1336,
+      "step": 60973
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1169405815423515e-07,
+      "loss": 0.9156,
+      "step": 60974
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1164664981036663e-07,
+      "loss": 2.3461,
+      "step": 60975
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1159924146649811e-07,
+      "loss": 1.8796,
+      "step": 60976
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1155183312262959e-07,
+      "loss": 1.2791,
+      "step": 60977
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1150442477876107e-07,
+      "loss": 3.0916,
+      "step": 60978
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1145701643489254e-07,
+      "loss": 3.001,
+      "step": 60979
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1140960809102402e-07,
+      "loss": 2.848,
+      "step": 60980
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1136219974715549e-07,
+      "loss": 2.8814,
+      "step": 60981
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1131479140328697e-07,
+      "loss": 1.3364,
+      "step": 60982
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1126738305941847e-07,
+      "loss": 1.2908,
+      "step": 60983
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1121997471554995e-07,
+      "loss": 2.527,
+      "step": 60984
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1117256637168142e-07,
+      "loss": 1.1324,
+      "step": 60985
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.111251580278129e-07,
+      "loss": 1.0683,
+      "step": 60986
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1107774968394438e-07,
+      "loss": 1.8738,
+      "step": 60987
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1103034134007586e-07,
+      "loss": 2.0005,
+      "step": 60988
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1098293299620733e-07,
+      "loss": 2.2935,
+      "step": 60989
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1093552465233881e-07,
+      "loss": 1.5967,
+      "step": 60990
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1088811630847029e-07,
+      "loss": 1.0925,
+      "step": 60991
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1084070796460177e-07,
+      "loss": 1.8998,
+      "step": 60992
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1079329962073324e-07,
+      "loss": 2.3895,
+      "step": 60993
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1074589127686474e-07,
+      "loss": 1.9418,
+      "step": 60994
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1069848293299622e-07,
+      "loss": 1.3531,
+      "step": 60995
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.106510745891277e-07,
+      "loss": 1.2942,
+      "step": 60996
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1060366624525918e-07,
+      "loss": 2.1777,
+      "step": 60997
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1055625790139064e-07,
+      "loss": 2.5864,
+      "step": 60998
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1050884955752212e-07,
+      "loss": 1.2621,
+      "step": 60999
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.104614412136536e-07,
+      "loss": 1.4227,
+      "step": 61000
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1041403286978508e-07,
+      "loss": 2.9235,
+      "step": 61001
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1036662452591656e-07,
+      "loss": 0.959,
+      "step": 61002
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1031921618204804e-07,
+      "loss": 1.9549,
+      "step": 61003
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1027180783817952e-07,
+      "loss": 1.1249,
+      "step": 61004
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.10224399494311e-07,
+      "loss": 0.9243,
+      "step": 61005
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1017699115044249e-07,
+      "loss": 1.7067,
+      "step": 61006
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1012958280657397e-07,
+      "loss": 1.9061,
+      "step": 61007
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1008217446270545e-07,
+      "loss": 1.3054,
+      "step": 61008
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.1003476611883693e-07,
+      "loss": 2.9641,
+      "step": 61009
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.099873577749684e-07,
+      "loss": 2.7425,
+      "step": 61010
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0993994943109987e-07,
+      "loss": 1.1739,
+      "step": 61011
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0989254108723135e-07,
+      "loss": 1.2194,
+      "step": 61012
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0984513274336283e-07,
+      "loss": 1.7977,
+      "step": 61013
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0979772439949431e-07,
+      "loss": 1.4969,
+      "step": 61014
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0975031605562579e-07,
+      "loss": 3.5627,
+      "step": 61015
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0970290771175727e-07,
+      "loss": 2.1304,
+      "step": 61016
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0965549936788876e-07,
+      "loss": 2.9007,
+      "step": 61017
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 1.0960809102402024e-07,
+      "loss": 1.3376,
+      "step": 61018
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0956068268015172e-07,
+      "loss": 0.7808,
+      "step": 61019
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.095132743362832e-07,
+      "loss": 1.7996,
+      "step": 61020
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0946586599241467e-07,
+      "loss": 2.9579,
+      "step": 61021
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0941845764854615e-07,
+      "loss": 3.0613,
+      "step": 61022
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0937104930467762e-07,
+      "loss": 1.3001,
+      "step": 61023
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.093236409608091e-07,
+      "loss": 1.6357,
+      "step": 61024
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0927623261694058e-07,
+      "loss": 2.1558,
+      "step": 61025
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0922882427307206e-07,
+      "loss": 2.0235,
+      "step": 61026
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0918141592920353e-07,
+      "loss": 2.4097,
+      "step": 61027
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0913400758533503e-07,
+      "loss": 0.9747,
+      "step": 61028
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.090865992414665e-07,
+      "loss": 2.0457,
+      "step": 61029
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0903919089759798e-07,
+      "loss": 2.8156,
+      "step": 61030
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0899178255372946e-07,
+      "loss": 0.6399,
+      "step": 61031
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0894437420986094e-07,
+      "loss": 2.2606,
+      "step": 61032
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0889696586599242e-07,
+      "loss": 1.5111,
+      "step": 61033
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.088495575221239e-07,
+      "loss": 1.1503,
+      "step": 61034
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0880214917825538e-07,
+      "loss": 1.7132,
+      "step": 61035
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0875474083438684e-07,
+      "loss": 3.3485,
+      "step": 61036
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0870733249051832e-07,
+      "loss": 4.5397,
+      "step": 61037
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.086599241466498e-07,
+      "loss": 2.7436,
+      "step": 61038
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0861251580278131e-07,
+      "loss": 1.5534,
+      "step": 61039
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0856510745891277e-07,
+      "loss": 0.9298,
+      "step": 61040
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0851769911504425e-07,
+      "loss": 2.7496,
+      "step": 61041
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0847029077117573e-07,
+      "loss": 1.8107,
+      "step": 61042
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0842288242730721e-07,
+      "loss": 1.7183,
+      "step": 61043
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0837547408343869e-07,
+      "loss": 1.4773,
+      "step": 61044
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0832806573957017e-07,
+      "loss": 2.5344,
+      "step": 61045
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0828065739570165e-07,
+      "loss": 0.9159,
+      "step": 61046
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0823324905183313e-07,
+      "loss": 3.3567,
+      "step": 61047
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0818584070796459e-07,
+      "loss": 1.8482,
+      "step": 61048
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0813843236409607e-07,
+      "loss": 2.1573,
+      "step": 61049
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0809102402022758e-07,
+      "loss": 2.5325,
+      "step": 61050
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0804361567635906e-07,
+      "loss": 1.8384,
+      "step": 61051
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0799620733249053e-07,
+      "loss": 1.6793,
+      "step": 61052
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.07948798988622e-07,
+      "loss": 2.7457,
+      "step": 61053
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0790139064475348e-07,
+      "loss": 2.3505,
+      "step": 61054
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0785398230088496e-07,
+      "loss": 2.65,
+      "step": 61055
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0780657395701644e-07,
+      "loss": 2.292,
+      "step": 61056
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0775916561314792e-07,
+      "loss": 1.8184,
+      "step": 61057
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.077117572692794e-07,
+      "loss": 2.9847,
+      "step": 61058
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0766434892541087e-07,
+      "loss": 2.4933,
+      "step": 61059
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0761694058154235e-07,
+      "loss": 1.0622,
+      "step": 61060
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0756953223767382e-07,
+      "loss": 2.5532,
+      "step": 61061
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0752212389380532e-07,
+      "loss": 1.7828,
+      "step": 61062
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.074747155499368e-07,
+      "loss": 2.6249,
+      "step": 61063
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0742730720606828e-07,
+      "loss": 1.8088,
+      "step": 61064
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0737989886219975e-07,
+      "loss": 1.9061,
+      "step": 61065
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0733249051833123e-07,
+      "loss": 1.7766,
+      "step": 61066
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.072850821744627e-07,
+      "loss": 0.917,
+      "step": 61067
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0723767383059418e-07,
+      "loss": 1.5657,
+      "step": 61068
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0719026548672566e-07,
+      "loss": 2.6107,
+      "step": 61069
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0714285714285714e-07,
+      "loss": 1.636,
+      "step": 61070
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0709544879898862e-07,
+      "loss": 0.3993,
+      "step": 61071
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.070480404551201e-07,
+      "loss": 1.3938,
+      "step": 61072
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0700063211125159e-07,
+      "loss": 0.9893,
+      "step": 61073
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0695322376738307e-07,
+      "loss": 1.7194,
+      "step": 61074
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0690581542351455e-07,
+      "loss": 2.0328,
+      "step": 61075
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0685840707964603e-07,
+      "loss": 1.6264,
+      "step": 61076
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0681099873577751e-07,
+      "loss": 2.1268,
+      "step": 61077
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0676359039190897e-07,
+      "loss": 1.7201,
+      "step": 61078
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0671618204804045e-07,
+      "loss": 1.3376,
+      "step": 61079
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0666877370417193e-07,
+      "loss": 0.9233,
+      "step": 61080
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 1.0662136536030341e-07,
+      "loss": 2.2912,
+      "step": 61081
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0657395701643489e-07,
+      "loss": 1.3435,
+      "step": 61082
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0652654867256637e-07,
+      "loss": 2.1927,
+      "step": 61083
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0647914032869786e-07,
+      "loss": 2.2425,
+      "step": 61084
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0643173198482934e-07,
+      "loss": 1.4145,
+      "step": 61085
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0638432364096082e-07,
+      "loss": 1.066,
+      "step": 61086
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.063369152970923e-07,
+      "loss": 3.1365,
+      "step": 61087
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0628950695322378e-07,
+      "loss": 1.8732,
+      "step": 61088
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0624209860935525e-07,
+      "loss": 2.6421,
+      "step": 61089
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0619469026548672e-07,
+      "loss": 2.8321,
+      "step": 61090
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.061472819216182e-07,
+      "loss": 1.7054,
+      "step": 61091
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0609987357774968e-07,
+      "loss": 2.3273,
+      "step": 61092
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0605246523388116e-07,
+      "loss": 1.949,
+      "step": 61093
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0600505689001264e-07,
+      "loss": 2.4553,
+      "step": 61094
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0595764854614413e-07,
+      "loss": 2.2124,
+      "step": 61095
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0591024020227561e-07,
+      "loss": 0.1526,
+      "step": 61096
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0586283185840709e-07,
+      "loss": 2.5765,
+      "step": 61097
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0581542351453857e-07,
+      "loss": 1.9207,
+      "step": 61098
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0576801517067004e-07,
+      "loss": 0.9226,
+      "step": 61099
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0572060682680152e-07,
+      "loss": 1.73,
+      "step": 61100
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.05673198482933e-07,
+      "loss": 1.5779,
+      "step": 61101
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0562579013906448e-07,
+      "loss": 0.9908,
+      "step": 61102
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0557838179519595e-07,
+      "loss": 2.3109,
+      "step": 61103
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0553097345132743e-07,
+      "loss": 0.784,
+      "step": 61104
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.054835651074589e-07,
+      "loss": 2.254,
+      "step": 61105
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0543615676359038e-07,
+      "loss": 1.1355,
+      "step": 61106
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0538874841972188e-07,
+      "loss": 1.354,
+      "step": 61107
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0534134007585335e-07,
+      "loss": 2.9351,
+      "step": 61108
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0529393173198483e-07,
+      "loss": 2.7397,
+      "step": 61109
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0524652338811631e-07,
+      "loss": 3.3027,
+      "step": 61110
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0519911504424779e-07,
+      "loss": 1.2101,
+      "step": 61111
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0515170670037927e-07,
+      "loss": 1.3209,
+      "step": 61112
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0510429835651075e-07,
+      "loss": 1.6222,
+      "step": 61113
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0505689001264223e-07,
+      "loss": 1.9967,
+      "step": 61114
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.050094816687737e-07,
+      "loss": 2.5766,
+      "step": 61115
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0496207332490517e-07,
+      "loss": 2.089,
+      "step": 61116
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0491466498103665e-07,
+      "loss": 1.3281,
+      "step": 61117
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0486725663716816e-07,
+      "loss": 1.3783,
+      "step": 61118
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0481984829329964e-07,
+      "loss": 1.8672,
+      "step": 61119
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.047724399494311e-07,
+      "loss": 0.7056,
+      "step": 61120
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0472503160556258e-07,
+      "loss": 1.6579,
+      "step": 61121
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0467762326169406e-07,
+      "loss": 1.3483,
+      "step": 61122
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0463021491782554e-07,
+      "loss": 1.264,
+      "step": 61123
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0458280657395702e-07,
+      "loss": 2.4031,
+      "step": 61124
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.045353982300885e-07,
+      "loss": 2.6622,
+      "step": 61125
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0448798988621998e-07,
+      "loss": 1.8531,
+      "step": 61126
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0444058154235145e-07,
+      "loss": 4.1451,
+      "step": 61127
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0439317319848292e-07,
+      "loss": 2.9013,
+      "step": 61128
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0434576485461443e-07,
+      "loss": 2.4041,
+      "step": 61129
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.042983565107459e-07,
+      "loss": 0.8881,
+      "step": 61130
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0425094816687738e-07,
+      "loss": 1.7318,
+      "step": 61131
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0420353982300885e-07,
+      "loss": 1.3878,
+      "step": 61132
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0415613147914033e-07,
+      "loss": 1.2383,
+      "step": 61133
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0410872313527181e-07,
+      "loss": 1.836,
+      "step": 61134
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0406131479140329e-07,
+      "loss": 1.3087,
+      "step": 61135
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0401390644753477e-07,
+      "loss": 2.0463,
+      "step": 61136
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0396649810366624e-07,
+      "loss": 2.9042,
+      "step": 61137
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0391908975979772e-07,
+      "loss": 3.5368,
+      "step": 61138
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.038716814159292e-07,
+      "loss": 1.8638,
+      "step": 61139
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.038242730720607e-07,
+      "loss": 3.5064,
+      "step": 61140
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0377686472819217e-07,
+      "loss": 1.5463,
+      "step": 61141
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0372945638432365e-07,
+      "loss": 3.6766,
+      "step": 61142
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0368204804045513e-07,
+      "loss": 2.1171,
+      "step": 61143
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0363463969658661e-07,
+      "loss": 1.6121,
+      "step": 61144
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 1.0358723135271808e-07,
+      "loss": 1.7118,
+      "step": 61145
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0353982300884955e-07,
+      "loss": 0.6986,
+      "step": 61146
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0349241466498103e-07,
+      "loss": 2.9889,
+      "step": 61147
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0344500632111251e-07,
+      "loss": 0.8437,
+      "step": 61148
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0339759797724399e-07,
+      "loss": 1.58,
+      "step": 61149
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0335018963337547e-07,
+      "loss": 1.3195,
+      "step": 61150
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0330278128950696e-07,
+      "loss": 1.191,
+      "step": 61151
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0325537294563844e-07,
+      "loss": 1.7455,
+      "step": 61152
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0320796460176992e-07,
+      "loss": 1.5627,
+      "step": 61153
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.031605562579014e-07,
+      "loss": 0.8908,
+      "step": 61154
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0311314791403288e-07,
+      "loss": 1.5543,
+      "step": 61155
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0306573957016436e-07,
+      "loss": 0.7542,
+      "step": 61156
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0301833122629584e-07,
+      "loss": 1.373,
+      "step": 61157
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.029709228824273e-07,
+      "loss": 1.7569,
+      "step": 61158
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0292351453855878e-07,
+      "loss": 1.2743,
+      "step": 61159
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0287610619469026e-07,
+      "loss": 1.6702,
+      "step": 61160
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0282869785082174e-07,
+      "loss": 0.6304,
+      "step": 61161
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0278128950695322e-07,
+      "loss": 2.779,
+      "step": 61162
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0273388116308471e-07,
+      "loss": 2.361,
+      "step": 61163
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0268647281921619e-07,
+      "loss": 1.0728,
+      "step": 61164
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0263906447534767e-07,
+      "loss": 1.8944,
+      "step": 61165
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0259165613147915e-07,
+      "loss": 1.221,
+      "step": 61166
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0254424778761063e-07,
+      "loss": 2.4999,
+      "step": 61167
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.024968394437421e-07,
+      "loss": 2.4986,
+      "step": 61168
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0244943109987358e-07,
+      "loss": 1.6687,
+      "step": 61169
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0240202275600505e-07,
+      "loss": 2.5354,
+      "step": 61170
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0235461441213653e-07,
+      "loss": 4.6363,
+      "step": 61171
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0230720606826801e-07,
+      "loss": 1.7093,
+      "step": 61172
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0225979772439949e-07,
+      "loss": 0.4036,
+      "step": 61173
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0221238938053099e-07,
+      "loss": 1.2273,
+      "step": 61174
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0216498103666246e-07,
+      "loss": 2.1266,
+      "step": 61175
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0211757269279394e-07,
+      "loss": 1.0272,
+      "step": 61176
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0207016434892542e-07,
+      "loss": 1.8861,
+      "step": 61177
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.020227560050569e-07,
+      "loss": 1.6891,
+      "step": 61178
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0197534766118837e-07,
+      "loss": 1.0584,
+      "step": 61179
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0192793931731985e-07,
+      "loss": 2.7053,
+      "step": 61180
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0188053097345133e-07,
+      "loss": 2.369,
+      "step": 61181
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0183312262958281e-07,
+      "loss": 1.2878,
+      "step": 61182
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0178571428571428e-07,
+      "loss": 1.468,
+      "step": 61183
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0173830594184575e-07,
+      "loss": 0.7967,
+      "step": 61184
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0169089759797726e-07,
+      "loss": 1.915,
+      "step": 61185
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0164348925410874e-07,
+      "loss": 1.7659,
+      "step": 61186
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.015960809102402e-07,
+      "loss": 1.1954,
+      "step": 61187
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0154867256637168e-07,
+      "loss": 2.1632,
+      "step": 61188
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0150126422250316e-07,
+      "loss": 1.5539,
+      "step": 61189
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0145385587863464e-07,
+      "loss": 1.479,
+      "step": 61190
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0140644753476612e-07,
+      "loss": 1.7168,
+      "step": 61191
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.013590391908976e-07,
+      "loss": 1.8442,
+      "step": 61192
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0131163084702908e-07,
+      "loss": 2.7831,
+      "step": 61193
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0126422250316056e-07,
+      "loss": 0.7321,
+      "step": 61194
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0121681415929202e-07,
+      "loss": 2.7133,
+      "step": 61195
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0116940581542353e-07,
+      "loss": 1.2147,
+      "step": 61196
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0112199747155501e-07,
+      "loss": 1.6218,
+      "step": 61197
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0107458912768649e-07,
+      "loss": 1.1391,
+      "step": 61198
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0102718078381796e-07,
+      "loss": 1.9171,
+      "step": 61199
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0097977243994943e-07,
+      "loss": 4.06,
+      "step": 61200
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0093236409608091e-07,
+      "loss": 0.7035,
+      "step": 61201
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0088495575221239e-07,
+      "loss": 2.217,
+      "step": 61202
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0083754740834387e-07,
+      "loss": 2.4069,
+      "step": 61203
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0079013906447535e-07,
+      "loss": 3.3728,
+      "step": 61204
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0074273072060683e-07,
+      "loss": 1.3612,
+      "step": 61205
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.006953223767383e-07,
+      "loss": 2.8033,
+      "step": 61206
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.006479140328698e-07,
+      "loss": 1.711,
+      "step": 61207
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 1.0060050568900128e-07,
+      "loss": 2.0876,
+      "step": 61208
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0055309734513275e-07,
+      "loss": 1.6285,
+      "step": 61209
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0050568900126423e-07,
+      "loss": 2.8469,
+      "step": 61210
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0045828065739571e-07,
+      "loss": 2.6671,
+      "step": 61211
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0041087231352718e-07,
+      "loss": 0.8634,
+      "step": 61212
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0036346396965866e-07,
+      "loss": 2.5708,
+      "step": 61213
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0031605562579014e-07,
+      "loss": 2.0708,
+      "step": 61214
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0026864728192161e-07,
+      "loss": 2.122,
+      "step": 61215
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.002212389380531e-07,
+      "loss": 0.9197,
+      "step": 61216
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0017383059418457e-07,
+      "loss": 2.3266,
+      "step": 61217
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0012642225031605e-07,
+      "loss": 1.5282,
+      "step": 61218
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0007901390644754e-07,
+      "loss": 1.5261,
+      "step": 61219
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 1.0003160556257902e-07,
+      "loss": 1.0414,
+      "step": 61220
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.99841972187105e-08,
+      "loss": 2.469,
+      "step": 61221
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.993678887484198e-08,
+      "loss": 2.279,
+      "step": 61222
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.988938053097346e-08,
+      "loss": 1.3479,
+      "step": 61223
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.984197218710494e-08,
+      "loss": 1.285,
+      "step": 61224
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.97945638432364e-08,
+      "loss": 1.0489,
+      "step": 61225
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.974715549936788e-08,
+      "loss": 1.8724,
+      "step": 61226
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.969974715549936e-08,
+      "loss": 2.6014,
+      "step": 61227
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.965233881163084e-08,
+      "loss": 1.8353,
+      "step": 61228
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.960493046776232e-08,
+      "loss": 0.2881,
+      "step": 61229
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.955752212389381e-08,
+      "loss": 0.6509,
+      "step": 61230
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.951011378002529e-08,
+      "loss": 0.8411,
+      "step": 61231
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.946270543615677e-08,
+      "loss": 1.2915,
+      "step": 61232
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.941529709228825e-08,
+      "loss": 2.3802,
+      "step": 61233
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.936788874841973e-08,
+      "loss": 2.3364,
+      "step": 61234
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.932048040455121e-08,
+      "loss": 2.6692,
+      "step": 61235
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.927307206068269e-08,
+      "loss": 2.0259,
+      "step": 61236
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.922566371681415e-08,
+      "loss": 1.9438,
+      "step": 61237
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.917825537294563e-08,
+      "loss": 2.8087,
+      "step": 61238
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.913084702907711e-08,
+      "loss": 2.2635,
+      "step": 61239
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.908343868520859e-08,
+      "loss": 2.2164,
+      "step": 61240
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.90360303413401e-08,
+      "loss": 1.3054,
+      "step": 61241
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.898862199747156e-08,
+      "loss": 2.3019,
+      "step": 61242
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.894121365360304e-08,
+      "loss": 1.5155,
+      "step": 61243
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.889380530973452e-08,
+      "loss": 1.0457,
+      "step": 61244
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.8846396965866e-08,
+      "loss": 2.5744,
+      "step": 61245
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.879898862199748e-08,
+      "loss": 1.4077,
+      "step": 61246
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.875158027812895e-08,
+      "loss": 2.0172,
+      "step": 61247
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.870417193426043e-08,
+      "loss": 1.8131,
+      "step": 61248
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.865676359039191e-08,
+      "loss": 1.0641,
+      "step": 61249
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.860935524652338e-08,
+      "loss": 2.5558,
+      "step": 61250
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.856194690265486e-08,
+      "loss": 2.4715,
+      "step": 61251
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.851453855878636e-08,
+      "loss": 2.7224,
+      "step": 61252
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.846713021491784e-08,
+      "loss": 1.8839,
+      "step": 61253
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.841972187104931e-08,
+      "loss": 0.7424,
+      "step": 61254
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.837231352718079e-08,
+      "loss": 1.8217,
+      "step": 61255
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.832490518331226e-08,
+      "loss": 0.9779,
+      "step": 61256
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.827749683944374e-08,
+      "loss": 1.2348,
+      "step": 61257
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.823008849557522e-08,
+      "loss": 1.6152,
+      "step": 61258
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.81826801517067e-08,
+      "loss": 0.9629,
+      "step": 61259
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.813527180783818e-08,
+      "loss": 1.7003,
+      "step": 61260
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.808786346396966e-08,
+      "loss": 0.871,
+      "step": 61261
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.804045512010113e-08,
+      "loss": 2.8503,
+      "step": 61262
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.799304677623263e-08,
+      "loss": 1.5865,
+      "step": 61263
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.794563843236411e-08,
+      "loss": 1.9826,
+      "step": 61264
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.789823008849559e-08,
+      "loss": 1.3644,
+      "step": 61265
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.785082174462707e-08,
+      "loss": 1.0177,
+      "step": 61266
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.780341340075853e-08,
+      "loss": 1.7696,
+      "step": 61267
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.775600505689001e-08,
+      "loss": 2.2295,
+      "step": 61268
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.770859671302149e-08,
+      "loss": 3.8041,
+      "step": 61269
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.766118836915297e-08,
+      "loss": 1.9241,
+      "step": 61270
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.761378002528445e-08,
+      "loss": 1.2684,
+      "step": 61271
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.756637168141593e-08,
+      "loss": 1.4957,
+      "step": 61272
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.751896333754741e-08,
+      "loss": 1.4357,
+      "step": 61273
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.747155499367889e-08,
+      "loss": 1.2401,
+      "step": 61274
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.742414664981038e-08,
+      "loss": 1.388,
+      "step": 61275
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.737673830594186e-08,
+      "loss": 1.9295,
+      "step": 61276
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.732932996207334e-08,
+      "loss": 3.1891,
+      "step": 61277
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.728192161820481e-08,
+      "loss": 1.4776,
+      "step": 61278
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.723451327433628e-08,
+      "loss": 2.8458,
+      "step": 61279
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.718710493046776e-08,
+      "loss": 2.2611,
+      "step": 61280
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.713969658659924e-08,
+      "loss": 1.1982,
+      "step": 61281
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.709228824273072e-08,
+      "loss": 0.9865,
+      "step": 61282
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.70448798988622e-08,
+      "loss": 1.343,
+      "step": 61283
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.699747155499368e-08,
+      "loss": 2.133,
+      "step": 61284
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.695006321112515e-08,
+      "loss": 2.381,
+      "step": 61285
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.690265486725665e-08,
+      "loss": 1.7828,
+      "step": 61286
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.685524652338813e-08,
+      "loss": 0.9014,
+      "step": 61287
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.68078381795196e-08,
+      "loss": 3.045,
+      "step": 61288
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.676042983565108e-08,
+      "loss": 1.3077,
+      "step": 61289
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.671302149178256e-08,
+      "loss": 1.6838,
+      "step": 61290
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.666561314791404e-08,
+      "loss": 2.5247,
+      "step": 61291
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.66182048040455e-08,
+      "loss": 1.5712,
+      "step": 61292
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.657079646017699e-08,
+      "loss": 0.9021,
+      "step": 61293
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.652338811630846e-08,
+      "loss": 0.7436,
+      "step": 61294
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.647597977243994e-08,
+      "loss": 1.6707,
+      "step": 61295
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.642857142857142e-08,
+      "loss": 2.3082,
+      "step": 61296
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.638116308470291e-08,
+      "loss": 2.9452,
+      "step": 61297
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.63337547408344e-08,
+      "loss": 0.9573,
+      "step": 61298
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.628634639696587e-08,
+      "loss": 2.0403,
+      "step": 61299
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.623893805309735e-08,
+      "loss": 1.0769,
+      "step": 61300
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.619152970922883e-08,
+      "loss": 1.9662,
+      "step": 61301
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.614412136536031e-08,
+      "loss": 1.7275,
+      "step": 61302
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.609671302149179e-08,
+      "loss": 3.7574,
+      "step": 61303
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.604930467762327e-08,
+      "loss": 1.1071,
+      "step": 61304
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.600189633375473e-08,
+      "loss": 1.847,
+      "step": 61305
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.595448798988621e-08,
+      "loss": 3.2941,
+      "step": 61306
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.590707964601769e-08,
+      "loss": 1.2768,
+      "step": 61307
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.58596713021492e-08,
+      "loss": 2.4428,
+      "step": 61308
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.581226295828066e-08,
+      "loss": 1.3008,
+      "step": 61309
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.576485461441214e-08,
+      "loss": 0.9057,
+      "step": 61310
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.571744627054362e-08,
+      "loss": 1.9165,
+      "step": 61311
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.56700379266751e-08,
+      "loss": 1.6566,
+      "step": 61312
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.562262958280658e-08,
+      "loss": 0.768,
+      "step": 61313
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.557522123893806e-08,
+      "loss": 1.5817,
+      "step": 61314
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.552781289506954e-08,
+      "loss": 0.6972,
+      "step": 61315
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.548040455120101e-08,
+      "loss": 2.2501,
+      "step": 61316
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.543299620733248e-08,
+      "loss": 1.6006,
+      "step": 61317
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.538558786346396e-08,
+      "loss": 2.7887,
+      "step": 61318
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.533817951959546e-08,
+      "loss": 1.3971,
+      "step": 61319
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.529077117572694e-08,
+      "loss": 1.4069,
+      "step": 61320
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.524336283185842e-08,
+      "loss": 3.8695,
+      "step": 61321
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.519595448798989e-08,
+      "loss": 2.4699,
+      "step": 61322
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.514854614412137e-08,
+      "loss": 2.5625,
+      "step": 61323
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.510113780025285e-08,
+      "loss": 1.0352,
+      "step": 61324
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.505372945638432e-08,
+      "loss": 2.1752,
+      "step": 61325
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.50063211125158e-08,
+      "loss": 3.817,
+      "step": 61326
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.495891276864728e-08,
+      "loss": 2.2701,
+      "step": 61327
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.491150442477876e-08,
+      "loss": 3.6745,
+      "step": 61328
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.486409608091024e-08,
+      "loss": 1.0311,
+      "step": 61329
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.48166877370417e-08,
+      "loss": 2.161,
+      "step": 61330
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.476927939317321e-08,
+      "loss": 2.0002,
+      "step": 61331
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.472187104930469e-08,
+      "loss": 4.6509,
+      "step": 61332
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.467446270543617e-08,
+      "loss": 3.8239,
+      "step": 61333
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.462705436156764e-08,
+      "loss": 2.1695,
+      "step": 61334
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.457964601769911e-08,
+      "loss": 2.1561,
+      "step": 61335
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.453223767383059e-08,
+      "loss": 1.806,
+      "step": 61336
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.448482932996207e-08,
+      "loss": 1.5747,
+      "step": 61337
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.443742098609355e-08,
+      "loss": 2.4962,
+      "step": 61338
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.439001264222503e-08,
+      "loss": 2.8832,
+      "step": 61339
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.434260429835651e-08,
+      "loss": 1.6841,
+      "step": 61340
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.429519595448799e-08,
+      "loss": 1.3788,
+      "step": 61341
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.424778761061948e-08,
+      "loss": 1.9292,
+      "step": 61342
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.420037926675096e-08,
+      "loss": 2.3746,
+      "step": 61343
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.415297092288244e-08,
+      "loss": 1.0205,
+      "step": 61344
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.410556257901392e-08,
+      "loss": 1.8345,
+      "step": 61345
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.40581542351454e-08,
+      "loss": 2.176,
+      "step": 61346
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.401074589127686e-08,
+      "loss": 1.0447,
+      "step": 61347
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.396333754740834e-08,
+      "loss": 2.2273,
+      "step": 61348
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.391592920353982e-08,
+      "loss": 2.3514,
+      "step": 61349
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.38685208596713e-08,
+      "loss": 1.4468,
+      "step": 61350
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.382111251580278e-08,
+      "loss": 0.7212,
+      "step": 61351
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.377370417193426e-08,
+      "loss": 2.2217,
+      "step": 61352
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.372629582806574e-08,
+      "loss": 1.909,
+      "step": 61353
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.367888748419721e-08,
+      "loss": 2.6603,
+      "step": 61354
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.363147914032869e-08,
+      "loss": 1.676,
+      "step": 61355
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.358407079646019e-08,
+      "loss": 1.1765,
+      "step": 61356
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.353666245259166e-08,
+      "loss": 2.2574,
+      "step": 61357
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.348925410872314e-08,
+      "loss": 3.3611,
+      "step": 61358
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.344184576485461e-08,
+      "loss": 2.4835,
+      "step": 61359
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.339443742098609e-08,
+      "loss": 1.2051,
+      "step": 61360
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.334702907711758e-08,
+      "loss": 1.6381,
+      "step": 61361
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.329962073324906e-08,
+      "loss": 0.8735,
+      "step": 61362
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.325221238938054e-08,
+      "loss": 0.4072,
+      "step": 61363
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.320480404551202e-08,
+      "loss": 2.3039,
+      "step": 61364
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.31573957016435e-08,
+      "loss": 2.5166,
+      "step": 61365
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.310998735777496e-08,
+      "loss": 2.944,
+      "step": 61366
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.306257901390645e-08,
+      "loss": 1.1852,
+      "step": 61367
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.301517067003793e-08,
+      "loss": 0.8828,
+      "step": 61368
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.296776232616941e-08,
+      "loss": 2.4003,
+      "step": 61369
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.292035398230089e-08,
+      "loss": 1.2141,
+      "step": 61370
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.287294563843237e-08,
+      "loss": 3.0863,
+      "step": 61371
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.282553729456384e-08,
+      "loss": 1.7032,
+      "step": 61372
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.277812895069533e-08,
+      "loss": 2.2948,
+      "step": 61373
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.27307206068268e-08,
+      "loss": 0.9725,
+      "step": 61374
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.268331226295829e-08,
+      "loss": 0.9437,
+      "step": 61375
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.263590391908976e-08,
+      "loss": 2.0216,
+      "step": 61376
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.258849557522124e-08,
+      "loss": 2.268,
+      "step": 61377
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.254108723135272e-08,
+      "loss": 2.6751,
+      "step": 61378
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.24936788874842e-08,
+      "loss": 0.9112,
+      "step": 61379
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.244627054361568e-08,
+      "loss": 1.0498,
+      "step": 61380
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.239886219974716e-08,
+      "loss": 1.524,
+      "step": 61381
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.235145385587864e-08,
+      "loss": 2.7766,
+      "step": 61382
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.230404551201012e-08,
+      "loss": 2.1905,
+      "step": 61383
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.22566371681416e-08,
+      "loss": 0.918,
+      "step": 61384
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.220922882427307e-08,
+      "loss": 1.3021,
+      "step": 61385
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.216182048040455e-08,
+      "loss": 2.2653,
+      "step": 61386
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.211441213653603e-08,
+      "loss": 0.7758,
+      "step": 61387
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.206700379266751e-08,
+      "loss": 2.7032,
+      "step": 61388
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.201959544879899e-08,
+      "loss": 0.9303,
+      "step": 61389
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.197218710493047e-08,
+      "loss": 1.2848,
+      "step": 61390
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.192477876106195e-08,
+      "loss": 2.0906,
+      "step": 61391
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.187737041719343e-08,
+      "loss": 2.6696,
+      "step": 61392
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.18299620733249e-08,
+      "loss": 2.4222,
+      "step": 61393
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.178255372945639e-08,
+      "loss": 2.9787,
+      "step": 61394
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.173514538558786e-08,
+      "loss": 1.6791,
+      "step": 61395
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.168773704171934e-08,
+      "loss": 2.1866,
+      "step": 61396
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.164032869785082e-08,
+      "loss": 3.0382,
+      "step": 61397
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.15929203539823e-08,
+      "loss": 1.7432,
+      "step": 61398
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.154551201011378e-08,
+      "loss": 1.327,
+      "step": 61399
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.149810366624526e-08,
+      "loss": 1.7931,
+      "step": 61400
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.145069532237674e-08,
+      "loss": 2.8185,
+      "step": 61401
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.140328697850822e-08,
+      "loss": 1.9252,
+      "step": 61402
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.13558786346397e-08,
+      "loss": 1.2484,
+      "step": 61403
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.130847029077117e-08,
+      "loss": 2.0689,
+      "step": 61404
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.126106194690265e-08,
+      "loss": 1.1784,
+      "step": 61405
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.121365360303415e-08,
+      "loss": 1.8909,
+      "step": 61406
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.116624525916562e-08,
+      "loss": 1.7102,
+      "step": 61407
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.111883691529709e-08,
+      "loss": 2.5879,
+      "step": 61408
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.107142857142857e-08,
+      "loss": 1.2362,
+      "step": 61409
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.102402022756005e-08,
+      "loss": 1.1307,
+      "step": 61410
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.097661188369153e-08,
+      "loss": 0.9034,
+      "step": 61411
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.092920353982302e-08,
+      "loss": 0.6348,
+      "step": 61412
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.08817951959545e-08,
+      "loss": 2.1387,
+      "step": 61413
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.083438685208596e-08,
+      "loss": 2.2588,
+      "step": 61414
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.078697850821744e-08,
+      "loss": 1.9846,
+      "step": 61415
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.073957016434892e-08,
+      "loss": 2.0514,
+      "step": 61416
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.069216182048041e-08,
+      "loss": 1.0178,
+      "step": 61417
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.064475347661189e-08,
+      "loss": 1.7691,
+      "step": 61418
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.059734513274337e-08,
+      "loss": 1.4164,
+      "step": 61419
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.054993678887484e-08,
+      "loss": 1.0087,
+      "step": 61420
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.050252844500632e-08,
+      "loss": 2.25,
+      "step": 61421
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.04551201011378e-08,
+      "loss": 2.5077,
+      "step": 61422
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.040771175726929e-08,
+      "loss": 1.9903,
+      "step": 61423
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.036030341340077e-08,
+      "loss": 1.826,
+      "step": 61424
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.031289506953225e-08,
+      "loss": 0.9546,
+      "step": 61425
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.026548672566372e-08,
+      "loss": 2.1236,
+      "step": 61426
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.021807838179519e-08,
+      "loss": 2.2772,
+      "step": 61427
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.017067003792667e-08,
+      "loss": 2.2441,
+      "step": 61428
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.012326169405816e-08,
+      "loss": 1.3273,
+      "step": 61429
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.007585335018964e-08,
+      "loss": 3.2777,
+      "step": 61430
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.002844500632112e-08,
+      "loss": 1.3331,
+      "step": 61431
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.99810366624526e-08,
+      "loss": 2.9309,
+      "step": 61432
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.993362831858406e-08,
+      "loss": 1.5923,
+      "step": 61433
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.988621997471556e-08,
+      "loss": 2.9905,
+      "step": 61434
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.983881163084703e-08,
+      "loss": 1.6534,
+      "step": 61435
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.979140328697851e-08,
+      "loss": 1.6099,
+      "step": 61436
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.974399494310999e-08,
+      "loss": 2.2722,
+      "step": 61437
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.969658659924147e-08,
+      "loss": 2.6985,
+      "step": 61438
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.964917825537294e-08,
+      "loss": 1.3491,
+      "step": 61439
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.960176991150443e-08,
+      "loss": 3.2603,
+      "step": 61440
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.955436156763591e-08,
+      "loss": 0.9164,
+      "step": 61441
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.950695322376739e-08,
+      "loss": 1.1206,
+      "step": 61442
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.945954487989887e-08,
+      "loss": 2.1358,
+      "step": 61443
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.941213653603035e-08,
+      "loss": 1.6893,
+      "step": 61444
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.936472819216182e-08,
+      "loss": 2.2652,
+      "step": 61445
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.93173198482933e-08,
+      "loss": 1.4701,
+      "step": 61446
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.926991150442478e-08,
+      "loss": 1.4663,
+      "step": 61447
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.922250316055626e-08,
+      "loss": 1.6094,
+      "step": 61448
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.917509481668774e-08,
+      "loss": 2.8065,
+      "step": 61449
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.912768647281922e-08,
+      "loss": 3.1115,
+      "step": 61450
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.90802781289507e-08,
+      "loss": 2.5395,
+      "step": 61451
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.903286978508218e-08,
+      "loss": 2.4301,
+      "step": 61452
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.898546144121366e-08,
+      "loss": 1.9557,
+      "step": 61453
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.893805309734513e-08,
+      "loss": 1.164,
+      "step": 61454
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.889064475347661e-08,
+      "loss": 0.4392,
+      "step": 61455
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.884323640960809e-08,
+      "loss": 0.9376,
+      "step": 61456
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.879582806573957e-08,
+      "loss": 3.1748,
+      "step": 61457
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.874841972187105e-08,
+      "loss": 0.9413,
+      "step": 61458
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.870101137800253e-08,
+      "loss": 1.6988,
+      "step": 61459
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.865360303413401e-08,
+      "loss": 2.5753,
+      "step": 61460
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 8.860619469026549e-08,
+      "loss": 1.8518,
+      "step": 61461
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.855878634639697e-08,
+      "loss": 1.079,
+      "step": 61462
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.851137800252845e-08,
+      "loss": 0.3927,
+      "step": 61463
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.846396965865992e-08,
+      "loss": 1.7532,
+      "step": 61464
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.84165613147914e-08,
+      "loss": 2.2214,
+      "step": 61465
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.836915297092288e-08,
+      "loss": 1.7008,
+      "step": 61466
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.832174462705436e-08,
+      "loss": 2.8668,
+      "step": 61467
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.827433628318585e-08,
+      "loss": 2.0931,
+      "step": 61468
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.822692793931732e-08,
+      "loss": 2.1146,
+      "step": 61469
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.81795195954488e-08,
+      "loss": 1.9258,
+      "step": 61470
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.813211125158028e-08,
+      "loss": 0.8716,
+      "step": 61471
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.808470290771176e-08,
+      "loss": 1.3144,
+      "step": 61472
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.803729456384325e-08,
+      "loss": 1.6642,
+      "step": 61473
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.798988621997473e-08,
+      "loss": 1.3097,
+      "step": 61474
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.794247787610619e-08,
+      "loss": 1.3238,
+      "step": 61475
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.789506953223767e-08,
+      "loss": 0.956,
+      "step": 61476
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.784766118836915e-08,
+      "loss": 3.1825,
+      "step": 61477
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.780025284450063e-08,
+      "loss": 1.2065,
+      "step": 61478
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.775284450063212e-08,
+      "loss": 2.7282,
+      "step": 61479
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.77054361567636e-08,
+      "loss": 1.6766,
+      "step": 61480
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.765802781289507e-08,
+      "loss": 1.2875,
+      "step": 61481
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.761061946902655e-08,
+      "loss": 1.6261,
+      "step": 61482
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.756321112515802e-08,
+      "loss": 1.0282,
+      "step": 61483
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.75158027812895e-08,
+      "loss": 1.9504,
+      "step": 61484
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.7468394437421e-08,
+      "loss": 1.3795,
+      "step": 61485
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.742098609355247e-08,
+      "loss": 2.6484,
+      "step": 61486
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.737357774968394e-08,
+      "loss": 0.721,
+      "step": 61487
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.732616940581542e-08,
+      "loss": 1.2332,
+      "step": 61488
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.72787610619469e-08,
+      "loss": 0.9876,
+      "step": 61489
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.723135271807839e-08,
+      "loss": 1.4566,
+      "step": 61490
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.718394437420987e-08,
+      "loss": 1.9803,
+      "step": 61491
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.713653603034135e-08,
+      "loss": 2.9284,
+      "step": 61492
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.708912768647283e-08,
+      "loss": 1.3849,
+      "step": 61493
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.704171934260429e-08,
+      "loss": 2.1952,
+      "step": 61494
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.699431099873577e-08,
+      "loss": 1.4174,
+      "step": 61495
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.694690265486726e-08,
+      "loss": 1.6769,
+      "step": 61496
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.689949431099874e-08,
+      "loss": 2.0169,
+      "step": 61497
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.685208596713022e-08,
+      "loss": 1.05,
+      "step": 61498
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.68046776232617e-08,
+      "loss": 1.8213,
+      "step": 61499
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.675726927939317e-08,
+      "loss": 1.1184,
+      "step": 61500
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.670986093552466e-08,
+      "loss": 1.9593,
+      "step": 61501
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.666245259165614e-08,
+      "loss": 1.2967,
+      "step": 61502
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.661504424778762e-08,
+      "loss": 1.4435,
+      "step": 61503
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.65676359039191e-08,
+      "loss": 1.1709,
+      "step": 61504
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.652022756005057e-08,
+      "loss": 0.9905,
+      "step": 61505
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.647281921618204e-08,
+      "loss": 2.1983,
+      "step": 61506
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.642541087231353e-08,
+      "loss": 1.2705,
+      "step": 61507
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.637800252844501e-08,
+      "loss": 2.2741,
+      "step": 61508
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.633059418457649e-08,
+      "loss": 1.4921,
+      "step": 61509
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.628318584070797e-08,
+      "loss": 1.2491,
+      "step": 61510
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.623577749683945e-08,
+      "loss": 1.3788,
+      "step": 61511
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.618836915297093e-08,
+      "loss": 2.7174,
+      "step": 61512
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.61409608091024e-08,
+      "loss": 2.013,
+      "step": 61513
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.609355246523388e-08,
+      "loss": 1.0004,
+      "step": 61514
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.604614412136536e-08,
+      "loss": 1.7143,
+      "step": 61515
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.599873577749684e-08,
+      "loss": 1.5115,
+      "step": 61516
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.595132743362832e-08,
+      "loss": 1.1767,
+      "step": 61517
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.59039190897598e-08,
+      "loss": 0.9132,
+      "step": 61518
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.585651074589128e-08,
+      "loss": 2.5403,
+      "step": 61519
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.580910240202276e-08,
+      "loss": 2.2054,
+      "step": 61520
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.576169405815424e-08,
+      "loss": 1.7201,
+      "step": 61521
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.571428571428572e-08,
+      "loss": 1.3754,
+      "step": 61522
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.56668773704172e-08,
+      "loss": 1.7181,
+      "step": 61523
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.561946902654867e-08,
+      "loss": 1.5195,
+      "step": 61524
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 8.557206068268015e-08,
+      "loss": 1.4205,
+      "step": 61525
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.552465233881163e-08,
+      "loss": 1.5199,
+      "step": 61526
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.547724399494311e-08,
+      "loss": 0.695,
+      "step": 61527
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.542983565107459e-08,
+      "loss": 1.6934,
+      "step": 61528
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.538242730720608e-08,
+      "loss": 4.5206,
+      "step": 61529
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.533501896333755e-08,
+      "loss": 1.1228,
+      "step": 61530
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.528761061946903e-08,
+      "loss": 2.1306,
+      "step": 61531
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.52402022756005e-08,
+      "loss": 0.7607,
+      "step": 61532
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.519279393173198e-08,
+      "loss": 1.9821,
+      "step": 61533
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.514538558786346e-08,
+      "loss": 2.7286,
+      "step": 61534
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.509797724399496e-08,
+      "loss": 1.777,
+      "step": 61535
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.505056890012642e-08,
+      "loss": 2.2427,
+      "step": 61536
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.50031605562579e-08,
+      "loss": 2.9869,
+      "step": 61537
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.495575221238938e-08,
+      "loss": 2.1561,
+      "step": 61538
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.490834386852086e-08,
+      "loss": 1.4188,
+      "step": 61539
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.486093552465234e-08,
+      "loss": 1.6676,
+      "step": 61540
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.481352718078383e-08,
+      "loss": 2.4458,
+      "step": 61541
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.47661188369153e-08,
+      "loss": 0.9414,
+      "step": 61542
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.471871049304677e-08,
+      "loss": 3.2226,
+      "step": 61543
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.467130214917825e-08,
+      "loss": 0.7211,
+      "step": 61544
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.462389380530973e-08,
+      "loss": 1.6089,
+      "step": 61545
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.457648546144122e-08,
+      "loss": 2.728,
+      "step": 61546
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.45290771175727e-08,
+      "loss": 2.6201,
+      "step": 61547
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.448166877370417e-08,
+      "loss": 2.6197,
+      "step": 61548
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.443426042983565e-08,
+      "loss": 1.7863,
+      "step": 61549
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.438685208596713e-08,
+      "loss": 1.0487,
+      "step": 61550
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.43394437420986e-08,
+      "loss": 2.2527,
+      "step": 61551
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.42920353982301e-08,
+      "loss": 1.7994,
+      "step": 61552
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.424462705436158e-08,
+      "loss": 2.2767,
+      "step": 61553
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.419721871049306e-08,
+      "loss": 0.6135,
+      "step": 61554
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.414981036662452e-08,
+      "loss": 0.9254,
+      "step": 61555
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.4102402022756e-08,
+      "loss": 1.0113,
+      "step": 61556
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.405499367888749e-08,
+      "loss": 2.1348,
+      "step": 61557
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.400758533501897e-08,
+      "loss": 0.8414,
+      "step": 61558
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.396017699115045e-08,
+      "loss": 1.9591,
+      "step": 61559
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.391276864728193e-08,
+      "loss": 1.9448,
+      "step": 61560
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.38653603034134e-08,
+      "loss": 1.4077,
+      "step": 61561
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.381795195954487e-08,
+      "loss": 2.8615,
+      "step": 61562
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.377054361567637e-08,
+      "loss": 2.7732,
+      "step": 61563
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.372313527180784e-08,
+      "loss": 1.9809,
+      "step": 61564
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.367572692793932e-08,
+      "loss": 1.2209,
+      "step": 61565
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.36283185840708e-08,
+      "loss": 1.5368,
+      "step": 61566
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.358091024020227e-08,
+      "loss": 2.2097,
+      "step": 61567
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.353350189633375e-08,
+      "loss": 1.8606,
+      "step": 61568
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.348609355246524e-08,
+      "loss": 1.6726,
+      "step": 61569
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.343868520859672e-08,
+      "loss": 2.3597,
+      "step": 61570
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.33912768647282e-08,
+      "loss": 1.8974,
+      "step": 61571
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.334386852085968e-08,
+      "loss": 2.0236,
+      "step": 61572
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.329646017699116e-08,
+      "loss": 3.1763,
+      "step": 61573
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.324905183312263e-08,
+      "loss": 2.8497,
+      "step": 61574
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.320164348925411e-08,
+      "loss": 3.1131,
+      "step": 61575
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.315423514538559e-08,
+      "loss": 1.697,
+      "step": 61576
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.310682680151707e-08,
+      "loss": 0.9634,
+      "step": 61577
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.305941845764855e-08,
+      "loss": 2.1304,
+      "step": 61578
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.301201011378003e-08,
+      "loss": 0.914,
+      "step": 61579
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.296460176991151e-08,
+      "loss": 2.8325,
+      "step": 61580
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.291719342604299e-08,
+      "loss": 1.7973,
+      "step": 61581
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.286978508217447e-08,
+      "loss": 1.8001,
+      "step": 61582
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.282237673830594e-08,
+      "loss": 2.3889,
+      "step": 61583
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.277496839443742e-08,
+      "loss": 0.5676,
+      "step": 61584
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.27275600505689e-08,
+      "loss": 1.1592,
+      "step": 61585
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.268015170670038e-08,
+      "loss": 2.2758,
+      "step": 61586
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.263274336283186e-08,
+      "loss": 2.7234,
+      "step": 61587
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 8.258533501896334e-08,
+      "loss": 0.5332,
+      "step": 61588
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.253792667509482e-08,
+      "loss": 2.188,
+      "step": 61589
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.24905183312263e-08,
+      "loss": 1.0885,
+      "step": 61590
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.244310998735778e-08,
+      "loss": 1.1962,
+      "step": 61591
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.239570164348926e-08,
+      "loss": 1.7506,
+      "step": 61592
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.234829329962073e-08,
+      "loss": 1.7501,
+      "step": 61593
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.230088495575221e-08,
+      "loss": 2.0811,
+      "step": 61594
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.225347661188369e-08,
+      "loss": 1.8419,
+      "step": 61595
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.220606826801517e-08,
+      "loss": 0.6341,
+      "step": 61596
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.215865992414665e-08,
+      "loss": 1.4504,
+      "step": 61597
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.211125158027813e-08,
+      "loss": 2.6687,
+      "step": 61598
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.206384323640961e-08,
+      "loss": 3.4812,
+      "step": 61599
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.201643489254109e-08,
+      "loss": 1.9419,
+      "step": 61600
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.196902654867257e-08,
+      "loss": 1.3321,
+      "step": 61601
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.192161820480406e-08,
+      "loss": 0.9057,
+      "step": 61602
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.187420986093552e-08,
+      "loss": 2.4633,
+      "step": 61603
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.1826801517067e-08,
+      "loss": 2.6149,
+      "step": 61604
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.177939317319848e-08,
+      "loss": 3.4889,
+      "step": 61605
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.173198482932996e-08,
+      "loss": 2.446,
+      "step": 61606
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.168457648546144e-08,
+      "loss": 0.779,
+      "step": 61607
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.163716814159293e-08,
+      "loss": 1.4004,
+      "step": 61608
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.15897597977244e-08,
+      "loss": 1.0846,
+      "step": 61609
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.154235145385588e-08,
+      "loss": 0.6378,
+      "step": 61610
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.149494310998736e-08,
+      "loss": 3.43,
+      "step": 61611
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.144753476611883e-08,
+      "loss": 2.6319,
+      "step": 61612
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.140012642225033e-08,
+      "loss": 1.0907,
+      "step": 61613
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.13527180783818e-08,
+      "loss": 1.204,
+      "step": 61614
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.130530973451328e-08,
+      "loss": 0.943,
+      "step": 61615
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.125790139064475e-08,
+      "loss": 1.7365,
+      "step": 61616
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.121049304677623e-08,
+      "loss": 1.8547,
+      "step": 61617
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.116308470290771e-08,
+      "loss": 2.0278,
+      "step": 61618
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.11156763590392e-08,
+      "loss": 1.2288,
+      "step": 61619
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.106826801517068e-08,
+      "loss": 3.2408,
+      "step": 61620
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.102085967130216e-08,
+      "loss": 2.0662,
+      "step": 61621
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.097345132743362e-08,
+      "loss": 1.4638,
+      "step": 61622
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.09260429835651e-08,
+      "loss": 1.8133,
+      "step": 61623
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.087863463969658e-08,
+      "loss": 1.3723,
+      "step": 61624
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.083122629582807e-08,
+      "loss": 3.2707,
+      "step": 61625
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.078381795195955e-08,
+      "loss": 2.3267,
+      "step": 61626
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.073640960809103e-08,
+      "loss": 2.7361,
+      "step": 61627
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.06890012642225e-08,
+      "loss": 2.02,
+      "step": 61628
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.064159292035398e-08,
+      "loss": 1.1099,
+      "step": 61629
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.059418457648547e-08,
+      "loss": 1.145,
+      "step": 61630
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.054677623261695e-08,
+      "loss": 2.0103,
+      "step": 61631
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.049936788874843e-08,
+      "loss": 1.9296,
+      "step": 61632
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.04519595448799e-08,
+      "loss": 1.4476,
+      "step": 61633
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.040455120101138e-08,
+      "loss": 1.0609,
+      "step": 61634
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.035714285714285e-08,
+      "loss": 1.7488,
+      "step": 61635
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.030973451327434e-08,
+      "loss": 2.1177,
+      "step": 61636
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.026232616940582e-08,
+      "loss": 0.9168,
+      "step": 61637
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.02149178255373e-08,
+      "loss": 1.2829,
+      "step": 61638
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.016750948166878e-08,
+      "loss": 1.8877,
+      "step": 61639
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.012010113780026e-08,
+      "loss": 1.3962,
+      "step": 61640
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.007269279393172e-08,
+      "loss": 2.7552,
+      "step": 61641
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 8.002528445006322e-08,
+      "loss": 2.5938,
+      "step": 61642
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.99778761061947e-08,
+      "loss": 0.933,
+      "step": 61643
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.993046776232617e-08,
+      "loss": 2.3278,
+      "step": 61644
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.988305941845765e-08,
+      "loss": 3.0797,
+      "step": 61645
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.983565107458913e-08,
+      "loss": 2.766,
+      "step": 61646
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.978824273072061e-08,
+      "loss": 0.9403,
+      "step": 61647
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.974083438685209e-08,
+      "loss": 1.2485,
+      "step": 61648
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.969342604298357e-08,
+      "loss": 2.372,
+      "step": 61649
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.964601769911505e-08,
+      "loss": 2.0295,
+      "step": 61650
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 7.959860935524653e-08,
+      "loss": 0.8073,
+      "step": 61651
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.9551201011378e-08,
+      "loss": 4.0523,
+      "step": 61652
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.950379266750948e-08,
+      "loss": 2.6422,
+      "step": 61653
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.945638432364096e-08,
+      "loss": 2.6212,
+      "step": 61654
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.940897597977244e-08,
+      "loss": 1.3481,
+      "step": 61655
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.936156763590392e-08,
+      "loss": 3.2877,
+      "step": 61656
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.93141592920354e-08,
+      "loss": 0.5916,
+      "step": 61657
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.926675094816688e-08,
+      "loss": 0.8528,
+      "step": 61658
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.921934260429836e-08,
+      "loss": 1.0762,
+      "step": 61659
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.917193426042984e-08,
+      "loss": 0.9514,
+      "step": 61660
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.912452591656132e-08,
+      "loss": 0.5944,
+      "step": 61661
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.90771175726928e-08,
+      "loss": 1.3981,
+      "step": 61662
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.902970922882427e-08,
+      "loss": 2.1576,
+      "step": 61663
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.898230088495575e-08,
+      "loss": 2.0976,
+      "step": 61664
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.893489254108723e-08,
+      "loss": 1.5026,
+      "step": 61665
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.888748419721871e-08,
+      "loss": 0.8767,
+      "step": 61666
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.884007585335019e-08,
+      "loss": 3.364,
+      "step": 61667
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.879266750948167e-08,
+      "loss": 2.2902,
+      "step": 61668
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.874525916561315e-08,
+      "loss": 1.3673,
+      "step": 61669
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.869785082174463e-08,
+      "loss": 2.8581,
+      "step": 61670
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.86504424778761e-08,
+      "loss": 0.6322,
+      "step": 61671
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.860303413400758e-08,
+      "loss": 1.989,
+      "step": 61672
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.855562579013906e-08,
+      "loss": 1.505,
+      "step": 61673
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.850821744627054e-08,
+      "loss": 1.8046,
+      "step": 61674
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.846080910240203e-08,
+      "loss": 2.007,
+      "step": 61675
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.841340075853351e-08,
+      "loss": 2.496,
+      "step": 61676
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.836599241466498e-08,
+      "loss": 2.3934,
+      "step": 61677
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.831858407079646e-08,
+      "loss": 1.2861,
+      "step": 61678
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.827117572692794e-08,
+      "loss": 1.1091,
+      "step": 61679
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.822376738305942e-08,
+      "loss": 1.4815,
+      "step": 61680
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.817635903919091e-08,
+      "loss": 0.5043,
+      "step": 61681
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.812895069532239e-08,
+      "loss": 1.5697,
+      "step": 61682
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.808154235145385e-08,
+      "loss": 3.3771,
+      "step": 61683
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.803413400758533e-08,
+      "loss": 0.7565,
+      "step": 61684
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.798672566371681e-08,
+      "loss": 1.7452,
+      "step": 61685
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.79393173198483e-08,
+      "loss": 1.8634,
+      "step": 61686
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.789190897597978e-08,
+      "loss": 1.6067,
+      "step": 61687
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.784450063211126e-08,
+      "loss": 1.0835,
+      "step": 61688
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.779709228824273e-08,
+      "loss": 1.6718,
+      "step": 61689
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.77496839443742e-08,
+      "loss": 1.9472,
+      "step": 61690
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.770227560050568e-08,
+      "loss": 1.9386,
+      "step": 61691
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.765486725663718e-08,
+      "loss": 1.2776,
+      "step": 61692
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.760745891276865e-08,
+      "loss": 1.8364,
+      "step": 61693
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.756005056890013e-08,
+      "loss": 2.2191,
+      "step": 61694
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.75126422250316e-08,
+      "loss": 2.0327,
+      "step": 61695
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.746523388116308e-08,
+      "loss": 1.7225,
+      "step": 61696
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.741782553729456e-08,
+      "loss": 2.4835,
+      "step": 61697
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.737041719342605e-08,
+      "loss": 1.7897,
+      "step": 61698
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.732300884955753e-08,
+      "loss": 0.2323,
+      "step": 61699
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.727560050568901e-08,
+      "loss": 2.4264,
+      "step": 61700
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.722819216182049e-08,
+      "loss": 3.2894,
+      "step": 61701
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.718078381795195e-08,
+      "loss": 0.7666,
+      "step": 61702
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.713337547408344e-08,
+      "loss": 0.5318,
+      "step": 61703
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.708596713021492e-08,
+      "loss": 0.3743,
+      "step": 61704
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.70385587863464e-08,
+      "loss": 1.6771,
+      "step": 61705
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.699115044247788e-08,
+      "loss": 2.5952,
+      "step": 61706
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.694374209860936e-08,
+      "loss": 0.6872,
+      "step": 61707
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.689633375474083e-08,
+      "loss": 1.5004,
+      "step": 61708
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.684892541087232e-08,
+      "loss": 2.9904,
+      "step": 61709
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.68015170670038e-08,
+      "loss": 2.5473,
+      "step": 61710
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.675410872313528e-08,
+      "loss": 1.743,
+      "step": 61711
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.670670037926675e-08,
+      "loss": 0.3019,
+      "step": 61712
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.665929203539823e-08,
+      "loss": 1.3204,
+      "step": 61713
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.661188369152971e-08,
+      "loss": 0.4996,
+      "step": 61714
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 7.656447534766119e-08,
+      "loss": 1.8099,
+      "step": 61715
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.651706700379267e-08,
+      "loss": 2.5115,
+      "step": 61716
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.646965865992415e-08,
+      "loss": 1.6521,
+      "step": 61717
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.642225031605563e-08,
+      "loss": 1.4037,
+      "step": 61718
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.637484197218711e-08,
+      "loss": 2.2087,
+      "step": 61719
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.632743362831859e-08,
+      "loss": 1.2926,
+      "step": 61720
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.628002528445007e-08,
+      "loss": 1.8245,
+      "step": 61721
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.623261694058154e-08,
+      "loss": 1.1442,
+      "step": 61722
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.618520859671302e-08,
+      "loss": 2.0471,
+      "step": 61723
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.61378002528445e-08,
+      "loss": 3.0277,
+      "step": 61724
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.609039190897598e-08,
+      "loss": 3.033,
+      "step": 61725
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.604298356510746e-08,
+      "loss": 2.4046,
+      "step": 61726
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.599557522123894e-08,
+      "loss": 1.1277,
+      "step": 61727
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.594816687737042e-08,
+      "loss": 1.8708,
+      "step": 61728
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.59007585335019e-08,
+      "loss": 1.2556,
+      "step": 61729
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.585335018963338e-08,
+      "loss": 1.0942,
+      "step": 61730
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.580594184576485e-08,
+      "loss": 1.8783,
+      "step": 61731
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.575853350189633e-08,
+      "loss": 1.0386,
+      "step": 61732
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.571112515802781e-08,
+      "loss": 0.4765,
+      "step": 61733
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.566371681415929e-08,
+      "loss": 2.5506,
+      "step": 61734
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.561630847029077e-08,
+      "loss": 3.5671,
+      "step": 61735
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.556890012642225e-08,
+      "loss": 1.5912,
+      "step": 61736
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.552149178255374e-08,
+      "loss": 1.3003,
+      "step": 61737
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.547408343868521e-08,
+      "loss": 1.5608,
+      "step": 61738
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.542667509481669e-08,
+      "loss": 1.7385,
+      "step": 61739
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.537926675094817e-08,
+      "loss": 1.8987,
+      "step": 61740
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.533185840707964e-08,
+      "loss": 0.7212,
+      "step": 61741
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.528445006321114e-08,
+      "loss": 2.5293,
+      "step": 61742
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.523704171934262e-08,
+      "loss": 1.7434,
+      "step": 61743
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.518963337547408e-08,
+      "loss": 2.2807,
+      "step": 61744
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.514222503160556e-08,
+      "loss": 1.1225,
+      "step": 61745
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.509481668773704e-08,
+      "loss": 2.712,
+      "step": 61746
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.504740834386852e-08,
+      "loss": 1.6844,
+      "step": 61747
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.500000000000001e-08,
+      "loss": 1.0663,
+      "step": 61748
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.495259165613149e-08,
+      "loss": 1.5722,
+      "step": 61749
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.490518331226295e-08,
+      "loss": 2.0531,
+      "step": 61750
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.485777496839443e-08,
+      "loss": 0.8624,
+      "step": 61751
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.481036662452591e-08,
+      "loss": 0.6188,
+      "step": 61752
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.476295828065739e-08,
+      "loss": 2.5265,
+      "step": 61753
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.471554993678888e-08,
+      "loss": 2.2385,
+      "step": 61754
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.466814159292036e-08,
+      "loss": 2.1989,
+      "step": 61755
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.462073324905183e-08,
+      "loss": 0.9465,
+      "step": 61756
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.457332490518331e-08,
+      "loss": 2.2047,
+      "step": 61757
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.452591656131479e-08,
+      "loss": 2.6114,
+      "step": 61758
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.447850821744628e-08,
+      "loss": 1.239,
+      "step": 61759
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.443109987357776e-08,
+      "loss": 1.2658,
+      "step": 61760
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.438369152970924e-08,
+      "loss": 3.427,
+      "step": 61761
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.433628318584072e-08,
+      "loss": 0.6216,
+      "step": 61762
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.428887484197218e-08,
+      "loss": 1.3891,
+      "step": 61763
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.424146649810366e-08,
+      "loss": 1.3904,
+      "step": 61764
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.419405815423515e-08,
+      "loss": 1.8702,
+      "step": 61765
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.414664981036663e-08,
+      "loss": 2.6086,
+      "step": 61766
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.409924146649811e-08,
+      "loss": 2.0904,
+      "step": 61767
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.405183312262959e-08,
+      "loss": 0.8548,
+      "step": 61768
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.400442477876105e-08,
+      "loss": 0.8814,
+      "step": 61769
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.395701643489255e-08,
+      "loss": 1.9287,
+      "step": 61770
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.390960809102403e-08,
+      "loss": 1.9959,
+      "step": 61771
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.38621997471555e-08,
+      "loss": 0.8274,
+      "step": 61772
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.381479140328698e-08,
+      "loss": 2.0553,
+      "step": 61773
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.376738305941846e-08,
+      "loss": 3.1317,
+      "step": 61774
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.371997471554993e-08,
+      "loss": 1.2114,
+      "step": 61775
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.367256637168142e-08,
+      "loss": 2.2538,
+      "step": 61776
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.36251580278129e-08,
+      "loss": 3.2321,
+      "step": 61777
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 7.357774968394438e-08,
+      "loss": 1.3941,
+      "step": 61778
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.353034134007586e-08,
+      "loss": 1.4686,
+      "step": 61779
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.348293299620734e-08,
+      "loss": 1.9488,
+      "step": 61780
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.343552465233881e-08,
+      "loss": 2.592,
+      "step": 61781
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.33881163084703e-08,
+      "loss": 0.8746,
+      "step": 61782
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.334070796460177e-08,
+      "loss": 1.7586,
+      "step": 61783
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.329329962073325e-08,
+      "loss": 1.2169,
+      "step": 61784
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.324589127686473e-08,
+      "loss": 2.576,
+      "step": 61785
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.319848293299621e-08,
+      "loss": 1.1997,
+      "step": 61786
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.315107458912769e-08,
+      "loss": 0.8787,
+      "step": 61787
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.310366624525917e-08,
+      "loss": 0.9685,
+      "step": 61788
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.305625790139065e-08,
+      "loss": 1.9614,
+      "step": 61789
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.300884955752213e-08,
+      "loss": 1.4064,
+      "step": 61790
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.29614412136536e-08,
+      "loss": 1.9576,
+      "step": 61791
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.291403286978508e-08,
+      "loss": 2.3916,
+      "step": 61792
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.286662452591656e-08,
+      "loss": 2.5899,
+      "step": 61793
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.281921618204804e-08,
+      "loss": 1.5097,
+      "step": 61794
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.277180783817952e-08,
+      "loss": 1.1923,
+      "step": 61795
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.2724399494311e-08,
+      "loss": 1.7524,
+      "step": 61796
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.267699115044248e-08,
+      "loss": 2.2479,
+      "step": 61797
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.262958280657397e-08,
+      "loss": 3.1711,
+      "step": 61798
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.258217446270544e-08,
+      "loss": 1.4278,
+      "step": 61799
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.253476611883691e-08,
+      "loss": 2.4254,
+      "step": 61800
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.24873577749684e-08,
+      "loss": 2.1048,
+      "step": 61801
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.243994943109987e-08,
+      "loss": 2.6887,
+      "step": 61802
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.239254108723135e-08,
+      "loss": 3.5036,
+      "step": 61803
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.234513274336284e-08,
+      "loss": 1.9802,
+      "step": 61804
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.229772439949431e-08,
+      "loss": 2.3895,
+      "step": 61805
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.225031605562579e-08,
+      "loss": 2.7293,
+      "step": 61806
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.220290771175727e-08,
+      "loss": 1.7433,
+      "step": 61807
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.215549936788875e-08,
+      "loss": 0.1904,
+      "step": 61808
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.210809102402023e-08,
+      "loss": 1.4407,
+      "step": 61809
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.206068268015172e-08,
+      "loss": 1.0552,
+      "step": 61810
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.201327433628318e-08,
+      "loss": 0.6586,
+      "step": 61811
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.196586599241466e-08,
+      "loss": 1.5074,
+      "step": 61812
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.191845764854614e-08,
+      "loss": 2.8664,
+      "step": 61813
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.187104930467762e-08,
+      "loss": 1.7059,
+      "step": 61814
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.182364096080911e-08,
+      "loss": 1.0519,
+      "step": 61815
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.177623261694059e-08,
+      "loss": 2.316,
+      "step": 61816
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.172882427307206e-08,
+      "loss": 1.6868,
+      "step": 61817
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.168141592920354e-08,
+      "loss": 1.7685,
+      "step": 61818
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.163400758533501e-08,
+      "loss": 1.6988,
+      "step": 61819
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.15865992414665e-08,
+      "loss": 1.3592,
+      "step": 61820
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.153919089759799e-08,
+      "loss": 1.6956,
+      "step": 61821
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.149178255372946e-08,
+      "loss": 1.7298,
+      "step": 61822
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.144437420986094e-08,
+      "loss": 1.2706,
+      "step": 61823
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.139696586599241e-08,
+      "loss": 0.8029,
+      "step": 61824
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.134955752212389e-08,
+      "loss": 2.1499,
+      "step": 61825
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.130214917825538e-08,
+      "loss": 2.274,
+      "step": 61826
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.125474083438686e-08,
+      "loss": 0.3558,
+      "step": 61827
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.120733249051834e-08,
+      "loss": 0.8786,
+      "step": 61828
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.115992414664982e-08,
+      "loss": 1.3074,
+      "step": 61829
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.111251580278128e-08,
+      "loss": 2.4371,
+      "step": 61830
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.106510745891276e-08,
+      "loss": 2.1301,
+      "step": 61831
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.101769911504425e-08,
+      "loss": 1.7959,
+      "step": 61832
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.097029077117573e-08,
+      "loss": 1.817,
+      "step": 61833
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.092288242730721e-08,
+      "loss": 2.012,
+      "step": 61834
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.087547408343869e-08,
+      "loss": 0.9894,
+      "step": 61835
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.082806573957016e-08,
+      "loss": 3.0548,
+      "step": 61836
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.078065739570164e-08,
+      "loss": 0.755,
+      "step": 61837
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.073324905183313e-08,
+      "loss": 2.0079,
+      "step": 61838
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.068584070796461e-08,
+      "loss": 0.7768,
+      "step": 61839
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.063843236409609e-08,
+      "loss": 0.9103,
+      "step": 61840
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 7.059102402022756e-08,
+      "loss": 2.0256,
+      "step": 61841
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.054361567635904e-08,
+      "loss": 1.3139,
+      "step": 61842
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.049620733249052e-08,
+      "loss": 1.086,
+      "step": 61843
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.0448798988622e-08,
+      "loss": 1.3102,
+      "step": 61844
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.040139064475348e-08,
+      "loss": 2.078,
+      "step": 61845
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.035398230088496e-08,
+      "loss": 3.3351,
+      "step": 61846
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.030657395701644e-08,
+      "loss": 0.9299,
+      "step": 61847
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.025916561314792e-08,
+      "loss": 0.8733,
+      "step": 61848
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.02117572692794e-08,
+      "loss": 2.0438,
+      "step": 61849
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.016434892541088e-08,
+      "loss": 1.5541,
+      "step": 61850
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.011694058154235e-08,
+      "loss": 1.4772,
+      "step": 61851
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.006953223767383e-08,
+      "loss": 2.7163,
+      "step": 61852
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 7.002212389380531e-08,
+      "loss": 1.3112,
+      "step": 61853
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.997471554993679e-08,
+      "loss": 1.1514,
+      "step": 61854
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.992730720606827e-08,
+      "loss": 3.129,
+      "step": 61855
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.987989886219975e-08,
+      "loss": 0.6688,
+      "step": 61856
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.983249051833123e-08,
+      "loss": 0.98,
+      "step": 61857
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.978508217446271e-08,
+      "loss": 1.8403,
+      "step": 61858
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.973767383059419e-08,
+      "loss": 1.2078,
+      "step": 61859
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.969026548672566e-08,
+      "loss": 1.4937,
+      "step": 61860
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.964285714285714e-08,
+      "loss": 2.3755,
+      "step": 61861
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.959544879898862e-08,
+      "loss": 1.5425,
+      "step": 61862
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.95480404551201e-08,
+      "loss": 1.9006,
+      "step": 61863
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.950063211125158e-08,
+      "loss": 2.0796,
+      "step": 61864
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.945322376738306e-08,
+      "loss": 0.6019,
+      "step": 61865
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.940581542351454e-08,
+      "loss": 3.4607,
+      "step": 61866
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.935840707964602e-08,
+      "loss": 2.7891,
+      "step": 61867
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.93109987357775e-08,
+      "loss": 0.8831,
+      "step": 61868
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.926359039190898e-08,
+      "loss": 0.9603,
+      "step": 61869
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.921618204804045e-08,
+      "loss": 1.1909,
+      "step": 61870
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.916877370417195e-08,
+      "loss": 3.1567,
+      "step": 61871
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.912136536030341e-08,
+      "loss": 1.279,
+      "step": 61872
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.907395701643489e-08,
+      "loss": 2.4561,
+      "step": 61873
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.902654867256637e-08,
+      "loss": 1.8515,
+      "step": 61874
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.897914032869785e-08,
+      "loss": 2.439,
+      "step": 61875
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.893173198482933e-08,
+      "loss": 1.4152,
+      "step": 61876
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.888432364096082e-08,
+      "loss": 1.9026,
+      "step": 61877
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.883691529709229e-08,
+      "loss": 1.757,
+      "step": 61878
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.878950695322376e-08,
+      "loss": 2.5506,
+      "step": 61879
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.874209860935524e-08,
+      "loss": 1.3649,
+      "step": 61880
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.869469026548672e-08,
+      "loss": 2.3011,
+      "step": 61881
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.864728192161821e-08,
+      "loss": 2.0615,
+      "step": 61882
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.85998735777497e-08,
+      "loss": 2.5959,
+      "step": 61883
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.855246523388117e-08,
+      "loss": 0.8728,
+      "step": 61884
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.850505689001264e-08,
+      "loss": 2.704,
+      "step": 61885
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.845764854614412e-08,
+      "loss": 0.3914,
+      "step": 61886
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.84102402022756e-08,
+      "loss": 2.5519,
+      "step": 61887
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.836283185840709e-08,
+      "loss": 1.9488,
+      "step": 61888
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.831542351453857e-08,
+      "loss": 1.934,
+      "step": 61889
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.826801517067005e-08,
+      "loss": 1.5375,
+      "step": 61890
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.822060682680151e-08,
+      "loss": 1.6628,
+      "step": 61891
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.817319848293299e-08,
+      "loss": 2.3646,
+      "step": 61892
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.812579013906447e-08,
+      "loss": 2.2507,
+      "step": 61893
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.807838179519596e-08,
+      "loss": 2.0518,
+      "step": 61894
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.803097345132744e-08,
+      "loss": 1.8072,
+      "step": 61895
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.798356510745892e-08,
+      "loss": 1.2094,
+      "step": 61896
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.793615676359039e-08,
+      "loss": 1.989,
+      "step": 61897
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.788874841972186e-08,
+      "loss": 0.739,
+      "step": 61898
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.784134007585336e-08,
+      "loss": 3.8988,
+      "step": 61899
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.779393173198484e-08,
+      "loss": 1.9803,
+      "step": 61900
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.774652338811631e-08,
+      "loss": 1.527,
+      "step": 61901
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.769911504424779e-08,
+      "loss": 2.6507,
+      "step": 61902
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.765170670037927e-08,
+      "loss": 1.0538,
+      "step": 61903
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.760429835651074e-08,
+      "loss": 2.8206,
+      "step": 61904
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 6.755689001264223e-08,
+      "loss": 1.6131,
+      "step": 61905
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.750948166877371e-08,
+      "loss": 1.2022,
+      "step": 61906
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.746207332490519e-08,
+      "loss": 1.3015,
+      "step": 61907
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.741466498103667e-08,
+      "loss": 2.7346,
+      "step": 61908
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.736725663716815e-08,
+      "loss": 2.6481,
+      "step": 61909
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.731984829329962e-08,
+      "loss": 2.1511,
+      "step": 61910
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.72724399494311e-08,
+      "loss": 1.1766,
+      "step": 61911
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.722503160556258e-08,
+      "loss": 2.1173,
+      "step": 61912
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.717762326169406e-08,
+      "loss": 1.0396,
+      "step": 61913
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.713021491782554e-08,
+      "loss": 2.4189,
+      "step": 61914
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.708280657395702e-08,
+      "loss": 2.1242,
+      "step": 61915
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.70353982300885e-08,
+      "loss": 2.7236,
+      "step": 61916
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.698798988621998e-08,
+      "loss": 2.9755,
+      "step": 61917
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.694058154235146e-08,
+      "loss": 2.2815,
+      "step": 61918
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.689317319848294e-08,
+      "loss": 2.0218,
+      "step": 61919
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.684576485461441e-08,
+      "loss": 2.5369,
+      "step": 61920
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.679835651074589e-08,
+      "loss": 4.3829,
+      "step": 61921
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.675094816687737e-08,
+      "loss": 1.2874,
+      "step": 61922
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.670353982300885e-08,
+      "loss": 0.2821,
+      "step": 61923
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.665613147914033e-08,
+      "loss": 2.2654,
+      "step": 61924
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.660872313527181e-08,
+      "loss": 2.2986,
+      "step": 61925
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.656131479140329e-08,
+      "loss": 3.0259,
+      "step": 61926
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.651390644753477e-08,
+      "loss": 1.3227,
+      "step": 61927
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.646649810366625e-08,
+      "loss": 2.7895,
+      "step": 61928
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.641908975979772e-08,
+      "loss": 1.4453,
+      "step": 61929
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.63716814159292e-08,
+      "loss": 1.724,
+      "step": 61930
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.632427307206068e-08,
+      "loss": 1.6152,
+      "step": 61931
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.627686472819216e-08,
+      "loss": 1.6679,
+      "step": 61932
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.622945638432364e-08,
+      "loss": 0.8582,
+      "step": 61933
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.618204804045512e-08,
+      "loss": 1.5919,
+      "step": 61934
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.61346396965866e-08,
+      "loss": 2.7662,
+      "step": 61935
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.608723135271808e-08,
+      "loss": 1.2032,
+      "step": 61936
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.603982300884956e-08,
+      "loss": 1.7408,
+      "step": 61937
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.599241466498105e-08,
+      "loss": 1.5366,
+      "step": 61938
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.594500632111251e-08,
+      "loss": 1.6651,
+      "step": 61939
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.589759797724399e-08,
+      "loss": 2.1782,
+      "step": 61940
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.585018963337547e-08,
+      "loss": 2.1326,
+      "step": 61941
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.580278128950695e-08,
+      "loss": 3.1278,
+      "step": 61942
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.575537294563843e-08,
+      "loss": 2.3289,
+      "step": 61943
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.570796460176992e-08,
+      "loss": 1.5117,
+      "step": 61944
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.56605562579014e-08,
+      "loss": 1.6379,
+      "step": 61945
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.561314791403287e-08,
+      "loss": 2.369,
+      "step": 61946
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.556573957016435e-08,
+      "loss": 3.384,
+      "step": 61947
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.551833122629582e-08,
+      "loss": 2.8813,
+      "step": 61948
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.54709228824273e-08,
+      "loss": 1.6541,
+      "step": 61949
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.54235145385588e-08,
+      "loss": 2.5705,
+      "step": 61950
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.537610619469027e-08,
+      "loss": 2.0336,
+      "step": 61951
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.532869785082174e-08,
+      "loss": 2.1766,
+      "step": 61952
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.528128950695322e-08,
+      "loss": 2.3614,
+      "step": 61953
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.52338811630847e-08,
+      "loss": 0.6707,
+      "step": 61954
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.518647281921619e-08,
+      "loss": 3.8567,
+      "step": 61955
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.513906447534767e-08,
+      "loss": 1.0144,
+      "step": 61956
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.509165613147915e-08,
+      "loss": 1.2488,
+      "step": 61957
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.504424778761061e-08,
+      "loss": 1.1627,
+      "step": 61958
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.499683944374209e-08,
+      "loss": 2.5383,
+      "step": 61959
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.494943109987357e-08,
+      "loss": 0.6311,
+      "step": 61960
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.490202275600506e-08,
+      "loss": 1.4647,
+      "step": 61961
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.485461441213654e-08,
+      "loss": 1.7342,
+      "step": 61962
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.480720606826802e-08,
+      "loss": 2.1942,
+      "step": 61963
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.475979772439949e-08,
+      "loss": 4.0556,
+      "step": 61964
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.471238938053097e-08,
+      "loss": 0.7413,
+      "step": 61965
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.466498103666246e-08,
+      "loss": 1.7084,
+      "step": 61966
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.461757269279394e-08,
+      "loss": 2.3304,
+      "step": 61967
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 6.457016434892542e-08,
+      "loss": 1.637,
+      "step": 61968
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.45227560050569e-08,
+      "loss": 2.4728,
+      "step": 61969
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.447534766118837e-08,
+      "loss": 2.619,
+      "step": 61970
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.442793931731984e-08,
+      "loss": 2.561,
+      "step": 61971
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.438053097345133e-08,
+      "loss": 4.2261,
+      "step": 61972
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.433312262958281e-08,
+      "loss": 0.7463,
+      "step": 61973
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.428571428571429e-08,
+      "loss": 2.2656,
+      "step": 61974
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.423830594184577e-08,
+      "loss": 3.9865,
+      "step": 61975
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.419089759797725e-08,
+      "loss": 1.5408,
+      "step": 61976
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.414348925410871e-08,
+      "loss": 0.9244,
+      "step": 61977
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.40960809102402e-08,
+      "loss": 2.8748,
+      "step": 61978
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.404867256637169e-08,
+      "loss": 2.8936,
+      "step": 61979
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.400126422250316e-08,
+      "loss": 1.5358,
+      "step": 61980
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.395385587863464e-08,
+      "loss": 1.9104,
+      "step": 61981
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.390644753476612e-08,
+      "loss": 2.2381,
+      "step": 61982
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.38590391908976e-08,
+      "loss": 0.8805,
+      "step": 61983
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.381163084702908e-08,
+      "loss": 1.774,
+      "step": 61984
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.376422250316056e-08,
+      "loss": 1.167,
+      "step": 61985
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.371681415929204e-08,
+      "loss": 2.4609,
+      "step": 61986
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.366940581542352e-08,
+      "loss": 1.585,
+      "step": 61987
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.3621997471555e-08,
+      "loss": 1.5752,
+      "step": 61988
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.357458912768647e-08,
+      "loss": 3.7873,
+      "step": 61989
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.352718078381795e-08,
+      "loss": 2.3335,
+      "step": 61990
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.347977243994943e-08,
+      "loss": 0.7211,
+      "step": 61991
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.343236409608091e-08,
+      "loss": 1.9241,
+      "step": 61992
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.338495575221239e-08,
+      "loss": 2.559,
+      "step": 61993
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.333754740834387e-08,
+      "loss": 2.1966,
+      "step": 61994
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.329013906447535e-08,
+      "loss": 2.0281,
+      "step": 61995
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.324273072060683e-08,
+      "loss": 1.7324,
+      "step": 61996
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.31953223767383e-08,
+      "loss": 2.2387,
+      "step": 61997
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.314791403286978e-08,
+      "loss": 1.1075,
+      "step": 61998
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.310050568900126e-08,
+      "loss": 2.4556,
+      "step": 61999
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.305309734513274e-08,
+      "loss": 2.3788,
+      "step": 62000
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.300568900126422e-08,
+      "loss": 0.6986,
+      "step": 62001
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.29582806573957e-08,
+      "loss": 1.7332,
+      "step": 62002
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.291087231352718e-08,
+      "loss": 0.9003,
+      "step": 62003
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.286346396965866e-08,
+      "loss": 2.2098,
+      "step": 62004
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.281605562579014e-08,
+      "loss": 0.9402,
+      "step": 62005
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.276864728192163e-08,
+      "loss": 2.822,
+      "step": 62006
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.27212389380531e-08,
+      "loss": 1.8866,
+      "step": 62007
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.267383059418457e-08,
+      "loss": 2.2602,
+      "step": 62008
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.262642225031605e-08,
+      "loss": 0.9328,
+      "step": 62009
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.257901390644753e-08,
+      "loss": 1.9959,
+      "step": 62010
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.253160556257902e-08,
+      "loss": 2.0435,
+      "step": 62011
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.24841972187105e-08,
+      "loss": 3.2073,
+      "step": 62012
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.243678887484197e-08,
+      "loss": 1.7754,
+      "step": 62013
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.238938053097345e-08,
+      "loss": 3.3785,
+      "step": 62014
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.234197218710493e-08,
+      "loss": 1.2984,
+      "step": 62015
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.22945638432364e-08,
+      "loss": 1.3643,
+      "step": 62016
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.22471554993679e-08,
+      "loss": 0.5673,
+      "step": 62017
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.219974715549938e-08,
+      "loss": 1.2611,
+      "step": 62018
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.215233881163084e-08,
+      "loss": 0.8197,
+      "step": 62019
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.210493046776232e-08,
+      "loss": 1.6663,
+      "step": 62020
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.20575221238938e-08,
+      "loss": 2.3054,
+      "step": 62021
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.201011378002529e-08,
+      "loss": 2.3416,
+      "step": 62022
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.196270543615677e-08,
+      "loss": 1.2024,
+      "step": 62023
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.191529709228825e-08,
+      "loss": 1.9643,
+      "step": 62024
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.186788874841972e-08,
+      "loss": 3.1486,
+      "step": 62025
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.18204804045512e-08,
+      "loss": 2.731,
+      "step": 62026
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.177307206068267e-08,
+      "loss": 1.1928,
+      "step": 62027
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.172566371681417e-08,
+      "loss": 1.933,
+      "step": 62028
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.167825537294565e-08,
+      "loss": 2.3685,
+      "step": 62029
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.163084702907712e-08,
+      "loss": 2.8455,
+      "step": 62030
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 6.15834386852086e-08,
+      "loss": 1.2965,
+      "step": 62031
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.153603034134007e-08,
+      "loss": 2.5533,
+      "step": 62032
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.148862199747155e-08,
+      "loss": 1.2732,
+      "step": 62033
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.144121365360304e-08,
+      "loss": 0.4719,
+      "step": 62034
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.139380530973452e-08,
+      "loss": 1.6461,
+      "step": 62035
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.1346396965866e-08,
+      "loss": 3.25,
+      "step": 62036
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.129898862199748e-08,
+      "loss": 2.0531,
+      "step": 62037
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.125158027812894e-08,
+      "loss": 1.7182,
+      "step": 62038
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.120417193426043e-08,
+      "loss": 1.923,
+      "step": 62039
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.115676359039191e-08,
+      "loss": 1.1881,
+      "step": 62040
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.110935524652339e-08,
+      "loss": 1.3984,
+      "step": 62041
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.106194690265487e-08,
+      "loss": 1.4504,
+      "step": 62042
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.101453855878635e-08,
+      "loss": 2.989,
+      "step": 62043
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.096713021491782e-08,
+      "loss": 0.7829,
+      "step": 62044
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.091972187104931e-08,
+      "loss": 1.1967,
+      "step": 62045
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.087231352718079e-08,
+      "loss": 1.7973,
+      "step": 62046
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.082490518331227e-08,
+      "loss": 3.0377,
+      "step": 62047
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.077749683944375e-08,
+      "loss": 2.9749,
+      "step": 62048
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.073008849557522e-08,
+      "loss": 3.0108,
+      "step": 62049
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.06826801517067e-08,
+      "loss": 1.4845,
+      "step": 62050
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.063527180783818e-08,
+      "loss": 2.0349,
+      "step": 62051
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.058786346396966e-08,
+      "loss": 2.11,
+      "step": 62052
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.054045512010114e-08,
+      "loss": 2.0714,
+      "step": 62053
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.049304677623262e-08,
+      "loss": 1.898,
+      "step": 62054
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.04456384323641e-08,
+      "loss": 1.0375,
+      "step": 62055
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.039823008849558e-08,
+      "loss": 1.211,
+      "step": 62056
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.035082174462706e-08,
+      "loss": 1.6729,
+      "step": 62057
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.030341340075853e-08,
+      "loss": 2.0754,
+      "step": 62058
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.025600505689001e-08,
+      "loss": 2.5415,
+      "step": 62059
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.020859671302149e-08,
+      "loss": 0.5846,
+      "step": 62060
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.016118836915297e-08,
+      "loss": 1.8844,
+      "step": 62061
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.011378002528445e-08,
+      "loss": 0.3672,
+      "step": 62062
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.006637168141593e-08,
+      "loss": 2.3982,
+      "step": 62063
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 6.001896333754741e-08,
+      "loss": 2.1383,
+      "step": 62064
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.997155499367889e-08,
+      "loss": 1.7942,
+      "step": 62065
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.992414664981037e-08,
+      "loss": 0.7674,
+      "step": 62066
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.987673830594186e-08,
+      "loss": 1.884,
+      "step": 62067
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.982932996207332e-08,
+      "loss": 2.5634,
+      "step": 62068
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.97819216182048e-08,
+      "loss": 1.6837,
+      "step": 62069
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.973451327433628e-08,
+      "loss": 0.8393,
+      "step": 62070
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.968710493046776e-08,
+      "loss": 1.2811,
+      "step": 62071
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.963969658659924e-08,
+      "loss": 1.6554,
+      "step": 62072
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.9592288242730725e-08,
+      "loss": 1.0212,
+      "step": 62073
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.9544879898862204e-08,
+      "loss": 2.6067,
+      "step": 62074
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.9497471554993683e-08,
+      "loss": 1.9951,
+      "step": 62075
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.9450063211125156e-08,
+      "loss": 1.0899,
+      "step": 62076
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.9402654867256635e-08,
+      "loss": 2.4642,
+      "step": 62077
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.935524652338812e-08,
+      "loss": 1.2156,
+      "step": 62078
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.93078381795196e-08,
+      "loss": 1.6698,
+      "step": 62079
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.926042983565108e-08,
+      "loss": 1.9348,
+      "step": 62080
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.921302149178256e-08,
+      "loss": 1.8859,
+      "step": 62081
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.916561314791403e-08,
+      "loss": 2.2128,
+      "step": 62082
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.911820480404551e-08,
+      "loss": 0.961,
+      "step": 62083
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.9070796460176994e-08,
+      "loss": 1.2307,
+      "step": 62084
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.902338811630847e-08,
+      "loss": 2.7037,
+      "step": 62085
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.897597977243995e-08,
+      "loss": 2.5785,
+      "step": 62086
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.892857142857143e-08,
+      "loss": 2.0965,
+      "step": 62087
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.88811630847029e-08,
+      "loss": 1.1535,
+      "step": 62088
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.883375474083438e-08,
+      "loss": 1.0235,
+      "step": 62089
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.8786346396965874e-08,
+      "loss": 2.7248,
+      "step": 62090
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.8738938053097346e-08,
+      "loss": 1.7018,
+      "step": 62091
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.8691529709228825e-08,
+      "loss": 1.6048,
+      "step": 62092
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.8644121365360304e-08,
+      "loss": 1.9968,
+      "step": 62093
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.859671302149178e-08,
+      "loss": 0.8508,
+      "step": 62094
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 5.854930467762327e-08,
+      "loss": 2.7801,
+      "step": 62095
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.850189633375475e-08,
+      "loss": 0.432,
+      "step": 62096
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.845448798988622e-08,
+      "loss": 0.9863,
+      "step": 62097
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.84070796460177e-08,
+      "loss": 1.9767,
+      "step": 62098
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.835967130214918e-08,
+      "loss": 1.7574,
+      "step": 62099
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.831226295828066e-08,
+      "loss": 1.2664,
+      "step": 62100
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.826485461441214e-08,
+      "loss": 0.5969,
+      "step": 62101
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.821744627054362e-08,
+      "loss": 0.8832,
+      "step": 62102
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.8170037926675094e-08,
+      "loss": 2.2919,
+      "step": 62103
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.812262958280657e-08,
+      "loss": 1.0568,
+      "step": 62104
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.807522123893805e-08,
+      "loss": 1.6233,
+      "step": 62105
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.802781289506954e-08,
+      "loss": 1.7819,
+      "step": 62106
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7980404551201016e-08,
+      "loss": 1.6344,
+      "step": 62107
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7932996207332495e-08,
+      "loss": 2.2969,
+      "step": 62108
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7885587863463974e-08,
+      "loss": 1.6236,
+      "step": 62109
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7838179519595446e-08,
+      "loss": 0.5419,
+      "step": 62110
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7790771175726925e-08,
+      "loss": 2.4982,
+      "step": 62111
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.774336283185841e-08,
+      "loss": 2.6889,
+      "step": 62112
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.769595448798989e-08,
+      "loss": 0.8869,
+      "step": 62113
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.764854614412137e-08,
+      "loss": 2.3476,
+      "step": 62114
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.760113780025285e-08,
+      "loss": 0.5995,
+      "step": 62115
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.755372945638432e-08,
+      "loss": 2.9987,
+      "step": 62116
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.75063211125158e-08,
+      "loss": 1.9622,
+      "step": 62117
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7458912768647284e-08,
+      "loss": 1.9298,
+      "step": 62118
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7411504424778763e-08,
+      "loss": 1.6471,
+      "step": 62119
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.736409608091024e-08,
+      "loss": 1.9104,
+      "step": 62120
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.731668773704172e-08,
+      "loss": 2.5151,
+      "step": 62121
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7269279393173194e-08,
+      "loss": 2.6291,
+      "step": 62122
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7221871049304686e-08,
+      "loss": 1.1463,
+      "step": 62123
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.717446270543616e-08,
+      "loss": 2.2836,
+      "step": 62124
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.712705436156764e-08,
+      "loss": 2.1807,
+      "step": 62125
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7079646017699116e-08,
+      "loss": 2.307,
+      "step": 62126
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.7032237673830595e-08,
+      "loss": 1.8617,
+      "step": 62127
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.698482932996207e-08,
+      "loss": 2.4387,
+      "step": 62128
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.693742098609356e-08,
+      "loss": 1.1565,
+      "step": 62129
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.689001264222504e-08,
+      "loss": 2.199,
+      "step": 62130
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.684260429835651e-08,
+      "loss": 3.0309,
+      "step": 62131
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.679519595448799e-08,
+      "loss": 1.4438,
+      "step": 62132
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.674778761061947e-08,
+      "loss": 2.242,
+      "step": 62133
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.6700379266750954e-08,
+      "loss": 0.9264,
+      "step": 62134
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.665297092288243e-08,
+      "loss": 0.6115,
+      "step": 62135
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.660556257901391e-08,
+      "loss": 2.6986,
+      "step": 62136
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.6558154235145384e-08,
+      "loss": 2.0417,
+      "step": 62137
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.651074589127686e-08,
+      "loss": 0.9028,
+      "step": 62138
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.646333754740834e-08,
+      "loss": 1.3136,
+      "step": 62139
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.641592920353983e-08,
+      "loss": 1.12,
+      "step": 62140
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.636852085967131e-08,
+      "loss": 0.8532,
+      "step": 62141
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.6321112515802786e-08,
+      "loss": 1.214,
+      "step": 62142
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.627370417193426e-08,
+      "loss": 1.7401,
+      "step": 62143
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.622629582806574e-08,
+      "loss": 1.2646,
+      "step": 62144
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.6178887484197216e-08,
+      "loss": 2.5844,
+      "step": 62145
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.61314791403287e-08,
+      "loss": 3.4821,
+      "step": 62146
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.608407079646018e-08,
+      "loss": 1.0092,
+      "step": 62147
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.603666245259166e-08,
+      "loss": 1.7958,
+      "step": 62148
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.598925410872313e-08,
+      "loss": 1.8764,
+      "step": 62149
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.594184576485461e-08,
+      "loss": 0.9268,
+      "step": 62150
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.58944374209861e-08,
+      "loss": 1.521,
+      "step": 62151
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.5847029077117575e-08,
+      "loss": 1.4238,
+      "step": 62152
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.5799620733249054e-08,
+      "loss": 2.5602,
+      "step": 62153
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.575221238938053e-08,
+      "loss": 1.151,
+      "step": 62154
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.570480404551201e-08,
+      "loss": 1.0217,
+      "step": 62155
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.5657395701643484e-08,
+      "loss": 1.941,
+      "step": 62156
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.5609987357774976e-08,
+      "loss": 1.7306,
+      "step": 62157
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 5.556257901390645e-08,
+      "loss": 1.1277,
+      "step": 62158
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.551517067003793e-08,
+      "loss": 1.7347,
+      "step": 62159
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.5467762326169407e-08,
+      "loss": 2.0498,
+      "step": 62160
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.5420353982300886e-08,
+      "loss": 1.0784,
+      "step": 62161
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.537294563843237e-08,
+      "loss": 1.2506,
+      "step": 62162
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.532553729456385e-08,
+      "loss": 1.3005,
+      "step": 62163
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.527812895069532e-08,
+      "loss": 0.9037,
+      "step": 62164
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.52307206068268e-08,
+      "loss": 0.9515,
+      "step": 62165
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.518331226295828e-08,
+      "loss": 0.6135,
+      "step": 62166
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.513590391908976e-08,
+      "loss": 2.5193,
+      "step": 62167
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.5088495575221245e-08,
+      "loss": 1.528,
+      "step": 62168
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.5041087231352724e-08,
+      "loss": 0.4653,
+      "step": 62169
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.49936788874842e-08,
+      "loss": 2.3058,
+      "step": 62170
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4946270543615675e-08,
+      "loss": 0.5829,
+      "step": 62171
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4898862199747154e-08,
+      "loss": 3.0006,
+      "step": 62172
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.485145385587863e-08,
+      "loss": 1.2246,
+      "step": 62173
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.480404551201012e-08,
+      "loss": 1.5057,
+      "step": 62174
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.47566371681416e-08,
+      "loss": 1.2865,
+      "step": 62175
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4709228824273076e-08,
+      "loss": 1.9443,
+      "step": 62176
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.466182048040455e-08,
+      "loss": 1.271,
+      "step": 62177
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.461441213653603e-08,
+      "loss": 1.7538,
+      "step": 62178
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.456700379266751e-08,
+      "loss": 2.7862,
+      "step": 62179
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.451959544879899e-08,
+      "loss": 1.8905,
+      "step": 62180
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.447218710493047e-08,
+      "loss": 1.5246,
+      "step": 62181
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.442477876106195e-08,
+      "loss": 1.1626,
+      "step": 62182
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.437737041719342e-08,
+      "loss": 1.4876,
+      "step": 62183
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.43299620733249e-08,
+      "loss": 1.1397,
+      "step": 62184
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.428255372945639e-08,
+      "loss": 1.6944,
+      "step": 62185
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4235145385587866e-08,
+      "loss": 1.5731,
+      "step": 62186
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4187737041719345e-08,
+      "loss": 2.0744,
+      "step": 62187
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4140328697850824e-08,
+      "loss": 1.317,
+      "step": 62188
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.4092920353982296e-08,
+      "loss": 1.1406,
+      "step": 62189
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.404551201011379e-08,
+      "loss": 0.5806,
+      "step": 62190
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.399810366624527e-08,
+      "loss": 1.8713,
+      "step": 62191
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.395069532237674e-08,
+      "loss": 0.9582,
+      "step": 62192
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.390328697850822e-08,
+      "loss": 1.4366,
+      "step": 62193
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.38558786346397e-08,
+      "loss": 1.8082,
+      "step": 62194
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.3808470290771176e-08,
+      "loss": 1.453,
+      "step": 62195
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.376106194690266e-08,
+      "loss": 1.2717,
+      "step": 62196
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.371365360303414e-08,
+      "loss": 1.0859,
+      "step": 62197
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.366624525916561e-08,
+      "loss": 2.7484,
+      "step": 62198
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.361883691529709e-08,
+      "loss": 1.2008,
+      "step": 62199
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.357142857142857e-08,
+      "loss": 2.9674,
+      "step": 62200
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.352402022756005e-08,
+      "loss": 2.8566,
+      "step": 62201
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.3476611883691535e-08,
+      "loss": 2.6576,
+      "step": 62202
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.3429203539823014e-08,
+      "loss": 2.0162,
+      "step": 62203
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.3381795195954487e-08,
+      "loss": 1.2238,
+      "step": 62204
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.3334386852085966e-08,
+      "loss": 1.4977,
+      "step": 62205
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.3286978508217445e-08,
+      "loss": 2.2421,
+      "step": 62206
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.323957016434893e-08,
+      "loss": 2.5847,
+      "step": 62207
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.319216182048041e-08,
+      "loss": 1.3255,
+      "step": 62208
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.314475347661189e-08,
+      "loss": 1.463,
+      "step": 62209
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.309734513274336e-08,
+      "loss": 2.2575,
+      "step": 62210
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.304993678887484e-08,
+      "loss": 2.579,
+      "step": 62211
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.300252844500632e-08,
+      "loss": 1.425,
+      "step": 62212
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.2955120101137804e-08,
+      "loss": 2.0706,
+      "step": 62213
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.290771175726928e-08,
+      "loss": 2.2521,
+      "step": 62214
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.286030341340076e-08,
+      "loss": 2.7033,
+      "step": 62215
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.281289506953224e-08,
+      "loss": 3.1397,
+      "step": 62216
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.276548672566371e-08,
+      "loss": 1.5235,
+      "step": 62217
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.271807838179519e-08,
+      "loss": 1.6818,
+      "step": 62218
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.267067003792668e-08,
+      "loss": 1.1014,
+      "step": 62219
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.2623261694058156e-08,
+      "loss": 2.2818,
+      "step": 62220
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 5.2575853350189635e-08,
+      "loss": 1.4624,
+      "step": 62221
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.2528445006321114e-08,
+      "loss": 2.3831,
+      "step": 62222
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.2481036662452587e-08,
+      "loss": 2.0515,
+      "step": 62223
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.243362831858408e-08,
+      "loss": 2.2745,
+      "step": 62224
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.238621997471555e-08,
+      "loss": 2.0308,
+      "step": 62225
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.233881163084703e-08,
+      "loss": 2.5433,
+      "step": 62226
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.229140328697851e-08,
+      "loss": 2.7372,
+      "step": 62227
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.224399494310999e-08,
+      "loss": 1.0384,
+      "step": 62228
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.219658659924146e-08,
+      "loss": 0.427,
+      "step": 62229
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.214917825537295e-08,
+      "loss": 1.6053,
+      "step": 62230
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.2101769911504425e-08,
+      "loss": 3.3478,
+      "step": 62231
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.2054361567635904e-08,
+      "loss": 1.8343,
+      "step": 62232
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.200695322376738e-08,
+      "loss": 1.4112,
+      "step": 62233
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.195954487989886e-08,
+      "loss": 3.0269,
+      "step": 62234
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.191213653603035e-08,
+      "loss": 2.0754,
+      "step": 62235
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1864728192161826e-08,
+      "loss": 2.8004,
+      "step": 62236
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1817319848293305e-08,
+      "loss": 2.1794,
+      "step": 62237
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.176991150442478e-08,
+      "loss": 2.8939,
+      "step": 62238
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1722503160556256e-08,
+      "loss": 3.3275,
+      "step": 62239
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1675094816687735e-08,
+      "loss": 1.0351,
+      "step": 62240
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.162768647281922e-08,
+      "loss": 1.4761,
+      "step": 62241
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.15802781289507e-08,
+      "loss": 2.3749,
+      "step": 62242
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.153286978508218e-08,
+      "loss": 2.6802,
+      "step": 62243
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.148546144121365e-08,
+      "loss": 2.3186,
+      "step": 62244
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.143805309734513e-08,
+      "loss": 1.1342,
+      "step": 62245
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.139064475347661e-08,
+      "loss": 1.0409,
+      "step": 62246
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1343236409608094e-08,
+      "loss": 3.8441,
+      "step": 62247
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1295828065739573e-08,
+      "loss": 0.7022,
+      "step": 62248
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.124841972187105e-08,
+      "loss": 2.2175,
+      "step": 62249
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1201011378002525e-08,
+      "loss": 2.0038,
+      "step": 62250
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1153603034134004e-08,
+      "loss": 1.1817,
+      "step": 62251
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.1106194690265496e-08,
+      "loss": 2.0997,
+      "step": 62252
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.105878634639697e-08,
+      "loss": 1.8901,
+      "step": 62253
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.101137800252845e-08,
+      "loss": 3.1235,
+      "step": 62254
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0963969658659926e-08,
+      "loss": 2.0701,
+      "step": 62255
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0916561314791405e-08,
+      "loss": 1.5517,
+      "step": 62256
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.086915297092288e-08,
+      "loss": 3.1847,
+      "step": 62257
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.082174462705437e-08,
+      "loss": 1.0943,
+      "step": 62258
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.077433628318584e-08,
+      "loss": 3.1238,
+      "step": 62259
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.072692793931732e-08,
+      "loss": 2.4904,
+      "step": 62260
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.06795195954488e-08,
+      "loss": 1.8603,
+      "step": 62261
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.063211125158028e-08,
+      "loss": 1.8356,
+      "step": 62262
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0584702907711764e-08,
+      "loss": 2.1064,
+      "step": 62263
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.053729456384324e-08,
+      "loss": 2.9353,
+      "step": 62264
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0489886219974715e-08,
+      "loss": 2.2487,
+      "step": 62265
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0442477876106194e-08,
+      "loss": 1.7404,
+      "step": 62266
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.039506953223767e-08,
+      "loss": 0.6481,
+      "step": 62267
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.034766118836915e-08,
+      "loss": 2.0333,
+      "step": 62268
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.030025284450064e-08,
+      "loss": 1.1157,
+      "step": 62269
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0252844500632117e-08,
+      "loss": 1.2791,
+      "step": 62270
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.020543615676359e-08,
+      "loss": 2.2241,
+      "step": 62271
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.015802781289507e-08,
+      "loss": 1.7281,
+      "step": 62272
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.011061946902655e-08,
+      "loss": 2.8192,
+      "step": 62273
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.0063211125158026e-08,
+      "loss": 3.0274,
+      "step": 62274
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 5.001580278128951e-08,
+      "loss": 2.0339,
+      "step": 62275
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.996839443742099e-08,
+      "loss": 1.1068,
+      "step": 62276
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.992098609355247e-08,
+      "loss": 1.6713,
+      "step": 62277
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.987357774968394e-08,
+      "loss": 1.7103,
+      "step": 62278
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.982616940581542e-08,
+      "loss": 2.4998,
+      "step": 62279
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.9778761061946906e-08,
+      "loss": 1.2285,
+      "step": 62280
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.9731352718078385e-08,
+      "loss": 1.2388,
+      "step": 62281
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.9683944374209864e-08,
+      "loss": 3.1448,
+      "step": 62282
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.963653603034134e-08,
+      "loss": 1.613,
+      "step": 62283
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.9589127686472815e-08,
+      "loss": 1.063,
+      "step": 62284
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 4.9541719342604294e-08,
+      "loss": 4.2171,
+      "step": 62285
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.949431099873578e-08,
+      "loss": 2.0486,
+      "step": 62286
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.944690265486726e-08,
+      "loss": 1.7605,
+      "step": 62287
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.939949431099874e-08,
+      "loss": 1.4709,
+      "step": 62288
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.9352085967130217e-08,
+      "loss": 1.9727,
+      "step": 62289
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.930467762326169e-08,
+      "loss": 2.5345,
+      "step": 62290
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.925726927939318e-08,
+      "loss": 1.462,
+      "step": 62291
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.9209860935524653e-08,
+      "loss": 1.2045,
+      "step": 62292
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.916245259165613e-08,
+      "loss": 2.3411,
+      "step": 62293
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.911504424778761e-08,
+      "loss": 2.0263,
+      "step": 62294
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.906763590391909e-08,
+      "loss": 1.7091,
+      "step": 62295
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.902022756005056e-08,
+      "loss": 3.1204,
+      "step": 62296
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8972819216182055e-08,
+      "loss": 1.6481,
+      "step": 62297
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8925410872313534e-08,
+      "loss": 1.9751,
+      "step": 62298
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8878002528445006e-08,
+      "loss": 1.6597,
+      "step": 62299
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8830594184576485e-08,
+      "loss": 1.3368,
+      "step": 62300
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8783185840707964e-08,
+      "loss": 1.9638,
+      "step": 62301
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.873577749683944e-08,
+      "loss": 2.4152,
+      "step": 62302
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.868836915297093e-08,
+      "loss": 1.9058,
+      "step": 62303
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.864096080910241e-08,
+      "loss": 1.6724,
+      "step": 62304
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.859355246523388e-08,
+      "loss": 1.1398,
+      "step": 62305
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.854614412136536e-08,
+      "loss": 3.4292,
+      "step": 62306
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.849873577749684e-08,
+      "loss": 1.9738,
+      "step": 62307
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.845132743362832e-08,
+      "loss": 1.7051,
+      "step": 62308
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.84039190897598e-08,
+      "loss": 2.6152,
+      "step": 62309
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.835651074589128e-08,
+      "loss": 0.9357,
+      "step": 62310
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.830910240202275e-08,
+      "loss": 3.6547,
+      "step": 62311
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.826169405815423e-08,
+      "loss": 2.2793,
+      "step": 62312
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.821428571428571e-08,
+      "loss": 2.099,
+      "step": 62313
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.81668773704172e-08,
+      "loss": 1.3031,
+      "step": 62314
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8119469026548676e-08,
+      "loss": 2.1213,
+      "step": 62315
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8072060682680155e-08,
+      "loss": 1.9044,
+      "step": 62316
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.8024652338811634e-08,
+      "loss": 1.5104,
+      "step": 62317
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.7977243994943106e-08,
+      "loss": 1.4498,
+      "step": 62318
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.79298356510746e-08,
+      "loss": 1.9873,
+      "step": 62319
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.788242730720607e-08,
+      "loss": 1.3503,
+      "step": 62320
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.783501896333755e-08,
+      "loss": 1.7853,
+      "step": 62321
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.778761061946903e-08,
+      "loss": 2.2992,
+      "step": 62322
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.774020227560051e-08,
+      "loss": 2.1671,
+      "step": 62323
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.769279393173198e-08,
+      "loss": 2.5828,
+      "step": 62324
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.764538558786347e-08,
+      "loss": 1.564,
+      "step": 62325
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.7597977243994944e-08,
+      "loss": 1.8857,
+      "step": 62326
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.755056890012642e-08,
+      "loss": 1.5303,
+      "step": 62327
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.75031605562579e-08,
+      "loss": 2.3023,
+      "step": 62328
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.745575221238938e-08,
+      "loss": 1.4153,
+      "step": 62329
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.740834386852085e-08,
+      "loss": 2.2455,
+      "step": 62330
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.7360935524652345e-08,
+      "loss": 2.8018,
+      "step": 62331
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.731352718078382e-08,
+      "loss": 0.5339,
+      "step": 62332
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.7266118836915297e-08,
+      "loss": 2.4263,
+      "step": 62333
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.7218710493046776e-08,
+      "loss": 3.7849,
+      "step": 62334
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.7171302149178254e-08,
+      "loss": 1.6232,
+      "step": 62335
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.712389380530974e-08,
+      "loss": 1.0593,
+      "step": 62336
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.707648546144122e-08,
+      "loss": 1.5998,
+      "step": 62337
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.70290771175727e-08,
+      "loss": 2.2297,
+      "step": 62338
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.698166877370417e-08,
+      "loss": 2.3286,
+      "step": 62339
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.693426042983565e-08,
+      "loss": 1.3408,
+      "step": 62340
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.688685208596713e-08,
+      "loss": 0.8731,
+      "step": 62341
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.683944374209861e-08,
+      "loss": 0.9863,
+      "step": 62342
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.679203539823009e-08,
+      "loss": 2.5308,
+      "step": 62343
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.674462705436157e-08,
+      "loss": 1.1458,
+      "step": 62344
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.6697218710493044e-08,
+      "loss": 1.7002,
+      "step": 62345
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.664981036662453e-08,
+      "loss": 1.7854,
+      "step": 62346
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.660240202275601e-08,
+      "loss": 1.3457,
+      "step": 62347
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 4.655499367888748e-08,
+      "loss": 2.3556,
+      "step": 62348
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.6507585335018966e-08,
+      "loss": 1.9189,
+      "step": 62349
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.6460176991150445e-08,
+      "loss": 1.3845,
+      "step": 62350
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.641276864728192e-08,
+      "loss": 0.9578,
+      "step": 62351
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.63653603034134e-08,
+      "loss": 1.5777,
+      "step": 62352
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.631795195954488e-08,
+      "loss": 2.5625,
+      "step": 62353
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.627054361567636e-08,
+      "loss": 1.5967,
+      "step": 62354
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.622313527180784e-08,
+      "loss": 0.2667,
+      "step": 62355
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.617572692793932e-08,
+      "loss": 3.1369,
+      "step": 62356
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.61283185840708e-08,
+      "loss": 1.5995,
+      "step": 62357
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.608091024020228e-08,
+      "loss": 2.872,
+      "step": 62358
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.6033501896333756e-08,
+      "loss": 2.0248,
+      "step": 62359
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.5986093552465235e-08,
+      "loss": 1.6258,
+      "step": 62360
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.5938685208596714e-08,
+      "loss": 2.4577,
+      "step": 62361
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.589127686472819e-08,
+      "loss": 2.9103,
+      "step": 62362
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.584386852085967e-08,
+      "loss": 2.4646,
+      "step": 62363
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.579646017699115e-08,
+      "loss": 0.5427,
+      "step": 62364
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.574905183312263e-08,
+      "loss": 1.416,
+      "step": 62365
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.570164348925411e-08,
+      "loss": 1.9686,
+      "step": 62366
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.565423514538559e-08,
+      "loss": 1.5938,
+      "step": 62367
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.560682680151707e-08,
+      "loss": 1.6706,
+      "step": 62368
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.5559418457648545e-08,
+      "loss": 1.5388,
+      "step": 62369
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.5512010113780024e-08,
+      "loss": 1.5097,
+      "step": 62370
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.546460176991151e-08,
+      "loss": 2.7353,
+      "step": 62371
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.541719342604298e-08,
+      "loss": 1.5973,
+      "step": 62372
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.536978508217446e-08,
+      "loss": 1.6967,
+      "step": 62373
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.5322376738305946e-08,
+      "loss": 2.0083,
+      "step": 62374
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.527496839443742e-08,
+      "loss": 1.0698,
+      "step": 62375
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.52275600505689e-08,
+      "loss": 0.5552,
+      "step": 62376
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.518015170670038e-08,
+      "loss": 0.6629,
+      "step": 62377
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.513274336283186e-08,
+      "loss": 1.7391,
+      "step": 62378
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.5085335018963335e-08,
+      "loss": 2.2257,
+      "step": 62379
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.503792667509482e-08,
+      "loss": 1.2374,
+      "step": 62380
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.49905183312263e-08,
+      "loss": 2.635,
+      "step": 62381
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.494310998735778e-08,
+      "loss": 1.9458,
+      "step": 62382
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.489570164348926e-08,
+      "loss": 3.4952,
+      "step": 62383
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4848293299620736e-08,
+      "loss": 2.6519,
+      "step": 62384
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4800884955752215e-08,
+      "loss": 1.5102,
+      "step": 62385
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4753476611883694e-08,
+      "loss": 3.1241,
+      "step": 62386
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.470606826801517e-08,
+      "loss": 3.9594,
+      "step": 62387
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.465865992414665e-08,
+      "loss": 1.7844,
+      "step": 62388
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.461125158027813e-08,
+      "loss": 1.0552,
+      "step": 62389
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.456384323640961e-08,
+      "loss": 1.8118,
+      "step": 62390
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.451643489254109e-08,
+      "loss": 1.8711,
+      "step": 62391
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.446902654867257e-08,
+      "loss": 1.6876,
+      "step": 62392
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4421618204804046e-08,
+      "loss": 1.0262,
+      "step": 62393
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4374209860935525e-08,
+      "loss": 1.8709,
+      "step": 62394
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4326801517067004e-08,
+      "loss": 0.585,
+      "step": 62395
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.427939317319848e-08,
+      "loss": 2.1942,
+      "step": 62396
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.423198482932996e-08,
+      "loss": 1.3097,
+      "step": 62397
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.418457648546144e-08,
+      "loss": 2.0465,
+      "step": 62398
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.4137168141592927e-08,
+      "loss": 1.7347,
+      "step": 62399
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.40897597977244e-08,
+      "loss": 2.8545,
+      "step": 62400
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.404235145385588e-08,
+      "loss": 1.9066,
+      "step": 62401
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.3994943109987363e-08,
+      "loss": 1.3216,
+      "step": 62402
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.3947534766118836e-08,
+      "loss": 1.0602,
+      "step": 62403
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.3900126422250315e-08,
+      "loss": 1.538,
+      "step": 62404
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.38527180783818e-08,
+      "loss": 1.5479,
+      "step": 62405
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.380530973451327e-08,
+      "loss": 1.1331,
+      "step": 62406
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.375790139064475e-08,
+      "loss": 1.9994,
+      "step": 62407
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.371049304677624e-08,
+      "loss": 1.7867,
+      "step": 62408
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.366308470290771e-08,
+      "loss": 0.8665,
+      "step": 62409
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.3615676359039195e-08,
+      "loss": 1.878,
+      "step": 62410
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 4.3568268015170674e-08,
+      "loss": 1.6383,
+      "step": 62411
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.3520859671302146e-08,
+      "loss": 2.8443,
+      "step": 62412
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.347345132743363e-08,
+      "loss": 0.9779,
+      "step": 62413
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.342604298356511e-08,
+      "loss": 1.5967,
+      "step": 62414
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.337863463969658e-08,
+      "loss": 1.5348,
+      "step": 62415
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.333122629582807e-08,
+      "loss": 2.1262,
+      "step": 62416
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.328381795195955e-08,
+      "loss": 1.5536,
+      "step": 62417
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.323640960809102e-08,
+      "loss": 2.6759,
+      "step": 62418
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.3189001264222505e-08,
+      "loss": 0.6026,
+      "step": 62419
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.3141592920353984e-08,
+      "loss": 2.0471,
+      "step": 62420
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.3094184576485463e-08,
+      "loss": 0.311,
+      "step": 62421
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.304677623261694e-08,
+      "loss": 1.7539,
+      "step": 62422
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.299936788874842e-08,
+      "loss": 1.1584,
+      "step": 62423
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.29519595448799e-08,
+      "loss": 1.8046,
+      "step": 62424
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.290455120101138e-08,
+      "loss": 1.7745,
+      "step": 62425
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.285714285714286e-08,
+      "loss": 1.3437,
+      "step": 62426
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.280973451327434e-08,
+      "loss": 0.9054,
+      "step": 62427
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2762326169405816e-08,
+      "loss": 1.4541,
+      "step": 62428
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2714917825537295e-08,
+      "loss": 1.974,
+      "step": 62429
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2667509481668774e-08,
+      "loss": 2.1381,
+      "step": 62430
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.262010113780025e-08,
+      "loss": 1.3999,
+      "step": 62431
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.257269279393173e-08,
+      "loss": 2.6639,
+      "step": 62432
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.252528445006321e-08,
+      "loss": 1.9641,
+      "step": 62433
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.247787610619469e-08,
+      "loss": 1.6243,
+      "step": 62434
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.243046776232617e-08,
+      "loss": 2.2748,
+      "step": 62435
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.238305941845765e-08,
+      "loss": 1.8216,
+      "step": 62436
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2335651074589126e-08,
+      "loss": 4.6839,
+      "step": 62437
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.228824273072061e-08,
+      "loss": 1.7349,
+      "step": 62438
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2240834386852084e-08,
+      "loss": 0.7344,
+      "step": 62439
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.219342604298356e-08,
+      "loss": 2.1966,
+      "step": 62440
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.214601769911505e-08,
+      "loss": 2.6201,
+      "step": 62441
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.209860935524653e-08,
+      "loss": 2.0035,
+      "step": 62442
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2051201011378e-08,
+      "loss": 1.7808,
+      "step": 62443
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.2003792667509486e-08,
+      "loss": 2.489,
+      "step": 62444
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1956384323640965e-08,
+      "loss": 0.7457,
+      "step": 62445
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.190897597977244e-08,
+      "loss": 2.1895,
+      "step": 62446
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.186156763590392e-08,
+      "loss": 1.3035,
+      "step": 62447
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.18141592920354e-08,
+      "loss": 1.7539,
+      "step": 62448
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1766750948166874e-08,
+      "loss": 3.4053,
+      "step": 62449
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.171934260429836e-08,
+      "loss": 2.1081,
+      "step": 62450
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.167193426042984e-08,
+      "loss": 1.9733,
+      "step": 62451
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.162452591656132e-08,
+      "loss": 2.0581,
+      "step": 62452
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1577117572692796e-08,
+      "loss": 1.7935,
+      "step": 62453
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1529709228824275e-08,
+      "loss": 1.288,
+      "step": 62454
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1482300884955754e-08,
+      "loss": 1.3944,
+      "step": 62455
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.143489254108723e-08,
+      "loss": 0.6934,
+      "step": 62456
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.138748419721871e-08,
+      "loss": 2.0756,
+      "step": 62457
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.134007585335019e-08,
+      "loss": 1.5983,
+      "step": 62458
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.129266750948167e-08,
+      "loss": 2.6616,
+      "step": 62459
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.124525916561315e-08,
+      "loss": 2.175,
+      "step": 62460
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.119785082174463e-08,
+      "loss": 1.3707,
+      "step": 62461
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1150442477876107e-08,
+      "loss": 2.0043,
+      "step": 62462
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1103034134007585e-08,
+      "loss": 1.8753,
+      "step": 62463
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1055625790139064e-08,
+      "loss": 2.5317,
+      "step": 62464
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.1008217446270543e-08,
+      "loss": 1.2203,
+      "step": 62465
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.096080910240203e-08,
+      "loss": 2.5461,
+      "step": 62466
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.09134007585335e-08,
+      "loss": 1.7461,
+      "step": 62467
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.086599241466498e-08,
+      "loss": 1.8477,
+      "step": 62468
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.0818584070796466e-08,
+      "loss": 1.5695,
+      "step": 62469
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.077117572692794e-08,
+      "loss": 1.9715,
+      "step": 62470
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.072376738305942e-08,
+      "loss": 0.8445,
+      "step": 62471
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.06763590391909e-08,
+      "loss": 1.0857,
+      "step": 62472
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.0628950695322375e-08,
+      "loss": 1.5855,
+      "step": 62473
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.0581542351453854e-08,
+      "loss": 2.9189,
+      "step": 62474
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 4.053413400758534e-08,
+      "loss": 3.2839,
+      "step": 62475
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.048672566371681e-08,
+      "loss": 2.7947,
+      "step": 62476
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.043931731984829e-08,
+      "loss": 2.3959,
+      "step": 62477
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.0391908975979776e-08,
+      "loss": 1.4495,
+      "step": 62478
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.034450063211125e-08,
+      "loss": 1.8053,
+      "step": 62479
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.0297092288242734e-08,
+      "loss": 1.6699,
+      "step": 62480
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.024968394437421e-08,
+      "loss": 1.7267,
+      "step": 62481
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.020227560050569e-08,
+      "loss": 0.7603,
+      "step": 62482
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.015486725663717e-08,
+      "loss": 4.0661,
+      "step": 62483
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.010745891276865e-08,
+      "loss": 0.5523,
+      "step": 62484
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.006005056890013e-08,
+      "loss": 1.0029,
+      "step": 62485
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 4.001264222503161e-08,
+      "loss": 2.8573,
+      "step": 62486
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.996523388116309e-08,
+      "loss": 2.5493,
+      "step": 62487
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9917825537294566e-08,
+      "loss": 2.6126,
+      "step": 62488
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9870417193426045e-08,
+      "loss": 1.764,
+      "step": 62489
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9823008849557524e-08,
+      "loss": 0.8226,
+      "step": 62490
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9775600505689e-08,
+      "loss": 2.4623,
+      "step": 62491
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.972819216182048e-08,
+      "loss": 1.0251,
+      "step": 62492
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.968078381795196e-08,
+      "loss": 2.7559,
+      "step": 62493
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.963337547408344e-08,
+      "loss": 1.2708,
+      "step": 62494
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.958596713021492e-08,
+      "loss": 1.1784,
+      "step": 62495
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.95385587863464e-08,
+      "loss": 2.88,
+      "step": 62496
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9491150442477876e-08,
+      "loss": 1.9392,
+      "step": 62497
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9443742098609355e-08,
+      "loss": 1.2492,
+      "step": 62498
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9396333754740834e-08,
+      "loss": 3.3251,
+      "step": 62499
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.934892541087231e-08,
+      "loss": 2.1325,
+      "step": 62500
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.930151706700379e-08,
+      "loss": 2.2874,
+      "step": 62501
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.925410872313527e-08,
+      "loss": 1.2973,
+      "step": 62502
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9206700379266756e-08,
+      "loss": 2.0062,
+      "step": 62503
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.915929203539823e-08,
+      "loss": 1.6743,
+      "step": 62504
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.911188369152971e-08,
+      "loss": 2.5956,
+      "step": 62505
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.906447534766119e-08,
+      "loss": 1.8014,
+      "step": 62506
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.9017067003792666e-08,
+      "loss": 3.3432,
+      "step": 62507
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.896965865992415e-08,
+      "loss": 1.9902,
+      "step": 62508
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.892225031605563e-08,
+      "loss": 1.2896,
+      "step": 62509
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.88748419721871e-08,
+      "loss": 1.36,
+      "step": 62510
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.882743362831859e-08,
+      "loss": 4.0674,
+      "step": 62511
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.878002528445007e-08,
+      "loss": 2.9673,
+      "step": 62512
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.873261694058154e-08,
+      "loss": 1.3822,
+      "step": 62513
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.8685208596713025e-08,
+      "loss": 1.1567,
+      "step": 62514
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.8637800252844504e-08,
+      "loss": 1.3727,
+      "step": 62515
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.8590391908975976e-08,
+      "loss": 1.9619,
+      "step": 62516
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.854298356510746e-08,
+      "loss": 2.3056,
+      "step": 62517
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.849557522123894e-08,
+      "loss": 1.3628,
+      "step": 62518
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.844816687737041e-08,
+      "loss": 2.2822,
+      "step": 62519
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.84007585335019e-08,
+      "loss": 1.2306,
+      "step": 62520
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.835335018963338e-08,
+      "loss": 4.9771,
+      "step": 62521
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.8305941845764856e-08,
+      "loss": 2.0681,
+      "step": 62522
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.8258533501896335e-08,
+      "loss": 1.6435,
+      "step": 62523
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.8211125158027814e-08,
+      "loss": 1.5655,
+      "step": 62524
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.816371681415929e-08,
+      "loss": 2.26,
+      "step": 62525
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.811630847029077e-08,
+      "loss": 1.6922,
+      "step": 62526
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.806890012642225e-08,
+      "loss": 2.1153,
+      "step": 62527
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.802149178255373e-08,
+      "loss": 1.4724,
+      "step": 62528
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.797408343868521e-08,
+      "loss": 0.4185,
+      "step": 62529
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.792667509481669e-08,
+      "loss": 0.9073,
+      "step": 62530
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.787926675094817e-08,
+      "loss": 0.8735,
+      "step": 62531
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.7831858407079646e-08,
+      "loss": 3.3184,
+      "step": 62532
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.7784450063211125e-08,
+      "loss": 3.2907,
+      "step": 62533
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.7737041719342604e-08,
+      "loss": 2.4998,
+      "step": 62534
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.768963337547408e-08,
+      "loss": 0.9174,
+      "step": 62535
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.764222503160557e-08,
+      "loss": 2.4916,
+      "step": 62536
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.759481668773704e-08,
+      "loss": 1.0983,
+      "step": 62537
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.754740834386852e-08,
+      "loss": 1.8818,
+      "step": 62538
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.7500000000000005e-08,
+      "loss": 0.6765,
+      "step": 62539
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.745259165613148e-08,
+      "loss": 2.5767,
+      "step": 62540
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.7405183312262956e-08,
+      "loss": 2.6687,
+      "step": 62541
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.735777496839444e-08,
+      "loss": 1.6136,
+      "step": 62542
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.7310366624525914e-08,
+      "loss": 2.0684,
+      "step": 62543
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.726295828065739e-08,
+      "loss": 2.312,
+      "step": 62544
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.721554993678888e-08,
+      "loss": 1.9418,
+      "step": 62545
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.716814159292036e-08,
+      "loss": 3.9957,
+      "step": 62546
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.712073324905183e-08,
+      "loss": 2.1673,
+      "step": 62547
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.7073324905183315e-08,
+      "loss": 1.0261,
+      "step": 62548
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.7025916561314794e-08,
+      "loss": 3.5902,
+      "step": 62549
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.697850821744627e-08,
+      "loss": 2.4592,
+      "step": 62550
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.693109987357775e-08,
+      "loss": 2.7957,
+      "step": 62551
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.688369152970923e-08,
+      "loss": 0.8483,
+      "step": 62552
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.683628318584071e-08,
+      "loss": 2.2274,
+      "step": 62553
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.678887484197219e-08,
+      "loss": 1.8889,
+      "step": 62554
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.674146649810367e-08,
+      "loss": 1.9293,
+      "step": 62555
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.669405815423515e-08,
+      "loss": 1.5115,
+      "step": 62556
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.6646649810366626e-08,
+      "loss": 2.4071,
+      "step": 62557
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.6599241466498105e-08,
+      "loss": 2.605,
+      "step": 62558
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.6551833122629584e-08,
+      "loss": 0.989,
+      "step": 62559
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.650442477876106e-08,
+      "loss": 1.1778,
+      "step": 62560
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.645701643489254e-08,
+      "loss": 3.4432,
+      "step": 62561
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.640960809102402e-08,
+      "loss": 0.9423,
+      "step": 62562
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.63621997471555e-08,
+      "loss": 1.9582,
+      "step": 62563
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.6314791403286985e-08,
+      "loss": 0.9467,
+      "step": 62564
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.626738305941846e-08,
+      "loss": 1.9759,
+      "step": 62565
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.6219974715549936e-08,
+      "loss": 0.7774,
+      "step": 62566
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.617256637168142e-08,
+      "loss": 2.0059,
+      "step": 62567
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.6125158027812894e-08,
+      "loss": 2.793,
+      "step": 62568
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.607774968394437e-08,
+      "loss": 0.919,
+      "step": 62569
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.603034134007586e-08,
+      "loss": 1.6447,
+      "step": 62570
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.598293299620733e-08,
+      "loss": 0.5726,
+      "step": 62571
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.593552465233881e-08,
+      "loss": 1.0433,
+      "step": 62572
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.5888116308470296e-08,
+      "loss": 1.1949,
+      "step": 62573
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.584070796460177e-08,
+      "loss": 1.0834,
+      "step": 62574
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.579329962073325e-08,
+      "loss": 3.1527,
+      "step": 62575
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.574589127686473e-08,
+      "loss": 0.5333,
+      "step": 62576
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.5698482932996205e-08,
+      "loss": 1.6975,
+      "step": 62577
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.565107458912769e-08,
+      "loss": 2.8691,
+      "step": 62578
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.560366624525917e-08,
+      "loss": 0.7112,
+      "step": 62579
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.555625790139064e-08,
+      "loss": 1.041,
+      "step": 62580
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.550884955752213e-08,
+      "loss": 0.9869,
+      "step": 62581
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.5461441213653606e-08,
+      "loss": 0.9556,
+      "step": 62582
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.541403286978508e-08,
+      "loss": 0.3536,
+      "step": 62583
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.5366624525916564e-08,
+      "loss": 0.7161,
+      "step": 62584
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.531921618204804e-08,
+      "loss": 1.0304,
+      "step": 62585
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.527180783817952e-08,
+      "loss": 2.1422,
+      "step": 62586
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.5224399494311e-08,
+      "loss": 2.4876,
+      "step": 62587
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.517699115044248e-08,
+      "loss": 0.9065,
+      "step": 62588
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.512958280657396e-08,
+      "loss": 1.5957,
+      "step": 62589
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.508217446270544e-08,
+      "loss": 1.7389,
+      "step": 62590
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.5034766118836917e-08,
+      "loss": 0.507,
+      "step": 62591
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.4987357774968395e-08,
+      "loss": 1.4745,
+      "step": 62592
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.4939949431099874e-08,
+      "loss": 2.794,
+      "step": 62593
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.4892541087231353e-08,
+      "loss": 1.4743,
+      "step": 62594
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.484513274336283e-08,
+      "loss": 3.4451,
+      "step": 62595
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.479772439949431e-08,
+      "loss": 1.8638,
+      "step": 62596
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.475031605562579e-08,
+      "loss": 1.9525,
+      "step": 62597
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.470290771175727e-08,
+      "loss": 2.043,
+      "step": 62598
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.465549936788875e-08,
+      "loss": 2.3969,
+      "step": 62599
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.460809102402023e-08,
+      "loss": 3.06,
+      "step": 62600
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.4560682680151706e-08,
+      "loss": 3.2816,
+      "step": 62601
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.4513274336283185e-08,
+      "loss": 3.3701,
+      "step": 62602
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.4465865992414664e-08,
+      "loss": 1.1117,
+      "step": 62603
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.441845764854614e-08,
+      "loss": 1.6878,
+      "step": 62604
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.437104930467762e-08,
+      "loss": 2.2401,
+      "step": 62605
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.432364096080911e-08,
+      "loss": 1.4508,
+      "step": 62606
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.4276232616940586e-08,
+      "loss": 1.1576,
+      "step": 62607
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.422882427307206e-08,
+      "loss": 2.7796,
+      "step": 62608
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.4181415929203544e-08,
+      "loss": 0.7946,
+      "step": 62609
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.413400758533502e-08,
+      "loss": 3.0886,
+      "step": 62610
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.4086599241466495e-08,
+      "loss": 2.8321,
+      "step": 62611
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.403919089759798e-08,
+      "loss": 0.4137,
+      "step": 62612
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.399178255372946e-08,
+      "loss": 0.6906,
+      "step": 62613
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.394437420986093e-08,
+      "loss": 0.6713,
+      "step": 62614
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.389696586599242e-08,
+      "loss": 3.0494,
+      "step": 62615
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.3849557522123897e-08,
+      "loss": 2.2381,
+      "step": 62616
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.380214917825537e-08,
+      "loss": 3.3113,
+      "step": 62617
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.3754740834386855e-08,
+      "loss": 1.6886,
+      "step": 62618
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.3707332490518333e-08,
+      "loss": 1.8518,
+      "step": 62619
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.365992414664981e-08,
+      "loss": 0.6549,
+      "step": 62620
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.361251580278129e-08,
+      "loss": 1.2765,
+      "step": 62621
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.356510745891277e-08,
+      "loss": 2.3068,
+      "step": 62622
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.351769911504425e-08,
+      "loss": 1.0658,
+      "step": 62623
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.347029077117573e-08,
+      "loss": 1.7604,
+      "step": 62624
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.342288242730721e-08,
+      "loss": 1.7458,
+      "step": 62625
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.3375474083438686e-08,
+      "loss": 2.848,
+      "step": 62626
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.3328065739570165e-08,
+      "loss": 1.8683,
+      "step": 62627
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.3280657395701644e-08,
+      "loss": 2.3744,
+      "step": 62628
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.323324905183312e-08,
+      "loss": 1.9012,
+      "step": 62629
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.31858407079646e-08,
+      "loss": 1.8799,
+      "step": 62630
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.313843236409608e-08,
+      "loss": 0.9261,
+      "step": 62631
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.309102402022756e-08,
+      "loss": 1.3936,
+      "step": 62632
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.304361567635904e-08,
+      "loss": 1.9135,
+      "step": 62633
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2996207332490524e-08,
+      "loss": 0.9591,
+      "step": 62634
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2948798988621997e-08,
+      "loss": 1.3451,
+      "step": 62635
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2901390644753475e-08,
+      "loss": 0.729,
+      "step": 62636
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.285398230088496e-08,
+      "loss": 1.3778,
+      "step": 62637
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2806573957016433e-08,
+      "loss": 1.0282,
+      "step": 62638
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.275916561314791e-08,
+      "loss": 2.2572,
+      "step": 62639
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.27117572692794e-08,
+      "loss": 3.1109,
+      "step": 62640
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.266434892541087e-08,
+      "loss": 1.5581,
+      "step": 62641
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.261694058154235e-08,
+      "loss": 2.8778,
+      "step": 62642
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2569532237673835e-08,
+      "loss": 1.9116,
+      "step": 62643
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.252212389380531e-08,
+      "loss": 2.4226,
+      "step": 62644
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2474715549936786e-08,
+      "loss": 0.8946,
+      "step": 62645
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.242730720606827e-08,
+      "loss": 2.6946,
+      "step": 62646
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2379898862199744e-08,
+      "loss": 1.0748,
+      "step": 62647
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.233249051833123e-08,
+      "loss": 3.1082,
+      "step": 62648
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.228508217446271e-08,
+      "loss": 0.7066,
+      "step": 62649
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.223767383059419e-08,
+      "loss": 0.4386,
+      "step": 62650
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2190265486725666e-08,
+      "loss": 3.3664,
+      "step": 62651
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2142857142857145e-08,
+      "loss": 2.3718,
+      "step": 62652
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.2095448798988624e-08,
+      "loss": 1.4943,
+      "step": 62653
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.20480404551201e-08,
+      "loss": 1.3246,
+      "step": 62654
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.200063211125158e-08,
+      "loss": 1.1977,
+      "step": 62655
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.195322376738306e-08,
+      "loss": 1.9294,
+      "step": 62656
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.190581542351454e-08,
+      "loss": 1.7597,
+      "step": 62657
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.185840707964602e-08,
+      "loss": 3.4067,
+      "step": 62658
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.18109987357775e-08,
+      "loss": 2.6235,
+      "step": 62659
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.176359039190898e-08,
+      "loss": 2.0239,
+      "step": 62660
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.1716182048040456e-08,
+      "loss": 1.4489,
+      "step": 62661
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.1668773704171935e-08,
+      "loss": 1.8383,
+      "step": 62662
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.1621365360303414e-08,
+      "loss": 1.5364,
+      "step": 62663
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.157395701643489e-08,
+      "loss": 2.3076,
+      "step": 62664
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.152654867256637e-08,
+      "loss": 0.5938,
+      "step": 62665
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.147914032869785e-08,
+      "loss": 2.3734,
+      "step": 62666
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.143173198482933e-08,
+      "loss": 1.2951,
+      "step": 62667
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.1384323640960815e-08,
+      "loss": 1.6027,
+      "step": 62668
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.133691529709229e-08,
+      "loss": 3.4454,
+      "step": 62669
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.1289506953223766e-08,
+      "loss": 1.6583,
+      "step": 62670
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.124209860935525e-08,
+      "loss": 1.6197,
+      "step": 62671
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.1194690265486724e-08,
+      "loss": 1.708,
+      "step": 62672
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.11472819216182e-08,
+      "loss": 1.1805,
+      "step": 62673
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.109987357774969e-08,
+      "loss": 2.6786,
+      "step": 62674
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.105246523388116e-08,
+      "loss": 2.4667,
+      "step": 62675
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.1005056890012646e-08,
+      "loss": 2.1474,
+      "step": 62676
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0957648546144125e-08,
+      "loss": 1.5088,
+      "step": 62677
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.09102402022756e-08,
+      "loss": 2.3782,
+      "step": 62678
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.086283185840708e-08,
+      "loss": 3.7028,
+      "step": 62679
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.081542351453856e-08,
+      "loss": 1.0215,
+      "step": 62680
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0768015170670034e-08,
+      "loss": 0.959,
+      "step": 62681
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.072060682680152e-08,
+      "loss": 1.6789,
+      "step": 62682
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0673198482933e-08,
+      "loss": 3.3186,
+      "step": 62683
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.062579013906447e-08,
+      "loss": 1.7948,
+      "step": 62684
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.057838179519596e-08,
+      "loss": 3.3742,
+      "step": 62685
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0530973451327436e-08,
+      "loss": 1.1817,
+      "step": 62686
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.048356510745891e-08,
+      "loss": 1.0118,
+      "step": 62687
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0436156763590394e-08,
+      "loss": 3.9029,
+      "step": 62688
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.038874841972187e-08,
+      "loss": 2.568,
+      "step": 62689
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.034134007585335e-08,
+      "loss": 1.7576,
+      "step": 62690
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.029393173198483e-08,
+      "loss": 1.8153,
+      "step": 62691
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.024652338811631e-08,
+      "loss": 1.3696,
+      "step": 62692
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.019911504424779e-08,
+      "loss": 0.9657,
+      "step": 62693
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.015170670037927e-08,
+      "loss": 2.8041,
+      "step": 62694
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0104298356510746e-08,
+      "loss": 1.4088,
+      "step": 62695
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0056890012642225e-08,
+      "loss": 2.4816,
+      "step": 62696
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.0009481668773704e-08,
+      "loss": 3.8396,
+      "step": 62697
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.996207332490518e-08,
+      "loss": 0.7292,
+      "step": 62698
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.991466498103666e-08,
+      "loss": 1.7119,
+      "step": 62699
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.986725663716814e-08,
+      "loss": 2.2164,
+      "step": 62700
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.981984829329962e-08,
+      "loss": 1.6696,
+      "step": 62701
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9772439949431102e-08,
+      "loss": 1.1637,
+      "step": 62702
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9725031605562578e-08,
+      "loss": 2.0916,
+      "step": 62703
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.967762326169406e-08,
+      "loss": 1.7772,
+      "step": 62704
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.963021491782554e-08,
+      "loss": 0.9838,
+      "step": 62705
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9582806573957015e-08,
+      "loss": 1.7586,
+      "step": 62706
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9535398230088497e-08,
+      "loss": 1.4952,
+      "step": 62707
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9487989886219976e-08,
+      "loss": 3.2661,
+      "step": 62708
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.944058154235145e-08,
+      "loss": 1.4184,
+      "step": 62709
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9393173198482937e-08,
+      "loss": 2.7702,
+      "step": 62710
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9345764854614413e-08,
+      "loss": 2.1474,
+      "step": 62711
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.929835651074589e-08,
+      "loss": 0.754,
+      "step": 62712
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9250948166877374e-08,
+      "loss": 1.1252,
+      "step": 62713
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.920353982300885e-08,
+      "loss": 1.7407,
+      "step": 62714
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.915613147914033e-08,
+      "loss": 1.5765,
+      "step": 62715
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.910872313527181e-08,
+      "loss": 1.9627,
+      "step": 62716
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.9061314791403286e-08,
+      "loss": 1.099,
+      "step": 62717
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.901390644753477e-08,
+      "loss": 0.6915,
+      "step": 62718
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8966498103666248e-08,
+      "loss": 1.6856,
+      "step": 62719
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8919089759797723e-08,
+      "loss": 0.6543,
+      "step": 62720
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8871681415929205e-08,
+      "loss": 3.7511,
+      "step": 62721
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8824273072060684e-08,
+      "loss": 1.685,
+      "step": 62722
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.877686472819216e-08,
+      "loss": 1.5758,
+      "step": 62723
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8729456384323642e-08,
+      "loss": 1.6878,
+      "step": 62724
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.868204804045512e-08,
+      "loss": 1.5112,
+      "step": 62725
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8634639696586597e-08,
+      "loss": 1.9286,
+      "step": 62726
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.858723135271808e-08,
+      "loss": 2.6177,
+      "step": 62727
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.8539823008849558e-08,
+      "loss": 2.2934,
+      "step": 62728
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.8492414664981034e-08,
+      "loss": 1.614,
+      "step": 62729
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.844500632111252e-08,
+      "loss": 1.1253,
+      "step": 62730
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.8397597977243995e-08,
+      "loss": 3.1528,
+      "step": 62731
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.8350189633375477e-08,
+      "loss": 1.5532,
+      "step": 62732
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.8302781289506956e-08,
+      "loss": 2.6212,
+      "step": 62733
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.825537294563843e-08,
+      "loss": 1.5656,
+      "step": 62734
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.8207964601769914e-08,
+      "loss": 0.8742,
+      "step": 62735
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.8160556257901393e-08,
+      "loss": 0.8117,
+      "step": 62736
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.811314791403287e-08,
+      "loss": 1.1426,
+      "step": 62737
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.806573957016435e-08,
+      "loss": 1.252,
+      "step": 62738
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.801833122629583e-08,
+      "loss": 1.8506,
+      "step": 62739
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7970922882427305e-08,
+      "loss": 2.6682,
+      "step": 62740
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7923514538558788e-08,
+      "loss": 1.6863,
+      "step": 62741
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7876106194690266e-08,
+      "loss": 0.7012,
+      "step": 62742
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7828697850821742e-08,
+      "loss": 2.3364,
+      "step": 62743
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7781289506953224e-08,
+      "loss": 0.6325,
+      "step": 62744
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7733881163084703e-08,
+      "loss": 1.6838,
+      "step": 62745
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7686472819216186e-08,
+      "loss": 2.4829,
+      "step": 62746
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.763906447534766e-08,
+      "loss": 1.6404,
+      "step": 62747
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.759165613147914e-08,
+      "loss": 1.3154,
+      "step": 62748
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7544247787610622e-08,
+      "loss": 1.1405,
+      "step": 62749
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.74968394437421e-08,
+      "loss": 2.1813,
+      "step": 62750
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7449431099873577e-08,
+      "loss": 1.716,
+      "step": 62751
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.740202275600506e-08,
+      "loss": 2.2057,
+      "step": 62752
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7354614412136538e-08,
+      "loss": 1.9201,
+      "step": 62753
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7307206068268014e-08,
+      "loss": 1.0232,
+      "step": 62754
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7259797724399496e-08,
+      "loss": 2.9359,
+      "step": 62755
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7212389380530975e-08,
+      "loss": 1.6443,
+      "step": 62756
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.716498103666245e-08,
+      "loss": 1.0455,
+      "step": 62757
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7117572692793933e-08,
+      "loss": 0.951,
+      "step": 62758
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7070164348925412e-08,
+      "loss": 1.6377,
+      "step": 62759
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.7022756005056894e-08,
+      "loss": 2.5803,
+      "step": 62760
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.697534766118837e-08,
+      "loss": 2.5943,
+      "step": 62761
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.692793931731985e-08,
+      "loss": 1.5306,
+      "step": 62762
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.688053097345133e-08,
+      "loss": 1.1911,
+      "step": 62763
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6833122629582807e-08,
+      "loss": 1.0805,
+      "step": 62764
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6785714285714285e-08,
+      "loss": 2.3626,
+      "step": 62765
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6738305941845768e-08,
+      "loss": 0.9862,
+      "step": 62766
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6690897597977243e-08,
+      "loss": 2.2416,
+      "step": 62767
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6643489254108722e-08,
+      "loss": 1.5741,
+      "step": 62768
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6596080910240205e-08,
+      "loss": 1.5669,
+      "step": 62769
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.654867256637168e-08,
+      "loss": 2.1828,
+      "step": 62770
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.650126422250316e-08,
+      "loss": 1.3153,
+      "step": 62771
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.645385587863464e-08,
+      "loss": 1.7847,
+      "step": 62772
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.640644753476612e-08,
+      "loss": 1.5233,
+      "step": 62773
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6359039190897596e-08,
+      "loss": 1.6625,
+      "step": 62774
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6311630847029078e-08,
+      "loss": 1.4374,
+      "step": 62775
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6264222503160557e-08,
+      "loss": 2.2868,
+      "step": 62776
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.621681415929204e-08,
+      "loss": 2.6312,
+      "step": 62777
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6169405815423515e-08,
+      "loss": 2.0858,
+      "step": 62778
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6121997471554994e-08,
+      "loss": 2.2979,
+      "step": 62779
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6074589127686476e-08,
+      "loss": 1.0501,
+      "step": 62780
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6027180783817952e-08,
+      "loss": 1.5288,
+      "step": 62781
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.597977243994943e-08,
+      "loss": 0.9653,
+      "step": 62782
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5932364096080913e-08,
+      "loss": 1.6448,
+      "step": 62783
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.588495575221239e-08,
+      "loss": 1.167,
+      "step": 62784
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5837547408343868e-08,
+      "loss": 1.8597,
+      "step": 62785
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.579013906447535e-08,
+      "loss": 0.8864,
+      "step": 62786
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5742730720606825e-08,
+      "loss": 0.8236,
+      "step": 62787
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5695322376738304e-08,
+      "loss": 3.3398,
+      "step": 62788
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5647914032869787e-08,
+      "loss": 1.0134,
+      "step": 62789
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5600505689001262e-08,
+      "loss": 1.6991,
+      "step": 62790
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5553097345132748e-08,
+      "loss": 1.1779,
+      "step": 62791
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5505689001264223e-08,
+      "loss": 3.0194,
+      "step": 62792
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5458280657395702e-08,
+      "loss": 2.6154,
+      "step": 62793
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5410872313527185e-08,
+      "loss": 1.0751,
+      "step": 62794
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.536346396965866e-08,
+      "loss": 1.6009,
+      "step": 62795
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.531605562579014e-08,
+      "loss": 1.6266,
+      "step": 62796
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.526864728192162e-08,
+      "loss": 1.4321,
+      "step": 62797
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5221238938053097e-08,
+      "loss": 1.6525,
+      "step": 62798
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5173830594184576e-08,
+      "loss": 0.779,
+      "step": 62799
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5126422250316058e-08,
+      "loss": 1.8655,
+      "step": 62800
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5079013906447534e-08,
+      "loss": 0.7292,
+      "step": 62801
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5031605562579013e-08,
+      "loss": 1.5469,
+      "step": 62802
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4984197218710495e-08,
+      "loss": 1.3807,
+      "step": 62803
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.493678887484197e-08,
+      "loss": 2.4525,
+      "step": 62804
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4889380530973453e-08,
+      "loss": 1.1219,
+      "step": 62805
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4841972187104932e-08,
+      "loss": 3.3563,
+      "step": 62806
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4794563843236408e-08,
+      "loss": 1.6824,
+      "step": 62807
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.474715549936789e-08,
+      "loss": 2.0715,
+      "step": 62808
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.469974715549937e-08,
+      "loss": 0.6488,
+      "step": 62809
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4652338811630844e-08,
+      "loss": 2.1072,
+      "step": 62810
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4604930467762327e-08,
+      "loss": 1.0503,
+      "step": 62811
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4557522123893806e-08,
+      "loss": 0.5065,
+      "step": 62812
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.451011378002528e-08,
+      "loss": 1.2259,
+      "step": 62813
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4462705436156767e-08,
+      "loss": 2.0534,
+      "step": 62814
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4415297092288242e-08,
+      "loss": 0.6203,
+      "step": 62815
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.436788874841972e-08,
+      "loss": 2.2011,
+      "step": 62816
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4320480404551204e-08,
+      "loss": 1.7361,
+      "step": 62817
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.427307206068268e-08,
+      "loss": 1.4522,
+      "step": 62818
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.422566371681416e-08,
+      "loss": 2.2647,
+      "step": 62819
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.417825537294564e-08,
+      "loss": 0.787,
+      "step": 62820
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4130847029077116e-08,
+      "loss": 2.9128,
+      "step": 62821
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.40834386852086e-08,
+      "loss": 2.667,
+      "step": 62822
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.4036030341340077e-08,
+      "loss": 1.7207,
+      "step": 62823
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3988621997471553e-08,
+      "loss": 2.8348,
+      "step": 62824
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3941213653603035e-08,
+      "loss": 2.1349,
+      "step": 62825
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3893805309734514e-08,
+      "loss": 2.3506,
+      "step": 62826
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.384639696586599e-08,
+      "loss": 3.6011,
+      "step": 62827
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3798988621997472e-08,
+      "loss": 2.4539,
+      "step": 62828
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.375158027812895e-08,
+      "loss": 1.574,
+      "step": 62829
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3704171934260427e-08,
+      "loss": 2.1002,
+      "step": 62830
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.365676359039191e-08,
+      "loss": 0.5537,
+      "step": 62831
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3609355246523388e-08,
+      "loss": 1.442,
+      "step": 62832
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.356194690265487e-08,
+      "loss": 1.1466,
+      "step": 62833
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.351453855878635e-08,
+      "loss": 3.2034,
+      "step": 62834
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3467130214917825e-08,
+      "loss": 1.5076,
+      "step": 62835
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3419721871049304e-08,
+      "loss": 1.4555,
+      "step": 62836
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3372313527180786e-08,
+      "loss": 3.3463,
+      "step": 62837
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3324905183312265e-08,
+      "loss": 0.554,
+      "step": 62838
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.327749683944374e-08,
+      "loss": 0.9003,
+      "step": 62839
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.3230088495575223e-08,
+      "loss": 2.4475,
+      "step": 62840
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.31826801517067e-08,
+      "loss": 1.0538,
+      "step": 62841
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.313527180783818e-08,
+      "loss": 2.4372,
+      "step": 62842
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.308786346396966e-08,
+      "loss": 1.3555,
+      "step": 62843
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.304045512010114e-08,
+      "loss": 1.4213,
+      "step": 62844
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2993046776232617e-08,
+      "loss": 1.5484,
+      "step": 62845
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2945638432364096e-08,
+      "loss": 0.7646,
+      "step": 62846
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2898230088495575e-08,
+      "loss": 1.3194,
+      "step": 62847
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2850821744627054e-08,
+      "loss": 1.3497,
+      "step": 62848
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2803413400758536e-08,
+      "loss": 2.3736,
+      "step": 62849
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2756005056890012e-08,
+      "loss": 2.4085,
+      "step": 62850
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.270859671302149e-08,
+      "loss": 1.3488,
+      "step": 62851
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.2661188369152973e-08,
+      "loss": 1.0287,
+      "step": 62852
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.261378002528445e-08,
+      "loss": 1.1022,
+      "step": 62853
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.256637168141593e-08,
+      "loss": 2.7717,
+      "step": 62854
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.251896333754741e-08,
+      "loss": 1.8802,
+      "step": 62855
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.247155499367889e-08,
+      "loss": 2.5375,
+      "step": 62856
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.2424146649810368e-08,
+      "loss": 1.1571,
+      "step": 62857
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.2376738305941847e-08,
+      "loss": 1.3244,
+      "step": 62858
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.2329329962073326e-08,
+      "loss": 1.5772,
+      "step": 62859
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.2281921618204805e-08,
+      "loss": 3.872,
+      "step": 62860
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.2234513274336284e-08,
+      "loss": 0.5746,
+      "step": 62861
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.2187104930467763e-08,
+      "loss": 2.0049,
+      "step": 62862
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.213969658659924e-08,
+      "loss": 1.4279,
+      "step": 62863
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.209228824273072e-08,
+      "loss": 1.05,
+      "step": 62864
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.20448798988622e-08,
+      "loss": 1.4133,
+      "step": 62865
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1997471554993682e-08,
+      "loss": 0.2523,
+      "step": 62866
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1950063211125157e-08,
+      "loss": 1.6018,
+      "step": 62867
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1902654867256636e-08,
+      "loss": 1.184,
+      "step": 62868
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.185524652338812e-08,
+      "loss": 0.7251,
+      "step": 62869
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1807838179519597e-08,
+      "loss": 1.899,
+      "step": 62870
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1760429835651073e-08,
+      "loss": 0.9116,
+      "step": 62871
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1713021491782555e-08,
+      "loss": 0.7922,
+      "step": 62872
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1665613147914034e-08,
+      "loss": 0.2892,
+      "step": 62873
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.161820480404551e-08,
+      "loss": 1.1897,
+      "step": 62874
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1570796460176992e-08,
+      "loss": 1.4187,
+      "step": 62875
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.152338811630847e-08,
+      "loss": 1.4439,
+      "step": 62876
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.147597977243995e-08,
+      "loss": 1.5226,
+      "step": 62877
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.142857142857143e-08,
+      "loss": 1.4967,
+      "step": 62878
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1381163084702908e-08,
+      "loss": 0.955,
+      "step": 62879
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1333754740834387e-08,
+      "loss": 0.7882,
+      "step": 62880
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1286346396965866e-08,
+      "loss": 2.323,
+      "step": 62881
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1238938053097345e-08,
+      "loss": 1.248,
+      "step": 62882
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1191529709228824e-08,
+      "loss": 2.748,
+      "step": 62883
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1144121365360306e-08,
+      "loss": 0.76,
+      "step": 62884
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.109671302149178e-08,
+      "loss": 2.2847,
+      "step": 62885
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1049304677623264e-08,
+      "loss": 2.3224,
+      "step": 62886
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.1001896333754743e-08,
+      "loss": 1.3807,
+      "step": 62887
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.095448798988622e-08,
+      "loss": 2.2859,
+      "step": 62888
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.09070796460177e-08,
+      "loss": 1.3293,
+      "step": 62889
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.085967130214918e-08,
+      "loss": 2.3342,
+      "step": 62890
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.081226295828066e-08,
+      "loss": 1.7553,
+      "step": 62891
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0764854614412138e-08,
+      "loss": 1.7297,
+      "step": 62892
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0717446270543616e-08,
+      "loss": 3.7425,
+      "step": 62893
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0670037926675095e-08,
+      "loss": 2.5692,
+      "step": 62894
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0622629582806574e-08,
+      "loss": 1.5578,
+      "step": 62895
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0575221238938053e-08,
+      "loss": 2.1387,
+      "step": 62896
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0527812895069532e-08,
+      "loss": 1.9013,
+      "step": 62897
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0480404551201014e-08,
+      "loss": 2.334,
+      "step": 62898
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.043299620733249e-08,
+      "loss": 2.5408,
+      "step": 62899
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.038558786346397e-08,
+      "loss": 3.8486,
+      "step": 62900
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.033817951959545e-08,
+      "loss": 1.7772,
+      "step": 62901
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0290771175726927e-08,
+      "loss": 2.6088,
+      "step": 62902
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0243362831858406e-08,
+      "loss": 1.5467,
+      "step": 62903
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0195954487989888e-08,
+      "loss": 1.4666,
+      "step": 62904
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0148546144121367e-08,
+      "loss": 1.0234,
+      "step": 62905
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0101137800252846e-08,
+      "loss": 1.4197,
+      "step": 62906
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0053729456384325e-08,
+      "loss": 1.4306,
+      "step": 62907
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.0006321112515804e-08,
+      "loss": 3.2708,
+      "step": 62908
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9958912768647283e-08,
+      "loss": 1.6702,
+      "step": 62909
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9911504424778762e-08,
+      "loss": 1.1159,
+      "step": 62910
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.986409608091024e-08,
+      "loss": 2.9851,
+      "step": 62911
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.981668773704172e-08,
+      "loss": 2.4639,
+      "step": 62912
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.97692793931732e-08,
+      "loss": 2.2644,
+      "step": 62913
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9721871049304678e-08,
+      "loss": 2.7716,
+      "step": 62914
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9674462705436156e-08,
+      "loss": 1.439,
+      "step": 62915
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9627054361567635e-08,
+      "loss": 2.354,
+      "step": 62916
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9579646017699114e-08,
+      "loss": 1.4735,
+      "step": 62917
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.9532237673830597e-08,
+      "loss": 1.2011,
+      "step": 62918
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9484829329962076e-08,
+      "loss": 1.6628,
+      "step": 62919
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.943742098609355e-08,
+      "loss": 2.1861,
+      "step": 62920
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9390012642225033e-08,
+      "loss": 2.3298,
+      "step": 62921
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9342604298356512e-08,
+      "loss": 1.3166,
+      "step": 62922
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9295195954487988e-08,
+      "loss": 1.2324,
+      "step": 62923
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.924778761061947e-08,
+      "loss": 1.272,
+      "step": 62924
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.920037926675095e-08,
+      "loss": 3.6825,
+      "step": 62925
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9152970922882428e-08,
+      "loss": 1.1792,
+      "step": 62926
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9105562579013907e-08,
+      "loss": 1.7236,
+      "step": 62927
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9058154235145386e-08,
+      "loss": 1.3166,
+      "step": 62928
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.9010745891276865e-08,
+      "loss": 2.2213,
+      "step": 62929
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8963337547408344e-08,
+      "loss": 1.3333,
+      "step": 62930
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8915929203539823e-08,
+      "loss": 1.8703,
+      "step": 62931
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8868520859671302e-08,
+      "loss": 2.3188,
+      "step": 62932
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8821112515802784e-08,
+      "loss": 1.602,
+      "step": 62933
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.877370417193426e-08,
+      "loss": 1.3303,
+      "step": 62934
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.872629582806574e-08,
+      "loss": 1.6287,
+      "step": 62935
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.867888748419722e-08,
+      "loss": 1.7567,
+      "step": 62936
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8631479140328697e-08,
+      "loss": 1.8818,
+      "step": 62937
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.858407079646018e-08,
+      "loss": 0.4173,
+      "step": 62938
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8536662452591658e-08,
+      "loss": 1.1174,
+      "step": 62939
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8489254108723137e-08,
+      "loss": 1.3812,
+      "step": 62940
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8441845764854616e-08,
+      "loss": 1.252,
+      "step": 62941
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8394437420986095e-08,
+      "loss": 2.2817,
+      "step": 62942
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8347029077117573e-08,
+      "loss": 3.4036,
+      "step": 62943
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8299620733249052e-08,
+      "loss": 1.7333,
+      "step": 62944
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.825221238938053e-08,
+      "loss": 1.0814,
+      "step": 62945
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.820480404551201e-08,
+      "loss": 1.7533,
+      "step": 62946
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8157395701643493e-08,
+      "loss": 1.8397,
+      "step": 62947
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8109987357774968e-08,
+      "loss": 0.4666,
+      "step": 62948
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.8062579013906447e-08,
+      "loss": 1.8286,
+      "step": 62949
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.801517067003793e-08,
+      "loss": 1.4426,
+      "step": 62950
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7967762326169405e-08,
+      "loss": 4.8862,
+      "step": 62951
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7920353982300884e-08,
+      "loss": 1.7898,
+      "step": 62952
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7872945638432366e-08,
+      "loss": 3.4705,
+      "step": 62953
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7825537294563845e-08,
+      "loss": 2.1508,
+      "step": 62954
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.777812895069532e-08,
+      "loss": 2.5931,
+      "step": 62955
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7730720606826803e-08,
+      "loss": 1.9171,
+      "step": 62956
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7683312262958282e-08,
+      "loss": 2.2612,
+      "step": 62957
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.763590391908976e-08,
+      "loss": 1.042,
+      "step": 62958
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.758849557522124e-08,
+      "loss": 2.1295,
+      "step": 62959
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.754108723135272e-08,
+      "loss": 1.8578,
+      "step": 62960
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7493678887484198e-08,
+      "loss": 1.7346,
+      "step": 62961
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7446270543615677e-08,
+      "loss": 1.4414,
+      "step": 62962
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7398862199747156e-08,
+      "loss": 1.5932,
+      "step": 62963
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7351453855878635e-08,
+      "loss": 1.8797,
+      "step": 62964
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7304045512010113e-08,
+      "loss": 0.7932,
+      "step": 62965
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7256637168141592e-08,
+      "loss": 0.4169,
+      "step": 62966
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.720922882427307e-08,
+      "loss": 1.555,
+      "step": 62967
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.7161820480404554e-08,
+      "loss": 2.7219,
+      "step": 62968
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.711441213653603e-08,
+      "loss": 1.0695,
+      "step": 62969
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.706700379266751e-08,
+      "loss": 0.9284,
+      "step": 62970
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.701959544879899e-08,
+      "loss": 1.8148,
+      "step": 62971
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6972187104930466e-08,
+      "loss": 1.9261,
+      "step": 62972
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6924778761061948e-08,
+      "loss": 1.9063,
+      "step": 62973
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6877370417193427e-08,
+      "loss": 3.2496,
+      "step": 62974
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6829962073324906e-08,
+      "loss": 2.1214,
+      "step": 62975
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6782553729456385e-08,
+      "loss": 1.3251,
+      "step": 62976
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6735145385587864e-08,
+      "loss": 0.9937,
+      "step": 62977
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6687737041719343e-08,
+      "loss": 1.8046,
+      "step": 62978
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.6640328697850822e-08,
+      "loss": 2.0503,
+      "step": 62979
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.65929203539823e-08,
+      "loss": 1.1184,
+      "step": 62980
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.654551201011378e-08,
+      "loss": 1.7535,
+      "step": 62981
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6498103666245262e-08,
+      "loss": 3.577,
+      "step": 62982
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6450695322376738e-08,
+      "loss": 1.1633,
+      "step": 62983
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6403286978508217e-08,
+      "loss": 3.9392,
+      "step": 62984
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.63558786346397e-08,
+      "loss": 1.3094,
+      "step": 62985
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6308470290771175e-08,
+      "loss": 1.4756,
+      "step": 62986
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6261061946902654e-08,
+      "loss": 2.0198,
+      "step": 62987
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6213653603034136e-08,
+      "loss": 1.9622,
+      "step": 62988
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6166245259165615e-08,
+      "loss": 1.9187,
+      "step": 62989
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6118836915297094e-08,
+      "loss": 1.425,
+      "step": 62990
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.6071428571428573e-08,
+      "loss": 1.4496,
+      "step": 62991
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.602402022756005e-08,
+      "loss": 2.2711,
+      "step": 62992
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.597661188369153e-08,
+      "loss": 0.5729,
+      "step": 62993
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.592920353982301e-08,
+      "loss": 1.9228,
+      "step": 62994
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.588179519595449e-08,
+      "loss": 2.089,
+      "step": 62995
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5834386852085967e-08,
+      "loss": 1.051,
+      "step": 62996
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5786978508217446e-08,
+      "loss": 1.4589,
+      "step": 62997
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5739570164348925e-08,
+      "loss": 2.1174,
+      "step": 62998
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5692161820480407e-08,
+      "loss": 2.7169,
+      "step": 62999
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5644753476611883e-08,
+      "loss": 3.3388,
+      "step": 63000
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5597345132743362e-08,
+      "loss": 2.798,
+      "step": 63001
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5549936788874844e-08,
+      "loss": 0.8616,
+      "step": 63002
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5502528445006323e-08,
+      "loss": 2.4215,
+      "step": 63003
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.54551201011378e-08,
+      "loss": 1.5145,
+      "step": 63004
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.540771175726928e-08,
+      "loss": 2.0073,
+      "step": 63005
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.536030341340076e-08,
+      "loss": 2.8361,
+      "step": 63006
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5312895069532236e-08,
+      "loss": 0.8967,
+      "step": 63007
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5265486725663718e-08,
+      "loss": 1.059,
+      "step": 63008
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5218078381795197e-08,
+      "loss": 1.6226,
+      "step": 63009
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5170670037926676e-08,
+      "loss": 1.7498,
+      "step": 63010
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5123261694058155e-08,
+      "loss": 1.2124,
+      "step": 63011
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5075853350189634e-08,
+      "loss": 1.9091,
+      "step": 63012
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.5028445006321113e-08,
+      "loss": 0.3746,
+      "step": 63013
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.498103666245259e-08,
+      "loss": 1.2587,
+      "step": 63014
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.493362831858407e-08,
+      "loss": 1.6426,
+      "step": 63015
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4886219974715551e-08,
+      "loss": 0.9318,
+      "step": 63016
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.483881163084703e-08,
+      "loss": 1.3188,
+      "step": 63017
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4791403286978507e-08,
+      "loss": 2.7856,
+      "step": 63018
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4743994943109988e-08,
+      "loss": 1.4674,
+      "step": 63019
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4696586599241469e-08,
+      "loss": 1.9544,
+      "step": 63020
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4649178255372946e-08,
+      "loss": 2.1809,
+      "step": 63021
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4601769911504425e-08,
+      "loss": 2.7256,
+      "step": 63022
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4554361567635905e-08,
+      "loss": 2.3401,
+      "step": 63023
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4506953223767384e-08,
+      "loss": 2.3337,
+      "step": 63024
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4459544879898862e-08,
+      "loss": 1.4782,
+      "step": 63025
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4412136536030342e-08,
+      "loss": 1.3475,
+      "step": 63026
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4364728192161821e-08,
+      "loss": 1.7875,
+      "step": 63027
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4317319848293298e-08,
+      "loss": 1.483,
+      "step": 63028
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4269911504424779e-08,
+      "loss": 2.7175,
+      "step": 63029
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.422250316055626e-08,
+      "loss": 2.2134,
+      "step": 63030
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4175094816687739e-08,
+      "loss": 0.5965,
+      "step": 63031
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4127686472819216e-08,
+      "loss": 2.8228,
+      "step": 63032
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4080278128950696e-08,
+      "loss": 2.8425,
+      "step": 63033
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.4032869785082175e-08,
+      "loss": 2.2449,
+      "step": 63034
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3985461441213653e-08,
+      "loss": 1.1753,
+      "step": 63035
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3938053097345133e-08,
+      "loss": 2.3341,
+      "step": 63036
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3890644753476612e-08,
+      "loss": 2.0084,
+      "step": 63037
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3843236409608093e-08,
+      "loss": 0.5203,
+      "step": 63038
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.379582806573957e-08,
+      "loss": 2.0233,
+      "step": 63039
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.374841972187105e-08,
+      "loss": 1.3083,
+      "step": 63040
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.370101137800253e-08,
+      "loss": 1.5954,
+      "step": 63041
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3653603034134007e-08,
+      "loss": 2.1865,
+      "step": 63042
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3606194690265487e-08,
+      "loss": 1.2353,
+      "step": 63043
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3558786346396966e-08,
+      "loss": 3.5607,
+      "step": 63044
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.3511378002528447e-08,
+      "loss": 2.0334,
+      "step": 63045
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3463969658659924e-08,
+      "loss": 1.441,
+      "step": 63046
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3416561314791403e-08,
+      "loss": 1.6178,
+      "step": 63047
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3369152970922884e-08,
+      "loss": 1.3523,
+      "step": 63048
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3321744627054361e-08,
+      "loss": 1.965,
+      "step": 63049
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.327433628318584e-08,
+      "loss": 2.6653,
+      "step": 63050
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.322692793931732e-08,
+      "loss": 1.9015,
+      "step": 63051
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3179519595448798e-08,
+      "loss": 2.2127,
+      "step": 63052
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3132111251580279e-08,
+      "loss": 3.1388,
+      "step": 63053
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3084702907711757e-08,
+      "loss": 1.2014,
+      "step": 63054
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.3037294563843238e-08,
+      "loss": 0.7274,
+      "step": 63055
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2989886219974715e-08,
+      "loss": 1.6503,
+      "step": 63056
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2942477876106194e-08,
+      "loss": 2.4527,
+      "step": 63057
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2895069532237675e-08,
+      "loss": 1.0072,
+      "step": 63058
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2847661188369152e-08,
+      "loss": 1.7963,
+      "step": 63059
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2800252844500631e-08,
+      "loss": 2.0694,
+      "step": 63060
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2752844500632112e-08,
+      "loss": 3.2702,
+      "step": 63061
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2705436156763592e-08,
+      "loss": 1.7068,
+      "step": 63062
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.265802781289507e-08,
+      "loss": 1.6609,
+      "step": 63063
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2610619469026549e-08,
+      "loss": 1.0514,
+      "step": 63064
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2563211125158029e-08,
+      "loss": 1.4502,
+      "step": 63065
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2515802781289506e-08,
+      "loss": 3.4032,
+      "step": 63066
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2468394437420985e-08,
+      "loss": 2.1991,
+      "step": 63067
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2420986093552466e-08,
+      "loss": 0.9458,
+      "step": 63068
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2373577749683945e-08,
+      "loss": 1.625,
+      "step": 63069
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2326169405815422e-08,
+      "loss": 2.7591,
+      "step": 63070
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2278761061946903e-08,
+      "loss": 1.6411,
+      "step": 63071
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.2231352718078383e-08,
+      "loss": 2.3154,
+      "step": 63072
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.218394437420986e-08,
+      "loss": 3.718,
+      "step": 63073
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.213653603034134e-08,
+      "loss": 2.4601,
+      "step": 63074
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.208912768647282e-08,
+      "loss": 2.4784,
+      "step": 63075
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.20417193426043e-08,
+      "loss": 0.3797,
+      "step": 63076
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1994310998735776e-08,
+      "loss": 0.9179,
+      "step": 63077
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1946902654867257e-08,
+      "loss": 1.8583,
+      "step": 63078
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1899494310998736e-08,
+      "loss": 1.0952,
+      "step": 63079
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1852085967130213e-08,
+      "loss": 1.156,
+      "step": 63080
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1804677623261694e-08,
+      "loss": 1.1708,
+      "step": 63081
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1757269279393174e-08,
+      "loss": 2.3279,
+      "step": 63082
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1709860935524652e-08,
+      "loss": 2.1934,
+      "step": 63083
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1662452591656132e-08,
+      "loss": 2.5597,
+      "step": 63084
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1615044247787611e-08,
+      "loss": 2.68,
+      "step": 63085
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.156763590391909e-08,
+      "loss": 1.3242,
+      "step": 63086
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.152022756005057e-08,
+      "loss": 2.5929,
+      "step": 63087
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1472819216182048e-08,
+      "loss": 2.5622,
+      "step": 63088
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1425410872313527e-08,
+      "loss": 3.4539,
+      "step": 63089
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1378002528445006e-08,
+      "loss": 2.2046,
+      "step": 63090
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1330594184576487e-08,
+      "loss": 2.0796,
+      "step": 63091
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1283185840707966e-08,
+      "loss": 1.8643,
+      "step": 63092
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1235777496839444e-08,
+      "loss": 1.5079,
+      "step": 63093
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1188369152970923e-08,
+      "loss": 1.478,
+      "step": 63094
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1140960809102402e-08,
+      "loss": 3.0362,
+      "step": 63095
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.1093552465233881e-08,
+      "loss": 0.7505,
+      "step": 63096
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.104614412136536e-08,
+      "loss": 1.3145,
+      "step": 63097
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0998735777496841e-08,
+      "loss": 3.5108,
+      "step": 63098
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0951327433628318e-08,
+      "loss": 2.4438,
+      "step": 63099
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0903919089759799e-08,
+      "loss": 2.3631,
+      "step": 63100
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0856510745891278e-08,
+      "loss": 1.195,
+      "step": 63101
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0809102402022755e-08,
+      "loss": 2.2911,
+      "step": 63102
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0761694058154236e-08,
+      "loss": 1.757,
+      "step": 63103
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0714285714285715e-08,
+      "loss": 2.5716,
+      "step": 63104
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0666877370417193e-08,
+      "loss": 1.5489,
+      "step": 63105
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0619469026548672e-08,
+      "loss": 1.6982,
+      "step": 63106
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0572060682680153e-08,
+      "loss": 1.3863,
+      "step": 63107
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 1.0524652338811632e-08,
+      "loss": 2.389,
+      "step": 63108
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.047724399494311e-08,
+      "loss": 1.8131,
+      "step": 63109
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.042983565107459e-08,
+      "loss": 2.165,
+      "step": 63110
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0382427307206069e-08,
+      "loss": 0.282,
+      "step": 63111
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0335018963337548e-08,
+      "loss": 1.7636,
+      "step": 63112
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0287610619469027e-08,
+      "loss": 2.0644,
+      "step": 63113
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0240202275600507e-08,
+      "loss": 1.5456,
+      "step": 63114
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0192793931731985e-08,
+      "loss": 0.928,
+      "step": 63115
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0145385587863463e-08,
+      "loss": 2.0866,
+      "step": 63116
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0097977243994944e-08,
+      "loss": 2.2822,
+      "step": 63117
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0050568900126423e-08,
+      "loss": 2.3183,
+      "step": 63118
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 1.0003160556257902e-08,
+      "loss": 2.943,
+      "step": 63119
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.955752212389381e-09,
+      "loss": 2.2114,
+      "step": 63120
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.90834386852086e-09,
+      "loss": 1.1592,
+      "step": 63121
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.860935524652339e-09,
+      "loss": 0.715,
+      "step": 63122
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.813527180783818e-09,
+      "loss": 1.9035,
+      "step": 63123
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.766118836915298e-09,
+      "loss": 2.8324,
+      "step": 63124
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.718710493046776e-09,
+      "loss": 2.1925,
+      "step": 63125
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.671302149178256e-09,
+      "loss": 1.876,
+      "step": 63126
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.623893805309735e-09,
+      "loss": 2.4486,
+      "step": 63127
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.576485461441214e-09,
+      "loss": 1.2402,
+      "step": 63128
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.529077117572693e-09,
+      "loss": 1.5034,
+      "step": 63129
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.481668773704172e-09,
+      "loss": 2.3872,
+      "step": 63130
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.434260429835651e-09,
+      "loss": 1.9104,
+      "step": 63131
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.38685208596713e-09,
+      "loss": 1.857,
+      "step": 63132
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.33944374209861e-09,
+      "loss": 1.7808,
+      "step": 63133
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.29203539823009e-09,
+      "loss": 1.7389,
+      "step": 63134
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.244627054361568e-09,
+      "loss": 0.8591,
+      "step": 63135
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.197218710493047e-09,
+      "loss": 2.3491,
+      "step": 63136
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.149810366624526e-09,
+      "loss": 1.8768,
+      "step": 63137
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.102402022756005e-09,
+      "loss": 2.9044,
+      "step": 63138
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.054993678887484e-09,
+      "loss": 0.9829,
+      "step": 63139
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.007585335018965e-09,
+      "loss": 1.1002,
+      "step": 63140
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.960176991150442e-09,
+      "loss": 0.9784,
+      "step": 63141
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.912768647281923e-09,
+      "loss": 1.4829,
+      "step": 63142
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.865360303413402e-09,
+      "loss": 1.631,
+      "step": 63143
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.81795195954488e-09,
+      "loss": 0.1839,
+      "step": 63144
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.77054361567636e-09,
+      "loss": 1.5728,
+      "step": 63145
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.723135271807838e-09,
+      "loss": 1.9972,
+      "step": 63146
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.675726927939317e-09,
+      "loss": 1.6189,
+      "step": 63147
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.628318584070796e-09,
+      "loss": 1.4823,
+      "step": 63148
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.580910240202277e-09,
+      "loss": 1.3508,
+      "step": 63149
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.533501896333756e-09,
+      "loss": 1.9155,
+      "step": 63150
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.486093552465233e-09,
+      "loss": 1.3348,
+      "step": 63151
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.438685208596714e-09,
+      "loss": 1.8546,
+      "step": 63152
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.391276864728193e-09,
+      "loss": 3.8126,
+      "step": 63153
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.343868520859672e-09,
+      "loss": 1.8346,
+      "step": 63154
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.29646017699115e-09,
+      "loss": 1.4321,
+      "step": 63155
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.249051833122631e-09,
+      "loss": 0.8984,
+      "step": 63156
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.201643489254108e-09,
+      "loss": 1.5,
+      "step": 63157
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.154235145385587e-09,
+      "loss": 1.3781,
+      "step": 63158
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.106826801517068e-09,
+      "loss": 2.5019,
+      "step": 63159
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.059418457648547e-09,
+      "loss": 1.7659,
+      "step": 63160
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 8.012010113780026e-09,
+      "loss": 2.043,
+      "step": 63161
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.964601769911505e-09,
+      "loss": 2.2969,
+      "step": 63162
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.917193426042984e-09,
+      "loss": 2.5672,
+      "step": 63163
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.869785082174463e-09,
+      "loss": 1.8164,
+      "step": 63164
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.822376738305942e-09,
+      "loss": 2.4605,
+      "step": 63165
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.774968394437422e-09,
+      "loss": 1.8983,
+      "step": 63166
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.7275600505689e-09,
+      "loss": 1.1291,
+      "step": 63167
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.68015170670038e-09,
+      "loss": 1.2179,
+      "step": 63168
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.632743362831859e-09,
+      "loss": 1.3297,
+      "step": 63169
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.585335018963338e-09,
+      "loss": 0.6329,
+      "step": 63170
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 7.537926675094817e-09,
+      "loss": 0.7677,
+      "step": 63171
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.490518331226296e-09,
+      "loss": 1.718,
+      "step": 63172
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.4431099873577755e-09,
+      "loss": 0.4286,
+      "step": 63173
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.395701643489254e-09,
+      "loss": 1.0256,
+      "step": 63174
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.348293299620734e-09,
+      "loss": 2.7535,
+      "step": 63175
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.300884955752212e-09,
+      "loss": 2.3854,
+      "step": 63176
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.253476611883692e-09,
+      "loss": 1.655,
+      "step": 63177
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.206068268015171e-09,
+      "loss": 1.7637,
+      "step": 63178
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.158659924146649e-09,
+      "loss": 2.4548,
+      "step": 63179
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.11125158027813e-09,
+      "loss": 1.9185,
+      "step": 63180
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.063843236409608e-09,
+      "loss": 1.7352,
+      "step": 63181
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 7.016434892541088e-09,
+      "loss": 1.5402,
+      "step": 63182
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.969026548672567e-09,
+      "loss": 0.5797,
+      "step": 63183
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.921618204804046e-09,
+      "loss": 1.8456,
+      "step": 63184
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.874209860935525e-09,
+      "loss": 1.3621,
+      "step": 63185
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.8268015170670034e-09,
+      "loss": 2.612,
+      "step": 63186
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.779393173198483e-09,
+      "loss": 2.1855,
+      "step": 63187
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.731984829329962e-09,
+      "loss": 0.4139,
+      "step": 63188
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.684576485461442e-09,
+      "loss": 2.5317,
+      "step": 63189
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.63716814159292e-09,
+      "loss": 3.2524,
+      "step": 63190
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.589759797724399e-09,
+      "loss": 1.7039,
+      "step": 63191
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.542351453855879e-09,
+      "loss": 3.202,
+      "step": 63192
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.494943109987358e-09,
+      "loss": 1.347,
+      "step": 63193
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.4475347661188375e-09,
+      "loss": 1.6123,
+      "step": 63194
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.4001264222503156e-09,
+      "loss": 1.1212,
+      "step": 63195
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.352718078381796e-09,
+      "loss": 3.5941,
+      "step": 63196
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.305309734513274e-09,
+      "loss": 0.874,
+      "step": 63197
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.257901390644753e-09,
+      "loss": 2.491,
+      "step": 63198
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.210493046776233e-09,
+      "loss": 1.8958,
+      "step": 63199
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.163084702907711e-09,
+      "loss": 1.863,
+      "step": 63200
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.115676359039192e-09,
+      "loss": 2.3875,
+      "step": 63201
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.06826801517067e-09,
+      "loss": 1.0851,
+      "step": 63202
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 6.02085967130215e-09,
+      "loss": 2.2227,
+      "step": 63203
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.9734513274336285e-09,
+      "loss": 2.1772,
+      "step": 63204
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.926042983565107e-09,
+      "loss": 1.3128,
+      "step": 63205
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.878634639696587e-09,
+      "loss": 2.2992,
+      "step": 63206
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.831226295828066e-09,
+      "loss": 1.9146,
+      "step": 63207
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.783817951959545e-09,
+      "loss": 1.7529,
+      "step": 63208
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.736409608091024e-09,
+      "loss": 2.0809,
+      "step": 63209
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.689001264222503e-09,
+      "loss": 2.1029,
+      "step": 63210
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.641592920353983e-09,
+      "loss": 2.5299,
+      "step": 63211
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.594184576485462e-09,
+      "loss": 2.843,
+      "step": 63212
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.546776232616941e-09,
+      "loss": 2.1021,
+      "step": 63213
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.4993678887484204e-09,
+      "loss": 0.9389,
+      "step": 63214
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.451959544879899e-09,
+      "loss": 1.6679,
+      "step": 63215
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.4045512010113775e-09,
+      "loss": 2.5506,
+      "step": 63216
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.357142857142857e-09,
+      "loss": 2.6185,
+      "step": 63217
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.309734513274336e-09,
+      "loss": 0.8915,
+      "step": 63218
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.262326169405816e-09,
+      "loss": 2.1274,
+      "step": 63219
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.214917825537295e-09,
+      "loss": 2.3695,
+      "step": 63220
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.167509481668774e-09,
+      "loss": 1.552,
+      "step": 63221
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.120101137800254e-09,
+      "loss": 2.6141,
+      "step": 63222
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.072692793931732e-09,
+      "loss": 2.3696,
+      "step": 63223
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 5.0252844500632115e-09,
+      "loss": 2.8162,
+      "step": 63224
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.9778761061946904e-09,
+      "loss": 0.8415,
+      "step": 63225
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.930467762326169e-09,
+      "loss": 1.1185,
+      "step": 63226
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.883059418457649e-09,
+      "loss": 1.6307,
+      "step": 63227
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.835651074589128e-09,
+      "loss": 2.5888,
+      "step": 63228
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.788242730720607e-09,
+      "loss": 0.9309,
+      "step": 63229
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.740834386852086e-09,
+      "loss": 2.8859,
+      "step": 63230
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.693426042983565e-09,
+      "loss": 2.4322,
+      "step": 63231
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.646017699115045e-09,
+      "loss": 1.021,
+      "step": 63232
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.598609355246524e-09,
+      "loss": 1.5526,
+      "step": 63233
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.5512010113780026e-09,
+      "loss": 2.452,
+      "step": 63234
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.503792667509482e-09,
+      "loss": 1.0611,
+      "step": 63235
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.456384323640961e-09,
+      "loss": 1.577,
+      "step": 63236
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.40897597977244e-09,
+      "loss": 3.2338,
+      "step": 63237
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.361567635903919e-09,
+      "loss": 1.2769,
+      "step": 63238
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.314159292035398e-09,
+      "loss": 1.3225,
+      "step": 63239
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.266750948166878e-09,
+      "loss": 1.3845,
+      "step": 63240
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.219342604298357e-09,
+      "loss": 2.197,
+      "step": 63241
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.171934260429836e-09,
+      "loss": 3.4153,
+      "step": 63242
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.1245259165613155e-09,
+      "loss": 1.9171,
+      "step": 63243
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.077117572692794e-09,
+      "loss": 2.1828,
+      "step": 63244
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 4.029709228824273e-09,
+      "loss": 1.4369,
+      "step": 63245
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.982300884955752e-09,
+      "loss": 1.4481,
+      "step": 63246
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.934892541087231e-09,
+      "loss": 1.414,
+      "step": 63247
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.887484197218711e-09,
+      "loss": 2.1112,
+      "step": 63248
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.84007585335019e-09,
+      "loss": 2.8895,
+      "step": 63249
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.792667509481669e-09,
+      "loss": 2.0772,
+      "step": 63250
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.745259165613148e-09,
+      "loss": 1.8755,
+      "step": 63251
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.697850821744627e-09,
+      "loss": 1.0712,
+      "step": 63252
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.650442477876106e-09,
+      "loss": 1.4104,
+      "step": 63253
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.6030341340075855e-09,
+      "loss": 0.628,
+      "step": 63254
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.555625790139065e-09,
+      "loss": 2.8916,
+      "step": 63255
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.508217446270544e-09,
+      "loss": 1.8917,
+      "step": 63256
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.460809102402023e-09,
+      "loss": 2.3805,
+      "step": 63257
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.4134007585335017e-09,
+      "loss": 1.6883,
+      "step": 63258
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.365992414664981e-09,
+      "loss": 1.3062,
+      "step": 63259
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.31858407079646e-09,
+      "loss": 2.3742,
+      "step": 63260
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.2711757269279394e-09,
+      "loss": 1.4095,
+      "step": 63261
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.2237673830594187e-09,
+      "loss": 3.3835,
+      "step": 63262
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.176359039190898e-09,
+      "loss": 1.7764,
+      "step": 63263
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.1289506953223766e-09,
+      "loss": 3.4554,
+      "step": 63264
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.0815423514538556e-09,
+      "loss": 4.0898,
+      "step": 63265
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.034134007585335e-09,
+      "loss": 3.2914,
+      "step": 63266
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.9867256637168143e-09,
+      "loss": 0.9911,
+      "step": 63267
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.9393173198482936e-09,
+      "loss": 0.8351,
+      "step": 63268
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.8919089759797726e-09,
+      "loss": 1.7496,
+      "step": 63269
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.8445006321112515e-09,
+      "loss": 1.2958,
+      "step": 63270
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.797092288242731e-09,
+      "loss": 2.279,
+      "step": 63271
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.7496839443742102e-09,
+      "loss": 2.1986,
+      "step": 63272
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.7022756005056887e-09,
+      "loss": 0.5162,
+      "step": 63273
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.654867256637168e-09,
+      "loss": 1.0283,
+      "step": 63274
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.6074589127686475e-09,
+      "loss": 1.7458,
+      "step": 63275
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.560050568900127e-09,
+      "loss": 0.6531,
+      "step": 63276
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.5126422250316058e-09,
+      "loss": 1.8966,
+      "step": 63277
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.4652338811630847e-09,
+      "loss": 2.3281,
+      "step": 63278
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.417825537294564e-09,
+      "loss": 1.4451,
+      "step": 63279
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.370417193426043e-09,
+      "loss": 1.6081,
+      "step": 63280
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.3230088495575223e-09,
+      "loss": 1.6086,
+      "step": 63281
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.2756005056890013e-09,
+      "loss": 2.0986,
+      "step": 63282
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.2281921618204806e-09,
+      "loss": 1.3006,
+      "step": 63283
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.1807838179519596e-09,
+      "loss": 1.4613,
+      "step": 63284
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.133375474083439e-09,
+      "loss": 2.4405,
+      "step": 63285
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.085967130214918e-09,
+      "loss": 1.6871,
+      "step": 63286
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.038558786346397e-09,
+      "loss": 2.0639,
+      "step": 63287
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.991150442477876e-09,
+      "loss": 4.141,
+      "step": 63288
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.9437420986093555e-09,
+      "loss": 1.5327,
+      "step": 63289
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.8963337547408345e-09,
+      "loss": 0.8173,
+      "step": 63290
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.8489254108723134e-09,
+      "loss": 0.7902,
+      "step": 63291
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.8015170670037928e-09,
+      "loss": 1.4029,
+      "step": 63292
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.754108723135272e-09,
+      "loss": 1.1798,
+      "step": 63293
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.7067003792667509e-09,
+      "loss": 0.9581,
+      "step": 63294
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.65929203539823e-09,
+      "loss": 2.358,
+      "step": 63295
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.6118836915297094e-09,
+      "loss": 3.2422,
+      "step": 63296
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.5644753476611883e-09,
+      "loss": 3.3045,
+      "step": 63297
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.5170670037926675e-09,
+      "loss": 2.299,
+      "step": 63298
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.4696586599241468e-09,
+      "loss": 2.4114,
+      "step": 63299
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.4222503160556258e-09,
+      "loss": 2.0888,
+      "step": 63300
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.3748419721871051e-09,
+      "loss": 1.2849,
+      "step": 63301
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.327433628318584e-09,
+      "loss": 2.1935,
+      "step": 63302
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.2800252844500634e-09,
+      "loss": 2.1652,
+      "step": 63303
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.2326169405815423e-09,
+      "loss": 1.4487,
+      "step": 63304
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.1852085967130215e-09,
+      "loss": 1.6594,
+      "step": 63305
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.1378002528445006e-09,
+      "loss": 1.0412,
+      "step": 63306
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.0903919089759798e-09,
+      "loss": 2.2555,
+      "step": 63307
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.042983565107459e-09,
+      "loss": 1.6409,
+      "step": 63308
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.95575221238938e-10,
+      "loss": 2.6471,
+      "step": 63309
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.481668773704172e-10,
+      "loss": 2.2372,
+      "step": 63310
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.007585335018964e-10,
+      "loss": 0.9482,
+      "step": 63311
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 8.533501896333754e-10,
+      "loss": 2.1451,
+      "step": 63312
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 8.059418457648547e-10,
+      "loss": 2.276,
+      "step": 63313
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 7.585335018963337e-10,
+      "loss": 2.1036,
+      "step": 63314
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 7.111251580278129e-10,
+      "loss": 2.8809,
+      "step": 63315
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 6.63716814159292e-10,
+      "loss": 1.8879,
+      "step": 63316
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 6.163084702907712e-10,
+      "loss": 1.2604,
+      "step": 63317
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 5.689001264222503e-10,
+      "loss": 1.5833,
+      "step": 63318
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 5.214917825537295e-10,
+      "loss": 1.4306,
+      "step": 63319
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 4.740834386852086e-10,
+      "loss": 3.2692,
+      "step": 63320
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 4.266750948166877e-10,
+      "loss": 3.6613,
+      "step": 63321
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 3.7926675094816686e-10,
+      "loss": 0.968,
+      "step": 63322
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 3.31858407079646e-10,
+      "loss": 1.7123,
+      "step": 63323
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.8445006321112516e-10,
+      "loss": 0.7514,
+      "step": 63324
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.370417193426043e-10,
+      "loss": 1.3728,
+      "step": 63325
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.8963337547408343e-10,
+      "loss": 0.4197,
+      "step": 63326
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 1.4222503160556258e-10,
+      "loss": 1.1008,
+      "step": 63327
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.481668773704172e-11,
+      "loss": 0.8998,
+      "step": 63328
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 4.740834386852086e-11,
+      "loss": 1.1769,
+      "step": 63329
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0,
+      "loss": 2.0118,
+      "step": 63330
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.5825932504440497,
+      "eval_loss": 1.7745146751403809,
+      "eval_runtime": 163.2682,
+      "eval_samples_per_second": 17.242,
+      "eval_steps_per_second": 4.312,
+      "step": 63330
+    }
+  ],
+  "max_steps": 63330,
+  "num_train_epochs": 10,
+  "total_flos": 8.79290602092102e+18,
+  "trial_name": null,
+  "trial_params": null
+}